language acquisition theories essay

Common Theories of Language Acquisition Essay (Critical Writing)

• To find inspiration for your paper and overcome writer’s block
• As a source of information (ensure proper referencing)
• As a template for you assignment

Literature Review

Behaviourist perspective, innatist perspective, cognitive perspective, works cited.

Today, the most common theories of language acquisition are behaviorist perspective, innatist perspective, and cognitive perspective. Each of these theories incorporates numerous perspectives, smaller theories, approaches, and hypotheses to explain how the first and second language learning process develops, and what the regularities and consistent patterns of this process are. The following literature review aims at observing the main aspects of these theories.

In the book “Language”, Sapir has defined the behaviorist perspective as the theory that views language acquisition as the process of imitation, habit formation, and reinforcement (27). Sapir has also stated that language acquisition occurs through a process of a habit formation (29). Overall, Sapir’s work describes behaviorist perspective as the theory that focuses on the process of first language acquisition by infants who master their first linguistic skills by repeating after their parents. Lightbown and Spada have stated that the behaviorist perspective became the theoretical foundation of the Audiolingual method of second language teaching (58).

Innatist perspective is described by Ellis and Shintani in their book as the theory having its foundation on the principle that humans are born with innate awareness of the principle of Universal Grammar (147). The concept of Chomsky’s Universal Grammar is explained by these scholars as the human cognitive ability to learn grammar intuitively. Krashen’s Acquisition theory agrees with Chomsky’s Universal Grammar on the idea that human language acquisition does not require learning conscious grammatical rules (Gass and Mackey 94).

Comprehensible Input concept implicates that listeners can understand the message of a piece in a foreign language even if they do not understand all words mentioned in it (Gass and Mackey 97). The Natural Order Hypothesis argues that acquisition of grammar structures both in the mother tongue and foreign language learning occurs in a predictable order (Gass and Mackey 95). Monitor hypothesis assumes that the Monitor or the inside Editor helps the new language learner alter one’s utterances based on the learned knowledge (Gass and Mackey 95). Affective Filter hypothesis states that the second language acquisition can be prevented through the system of filters such as boredom, fear, anxiety, and resistance to change (Gass and Mackey 98).

Cognitive perspective is defined by Mitchell, Myles, and Marsden in their book as the theory that states that second language acquisition is a conscious cognitive process that involves the purposeful use of learning strategies (48). The Interaction hypothesis is one of the major theories that falls in the cognitive perspective theory group. It states that second language learning requires the face to face communication with the fluent language speaker. The Noticing hypothesis is another influential theory in this group. It argues that the second language acquisition requires that the learner notices the new language grammar structures before one can proceed to make progress in learning (Larsen-Freeman and Long 67). The role of practice is central according to this group of theories as Larsen-Freeman and Long have stated (68).

In conclusion, the theories of behaviorist perspective, innatist perspective, and cognitive perspective overview the aspect of the language acquisition process. The majority of these theories have provided important theoretical foundation for the development of the modern-day language learning methodologies. This paper has reviewed six scholarly sources to observe the three theories along with the major hypotheses that these theories comprise.

Ellis, Rod, and Natsuko Shintani. Exploring Language Pedagogy through Second Language Acquisition Research . London: Routledge, 2013. Print.

Gass, Susan M., and Alison Mackey. The Routledge Handbook of Second Language Acquisition . London: Routledge, 2013. Print.

Larsen-Freeman, Diane, and Michael Long. An Introduction to Second Language Acquisition Research . London: Routledge, 2014. Print.

Lightbown, Patsy M., and Nina Spada. How Languages are Learned. 4th ed. Oxford: Oxford University Press, 2013. Print.

Mitchell, Rosamond, Florence Myles, and Emma Marsden. Second language Learning Theories . London: Routledge, 2013. Print.

Sapir, Edward. Language , Cambridge: Cambridge University Press, 2014. Print.

• What Exactly Does It Mean To Be A Behaviorist?
• Linguistic Determinism and Linguistic Relativity
• Diverse Cultural Background in the School
• Figurative Language in English Language Learning
• Peer Tutoring and English Language Learning
• Technology in Second Language Acquisition
• Community Speech Norms' Acquisition by Asian Immigrants
• Sociocultural Perspective on Second Language Acquisition
• Chicago (A-D)
• Chicago (N-B)

IvyPanda. (2020, August 25). Common Theories of Language Acquisition. https://ivypanda.com/essays/common-theories-of-language-acquisition/

"Common Theories of Language Acquisition." IvyPanda , 25 Aug. 2020, ivypanda.com/essays/common-theories-of-language-acquisition/.

IvyPanda . (2020) 'Common Theories of Language Acquisition'. 25 August.

IvyPanda . 2020. "Common Theories of Language Acquisition." August 25, 2020. https://ivypanda.com/essays/common-theories-of-language-acquisition/.

1. IvyPanda . "Common Theories of Language Acquisition." August 25, 2020. https://ivypanda.com/essays/common-theories-of-language-acquisition/.

Bibliography

IvyPanda . "Common Theories of Language Acquisition." August 25, 2020. https://ivypanda.com/essays/common-theories-of-language-acquisition/.

Language Acquisition Concept and Theories

Introduction.

One of the most important topics in cognitive studies is language acquisition. A number of theories have attempted to explore the different conceptualization of language as a fundamental uniqueness that separates humans from other animals and non-living things (Pinker& Bloom, 1990). Similarly, Pinker (1994) recognizes language as a vehicle which engineers humans to know other people’s thoughts, and therefore, he reasons that the two (language and thought) are closely related. He adds that when one speaks his/her thoughts, he depicts some language. Else, he notes that a child’s first language is often times learnt well enough in the earlier periods of his life without having to be taught in school. With this astonishment, he believes children language acquisition has received a lot of attention in scholarly circles and debates (Pinker, 1994).

Indeed, accordingly, acquisition of language goes beyond it being interesting, but is an answer to the study of cognitive science. The recognition here is the many facets that language acquisition studies come with. These include Modularity, Human Uniqueness, Language and Thought, and Language and Innateness.

Historically, the scientific study of language and the way it is learnt began in the late 1950s, supposedly the time around which cognitive studies were launched. Pinker observes that the anchor of this was when Noam Chomsky reviewed Skinner’s verbal behavior (Pinker, 1994).

Understanding Language Acquisition

Language acquisition can be understood biologically. The understanding here is that human language came to be based upon the unique adaptations that the body and mind developed during the process of evolution (Pinker& Bloom, 1990).

Language and Evolution

Pinker (2000) begins by indicating that human’s vocal tracks appear to have been modified to respond to the demands of evolution. In addition, this is the basis of speech. Pinker (2000), citing Lieberman (1984) argues that the larynx is at the base of the throats and that the vocal tracts have a sharp right angle bend that creates two independently modified cavities.

The process/course of Language Acquisition

Pink and Bloom (1990) assert that a number of scholars and thinkers alike, have kept diaries of children’s speech for a long time, and it was only later that children’s speech began to be analyzed in developmental psychology. Language acquisition begins at a very early stage in human’s life span. This usually stems initially with Sound Patterns. Pinker notes that within the earliest five years of an individual’s existence, children acquire control of speech musculature and sensitivity to the phonetic distinctions in the maiden mother tongue. In addition, children acquire these skills even before they know or understand any words, and therefore at this stage, they only relate sound to meaning (Kuhl, 1992).

When a child is almost hitting one-year age mark, he slowly begins to muster and understand words, and eventually produce them. Interestingly, at this stage they produce the word in ‘isolation’, that is one word at a time, with this period lasting two to twelve months. The words they produce at this stage are similar the world over and include words such as baba, baby among others, and others such as up, off, eat (Pinker& Bloom, 1990).

At about the time a child is 1 year and 6 months, two changes in language acquisition occurs. One is that there is an increase in vocabulary growth and two is that primitive syntax emerges. When Vocabulary growth increases, the child systematically starts learning “words at a rate of one every two waking hours, and will keep learning that rate or faster through adolescence” (Pinker, 1994). Primitive syntax on the other hand involves ‘two word strings’; examples of such include expressions such as ‘see baby’, ‘more hot‘, among others. These two-word expressions, Pinker notes, are similar the world over; for instance, everywhere, children reject and request for activities and therefore ask about who, what and where (Pinker& Bloom, 1990).

Overall, “these sequences already reflect the language being acquired: in 95% of them, the words are properly ordered” (Ingram, 1989). More interestingly is the fact that before they put the words, they can at this stage fathom a sentence by use of syntax. Notably is the fact that the struggle and output depends on the complexity of the sentence at this stage.

Between the time Children are almost going through year two up to mid of year three of age, language evolves to fluency and blossoms into good grammatical expressions and the reasons for this rapidity is still subject of research to today. At this stage, the length of the sentences that the children produce increase steadily and the number of syntax types increases steadily as well (Pinker& Bloom, 1990).

Pinker (1994), notes that children may differ in language development by a span of 1 year. Regardless, the stages they go through in language development remain the same and many children acquire and can speak complex sentences before their second age. At the stage of grammar explosion, the sentences get longer and more complex, even though at age three children’s may have grammatical challenges of one nature or the other (Pinker, 1994).

Language System and Its maturation

A number of scholars have observed that as language circuits mature in a child’s early years so is language acquisition, i.e. a child masters language development from the initial years of his/her birth and the process continues as the child’s brain develops during his/her life (Pinker, 1994). He notes that it is usually nerve cell degenerate shortly before birth, and it is also during this time that they are allocated to brain. However, he observes that an individuals “head size, brain weight, and thickness of the cerebral cortex” continue to rapidly increase over the first year of birth (Pinker, 1994).

White matter is not fully complete until after the child gets to nine months of age. The emergence of synapses will continue and reach climax when one is between 9 months to 2 years; however, this is usually dependent relative on the brain region. The development process continues and as the synapses wither, adolescence sets, with the individual showing signs of transforming from childhood to adulthood.

What accrues here, accordingly therefore is that perhaps “first words, and grammar require minimum levels of brain size, long-distance connections, or extra synapses, particularly in the language centers of the brain” (Pinker, 1994). In addition, the assumption can also be that these changes are the rationale behind the low ability to learn language overtime as people age over a lifespan (Pinker, 1994).

This probably explains why most people in their adulthood cannot master foreign languages especially in their native accent and especially in the language aspect of phonology, and this is what leads to what is now popularly called referred to as foreign accent. No teaching or amount of correction can usually undo the errors that characterizes ‘foreign accent’. However, as Pinker notes, there exists differences depending on one’s efforts, attitudes, degree of exposure, teaching quality, and sometimes, plain talent. However, there is no empirical evidence that adduces learning of words as people age (Pinker, 1994).

Explaining Language Acquisition: Learnerbility Theory

Several theories have been developed in understanding language acquisition. One such theory is the learnerbility theory. This is a computer mathematical theory of language, which deals with learning procedures for children in acquiring grammar, riding on language evidence and exposure. For instance, a learning procedure is taken as an infinite loop running through endless tings of inputs, which are grammatical as chosen from a particular language. This theory by Gold largely shows that innate knowledge of universal grammar assists in learning (Pullum, 2000).

Language acquisition is a complicated issue that needs an elaborate research and study; indeed, some of the tenets of this issue have been addressed in this paper. It is a very central issue in understanding human growth and development. It captures a number of conceptualizations that relate directly to the Universal versus Context Specific development modules, as well as nature versus nurture controversy. Moreover, attempts to understand language scientifically has brought a number of frustrations, with a number of break thoughts as well. All in all, it is important to note that language acquisition begins from the initial periods of a child’s development and continues as the child grows.

Kuhl, P. K. (1992). Brain Mechanisms in Early Language Acquisition. Neuron Review. Web.

Pinker, S. (1994). Language Leanerbility and Language Development . Cambridge: Havard University Press. Web.

Pinker, S. (2000). Language Acquisition . Massachusetts Institute of Technology. Web.

Pinker, S. & Bloom, P. (1990). Natural language and natural selection . Behavioral and Brain Science. Web.

Pullum, G. (2000) . Learnerbilty . New York. Web.

Cite this paper

• Chicago (N-B)
• Chicago (A-D)

StudyCorgi. (2020, August 25). Language Acquisition Concept and Theories. https://studycorgi.com/language-acquisition-concept-and-theories/

"Language Acquisition Concept and Theories." StudyCorgi , 25 Aug. 2020, studycorgi.com/language-acquisition-concept-and-theories/.

StudyCorgi . (2020) 'Language Acquisition Concept and Theories'. 25 August.

1. StudyCorgi . "Language Acquisition Concept and Theories." August 25, 2020. https://studycorgi.com/language-acquisition-concept-and-theories/.

Bibliography

StudyCorgi . "Language Acquisition Concept and Theories." August 25, 2020. https://studycorgi.com/language-acquisition-concept-and-theories/.

StudyCorgi . 2020. "Language Acquisition Concept and Theories." August 25, 2020. https://studycorgi.com/language-acquisition-concept-and-theories/.

This paper, “Language Acquisition Concept and Theories”, was written and voluntary submitted to our free essay database by a straight-A student. Please ensure you properly reference the paper if you're using it to write your assignment.

Before publication, the StudyCorgi editorial team proofread and checked the paper to make sure it meets the highest standards in terms of grammar, punctuation, style, fact accuracy, copyright issues, and inclusive language. Last updated: August 25, 2020 .

If you are the author of this paper and no longer wish to have it published on StudyCorgi, request the removal . Please use the “ Donate your paper ” form to submit an essay.

Theory of Second Language Acquisition

• Open Access
• First Online: 23 August 2023

Cite this chapter

You have full access to this open access chapter

• Maleika Krüger 2  

4834 Accesses

The scientific field of second language acquisition (SLA), as it emerged in the 1970s, is concerned with the conditions and circumstances in which second and foreign language learning occurs. Although sometimes used synonymously, the terms second language and foreign language describe two different aspects: a second language refers to the official language within the country of residence, which is not a person’s mother tongue. This is, for example, the case for immigrant children who learn the language of their parents’ homeland before or while learning the language of their country of residence as a second language in school or kindergarten.

You have full access to this open access chapter,  Download chapter PDF

The scientific field of second language acquisition (SLA), as it emerged in the 1970s, is concerned with the conditions and circumstances in which second and foreign language learning occurs. Although sometimes used synonymously, the terms second language and foreign language describe two different aspects: a second language refers to the official language within the country of residence, which is not a person’s mother tongue. This is, for example, the case for immigrant children who learn the language of their parents’ homeland before or while learning the language of their country of residence as a second language in school or kindergarten. By contrast, the term foreign language describes a language that is not an official language in the country of residence nor a persons’ mother tongue. A foreign language is usually learned through formal classroom instruction within the educational system (Hasebrink et al., 1997; Olsson, 2016; Sundqvist, 2009a). English is a foreign language in both Germany and Switzerland. Therefore, the focus of this study is English as a foreign language (EFL).

Over the years, SLA has produced a great variety of theoretical frameworks and methodology to cover the broad aspects of the field (Olsson, 2016). It is beyond the scope of this study to provide the reader with a comprehensible overview. In short, the different strands of theory can be subsumed under three main groups: formal properties of language learning, cognitive processes while learning a language, and social aspects of language learning. These three groups are not distinct, as there are various overlaps and interactions. Researchers might draw from one or more theory strands, depending on their research questions (Olsson, 2016). The present study will draw on the cognitive theoretical framework, i.e., the process of learning a foreign language, and the social framework, to investigate unplanned and unprompted language learning through media-related extramural English contacts and the influence of two important social factors on the learning process.

4.1 Incidental Language Learning Through Media-related Extramural English Contacts

As defined above, extramural English contacts are defined as any form of out-of-school contact with English as a foreign language arising from voluntary contact with and the use of authentic English media content. The term does not deny the possibility that learners might be aware of the beneficial effect of these contacts, yet the focus of these contacts lies in the appreciation for the media content or a desire to communicate with others (Sundqvist, 2009a, 2011).

While in contact with authentic media content in such a natural setting, learners will be less concerned with studying underlying rules and principles of a foreign language but will instead be focused on the social nature of the situation, on participation, observation, communication, and understanding (R. Ellis, 2008). As a result, any learning processes that might arise from these situations is most likely characterized by incidental, implicit, or explicit learning processes and will often be an unconscious process, without intent or active learning strategies by the learner (Elley, 1997; R. Ellis, 2008). Such incidental language learning processes are defined as the “[…] by-product of any activity not explicitly geared to […] learning” (Hulstijn, 2001, p. 271). Kekra (2000) also defines it as “unintentional or unplanned learning that results from other activities” (p. 3). Incidental language learning is thus a process “without the conscious intention to commit the element to memory” (Hulstijn, 2013, p. 1). In contrast, intentional learning is defined as “any activity aiming at committing lexical information to memory” (Hulstijn, 2001, p. 271).

These definitions of incidental learning are closely related to the definition of informal learning as provided by Stevens (2010):

“Learning resulting from daily life activities related to work, family or leisure. It is not structured (in terms of learning objectives, learning time or learning support) and typically does not lead to certification. Informal learning may be intentional but in most cases it is non-intentional (or ‘incidental’/random).” (Stevens, 2010, p. 12) .

Both definitions emphasize the subconscious nature of the process, which occurs while a person is engaging in everyday activities. Thus, incidental learning could also be referred to as a language acquisition process , as the term acquisition is commonly used to refer to the subconscious process in which children acquire their mother tongue. Usually, children are not consciously aware of the language acquisition nor the resulting language competences. Instead, they are focused on meaning as they interact with the people around them. As a result, children cannot ‘name the rules’ they have acquired, only that something ‘feels correct’. By contrast, learning usually describes a much more conscious process of committing information to memory (R. Ellis, 2008; Krashen, 1985; Sok, 2014).

Given these definitions, incidental learning could be seen as more closely related to the concept of acquisition, while intentional learning could be seen as being closer related to the definition of learning (R. Ellis, 2008). However, the fact that incidental language learning occurs as a by-product of another activity does not require the complete absence of consciousness (Rieder, 2003). Indeed, even though sometimes used synonymously, the distinction between implicit and explicit learning is not congruent with the distinction between incidental and intentional learning (N. C. Ellis, 1994).

The terms implicit and explicit learning refer to the level of awareness and attention a learner pays towards learning. Implicit learning is defined as “acquisition of knowledge about the underlying structure of a complex stimulus environment by a process which takes place naturally, simply and without conscious operation” (N. C. Ellis, 1994, p. 1). On the other hand, explicit learning is a “more conscious operation where the individual makes and test hypotheses in a search for structure” (N. C. Ellis, 1994, p. 1).

However, unconscious in this sense does not, as is often thought, refer to unintentional behavior, but rather to the fact that something is done without awareness and attention. Explicit learning is thus a conscious process in that learners are aware and pay attention to concept formation and linking. This can occur under instruction (e.g., in a classroom) or by understanding concepts and rules without instruction. On the other hand, implicit learning has a person paying attention to the stimulus but being unaware of the acquisition processes (N. C. Ellis, 1994; R. Ellis, 2008).

The result of explicit and implicit learning is explicit and implicit knowledge, which differ in their degree of awareness of rules and the possibility to verbalize them. Implicit knowledge is procedural and intuitive, while explicit knowledge is declarative and conscious. The former comes with the ability to use the language automatically, while the latter comes with the knowledge of underlying rules and regularities (Olsson, 2016). This is why formal instructions are often seen as crucial for grammar learning in a foreign language, as they explicitly teach grammatical rules and regulations (d'Ydewalle, 2002; d'Ydewalle & van de Poel, 1999). However, things learned implicitly at some point may be reflected upon explicitly at a later point in one’s language learning journey (Olsson, 2016).

In contrast to this distinction, the term consciousness within the framework of incidental and intentional learning usually refers to intentionality. Indeed, the definitions of incidental learning provided above do not exclude awareness of the learning process. The important distinction is that in intentional learning, learners are focused on the linguistic form. In incidental learning, the focus is on the meaning, yet a peripheral focus on form is not denied (R. Ellis, 1999). Incidental learning can therefore include implicit, i.e., unaware, learning processes, as well as explicit learning processes, i.e., processes that take place unintentionally but not without a learners’ awareness or (peripheral) attention, and hypothesis forming (Rieder, 2003; Sok, 2014) Footnote 1 .

For example, learners might engage in reading for pleasure, during which implicit learning processes will occur automatically, but they might also decide to engage in explicit learning processes (i.e., paying attention to form) by looking up an unknown word. In addition, they might actively test new words and phrases in a sentence, thus testing their hypothesis about the meaning (Letchumanan et al., 2015).

While the exact definition of these terms remains a matter of ongoing debate within the research community, and the terms are often used interchangeably (see for an overview for example Hulstijn, 2001, 2002, 2005, 2015; Laufer & Hulstijn, 2001), the present study will stay within the original terminology of the theoretical framework of incidental learning and define it as an unintentional or unplanned process, resulting as a by-product of another activity. This by-product can result from implicit processes but might also be accompanied by explicit processes, during which a person pays at least peripheral attention to certain language forms and engages in hypothesis forming and testing.

While often used within the framework of first language acquisition, the concept of incidental learning can also be related to the field of second and foreign language learning and is generally acknowledged in the research field of psychology and language learning (R. Ellis, 2008; Krashen, 1982). For Chomsky (1968, cited in Elley, 1997) learning a native language is in fact so deeply biologically programmed into the brain that children learn their native tongue simply by being exposed to it. In addition, there is little dispute that, except for the first few thousand most common words, which are usually learned intentionally, the vast majority of the vocabulary is acquired incidentally as a by-product of other activities (Hulstijn, 2001, 2003, 2013). Nagy and Anderson (1984) conclude that it would indeed be impossible to explain high school students’ knowledge of 25,000—50,000 words in their mother tongue otherwise. Most words, phrases, and grammar rules have to be ‘picked up’ from the context while engaging in other activities.

While acquiring a first language is not the same as learning a second or foreign language, some research suggests that the two processes are not that different. Moreover, while explicit instructions within the classroom have been proven to be an effective route to foreign language learning, teachers could simply not include enough vocabulary learning in the classroom to explain some learners’ language proficiency (Rieder, 2003). In his work, Krashen claims that the process of language acquisition is indeed not limited to children learning their first language, as adults do not lose the mental capacity for acquisition. According to him, language acquisition is an autonomous process outside of one’s conscious control, as humans cannot choose not to encode and store the information they encounter (Krashen, 1982). Therefore, his input hypothesis claims that as long as learners are presented with a high amount of comprehensible language input , incidental language learning will take place, even in the absence of explicit instructions and intentional learning activities (Krashen, 1985, 1989). Comprehensible input (i + 1) can derive from spoken words or through media channels (e.g., books, movies) and is input that is just slightly more complex (+1) than a person’s current level of competences (i). Under such conditions, a person can derive unknown words and grammatical structures from the surrounding context and thus acquire higher language competences (Krashen, 1982, 1985, 1989).

The learning process is mediated by a person’s resistance to process the input, i.e., the level of their affective filter , which is any kind of internal resistance to process the input. It functions as a mediator between the language input and the acquisition process. Even if sufficient comprehensible input is available, a high filter might lead to a reduced or total lack of acquisition. Under such circumstances, the information might be understood in the moment, but will not be processed for acquisition. Reasons for a high affective filter are often anxiousness, a lack of motivation or self-confidence. A person’s affective filter is low if one is not afraid of failure and feels self-confident in their role as a language speaker and member of the language community. Krashen suspects the filter to be lowest if a person does indeed forget that they are speaking another language and are instead entirely focused on the message at hand (Krashen, 1982, 1985, 1989).

Given a low enough filter, language acquisition will take place in the language acquisition device of the brain (LAD). According to this theory, learners will naturally progress to continuously higher levels of language competences, as long as they come into contact with enough comprehensible input (Krashen, 1982, 1985, 1989). Consequently, a lack of comprehensible input will slow down or stop this trajectory. This might then lead to fossilization , i.e., the learner will stop short of achieving a native speaker level (Krashen, 1985, p. 43). This can happen in two ways: First, learners might encounter input that is too easy and will not provide learners with new syntax or will only subject them to a limited range of vocabulary. Second, learners might encounter input that is too complex and the input will consequently prove to be too difficult for them to decipher. As a result, students will be unable to understand enough of the content to derive unknown words from the surrounding context. Both situations would result in diminished learning outcomes (Krashen, 1985).

Krashen finds empirical support for his hypothesis not only in children’s first language acquisition but also in several studies that show empirical evidence for incidental learning in second and foreign language learners through input from leisure time reading and free reading programs within the classroom, as well as from listening to stories being read out loud (for a summary see, for example, Krashen, 1989). Further empirical evidence for incidental learning processes from language input in natural settings will be discussed in Section  4.2 .

Despite his influence in the field, Krashen has been criticized for his strong focus on language input, and for ignoring the social nature of language and the importance of output production and interaction for language learning in general and for incidental language learning processes in particular. Other researchers have stressed the importance of social interaction for (incidental) language learning. These theories and studies have often drawn on Vygotsky’s sociocultural theory. According to Vygotksy , humans need social interaction and communication in order to levitate their natural biological mental capacities into higher-order mental functions. Only through interaction are these capacities modified and interwoven with cultural values and meaning. Through this process, individuals gain understanding and control over psychological tools, which helps them to moderate interaction with objects in their surroundings. Written and spoken utterances made in a foreign or second language are such objects of interaction (R. Ellis, 2008; Vygotsky, 1978). According to this theory, learners will not be able to interact directly with a language as the object of their attention at the beginning. Instead, they will rely on external assistance in the form of other-regulation via more advanced speakers or object-regulation via tools (e.g., dictionaries), which act as moderators for the interaction with the object ‘language’. Other-regulation through personal assistance in a verbal interaction can, for example, be provided in the form of waiting (giving the speaker time to think), prompting (repeating words in order to help the speaking person to continue), co - constructing (providing missing words or phrases), and explaining (addressing errors; often in the first language) (R. Ellis, 2008; Vygotsky, 1978). Through this assistance, learners will be able to perform tasks which lie within their zone of proximal development . Vygotsky defines this zone as

“[…] the distance between the actual developmental level as determined by independent problem solving and the level of potential development as determined through problem solving under adult guidance or in collaboration with more capable peers.” (Vygotsky, 1978, p. 86).

Hence, the zone of proximal development lies between tasks a person can already carry out by themselves ( level of actual development ) and tasks that a person could not perform, even if assistance is available (Vygotsky, 1978).

The interaction with another person or an object frees the novice of some of the cognitive load of the task at hand and allows them to reach their goal. At the same time, the interaction will provide them with behavior to imitate and internalize for future use. In time, learners will become able to perform these tasks or activities on their own and will rely less on outside regulation (Dunn & Lantolf, 1998; Lantolf, 2000, 2005, 2011; Swain, 2005; Vygotsky, 1978).

Eventually, language learners will reach a level of proficiency where they no longer need outside assistance and instead become self-regulating in their use of the language. In this state, a person can facilitate their own language resources through private (inner) speech to achieve and execute control over their mental processes and their interaction with the language. The process from other-regulation or object-regulation towards self-regulation is called internalization, and (verbal) communication is the crucial means by which such a process is achieved (R. Ellis, 2008). According to the theory, the highest level of proficiency in any language can thus only be achieved if learners interact with others and produce output as well as take in input (R. Ellis, 2008; Swain, 2000, 2005).

It might be tempting to equate Vygotsky’s Zone of Proximal Development with Krashen’s i + 1. However, as Dunn and Lantolf (1998) have pointed out, the two theories are incommensurable at their core. For Krashen, language learning takes place automatically within a person’s language acquisition device (LAD), given a sufficient amount of input within a person’s i + 1. If the affective filter is low enough and enough comprehensible input is available, language acquisition will be inevitable. As long as enough input is provided, the acquisition curve will be a steady, continuous, and linear process, moving from one stage to the next (Dunn & Lantolf, 1998; Krashen, 1982, 1985, 1989; Lantolf, 2005). Krashen does support a weak interaction hypothesis by acknowledging that dialogue and interaction can help to negotiate meaning and clarification, making input more comprehensible. However, he rejects the idea that output production and interaction are necessary factors for (incidental) language learning. For him, a true interaction hypothesis cannot explain cases in which learners have reached a high level of proficiency without interaction. In fact, he sees the value of interaction not in the amount of language spoken by the learner but in the amount of input provided by the interaction partner (Dunn & Lantolf, 1998; Krashen, 1982, 1985, 1989).

On the other hand, Vygotsky rejects the idea of an autonomous individual acquiring a language through an automatic cognitive process. Instead, he states that language development results from humans constantly developing to a higher state of control over their own mental activities by using the assistance of others or objects. Language development is thus not a linear but rather a historical process, rooted in a social context and acquired through interaction and imitation. As a result, interactional and material circumstances shape the form and outcome of each individual process (Dunn & Lantolf, 1998; Lantolf, 2005).

Dunn and Lantolf (1998) concluded that trying to converge the two theories means reading into Krashen something that is not there and taking the interactive core out of Vygotsky. Instead, they call for an acceptance of this incommensurability and peaceful coexistence, dialogue, and appreciation for their individual contributions to the field (Dunn & Lantolf, 1998). Following this call for dialogue, this study will draw on both theories in order to explain incidental language learning in the context of media-related extramural English contacts. As Section  4.2 . will show, there is empirical evidence for incidental learning processes through input only contact, as well as evidence for the (additional) benefit of interaction and output production.

Similar to Vygotsky, Swain also sees language as an inherently social and interactive artifact that humans use to interact with each other and their environment. In her output hypothesis , she emphasizes not only the interactive nature but also the need for active output production in order for learners to reach higher language competences.

While evaluating Canadian immersion programs in 1985, she found higher French test scores for the immersion students than for the non-immersive students. However, while the reading and listening scores of the immersion students were almost similar to native speakers, their performance for writing and speaking stayed behind those of their native counterparts. Since students in immersive classes are presented with a high amount of comprehensible input on a daily basis, her findings raised doubts about Krashen’s input hypothesis (Swain, 2000, 2005). Swain and her team argued that the important difference between native French speakers and immersion students was that students in the immersion classes were not pushed to produce a high amount of output. For Swain, the production of comprehensible output, i.e., output that is “grammatically accurate and socio-linguistically appropriate” (Swain, 2005, p. 472) for a given situation, and which allows the interaction partner to understand the speaker, goes far beyond simply providing an opportunity for enhancing fluency through practice (Swain, 2005). Instead, the output serves three functions:

First, producing language output can trigger noticing on different levels. Learners may notice a word or form because it is frequent or salient. However, they may also notice gaps and language problems in their own interlanguage, which hinders their ability to express themselves accurately. They might then seek to fill the gap by interacting with an interaction partner or an inanimate tool (e.g., dictionary, grammar book) or make a mental note to pay further attention to the relevant input in the future. In this way, through the recognition of problems, a mental conflict is triggered, and a cognitive process is initiated, leading to generating new or consolidating existing knowledge (Swain, 2000, 2005). Empirical research has shown evidence for such a process in learners after producing written or spoken language output in interaction with another student (for an overview see, for example, Swain, 2005).

Second, empirical findings suggest that output serves as an opportunity for testing one’s language hypothesis and provides the learner with an opportunity to alter and modify the output if the hypothesis proves to be incorrect (Swain, 2000, 2005). This becomes possible through feedback from the interaction partner. The feedback can be implicit or explicit. With implicit feedback, learners must infer the inaccuracy of their utterance, while explicit feedback clearly states where the learners’ utterance was correct and where it was incorrect (Carroll & Swain, 1993). Both implicit and explicit feedback can be positive or negative. Implicit or explicit positive feedback verbally or nonverbally confirms that an utterance was indeed correct (Carroll & Swain, 1992). Explicit negative feedback verbally or nonverbally states that a form does not belong to the target language. Implicit negative feedback occurs verbally in the form of error correction, corrective recast, and rephrasing of erroneous sentences or phrases, or through nonverbal communication (Aljaafreh & Lantolf, 1994; Carroll & Swain, 1992, 1993; R. Ellis, 2008; Long et al., 1998). Both forms of feedback are effective and can provide learners with information that input alone cannot provide. Empirical studies have, for example, shown that negative feedback, both implicit and explicit, can induce noticing of forms and phrases which are not as salient through comprehensible input alone or are rare or unlearnable through positive feedback (Aljaafreh & Lantolf, 1994; Carroll & Swain, 1992, 1993; R. Ellis, 2008; Long et al., 1998).

Therefore, feedback helps learners to test their hypothesis about the target language.

Strategies to solve language problems might include testing alternative hypotheses, applying existing knowledge to the context at hand, and internalizing newfound knowledge into one’s system. In fact, some errors or turns observed in learners’ written or spoken interactions may be seen as evidence for testing different hypotheses about the target language. By doing so, learners are engaged in deeper processing of the target language, ultimately resulting in increased control and automaticity in using the target language. This, in turn, releases cognitive resources for higher-level processes. Therefore, it can be argued that the process of modifying one’s own output represents language acquisition (Swain, 2000, 2005).

Third, the production of language output provides an opportunity for metalinguistic reflective functions (Swain, 2005). By putting thoughts into words, they become sharpened and transformed into an artificial form that is accessible to further reflection and response by oneself and others. Thus, speaking or writing represents both cognitive activity and the product of the activity itself.

In addition, output production triggers a deeper understanding and elaboration because it requires the speaker/writer to pay more attention to the elements of a message and their relationships to each other to connect and organize them into a coherent whole. Through this process, a more durable memory trace is established in learners’ minds, and language learning is facilitated (Swain, 2000, 2005).

All of these functions and benefits of output production are present in collaborative dialogue, in which speakers work together in order to solve linguistic problems and build linguistic knowledge (Swain, 2000, 2005). It is in output and interaction that learners have the opportunity to actually use the target language and stretch beyond their present stage of language competence. Swain, therefore, concludes that the production of comprehensible output is necessary for learners to reach the highest levels of proficiency (Swain, 2000, 2005). Language is thus interlinked with and fostered through social interaction, emerging as a result of “meaning-making processes” (Black, 2005, p. 120) within a specific social context (Black, 2005). In interaction with others, learners have the chance to test their hypotheses and to gain more control over their own language production.

Despite the somewhat incommensurability of the way the discussed theories model language learning and the importance language input, output, and interaction play in its process, the discussed theories suggest that regular contact with a foreign language and the chance to use it to interact with others can result in incidental language learning.

In terms of input and media-related extramural English contact, Chapter  2 has shown that technological developments in the last few decades have made regular access to English-language media content highly accessible for learners in Germany and Switzerland. By regularly engaging in English media content, learners are presented with a high amount of input for the most frequent English words and chunks, both written and auditory. In addition, learners can benefit from contact to less frequent and topic-specific vocabulary. Since learners choose the content themselves, it should be highly motivating and engaging, thus lowering the affective filter.

In addition, newer interactive media channels can further increase and deepen the learning process by allowing learners to interact and engage in dialogue. Here, Vygotksy’s theory and Swain’s output hypothesis provide a framework that might explain why a high level of interactive extramural contact with a foreign language might help students to navigate their interaction with the target language. The interaction will provide them with the necessary assistance to make complex input comprehensible, work through their zone of proximal development, and engage in mean-making processes within a given social context. Such interactions can be with a more advanced speaker of the target language (other-regulated), but also with other inanimate objects (object-regulated), such as additional information material, a dictionary, or other forms of technological tools. This might be especially important for less proficient learners. As learners progress in their control over the language, they might be more and more able to self-regulate and process even complex authentic input on their own. In addition, the chance to produce output and engage in dialogue will help learners to notice gaps in their own knowledge, reflect on their language use, and test hypothesis.

However, it should be mentioned that in the very beginning, learners might not be able to navigate authentic English-language media content even with the help of other-regulation and object-regulation. Instead, most learners will rely on in-classroom instructions at this stage. Even Krashen admits that for most learners, the first contact with a target language will most likely be through the educational system. This in-classroom instruction will provide learners with comprehensible learning material geared explicitly towards their competence level (Krashen, 1985). As such, formal instruction within the classroom will have a significant impact on students’ language development and will lay the base for any future language learning. Indeed, as Hulstijn (2001) points out, most teachers and scientist are well aware of the fact that even though incidental learning is a useful and powerful tool for language learning, it is important to teach learners the linguistic principles and lexical system of the target language, as well as making them aware of (vocabulary) learning tasks and teach them explicit strategies for doing so. Most teaching materials recognize this by including a vast number of techniques and activities to teach beginners and intermediate learners the necessary core vocabulary. This ensures that learners start their language journey with the study of a base vocabulary, learned to automaticity, while contextual learning does only play a role in later stages (Hulstijn, 2001).

In addition, formal instruction will also help learners to develop what Krashen calls the monitor . While a person’s ability to produce language derives from their unconscious knowledge and acquired competence, conscious learned knowledge about the target language serves as a monitor. This monitor helps to regulate and check output before it is uttered. For the monitor to work, learners need to be aware of the rules and be concerned with correctness (Krashen, 1985).

With time, learners will become more proficient and, as a result, will find it easier to find comprehensible media content outside of the classroom and engage in more complex interaction and dialogue with advanced learners and native speakers. Chapter  2 could show that newer interactive media channels do provide learners with said opportunity to produce and actively use English in natural settings. In this way, the media has created new assisted and interactive language learning opportunities outside of the educational setting, which provide more than just language input. New forms of interactive online communication tools, such as chatrooms, messaging apps, and message boards, can provide opportunities for extramural English contacts and activities through synchronous or asynchronous interaction with native and non-native speakers. By using these media channels, learners not only receive a high amount of input but can also actively produce output and engage with others in collaborative dialogue and interaction. In these interactive contexts, they will get immediate feedback on their language production. Here, advanced learners and native speakers can act as sources for other-regulated interaction, similar to a teacher in the classroom. They provide positive and negative feedback and help learners in the form of, for example, co-construction, explanations. Through these contacts, learners may even be provided with the opportunity for a high level of immersion within a language community. In this way, new words and phrases can be used and repeated regularly, which in turn fosters a higher conversion rate into long-term memory (Hulstijn, 2001).

The next chapter will summarize empirical evidence for incidental language learning occurring both from input-only as well as from more interactive media channels.

4.2 Empirical Evidence

Early research into incidental learning processes was often conducted within the field of psychology and concentrated on learning through input by reading or being read to by others. The studies were usually experimental in design and did not focus on language contact through extramural English contacts. In recent years, interest in incidental learning processes through media-related extramural English contacts in natural settings has grown significantly outside of the field of psychology. Extramural language contact in these natural settings might be provided through books or other written online and offline material or through music, podcasts, audiobooks, radio, movies, TV series, TV shows, online communities, and computer games. While the first of these media channels only provide language input, online communities (e.g., social media platforms) and computer games can also provide learners with opportunities for output production and synchronous and asynchronous social interaction. The following chapter will summarize important recent empirical findings for incidental language learning in natural settings among young learners (i.e., children and adolescents) through these channels.

As the media landscape changes rapidly, the summary will focus on newer findings to increase comparability with the present study. In addition, the summary will focus on studies about extramural English contacts in natural settings as this aligns with the focus of the present study. Key findings from experimental studies will be discussed only where they provide important insight otherwise missing (for a more detailed discussion on experimental studies in this area see, for example, Huckin & Coady, 1999; Ramos, 2015).

Studies investigating media exposure within the classroom and homework assignments were excluded as they do not focus on extramural English contacts. This also excludes the use of educational computer games, computer-assisted language learning, online learning platforms, and other forms of material specially developed for language learners.

4.2.1 Incidental Language Learning Through Reading

Books have been one of the traditional ways for extramural contact with English as a foreign language. One advantage of reading is that it provides learners with the possibility of repeatedly encountering unknown words and phrases, thus increasing the knowledge of those words and the chance of committing them to memory (Vidal, 2011). However, research has suggested that reading a book is a demanding activity as learners already need to have advanced language competences (Peters, 2018). According to Huckin and Coady (1999), readers need knowledge of at least 2,000 of the most common words in English to understand and use 84% of the words in most texts (and spoken language). For general text comprehension, readers must even be able to understand 95% of the words used in a given text. In order to be able to do so, people need to know the 3,000 most common words. Complete comprehension will not be reached until one understands 98% of the words in a text, which already requires a vocabulary of the 5,000 most common words. According to Sylvén and Sundqvist (2015), this might be the reason why learners in their study only reported low frequencies of leisure time reading in English. Nevertheless, even though 5,000 sounds like a relatively large number, Huckin and Coady argue that it is well within reach of the average language learner (Huckin & Coady, 1999).

Despite these challenges, empirical research suggests the effectiveness of extensive reading, especially for incidental vocabulary gain. To this author's knowledge, at the time of this study, there seem to be no studies looking exclusively into unprompted extramural reading and language competences in natural settings. Empirical evidence must therefore be drawn from studies investigating incidental language learning in experimental settings. However, it should be kept in mind that these settings do not strictly provide extramural contact as defined in this study.

Elley and Mangubhai (1983) conducted a study to examine the effect of extensive reading programs for children from Fujian primary schools learning English as a foreign language. In the experimental groups, teachers encouraged students to read as much as possible and provided age-appropriate books within the classroom. In one of the experimental conditions, teachers also discussed and followed up on the material. Compared to the control group, students in both experimental conditions showed increased language competences in the post-tests. Even though the study suffers from a lack of control over what happened in the classrooms (e.g., some teachers in the control groups read aloud to their students on a regular basis, even though they were instructed not to), the results all point towards the existence of incidental learning processes through extensive reading.

Pitts et al. (1989) conducted a study with 74 learners of English as a foreign language, who were asked to read an excerpt from Anthony Burgess’ book A Clockwork Orange . The book contains the artificial language nasdat and is thus ideal for testing, as students most likely did not know these words beforehand and could therefore not derive their meaning from any similar words in their native language. They were told they would be tested on the story's content afterward but were not told about any vocabulary testing. Two experimental groups were tested in addition to one control group. Experimental group 1 was given 60 minutes to read the text. Group 2 was additionally shown a short clip from the film before reading for 60 minutes. This was due to the high complexity of the text and the younger sample in group 2. The control group neither read the text nor watched the movie. Results from the subsequent vocabulary test showed a significant difference between the experimental and control groups, with group 2 scoring significantly higher than group 1.

Similar to these findings, Dupuy and Krashen (1993) found in their experimental study that even exposure to 40 minutes of reading showed significant gains in students’ vocabulary knowledge. They showed students of French as a foreign language a short clip of the film Trois hommes et un couffin, followed up by a 40-minute reading of an excerpt from the book. Results showed a significant language gain in the experimental group. The group of 3 rd year students of French as a foreign language even outperformed the advanced 4 th year language students in the second control group.

Both teams concluded that in the light of the significant, yet sometimes minor, gains in vocabulary, incidental language learning from reading can occur with foreign language learners, even in a short timeframe. In addition, subjects were only tested on a fraction of words, meaning that they could have learned other words incidentally as well, without it being represented in their test scores. Furthermore, subjects did not read the entire book, which would have provided them with the opportunity to encounter unknown words multiple times, thus increasing the chance of storing them to memory. Last, the chosen texts were quite difficult for readers in both experiments. Hence, it would be possible that more incidental learning would have taken place if learners had been able to understand more of the texts and thus infer more meaning of unknown words from the surrounding context (Dupuy & Krashen, 1993; Pitts et al., 1989).

In order to overcome the limitations of these earlier studies, Horst et al. (1998) conducted a pre-post-test experimental study in which subjects were asked to read a whole novel over a period of 14 weeks. Students read along while the text was read aloud in class. After each session, the texts were re-collected and stored in the school to prevent students from reading ahead or looking up unknown words at home. The results showed a significant gain in vocabulary by the subjects. The gain was higher than in Dupuy and Krashen (1993) or in Pitts et al. (1989), which the authors attributed to the longer exposure and the longer text. Prior knowledge seemed to have played a moderating role in students’ ability to pick up words, as higher knowledge allows students to infer the meaning of unknown words more easily from the surrounding context. Word frequency in the text also played a moderating role in the chance of words being picked up. In addition, nouns were picked up more often than other word types. In a follow-up interview, students reported being surprised that the words they were tested on in the post-test were actually in the novel. This is a strong indicator of the implicit knowledge built through incidental learning.

Despite the findings, the authors conclude that while reading might be a source for incidental learning, it seems to be a slow process. Learners, on average, picked up one word for every fifth word read. However, this result is much higher than for the previous studies, which found retention rates of around one in twelve (Horst et al., 1998).

Pigada and Schmitt (2006) investigated the influence of incidental vocabulary learning in a qualitative study design. They observed one intermediate learner of French as a foreign language. Even though the study used simplified reading material, not authentic texts, their results are still interesting, especially since they not only tested for increased knowledge about word meaning but also spelling and grammatical characteristics. This aids the understanding of the incidental learning process. As the authors and others have noted, the disadvantage of texts with a rich context is that the meaning of a single unknown word might not be necessary to understand the text as a whole. As a result, learners might not subconsciously try to infer the meaning of each unknown word and thus might not learn the meaning of these words. However, the exposure might still increase their knowledge about other aspects of a word, with spelling being the most affected characteristic. Their results revealed that their test subject was able to recall at least one of the word aspects in two-thirds of the target words. Moreover, while not all words were fully mastered by the subject, he was nevertheless capable of using them in productive writing. The highest number of exposures within the text was necessary for learning the meaning of nouns, and some words were still unclear after they appeared more than twenty times in the text. However, one exposure was enough for spelling in some instances. Results also suggest that the inference of meaning for some words was hindered by the interference of the subject’s native language and similar words in French (Pigada & Schmitt, 2006).

In a more recent study, Vidal (2011) showed significant gains in vocabulary knowledge for language learners through written academic texts. In comparison to auditory input, readers recalled more information overall, especially low proficiency learners. The author concluded that reading provided learners with ideal opportunities to dwell on unknown words and sentences. Repetition of words was an important factor for recollection, with readers needing significantly less repetition than listeners to store words to memory. However, the author also concluded that readers and listeners made more gains in words that were explicitly elaborated beforehand. This shows that explicit elaboration can foster robust connections between form and meaning.

Overall, the empirical findings show that incidental language learning from extensive reading does occur, albeit the process being slow and challenging for readers. In addition, some words (e.g., nouns) seem easier to pick up than others and repetition seems to be an important factor for recollection but does not guarantee a successful learning process.

All of the discussed studies used books or book excerpts for their research. However, the internet has also made new forms of written content available. While social media sites often only provide shorter texts, blogs might provide readers with longer English content from various areas of interest. It can thus be hypothesized that online reading activities will also lead to incidental learning processes. However, to this author's knowledge, there is no empirical data available for reading online in terms of incidental language learning, yet. Studies concerning online communities, including social media platforms, will be discussed separately in Section  4.2.4 , as they provide not only input but also enable output production and interaction.

4.2.2 Incidental Language Learning through listening

Music can be a valuable source for language learning, as the lyrics are highly repetitive, conversation-like, and slower-paced than spoken, non-musical discourse. In addition, people tend to listen to the same song multiple times (Toffoli & Sockett, 2014). It is therefore surprising that there seems to be little empirical evidence for incidental language learning from exposure to music, either in an experimental or in a natural setting. One of the few studies investigating the effect of extramural listening to pop music on students’ vocabulary competences is Schwarz (2013). In the study, 74 secondary students were tested on their word recognition for 14 common words from 10 popular songs. In addition to self-reported word recognition, students also had to use some words productively or provide a translation or synonym, thus making the results more reliable. Results showed a significant increase in vocabulary knowledge between the pre- and post-test. In addition, the qualitative analysis of the translation and synonyms also showed that some students already referred to the song lyrics in the pre-test, demonstrating that they already knew the lyrics and the words were processed in the context of the song. However, four students had inferred the wrong meaning of the target word from the song. The author did not investigate the differences between students with a high number of extramural contacts and students with lower extramural contacts. This is probably due to the small variance in the sample, as all students listened to English music every day (Schwarz, 2013).

Even though the sample size was small, and the data relies on students’ self-reported knowledge, the results showed a promising trend towards incidental language learning from exposure to English pop songs. However, similar to the findings for reading, the vocabulary gains were small (Schwarz, 2013). This once again supports the notion that incidental learning takes place in small increments and through repeated exposure.

In their experimental study,Pavia et al. (2019) investigated vocabulary gains from listening to music for 300 Taiwanese children ages 11 to 14. Their results showed significant gains in knowledge of spoken-form recognition for both single word items and collocations for the experimental groups between pretest and immediate post-test, but not for the control group. Repeated exposure significantly increased learning gains, starting around seven encounters. Overall, students’ learning gains were again small. Results for the delayed post-test could not solely be attributed to the treatment, as the control group also showed a significant increase in vocabulary. The authors attributed this to learning effects from the immediate post-tests or conscious discussions about words and collocations among the students after the test. Experimental and control groups did not differ in regard to their gains in form-meaning connections. This is in line with other empirical findings that showed learners retain spoken-form recognition before form-meaning connection, the latter needing more exposure than the former. As the authors note, this might be even more dominant in exposure to music since songs do not provide as much context as other forms of media content. Nevertheless, even though participants only listened to two songs and results only show learning gains for spoken-form recognition, the results are promising and show that incidental learning through music can occur even after a short exposure and even in learners at a beginner level.

Additional evidence for incidental language learning through music in older learners comes Toffoli and Sockett (2014). Results from their study with 207 Arts and Humanities students in France showed that French university students listen to a high amount of English music on a daily basis; some even listen exclusively to English music. Furthermore, the music was not just background noise, but students engaged in active listening strategies such as looking up song lyrics online or pausing and rewinding songs to understand the lyrics better. Learners were also asked to translate four excerpts from popular song lyrics in order to measure possible learning effects. Results showed that frequent listeners (at least once a week) outperformed non-frequent listeners for all four excerpts. Unfortunately, the language comprehension test only included four items in the form of four excerpts from song lyrics. What is more, it is not clear if learners had come across any of the words presented in the test before. As the authors noted, preferences for genre, artists, and songs varied considerably in the sample, making it difficult to choose lyrics for the test. In addition, the sample size was relatively small. Still, the results yield important insights in terms of the variety of music styles learners listen to, as well as the listening strategies employed by learners.

Apart from music, another form of auditory input is spoken auditory input, e.g., from reading aloud to learners. R. Ellis (1999) summarized findings for language learning by reading out loud to younger children in multiple experimental studies. The results show an increase in language competences for young learners in classes where students were being read to on a regular basis. Again, repetition was an important factor for learning gains. The author also stressed the significance of the opportunity for learners to ask questions and show their non-comprehension in face-to-face settings. These interactions will probably lead to additional input from the reader, specifically tailored to the individual learners’ language skills.

In addition to the reported learning gains from reading, Vidal (2011) also showed significant vocabulary gains for university students listening to academic texts (see also Section  4.2.1 ). However, listeners recalled less information in direct comparison to readers in the sample. Vidal concluded that listening to a speaker seems to be a rather challenging activity, especially for lower proficiency learners, as real-time language processing makes it harder to segment the spoken text into separate words and recognize unknown words or phrases from running speech. As a result of these challenges, listeners most likely needed more repetition in order to commit a word to memory than readers do. However, it is likely that, as learners’ proficiency increases, so will the ability to identify and process unknown words from listening to audio input (Vidal, 2011).

Furthermore, the results showed that listeners most likely cannot suppress the activation of knowledge from their native language. As a result, they often do not recognize the differences in cognates or false friends. They are also less likely to add new, formerly unknown meanings of polysemous words to memory. Instead, they were shown to stick with the meaning they already knew, even if it made no sense in the given context. Readers in the study suffered less from this problem (Vidal, 2011).

However, despite the challenging nature, Vidal concludes that listening to English audio content can aid learners in their language learning process since words heard auditorily are stored directly into the phonological memory. Words encountered in the written form still need to be recorded, a process that might be partially or entirely unsuccessful in some cases. Listening can thus help to foster more stable and long-lasting memory traces (Vidal, 2011).

Similar to Vidal, van Zeeland and Schmitt (2013) conducted a study with postgraduate students from an English university who learned English as a second language. While it has to be kept in mind that the study tested learners much older than in the context of this study, who also lived in a country where the target language (English) was the native language, the results still yield interesting insight into the complex nature of incidental vocabulary learning.

As opposed to earlier studies, this study did not only assess recognition and recall of meaning, but also form and grammar recognition. Thirty high-intermediate to advanced learners of English were asked to listen to a text passage read to them that contained several made-up words. They were told to concentrate on the meaning of the text as a whole. While 20 learners were tested immediately afterward, ten learners were tested with a delay of two weeks to identify long-term retention without confounding learning effects from the first post-test. Results showed a significant but again small learning gain for all knowledge dimensions. Overall, meaning recall showed the smallest gains. Learners scored highest in form recognition, followed by grammar recognition, for immediate and delayed post-test. The authors conclude that these results show that some vocabulary dimensions are picked-up later than others. Interestingly, what little meaning learners were able to gain incidentally was better recalled after two weeks than gains for form and grammar recognition.

Overall, the empirical evidence suggests the benefit of extramural audio contact to a foreign language. As music is a popular leisure-time activity and people tend to listen to their favorite songs repeatedly, extramural contacts through songs offer a beneficial way to learn a language.

English-language music has traditionally been easy to access, even before the advent of the internet in both Germany and Switzerland (see Chapter  2 ). Therefore, music has most likely already been an opportunity for incidental language learning for adolescents in past generations. However, the possibility of modern music streaming on online-based music platforms might provide learners with a greater locus of control over their listening experience. Being able to pause, rewind, and use the lyrics-on-screen function at their own discretion is likely to make input more comprehensible for learners (Toffoli & Sockett, 2014).

In addition, the empirical evidence for spoken language summarized in this chapter also highlights the learning opportunities provided by English audiobooks, radio programs, and podcasts. However, research about learning gains from extramural contacts in a natural setting is still scarce.

Similar to reading, learning gains from this kind of input seem to be small (van Zeeland & Schmitt, 2013; Vidal, 2011). This is likely also due to the fact that listening to authentic input is equally if not more challenging for learners. Learners need to know as many as 6,000 to 7,000 of the most common words to follow a spoken discourse (Nation, 2006). In addition, empirical evidence shows that especially low proficiency learners might have problems with the recognition and segmentation of words from running speech (van Zeeland & Schmitt, 2013; Vidal, 2011).

4.2.3 Incidental Language Learning Through Watching

English-language movies, TV series, and TV shows provide viewers with both auditory and visual input. New words are presented within a narrative context and supported by visual aids. If subtitles are added, written content is provided as well. Watching movies, TV series, TV shows with subtitles thus provide auditory, written, and visual information, with the latter providing rich contextual clues for the former two (d'Ydewalle, 2002; d'Ydewalle & van de Poel, 1999; Lindgren & Muñoz, 2013). In addition, Webb and Rodgers (2009) point out the beneficial characteristic of repetition for vocabulary learning, especially in TV series.

Earlier studies about incidental learning through watching audio-visual content usually investigated subtitled movies and TV series, as these were the options most accessible to viewers at the time. In their study, Neuman and Koskinen (1992) investigated the influence of subtitled TV programs on both language and topic knowledge for Asian minority students in the US. They found significantly higher results for the subtitled TV and the normal TV group in comparison to the two control groups (listen to audio and reading along; reading only). These results strongly support the claim that reading (subtitles) is not the only route for incidental learning processes and that visual content does, indeed, foster learning. In addition, the study also showed evidence that students’ prior vocabulary knowledge and a supportive context, in the form of video print, play an important moderating role in the incidental learning process. However, the authors pointed out that since the content is not produced with the language learner in mind, the content might be too complicated for beginners to follow. In addition, the pace of the spoken information in most TV series and movies might be too quick for some learners and subtitles are designed to keep pace with the scene on screen (Neuman & Koskinen, 1992).

d’Ydewalle and his team conducted several experimental studies investigating the effect of watching subtitled television on learners' language competences (an overview can be found in d'Ydewalle, 2002). Results showed evidence for the fact that reading and processing the input provided by subtitles is an automatic process beyond conscious control and that it triggers incidental learning processes (d'Ydewalle, 2002; d'Ydewalle & van de Poel, 1999).

Incidental learning proved to be even more effective when subtitling was reversed, i.e., when the foreign language was presented in subtitles and the native language in the audio track. The authors attributed this to the fact that processing the subtitles was the main activity for participants and thus, providing the foreign language in written form led to higher learning gains (d'Ydewalle & Pavakanun, 1997).

Furthermore, results from studies with different age groups showed that, in general, younger children pay less attention to subtitles and prefer dubbed movies and TV series. This is most likely due to their lower reading skills. However, a small part might also be influenced by the fact that younger children in the Netherlands (where the studies were carried out) are not as accustomed to watching subtitled television as older children and adults are. As a result, they benefit less from extramural English contact if subtitling is reversed and therefore show lower vocabulary gains (d'Ydewalle & van de Poel, 1999).

Results from the research group also showed that the similarity between a person’s native language and the foreign language in question plays a moderating role in the effectiveness of the incidental learning process (d'Ydewalle & van de Poel, 1999).

In addition to vocabulary, d’Ydewalle and colleagues are also one of the few teams to investigate the acquisition of grammar and syntax through incidental learning. While the initial studies failed to detect any effect, they were eventually able to show slight increases in grammatical competences. However, it should be mentioned that the increases were most significant when explicit rules were presented in advance. Therefore, the authors concluded that grammar might be too complicated to acquire solely from exposure to the target language (d'Ydewalle, 2002). Increases in grammar and syntax competence should thus only be expected after some form of formal instruction has taken place. They confirmed this in a study comparing children before and after they were first introduced to French as a second language within the school context (d'Ydewalle, 2002). In contrast, words, especially nouns, seem to be much easier to acquire incidentally (d'Ydewalle & van de Poel, 1999).

Apart from subtitled content d'Ydewalle and Pavakanun (1997) also found learning gains for experimental groups with only the foreign language in the audio track (without any subtitles) and concluded that watching the rich visual information provided by the movie enabled participants to derive the meaning of the story from the visual context. This was not the case when the foreign language was provided in the subtitles, and no audio track was played, which is probably due to the fact that participants missed important visual clues while concentrating on the subtitles.

In support of these findings, Araújo and da Costa (2013) could also show that advanced learners from the European Survey on Language Competences (ESLC) did not significantly benefit from movies with subtitles compared to movies without them. The reverse was true for students at the beginner level. The authors attributed these findings to the fact that learners need to reach a certain level of proficiency before being able to process non-subtitled audio-visual content efficiently. Once they reach that threshold, subtitling no longer contributes significantly to the learning process.

Kusyk and Sockett (2012) tested 43 French university students on their word recognition from audio-visual input. High-frequency watchers demonstrated a significantly higher rate of recognizing and ability to define the most frequent 4-word chunks tested in the vocabulary test than low-frequency watchers. In addition, the results showed a tendency for more frequent and more salient chunks to be recognized more easily. The results underscore the importance of previous knowledge for extramural contacts in natural settings. Most students situated themselves at a B1 level at the beginning of the study. As the authors point out, at this level, learners should be able to understand most of the spoken content in standard dialect on TV or radio. However, the results should be interpreted with caution due to the small sample and the fact that word comprehension was not measured by a comprehension test but by students’ self-evaluation.

Last, results from Lindgren and Muñoz (2013) also show that watching television is the second-best predictor for learners’ listening and reading comprehension.

Overall, the empirical evidence presented in this section shows the beneficial effect of audio-visual contact in the form of movies and TV series for foreign language learning. In contrast to audio-only input, watching a movie or TV series provides a rich visual context to help learners follow a story, even if they do not understand every word. Similar to music, the technical opportunities of streaming services provide learners with a greater locus of control over their viewing experience. Being able to pause and rewind, switch between native and foreign language audio tracks and use subtitling is likely to make input more comprehensible and help with listening comprehension overall (Toffoli & Sockett, 2014).

As with other forms of language input, learning gains from this kind of input seem to be small, most likely due to the challenge of decoding words and meaning while listening to authentic language input. Similar to audio-only material, learners need an extensive vocabulary in order to follow spoken discourse (Nation, 2006; Webb & Rodgers, 2009), and low proficiency learners will most likely struggle to recognize and segment running speech (van Zeeland & Schmitt, 2013; Vidal, 2011). However, even though the requirements for incidental learning through watching television might be quite high and the medium might therefore not automatically be suited for beginners, Sylvén and Sundqvist (2015) could show that even children as young as 11 or 12 might reach the appropriate level of prior knowledge. Motivation is probably a key factor since the children want to understand their favorite TV series, movies, and TV shows and thus tend to pay close attention to what is shown on screen (Sylvén & Sundqvist, 2015).

Apart from movies, TV shows, and TV series, online videos might be another source for audio-visual input. These videos are usually shared via video-sharing platforms, such as YouTube, and cover various topics, from makeup to gaming to lifestyle and mental health. These platforms have also given rise to a new form of celebrity: social influencers (see Section  2.1 for reference). Social influencers produce and upload videos of varying lengths to video-sharing platforms or social media platforms (e.g., Instagram). They often have millions of followers worldwide and post multiple videos per week or even per day. Most of the most popular influencers come from the US or the UK. In addition, influencers from other countries might also choose to produce their content in English to reach a broader audience. Video platforms, therefore, provide an increasingly rich amount of authentic audio-visual input in English, including different accents and dialects. These videos also give insight into different cultures. To this author's knowledge, there are no empirical studies for this form of extramural English contact and language learning, yet. This is surprising, given the large amount of input available and the popularity of these platforms among young people (MPFS, 2017; Waller et al., 2016). It is thus very likely that German and Swiss adolescents follow international English-speaking influencers who meet their interests on social media and video-sharing platforms. This will, in turn, provide them with yet another source of extramural English contact.

4.2.4 Incidental Language Learning Through Online Communication

With the rise of interactive online platforms, such as chatrooms, messenger boards, and social media sites, learners not only have the opportunity to take in a rich amount of language input but also to socialize and interact with other native and non-native speakers online (Thorne et al., 2009). The internet thus provides the opportunity for new, participatory forms of learning and interaction (Black, 2005; Thorne & Black, 2007; Thorne et al., 2009). However, empirical evidence in this area is still sparse. Among the various online communities, fan fiction communities have received the most attention for their potential for incidental learning. The following section will thus summarize findings for this form of participatory writing space and its learning potential, but the findings can most likely be generalized beyond the scope of this specific form of online community.

Fan fictions are “original works of fiction based on forms of popular media such as television, movies, books, music, and video games” (Black, 2005, p. 118). Within these communities, “native and non-native English speakers [have the opportunity] to use literacy skills to forge relationships with individuals who share their interests” (Black, 2005, p. 120).

Empirical evidence for incidental language learning from this kind of extramural contact can mostly be drawn from the work of Black (2005, 2009). The author used ethnographic and discourse analytic methods to estimate and understand how English learners interact and communicate on these platforms. Additional theoretical considerations and literature reviews can be found in Thorne (2008), Thorne and Black (2007), and Thorne et al. (2009). The results show that online (fan) communities offer learners the opportunity to use language in a social environment and in a way that is meaningful to a particular purpose. In order to participate in the community, users do more than type grammatically correct utterances; they use language to create communities and interact with each other (Thorne et al., 2009).

Through engaging in the community, learners get in contact with a rich amount of input of meaningful content, but also actively use language to produce various forms of output and engage in interaction with more experienced members of the community, thus increasing their language competences (Black, 2005; Thorne et al., 2009).

Within fan fiction communities, members can choose multiple levels of participation. First, members can be readers only, i.e., only read stories written by others and benefit from the vast amount of language input through extensive reading and familiarization with techniques and conventions of different genres of writing, without having to produce content themselves (Black, 2005; Thorne et al., 2009). Second, members can choose to contribute by writing reviews for other people’s stories, even though a reader might not be proficient enough in English to write their own stories, yet. By giving others (constructive) feedback, users are able to demonstrate their knowledge and expertise within a specific fandom (Black, 2005). Last, members might decide to write and publish their own stories. Writers can decide to publish in their native language or choose another language. For example, non-native writers might choose to publish their stories in English to reach a larger audience. Announcing one’s status as a non-native speaker might help those authors, as it tells readers to focus on the content rather than grammatical correctness. At the same time, more advanced readers and native speakers often offer extensive feedback on grammatical errors, spelling mistakes, and style issues (Black, 2005). In doing so, they aid novices on their journey to use language as an internal resource to control their own mental processes (R. Ellis, 2008). As Black shows, this form of support and feedback helps non-native writers increase their awareness for audience-specific composition issues and drastically improve their writing skills (Black, 2005; Thorne et al., 2009). Authors might also choose to find a beta reader , i.e., an official proofreader, for their story (Black, 2005; Thorne & Black, 2007).

The actual writing process is further aided by the fact that authors can draw on a rich body of characters and plotlines from the original material. It is also common (as long as it is acknowledged) to incorporate elements and plots from other works of fiction or create crossovers (Black, 2005; Thorne & Black, 2007). By doing so, fan fiction communities not only offer other-regulation in forms of support and help from the community but also object-regulation by artifacts such as existing plotlines, characters, and genre conventions provided by the source material (Thorne et al., 2009). Ultimately, this fosters learners to “move beyond the mechanical aspects of decoding and encoding in the target language.” (Black, 2005, p. 127).

Overall, the analyses have shown that different levels of involvement offer even novice learners an opportunity to be part of an online community and make fan fictions sites a perfect place for collaborative and participatory writing processes. Within the community, learners get constructive feedback from native or more advanced speakers in a supportive environment and have the opportunity to solve linguistic problems together as proposed by the sociocultural theory (Black, 2005; Thorne et al., 2009). Students can revise, edit, and redesign their texts by drawing on and incorporating input from a broad audience of reviewers, engaging in dialog-based interaction, and drawing on the meta resources available in the community. Fan fiction communities are thus ideal places for English learners to become accepted members of an English-speaking community, practice their language skills with native speakers (both receptive and productive), get constructive feedback, and eventually take on their own identity as an English speaker (Black, 2005).

While fan-fiction communities have drawn particular attention by researchers in the last few years, the findings can be expected to be expandable to other forms of online communications, such as forums or message boards and social media. Unfortunately, however, to this author’s knowledge, there is no empirical research on incidental language learning in that area. Nevertheless, it seems that online communities present users with an environment rich in authentic content as well as the opportunity to try out and develop one’s own identity as an English user within an international community. With these characteristics, online communities have long surpassed the simple input mode offered by traditional printed media.

4.2.5 Incidental Language Learning Through Gaming

Computer games have often been frowned upon as leisure time activities and have been suspected of causing violent and addictive behavior in adolescents and children (Graham, n.d.). However, research has shown that computer and video games can also have a positive effect on language learning, especially if they provide gamers with a complex narrative and offer the opportunity to interact with other gamers during the game.

Computer and video games differ in the degree to which they provide such a rich and interactive gaming environment. Following Graham (n.d.), games can be categorized into three levels of narrative complexity. Low narrative games  – e.g., puzzles, rhythm, or simulation games – do not follow a narrative and often have no endpoint or final goal. By comparison, narrative games  − e.g., sport and racing games − provide a narrative and require some background knowledge from the real world. High narrative games provide an even richer and more complex narrative story, in which the gamer has to perform a set of tasks and quests to win the game (Graham, n.d.). It can be expected that more complex narratives might provide a higher level of authentic and comprehensible input to gamers.

Narrative and high narrative games are designed to engulf the player within the inherent logic of the gaming world. While playing, gamers are presented with situations and decisions to choose from. As a result, the course of the story depends on the player’s preceding decisions. Players can thus be seen as co-creators of the game, not just mere users. By playing the game, they shape the game’s environment as much as it shapes them (Gee, 2005). However, similar to the real world, not all actions are available in all situations and to all players alike. Instead, players have to follow a specific set of rules and regulations, which they have to learn and master to succeed in the game (Gee, 2005).

Players get to know the world by wandering through it and solving tasks (i.e., quests) (Gee, 2005; Zheng et al., 2015). Depending on the game, quests can be solved alone or in collaboration with other players. By completing these quests, players build up their character’s abilities, skills, and equipment (Gee, 2005; Zheng et al., 2015). In order to solve quests, players will have to take risks and try out new ways or creative solutions. After successfully finishing a quest, a player moves on to new, slightly more challenging adventures. This forces the player to develop new solutions and communication strategies since the ones used in the level before might not be sufficient anymore. By continuously presenting the player with new and slightly more complicated, yet still solvable, tasks, game designers make sure that the games stay interesting yet rewarding enough for people to keep playing (Gee, 2005).

In such an environment, new information, words, and phrases are introduced at the exact time necessary and are embedded within a situated and communicative context. They are easy to process and do not overwhelm players at the beginning of the game. New words and phrases are also strongly linked to a gamer’s immediate purpose and goals, as the new information is needed immediately to solve the subsequent quest in the game (Gee, 2005). This makes computer and video games ideal for contextualized and situated language learning. By contrast, schools often introduce topics detached from people's goals and purposes, causing them to be more difficult to remember (Gee, 2003).

“People are quite poor at understanding and remembering information they have received out of context or too long before they can make use if it […]. Good games never do this to players, but find ways to put information inside the worlds the players move through, and make clear the meaning of such information and how it applies to that world.” (Gee, 2003, p. 2)

Due to these characteristics, Gee identifies 25 out of 36 learning principles related to language learning in modern gaming (Gee, 2005, 2007). These advantages of gaming for incidental language learning led some researchers to predict the rise of digital games as a game-changer in modern language teaching methodology. However, in recent years, the discussion has shifted somewhat away from how to convert digital games for educational purposes to the notion that digital games already come equipped with the ability to teach cognitive skills and promote problem-solving (Thomas, 2012).

In addition to these general advantages, some games also provide the opportunity to interact not only with the gaming engine but also with other players via written or audio chats. According to the Scale of Social Interaction (SSI) model, games can be categorized according to the level of interaction they allow for, i.e., how many players can play simultaneously. These differ in the way they allow language input and output from the gamers. The model distinguishes between single-player, multiplayer, and massive multiplayer online role-playing games. Single-player games are played alone and do not allow interaction with other gamers. As a result, they only offer language input and few to no opportunities for output production. Multiplayer games allow for the interaction of multiple players simultaneously. These players might be in the same room or might be connected online. These games provide the opportunity for authentic interaction with other players. As a result, they can provide more opportunities for incidental learning within the natural game setting. Massive multiplayer online role-playing games ( MMORPG) can be seen as the most advanced form of interactive gaming. Here a large number of gamers can be logged in to the games’ online servers and can play and interact with each other simultaneously (Sundqvist, 2013).

Within MMORPGs, players are usually encouraged to work together to solve quests. In doing so, players fall back on their social competences from the real world, building social connections, cooperating with each other, and even building communities (Gee, 2005; Piirainen-Marsh & Tainio, 2009; Zheng et al., 2015). Depending on the abilities and experiences of the player’s characters, these communities often form rather complex hierarchies and rules of interacting with each other, making sure that each player’s abilities and skills are utilized the best way possible. Novices are integrated into the group and can learn from other, more experienced players (Gee, 2005; Piirainen-Marsh & Tainio, 2009; Sylvén & Sundqvist, 2012b).

Gaming communities in these MMORPGs can consist of people from the same geographical region, who might know each other in real life, but there are many communities in which members do not live close to each other. In these communities, English is often the language of communication among group members (Piirainen-Marsh & Tainio, 2009; Sylvén & Sundqvist, 2015). Just as with the skills necessary for successfully participating in the quests, more experienced language users within these communities serve as role models and catalysts for the language socialization of novice English speakers (Thorne et al., 2009). As suggested by the sociocultural theory, social interaction and other-regulated activities help novice learners move towards a self-regulating state in their language and gaming trajectory. In this way, multiplayer games and MMORPGs offer an immersive environment with repeated exposure to the target language in an authentic communicative context and meaningful interaction. Gamers have to communicate, negotiate meaning, and get real-time feedback from their gaming partners. In addition, MMORPGs usually involve a high level of engagement, motivation, and commitment to the task and the people involved. According to Gee, these characteristics make MMORPGs a silver bullet for language learning in natural settings (Gee, 2003, 2005, 2007; M. Peterson, 2010; Sylvén & Sundqvist, 2015).

Sylvén and Sundqvist even argue that MMORPGs might be similar to content and language integrated learning (CLIL) Footnote 2 in school, as it forces learners to use their language skills to solve tasks, meet the given requirements in order to be successful gamers, as well as communicate and get immediate feedback from other gamers. Similar to learners in a CLIL classroom, gamers thus have a high motivation to understand new vocabulary and grammar in order to solve quests successfully and communicate with other players. Moreover, since the game is a voluntary, leisure time activity rather than a school requirement, gamers will probably be more motivated to put in endless hours to perfect their gaming and language skills than learners within a classroom (Sundqvist, 2011; Sylvén & Sundqvist, 2012b).

The two authors also investigated Gee’s statements about learning principles in relation to the MMORPG World of Warcraft . They conclude that the game does, in fact, provide eight of Gee’s 36 criteria, i.e., active and critical learning, psychosocial moratorium, identity, practice, regime of competence, subset, transfer, and affinity group (Sylvén & Sundqvist, 2012b). They also confirm Gee’s proposed similarities between MMORPGs and the CLIL classroom in terms of the authenticity of the materials, the integration within a language community, and learners’ motivation. They conclude that the advantages of playing MMORPGs might be responsible for the repeated empirical finding that boys outperform girls in vocabulary tests, even though girls tend to hold more positive attitudes towards languages and attend CLIL classes more often (Sylvén & Sundqvist, 2012b).

In a similar vein, Zheng et al. (2015) could show that MMORPGs provide learners with a rich input of social, historical, and cultural material to use as tools for their interactions with each other. Similar to Gee, the authors see these characteristics of games as highly beneficial, as they provide players a sense of embodiment by giving them a specific role, a goal, and the opportunity to experience the consequences of their actions. In addition, they found that gaming encourages learner agency and allows learners to transcend from the here and now of the situation to more general knowledge and use of the language (Zheng et al., 2015).

Further empirical support comes from Thorne (2008). In her qualitative study, she could show the fruitful way gamers communicate with each other and how language learning may occur. In her study, an American and a Ukrainian gamer began to communicate and chat within the MMORPG World of Warcraft. Their interaction showed forms of collaboration, negotiation of meaning, feedback, as well as other- and self-correction. In addition, the American gamer reported that the communication reduced inhibitions and insecurities and increased their motivation to further engage in language learning activities.

Similar to these findings, Rankin et al. (2006) showed increased vocabulary knowledge and enhanced output production for four participants in a pilot study. Students were asked to play the interactive game Ever Quest II for at least four hours per week. However, while more advanced learners seemed to benefit from the game-based interaction and communication, beginners seemed to struggle with cognitive overload from the game’s requirements.

In another study, Rankin et al. (2009) employed a pre-post-test experimental design to investigate gamers’ actual increase in vocabulary knowledge and conduct an in-depth analysis of their social interactions. Two experimental groups were established: in the first experimental group, six native Mandarin speakers were asked to play a video game among themselves. In the second experimental group, another group of six native Mandarin speakers played the game in interaction with a group of native English speakers. The six students in the control group did not play but instead received three hours of language instruction. Results showed that the two experimental groups outperformed the control group in the post-test regarding vocabulary knowledge in the context of the game. However, classroom instruction was more beneficial for participants’ scores on sentence usage. The authors attribute this finding to the fact that the employed test was very close to the classroom exercises students were exposed to before. It should be noted that the statistical results should be interpreted with caution due to the small sample size.

In-depth analysis of chat protocols revealed that the native speakers helped and guided the novice players through the all-English interface and the unfamiliar game. Results also showed that language use increased for Mandarin speakers over time. The protocols showed that these gamers started to produce more output as they grew more confident with the game (Rankin et al., 2009).

Results from M. Peterson (2012) also support the fact that gaming can help introduce language learners to specific language practices of a target group. The data showed that the six foreign language students in his sample adapted their interaction strategies in an online-based gaming environment and used time-saving techniques, such as abbreviations and emoticons. The data also shows how students engaged in continuous collaborative dialogue and interaction in the target language English.

Last, Piirainen-Marsh and Tainio (2009) conducted a qualitative study about the interaction of two teenage boys (10–14 years) regularly engaging in playing Final Fantasy X together. Although not an MMORPG, the study shows that even games with extensive (subtitled) dialogues offer a rich amount of linguistic input for the players, as well as a chance for playful and casual practice of language skills. The constant repetition throughout the game can lead to considerable learning effects.

While most studies reported here have focused on interactive gaming, Purushotma (2005) also found evidence for learning effects from non-interactive single-player games. In his article, the author analyzed the benefits of playing The Sims (a life simulation game played in single-player mode). While the game characters speak an artificial language, the game offers a wide range of text within the menu and in-game notifications. The vocabulary resembles an English beginner course, with a high number of everyday words and phrases. As with other games, players will get immediate feedback for their hypothesis of unknown words in the form of character's behavior in the game and the game environment. In addition, the newest version of the game offers the possibility to change the program code to show in-game messages in two languages (e.g., the native and a foreign language) and can offer translations for unknown words within a pop-up window. The analysis shows that even non-interactive games can offer opportunities for incidental language learning. With its high level of frequent vocabulary, games like The Sims might be especially suitable for beginners. The non-violent and fighting-free setting might also make it especially suitable for younger learners. Research has also shown empirical evidence that these non-violent games might be a more attractive gaming option for female students than many of the often violent or sports-centered interactive game options (MPFS, 2017).

Although the sample sizes in the reported studies were often small, the empirical findings in this chapter suggest that incidental language learning can occur from interactive and non-interactive gaming. Furthermore, interactive games can help learners move from other-regulated learning to a state of internalized self-regulation and control of language as a mediative tool, as proposed by Vygotsky’s sociocultural theory. However, as with other extramural contacts, interactive gaming in English can be challenging for beginners. However, collaborative dialogue and corrective feedback from other more advanced speakers can help bridge the gap, reduce inhibition, enhance motivation, and facilitate language learning. Overall, the immersive environment offered by modern interactive computer and video games is thought to offer an ideal platform for situated and incidental learning, thus bridging learning in and outside of the classroom (Reinders, 2012).

4.3 Incidental Language Learning Through Multi-channel Media Exposure

This last section will summarize empirical findings from studies that did not focus on a specific media channel but rather looked at learners’ overall frequency of extramural contacts across multiple media channels.

At the beginning of the 2000s, Hasebrink (2001) showed that the German participants in his studies claimed to have learned around 20% of their English competences outside of school through informal contact (Hasebrink et al., 1997, p. 163ff). However, as this is only a self-reported estimate and the study did not include a test on language competences, these results should only be seen as a rough estimate. However, the result points towards the occurrence of informal learning processes even before the advent of the internet.

The only other empirical evidence for Germany comes from the study Assessment of Student Achievements in German and English as a Foreign Language ( DESI). The study investigated 9 th graders in Germany and included some questions about media-related extramural English contacts via email, video, television, books, comics, manuals, and songs in the questionnaire. While these categories are by no means exhaustive in terms of modern online and offline media content, the results can still yield some interesting insights. Media-related extramural English contact activities showed a medium-sized correlation with students’ English test results and English grades. Students in the highest educational track (Gymnasium) reported higher frequencies of media-related extramural English contacts and a higher interest in reading (Helmke et al., 2008). Apart from these results, no further empirical evidence seems to exist for Germany or Switzerland.

For Sweden, Sundqvist (2009a) Footnote 3 showed significant and positive correlations for the overall frequency of media-related extramural English contacts, vocabulary competences, and oral proficiency (for details about the test procedure see Sundqvist, 2009a). While the effect of reading was especially strong for oral performance, gaming and surfing showed the highest correlation for the vocabulary tests in her study. Dividing students into user groups showed that high-frequency users received significantly better test results than low-frequency users (Sundqvist, 2009a). Interestingly, however, the author also found indications for the effects of extramural English contacts to be stronger for low-frequency than for high-frequency users. She interpreted the findings as an indication that the increase from no contact (0 hours) to some contact (e.g., 8 hours) might be more beneficial than the increase from 45 hours to 53 hours (Sundqvist, 2009a).

In addition, the positive correlation between extramural contacts and oral test results found in the data only holds for two of the four classes, while it is negative for the other two. Sundqvist assumes this could be due to the socio-economic composition of the classes or due to the teacher influence but did not elaborate further (Sundqvist, 2009a). While her sample is relatively small (n = 80), her study does give an interesting and compelling inside view into the field of media-related extramural English contact through the media and the relationship with learners’ competences. In addition, her use of language diaries provides a detailed, in-depth measurement of students’ actual frequency of extramural contacts that might be more reliable than some of the ex-post-facto questionnaires employed in other studies, including the present.

Forsman (2004, cited in Sundqvist, 2009a) also found a significant and positive relationship between the overall frequency of media-related extramural English contact and students’ tendency to use American words and phrases (in comparison to British ones) in his study with 330 Swedish-Finish students. The author attributes these findings to the dominance of American media content.

Lindgren and Muñoz (2013) could also show the positive effect of extramural exposure to a foreign language on children’s listening and reading comprehension in multiple European countries (aged 10 to 11). The results also showed a significant effect for the cognate distance between the native language and the foreign language: students with a native language closer related to the target language showed a significant higher learning effect.

Peters (2018) found a significant positive correlation between media-related extramural English contacts and language competences. Significant effects could be shown for reading books and magazines, surfing on English-language websites, and watching movies and TV series without subtitles, but the correlations were small in effect (except for browsing). Surprisingly the results showed a small negative correlation between vocabulary knowledge and listening to English-language songs, as well as no significant correlation for watching subtitled movies and TV series or for gaming. The study was conducted in the Flemish region in Belgium, which has a high level of non-Flemish and non-dubbed TV productions. The author attributes the lack of correlation between subtitled TV series and movies with test scores, therefore, to the fact that there is virtually no variance in her dataset since almost all students watch subtitled movies and TV series regularly (Peters, 2018).

In addition to the correlations, results from an analysis of variance with covariates also revealed the overall frequency of media-related extramural contact to be a positive predictor for students’ vocabulary knowledge. The effect explained with 13% more variance than the length of in-class English instruction (Peters, 2018, p. 159).

Olsson (2011) focused specifically on the effect of extramural English contacts on students’ writing skills. The author found a strong and positive significant correlation between overall media-related extramural English contacts and test results for a national mandatory writing test. Examining the individual media categories separately, she found a significant and positive correlation between extramural reading, writing, and watching television and the writing test scores. An in-depth analysis showed that students with a higher level of extramural contact on average wrote longer sentences and used longer and more complex words for some text types. In addition, she found that all students with at least moderate extramural contacts reached a pass with distinction or a pass with special distinction in their 9 th grade finals. The extramural contacts also showed a moderate, significant correlation with learners’ grades (Olsson, 2011).

In addition to the overall scores for writing, the study also looked at certain text features in more detail and found significant correlation effects for sentence length in the written mails and the use of infrequent vocabulary for the newspaper articles, but not the other way around. Moreover, even though all students showed a higher variation in vocabulary for the newspaper article than the mails, students with high frequent extramural contacts did show significantly more variation than non-users or low-frequency users. This points towards the fact that students with frequent extramural contacts might gain a more extensive and more diverse language register, which allows them to adapt their language to different text types (Olsson, 2011).

Despite these interesting findings, the results should be read with caution as Olsson’s sample is very small (n = 37). Still, the study gives an important insight into the relationship between extramural contacts and writing in English as a foreign language in general and different text features in particular.

In a longitudinal study, Olsson and Sylvén (2015) also investigated the effect of media-related extramural English contacts on the academic vocabulary of CLIL and non-CLIL students. As in Sundqvist’s study, students were asked to fill out a survey and keep a language diary. Students were then asked to write four argumentative and explanatory essays. The results reveal that CLIL students had slightly more extramural contacts and wrote and read English texts significantly more often outside of the classroom, which in turn seems to lead to a more positive attitude towards English. However, the frequency of extramural contacts did not significantly affect students’ test results and learning progress. The two authors even raise the question of whether or not extramural contacts might level the advantages in language learning for students attending CLIL classes. However, as the authors also note, the study does not answer how much vocabulary students are subjected to through extramural contacts (Olsson & Sylvén, 2015).

Sylvén (2019) further investigated the differences reported by Olsson and Sylvén (2015) with the same dataset. The language diaries from both measurement points again showed that CLIL students were exposed to a greater amount of media-related extramural English than non-CLIL students over time. In addition, the frequency of extramural contacts showed a positive correlation with sentence length and sentence types.

Results from Sylvén (2004, as cited in Sylvén & Sundqvist, 2015) support these findings. The data showed that Swedish CLIL students seem to not benefit as much from English within the classroom as from the use of English outside of school. In addition, although CLIL students on average scored higher than non-CLIL students, non-CLIL students who had a high level of media-related extramural English contacts scored higher than CLIL students who did not have frequent out-of-school exposure to English.

Two quasi-experimental studies further investigated the causal link between extramural English contacts and language competences. In his study, Kuppens (2010) recruited 374 primary students in the Netherlands, who did not have had any English instructions in school and did not have many extramural contacts with English before the study. The questionnaire included watching subtitled television, playing computer games, and listening to music as extramural categories. Non-subtitled movies, TV series, TV shows, websites, and radio were excluded since it could be assumed that a certain level of preexisting proficiency in English would have been necessary to utilize these media forms in a meaningful way. On the other hand, watching subtitled television does not require such a high level of proficiency, nor does listening to music or playing computer games. The results showed that students did use the mentioned media categories regularly. Watching subtitled television showed a significant influence on students’ language test results. Playing computer games also showed a significant effect but only for the English-to-Dutch test, not the other way around. Since the survey did not distinguish between different computer games, it is difficult to determine if variance regarding the preferred games might have influenced the results. The author also speculates that watching subtitled television might be functioning as a form of ‘gateway’ for eventually switching to monolingual television in English as well as the use of other media channels (e.g., fan sites, blogs) (Kuppens, 2010).

In their longitudinal study, Verspoor et al. (2011) compared a group of students who, for religious reasons, had minimal media-related extramural English contact (control group) with students who attended public schools and had the opportunity for regular extramural contact (experimental group). The data showed that lack of extramural contact had a long-term effect on students’ proficiency development. While the control group did not differ significantly in their language competences from the rest of the students at the beginning of the study, a significant difference was found after three years (Verspoor et al., 2011).

Overall, the results presented in this section strengthen the findings from studies focusing on specific media channels. A higher frequency of overall media-related extramural English contacts seems to be positively correlated with higher language competences. While some of these studies only reported correlative results, findings from Kuppens (2010) and Verspoor et al. (2011) lend support to the notion of a causal effect of these contacts on language competences. The results from these two studies also support the claim that extramural English contacts have a positive effect on language competence, even without additional in-class instruction.

4.4 Conclusion

This chapter began by arguing that regular media-related extramural English contact with English as a foreign language can lead to unprompted and unconscious language learning processes. When reading in English, listening to music, watching a movie, or playing a video game, learners usually do not have a dictionary at hand. Instead, they are concentrated on the content and need to derive the meaning of unknown words from the surrounding context. According to the input hypothesis, this will result in incidental language learning, as long as the input is comprehensible, i.e., slightly more complex than a person’s current level of competences. Under such conditions, learners can form plausible and practical hypotheses about the meaning of unknown words. This process is automatic, given that no significant cognitive obstacles or resistance are active (Krashen, 1982, 1985, 1989).

In addition, the chapter drew on the sociocultural theory and the output hypothesis and discussed the possibility of incidental language learning through output production, feedback, collaborative dialogue, interaction, and communication through interactive media platforms and games. According to the theory, learners will only reach the highest levels of language proficiency and self-regulated language use by interacting with other, more advanced learners or native speakers (Dunn & Lantolf, 1998; Lantolf, 2000, 2005, 2011; Swain, 2005; Vygotsky, 1978). Thus, frequent interactive extramural English contact can allow learners to increase their language competences as a by-product of other activities.

The empirical research presented in this chapter has supported the positive relationship between media-related extramural English contacts and learners’ language competences. In addition, newer studies on interactive online media activities, such as gaming or message boards, social media, or online communities, have also shown the advantages of interaction and output production for incidental language learning. While some studies can only report correlative findings, (quasi-) experimental studies have also provided evidence for the causal effect of extramural English contacts on language competences.

Together these findings suggest that learners should not only receive input but also produce, use and repeat new words and phrases on a regular basis in order to foster a higher conversion rate into long-term memory through repetition and forming links with other words within the mental lexicon (Hulstijn, 2001, 2013).

Despite these positive findings, the process of incidental language learning seems to be limited in terms of the scope and speed by which learning can take place. Most of the studies summarized above have focused on vocabulary gains. Studies that have tried to show increases in learners’ knowledge of grammar, morphology, or syntax have generally only reported a marginal effect or no effect at all. Indeed, studies have shown that presenting students with formal instruction before presenting them with an incidental learning opportunity produced larger learning effects for grammar tests (d'Ydewalle, 2002; d'Ydewalle & van de Poel, 1999; Elley, 1997; Vidal, 2011). These results indicate that not all aspects of a foreign language can be easily acquired incidentally. While vocabulary, especially nouns, seems to be easy to pick up as a by-product of other activities, grammar seems to be too complex of a topic for such an incidental process. Instead, formal instruction and feedback seem to be needed for learners to grasp important grammatical concepts in a foreign language (d'Ydewalle, 2002; d'Ydewalle & van de Poel, 1999). However, this does not diminish the importance of learning opportunities through incidental language learning. A rich and vast vocabulary is essential for language learners to master. In order to understand a message, learners must know the meaning and functions of words, as well as the conventional way in which they are used in the target language (Elley, 1997).

Empirical findings also indicate that incidental learning is a relatively slow process, with an unpredictable outcome, and prone to errors. Texts with 200,000 words or more are most likely needed for a person to learn 108 new words (Letchumanan et al., 2015; Sok, 2014; Webb & Rodgers, 2009), and learning gains from listening seem to be even smaller than gains from reading exposure (van Zeeland & Schmitt, 2013; Vidal, 2011). It is thus not surprising that some studies have shown that intentional learning is more effective and faster, even for vocabulary learning in direct comparison (R. Ellis, 1999).

In addition, several factors have been shown to influence the speed and success of incidental language learning. This includes word characteristics (e.g., distinctiveness, polymeny, length, imageability, and correlation between form and meaning), frequency of exposure, repetition, text type, input complexity, contextual clues, learners’ language proficiency, and ability to guess words, mother tongue and motivation. In addition, the proportion of words already known and the students’ background knowledge has also been shown to influence the incidental learning process (N. C. Ellis, 1994; R. Ellis, 1999; Huckin & Coady, 1999; Hulstijn, 2003; Letchumanan et al., 2015; Neuman & Koskinen, 1992; Ramos, 2015; Sok, 2014).

These last two factors also underline the fact that extramural contacts might not be suitable for all language learners alike. As empirical research has shown, this might be especially true for auditory and audio-visual input (d'Ydewalle, 2002; d'Ydewalle & van de Poel, 1999; Vidal, 2011). As movies and TV series were not made with the language learner in mind, the high pace, use of less frequent vocabulary, idioms, different dialects, and advanced syntax might simply be too difficult for beginners to follow. Listening to and watching authentic media content in English is, therefore, most likely not suited for low proficiency learners, as they lack the competence to distinguish words in running speech and cannot identify certain word characteristics correctly (d'Ydewalle, 2002; d'Ydewalle & van de Poel, 1999; Vidal, 2011). As a result, learners who have not yet reached the necessary threshold will probably not engage in watching movies and TV series on a regular basis, at least not without subtitles (Webb & Rodgers, 2009).

This problem might be less prominent in books or other forms of written material, in which the reader has more time to engage with the text. However, overall, learners seem to need to have reached a certain level of language proficiency (usually within an educational context) before they can enjoy more complex forms of media content. Otherwise, even the most compelling authentic input will just be incomprehensible noise (Krashen, 1982). This is also emphasized by Neuman and Koskinen (1992), who pointed out the importance of prior knowledge of vocabulary as a moderating variable for incidental learning outcomes. Similarly, Vidal (2011) also found both readers and listeners to benefit from explicit elaboration before the extramural contact. He concluded that explicit (classroom) instruction helps to foster robust connections between form and meaning. Olsson (2016) also suggests that form-focused instruction will enhance the quality and depth of learners’ vocabulary acquisition through incidental learning processes and might help with transforming receptive vocabulary knowledge into productive knowledge. Overall, the findings underline the importance of formal language instruction, especially in the beginning, in order to teach learners the most frequent vocabulary and linguistic principles of the target language (Hulstijn, 2001) as well as providing them comprehensible learning material for their competence level (Krashen, 1985).

In addition to these limitations of the incidental learning process, research has also yet to conclusively prove how incidental learning works within the brain. This is primarily due to the challenges in designing reliable, valid, and objective measurements, as it is difficult to measure what people do and how they deal with an unknown input while making sure that what is measured is, in fact, incidental learning.

Most research in the field of psychology has been experimental in nature, testing participants in a laboratory and sometimes using artificial language to avoid the problem of subjects’ prior knowledge of the language. As a result, findings from these studies cannot easily be generalized to naturalistic contexts (Hulstijn, 2003; Kuppens, 2010).

Most experiments were also only able to provide evidence for short-term effects since they tested participants shortly after exposure to the stimuli (Hulstijn, 2003; Kuppens, 2010). As Vidal points out, the findings might thus only represent the strength of memory traces due to exposure rather than real incidental learning in terms of new lexical entries (Vidal, 2011). Investigating long-term language acquisition would require frequent and intensive contact with a target language. Such intensive exposure is difficult to implement within the confinements of an experimental setting. Still, if people pick up vocabulary or grammar after only a short period of exposure, it is almost certain to assume that more prolonged exposure would result in similar, if not even greater language acquisition (Kuppens, 2010).

Furthermore, most experimental studies tend to have a problem with priming. In order to investigate incidental learning processes, participants cannot be told to read texts and try not to learn something, as that means ‘putting the elephant in the room’ (Bruton et al., 2011). Studies usually ask participants in the experimental group to read a text without telling them that they would be tested afterward, while they instructed the control groups to read a text and announced the post-test beforehand (Hulstijn, 2001, 2003). However, participation alone might be enough to prime participants to expect some kind of test (Sok, 2014). Newer studies usually instruct the experimental group that they will be tested about a certain stimulus and then test a different, second stimulus, for which no test was announced. However, even such experimental designs cannot ensure validity since it cannot be conclusively proven that participants did not have any outside motive to learn. Thus, it is rather difficult to implement a study that can indisputably claim to measure the effect of incidental learning (Hulstijn, 2001; Sok, 2014). Footnote 4

Studies outside of the field of psychology suffer to a lesser degree when it comes to these problems. Instead, they usually struggle to conclusively prove causality. While some studies have implemented quasi-experimental designs (e.g., Kuppens, 2010), most studies were carried out with learners who had already received years of classroom instruction in the target language. In addition, these studies often employed ex-post-facto study designs. It is thus difficult to determine how much of the increase in language competences over a certain period of time is due to extramural contacts and how much must be attributed to students’ prior knowledge and parallel classroom instructions.

Furthermore, while regular extramural English contact can be assumed to increase students’ language competence through incidental language learning processes, it is also very likely that students with high language competences are more likely to engage more frequently in media-related extramural English contact. This is further supported by findings suggesting that authentic media input might be especially challenging for beginners. Consequently, a learner’s language competence and their frequency of media-related extramural English contact will most likely influence each other. As a result of this unclear direction of causality, some of the studies presented above have only reported correlative effects. Thus, while high on ecological validity, most of these studies are relatively low on reliability.

In addition, frequency and form of students’ media-related extramural English contact were often measured via a self-report questionnaire in which students were asked to average their frequency of media contact. A detailed day-to-day analysis of media habits and the specific media content students encounter was therefore often not possible. Thus, some studies cannot assess the true nature and scope of language input students might have had, making definite conclusions about causality impossible. These last two limitations also apply to the present study.

Despite these shortcomings and open (research) questions, the empirical research summarized in this chapter has shown that media-related extramural English contact can have a positive relationship with learners’ language competences. Incidental language learning can most likely be a helpful and interesting route for language learning, especially for more advanced learners. Once students reach a certain level of language proficiency, they will be able to choose from various language sources outside of the educational system, enjoying them for their entertaining characteristics while increasing their language competences, without actively trying to store new information to memory. Learning effects are likely to be strongest for vocabulary, but other areas might also benefit. In addition, newer and more interactive forms of media content might allow learners to produce language output, form hypotheses, and test them in real-life interaction. Through this interaction, learners will also get feedback and assistance from advanced learners and native speakers.

The body of empirical studies summarized above was able to show positive effects for listening, reading, speaking, and even writing skills. Given the highly complex nature of writing in a foreign language, the latter is especially impressive. The present study will analyze the effect of extramural English contact simultaneously on students’ reading, writing, and listening skills (for details on language assessment, see the next chapter). Given the empirical results above, a positive effect of extramural English contacts on all three language skills can be expected. The final research hypothesis is, therefore:

H4: : The frequency of media-related extramural English contact will have a positive effect on students’ reading, listening, and writing skills.

In terms of how much of this process is implicit and how much is acquired through explicit processes, N. C. Ellis (1994) concludes that acquiring vocabulary (i.e., words, collocations, and grammatical class information) might mostly be an implicit process, while for the acquisition of sematic properties and mapping words from context explicit processes are more relevant, see also Rieder (2003). However there is still some doubt if learning without awareness is even possible (R. Ellis, 2008). Since the focus on this study is on incidental learning and not implicit/explicit learning, the distinction will not be discussed in detail here.

Content and language integrated learning can be defined as any form of classroom based instruction in which a foreign/minority or another state language is used as the language of instruction in a non-language related school subject, e.g., biology (Olsson (2016)).

Results reported here are from Sundqvist’s 2009 dissertation. The author has conducted several follow-up studies (Sundqvist, 2008, 2009b, 2011, 2012, 2013). Findings from these other publications will only be reported if they differ from the findings in the main thesis or if they add additional insight.

The same uncertainty seems to arise when it comes to the question of whether operationalizing implicit learning is, in fact, possible. On the other hand, there is consensus that it is possible to operationalize explicit knowledge (Hulstijn, 2002).

Author information

Authors and affiliations.

Berlin, Germany

Maleika Krüger

You can also search for this author in PubMed   Google Scholar

Rights and permissions

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License ( http://creativecommons.org/licenses/by/4.0/ ), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Reprints and permissions

Copyright information

© 2023 The Author(s)

About this chapter

Krüger, M. (2023). Theory of Second Language Acquisition. In: Media-Related Out-of-School Contact with English in Germany and Switzerland. Springer VS, Wiesbaden. https://doi.org/10.1007/978-3-658-42408-4_4

Download citation

DOI : https://doi.org/10.1007/978-3-658-42408-4_4

Published : 23 August 2023

Publisher Name : Springer VS, Wiesbaden

Print ISBN : 978-3-658-42407-7

Online ISBN : 978-3-658-42408-4

eBook Packages : Education and Social Work (German Language)

Share this chapter

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Publish with us

Policies and ethics

• Find a journal
• Track your research

Academia.edu no longer supports Internet Explorer.

To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to  upgrade your browser .

Enter the email address you signed up with and we'll email you a reset link.

• We're Hiring!
• Help Center

Main Theories of Language Acquisition

Related Papers

Leire Lopez

Esmaeil Hassannejad

Mary Ojebisi

In second language acquisition, acquiring English as a second language at an early stage in a bilingual or multilingual environment like Nigeria, has posited quite a number of difficulties over the years. It is even complicated when children with parents from different ethnic groups have to decipher which language is acceptable for communication in the home, school and environment. To determine if the environment influences children's acquisition of English as a second language, certain environmental variables were considered. Children between the ages of four and six who were already in school were selected. Data was collected by administering questionnaires to the teachers of the pupils and recording sessions with the children was observed. The result indicated that the level of competence or performance of a child’s acquisition of English as a second language, is hinged upon the child (if he/she is not mentally deformed or has speech deficiency), the peer group or play mate, the home and the school environment.

Theory and Practice in Language Studies

Davood Mashhadi Heidar

Humanising Language Teaching

Mohammad Amerian

Phenomenology and the Cognitive Sciences

Slobodan Perovic

Michele Merritt

Jonathan D. Nelson

Prelude| A long time ago, in the 1950s, the world of the social sciences was dominated by a fruitless paradigm known as behaviorism. The evil Burrhus Frederic Skinner was its champion, and he banned from the face of the earth all discussion of concepts, ideas, meaning, or mental entities of any sort. Then came Noam Chomsky, the epic linguist and father of cognitive science, who demonstrated the heretofore unseen complexity in human speech.

mannai chakib

Loading Preview

Sorry, preview is currently unavailable. You can download the paper by clicking the button above.

RELATED PAPERS

Ambi A . Izang

Zatusy Naim

Ziyad Rahman

Praca dyplomowa Paulina Woźniak

Paulina Klaudia Woźniak

José Luis Muñoz

wahyu kumalasari

Suren Naicker

Kathy Hirsh-pasek

International Journal of Language & Communication Disorders

Annette Karmiloff-Smith

Behavioral and Brain Sciences

Mohammed Shanawaz

English in the Malaysian Context

Airil Haimi Mohd Adnan

Brett Cumming

Ayokunle Ariyo

Dialogues on Bakhtin: Interdisciplinary readings

Mikael Leiman

Axel Schoelmerich

Daniel M Mukuni

Dr. Minoo Alemi , parisa Daftarifard

Hannele Dufva

joseph galasso

Hameedullah Khan

Christina Behme

William Bechtel

Midwest Studies in Philosophy

Jerry Samet

dayvid junio

Abdurrahman Gülbeyaz

Mind and Language, 17, 3, 233-265

Richard Samuels

mohammad rahimi

Yvan Tovarović

Ingmar Visser

Stephen J. Cowley

Imam Nuraziz

Lando Kirchmair

A Companion to Wittgenstein on Education: Pedagogical Investigations. Eds, M. A. Peters & J. Stickney (Springer, 2016)

Danièle Moyal-Sharrock

Brian MacWhinney

Rijalda Dizdarevic

Parviz Maftoon

KEY IDEAS IN LINGUISTICS

Ana Grace Comendador

Cognitive Development

Thomas Shultz

RELATED TOPICS

•   We're Hiring!
•   Help Center
• Find new research papers in:
• Health Sciences
• Earth Sciences
• Cognitive Science
• Mathematics
• Computer Science
• Academia ©2024

Jump to navigation Jump to Content

• Newsletters

• Supporting ELLs Through COVID-19
• New to Teaching ELLs?
• ELL Glossary
• Special Populations
• Resources by Grade
• ELL Resources by State
• ELL Policy & Research
• ELL News Headlines
• Serving and Supporting Immigrant Students
• Special Education and ELLs
• Social and Emotional Support for ELLs
• ELL School Enrollment
• Bilingual & Dual-Language Education: Overview
• COVID-19 & ELLs
• College Readiness for ELLs
• ELL Program Planning
• Events During the School Year
• Programs for Success
• School Libraries
• ELL Family Outreach
• For Administrators
• For Paraprofessionals
• Supporting Newcomer Students: Resource Gallery
• Creating a Welcoming Classroom
• ELL Classroom Strategy Library
• ELL Strategies & Best Practices
• Language & Vocabulary Instruction
• Literacy Instruction
• Content Instruction for ELLs
• Common Core
• Technology & ELLs
• The Role of ESL Teachers
• Navigating Tough Topics in the Classroom: Tips for ELL Educators
• Help Your Child Learn to Read
• Learning Together at Home
• The Preschool Years
• Schools and Families: An Important Partnership
• School Success
• Raising Bilingual Kids
• Technology at Home
• Resources for Parents of Teens
• Visiting the Public Library
• When Your Children Need Extra Help
• Colorín Colorado Book Finder
• Books for Young Children
• Books for Kids
• Books for Young Adults
• Books for Professionals
• Diverse Books: Resource Guide
• Pura Belpré: Her Life and Legacy
• Literacy Calendar
• Meet the Authors
• Classroom Videos
• Advice for New Teachers of ELLs
• Meet the Experts
• Meet the Administrator
• Facebook Live Series
• PBS Show: Becoming Bilingual
• Being Bilingual Is a Superpower!
• Family Literacy: Multilingual Video Series
• Supporting ELLs During COVID-19: Educator Voices
• Administrators
• Paraprofessionals
• Parents & Families
• ELL Resource Collections
• Guides & Toolkits
• Multilingual Tips for Families
• Reading Tips for Parents (Multilingual)
• Reading Tips for Educators
• Topics A to Z
• Frequently Asked Questions
• Research & Reports
• Web Resources
• Colorín Colorado Blog
• Natural Disaster Resources for Schools
• Responding to a Crisis

Add new comment

Language acquisition: an overview, on this page, stages of language acquisition, instructional strategies, recommendations.

"One generation plants the trees; another gets the shade." — Chinese Proverb

Wouldn't it be just great if learning a new language were that easy (despite the "yuck" factor)? While we do have some technology that provides translation into a variety of languages, it often fails to translate accurately due to the complexity of language. Effective communication requires so much more than just being able to translate vocabulary words — it requires knowledge of intonation, dialect, and intent, and a nuanced understanding of word use, expression, and a language's cultural context. For example, one online translation application I tried translated "Fall Events" as "fall down events" in Spanish because it didn't know that I was referring to events in autumn.

So, without a babel fish or perfect technology, we are left with the old-fashioned way of learning a new language, which requires time, effort, and patience. How much time, effort, and patience depends a lot on the individual who is learning, as well as the learning environment and situation, but language researchers have developed a general outline of language acquisition that helps explain the process that language learners go through to develop skills in a foreign language. In this article, I will provide an overview to the stages of language acquisition, and offer strategies designed to support ELL instruction at different stages of language acquisition.

Researchers define language acquisition into two categories: first-language acquisition and second-language acquisition. First-language acquisition is a universal process regardless of home language. Babies listen to the sounds around them, begin to imitate them, and eventually start producing words. Second-language acquisition assumes knowledge in a first language and encompasses the process an individual goes through as he or she learns the elements of a new language, such as vocabulary, phonological components, grammatical structures, and writing systems.

The Six Stages of Second-Language Acquisition

How long does it take for a language learner to go through these stages? Just as in any other learning situation, it depends on the individual. One of the major contributors to accelerated second language learning is the strength of first language skills. Language researchers such as Jim Cummins, Catherine Snow, Lily Wong Filmore and Stephen Krashen have studied this topic in a variety of ways for many years. The general consensus is that it takes between five to seven years for an individual to achieve advanced fluency. This generally applies to individuals who have strong first language and literacy skills. If an individual has not fully developed first language and literacy skills, it may take between seven to ten years to reach advanced fluency. It is very important to note that every ELL student comes with his or her own unique language and education background, and this will have an impact on their English learning process.

It is also important to keep in mind that the understood goal for American ELL students is Advanced Fluency, which includes fluency in academic contexts as well as social contexts. Teachers often get frustrated when ELL students appear to be fluent because they have strong social English skills, but then they do not participate well in academic projects and discussions. Teachers who are aware of ELL students' need to develop academic language fluency in English will be much better prepared to assist those students in becoming academically successful. (Learn more about academic language in Colorín Colorado's academic language resource section .)

If you have ELL students in your classroom, it is more than likely there will be students at a variety of stages in the language acquisition process. What can teachers do to differentiate instruction according to language level? Here are some suggestions for appropriate instructional strategies according to stages of language acquisition.

Scaffold instruction so students receive comprehensible input and are able to successfully complete tasks at their level. Instructional scaffolding works just like the scaffolding used in building. It holds you at the level needed until you are ready to take it down. Scaffolding includes asking students questions in formats that give them support in answering, such as yes/no questions, one-word identifications, or short answers. It also means providing the context for learning by having visuals or other hands-on items available to support content learning. Also, when practicing a new academic skill such as skimming, scaffolding involves using well-known material so the students aren't struggling with the information while they are trying to learn a new skill. Scaffolding includes whatever it takes to make the instruction meaningful for the student in order to provide a successful learning experience.

Use cognates to help Spanish speakers learn English and derive meaning from content. The Colorín Colorado website has a helpful list of common cognates in Spanish for teachers to reference. Teachers can explicitly point out cognates for Spanish speaking students so they begin to realize that this is a useful way for them to increase their English vocabulary.

Explicit vocabulary instruction is very important in accelerating ELL students' English language development. Textbooks include lists of new vocabulary words based on grade-level content, but ELL students need further vocabulary instruction. There are many words in a text that may affect the ELL student's comprehension of the text that a teacher may assume he or she knows. It is important for teachers to develop ways to help students identify the words they don't know, as well as strategies for getting their meaning. Of course it is also beneficial if teachers reinforce the language structures or common associations of vocabulary. For example, "squeak" is a sound that often goes with "mouse" or "door" and it may be stated as, "squeak, squeaky, squeaks, or squeaked."

Error correction should be done very intentionally and appropriately according to student language ability, as noted earlier in the article. Students who are just beginning to speak English are already nervous about using their new language skills and constant correction will not improve their ability; it will just make them want to withdraw. I inform students in advance of the type of errors I will correct, such as "missing articles" and "third person agreement," and then those are the only errors I check. In my class, I do not correct the errors; I circle the mistakes and return the paper to the student. They are responsible for correcting the errors and returning the paper to receive more points. Most of the time the students can make the corrections themselves when they see the area I've circled, but if they have difficulty, I guide them as they make the correction. In this way, I feel there is a manageable amount of correction information to work with and the student will actually learn from doing the correction.

Learning another language . If you learn the language(s) your students speak, they will be thrilled to hear you try it with them. I learned how to say "good morning" in Somali and had to practice for an hour before I felt comfortable saying it. When I did I was rewarded with the big grins of students as they entered the room. They were excited to teach me other phrases as well, and we discussed how much English they had learned since they arrived in the country. They were very proud to think of how much progress they'd made.

Seek the experts in your building or district who can offer you guidance on effective instructional strategies for your ELL students. There are many teachers who have taught ELL students in your content area, have taught a certain population of students, or are trained ESL or bilingual teachers who have a lot of advice and support to offer. Don't hesitate to look for support when you are challenged to reach students who are learning English. This can be especially true when you have a "pre-production" or "beginning level" student and you are responsible for grade level content instruction.

Visit the hotlinks section for this article for more information on specific research regarding language acquisition and recommended instructional strategies. You can also search the Colorín Colorado educator information for useful information and resources to assist you in meeting ELL student needs.

ELL teachers encounter students with a variety of backgrounds and abilities, and until the babel fish comes into existence, they will need to have flexibility, creativity and skill in order to help ELL students make meaning from the new language and content they are learning. An understanding of the language acquisition process and levels will help teachers tailor instruction to meet the needs of a diverse group of learners. Students will benefit from everything teachers do to support the development of their language skills while teaching them grade level content. Together teachers and students develop their understanding of each other, the world around them, and the language that connects us all.

The Stages of Second Language Acquisition

This chapter from Classroom Instruction That Works with English Language Learners Facilitator's Guide , by Jane D. Hill and Cynthia L. Björk, offers information on the second language acquisition process and effective ELL instruction. It includes a a simplified chart of language acquisition levels and the kinds of language teachers can use to help students at each level.

Simple description of the stages of language acquisition and recommendations for instructional strategies according to level.

Overview of Second Language Acquisition and Strategies

Downloadable booklet from the Northwest Regional Education Laboratory, "Strategies and Resources for Mainstream Teachers of English Language Learners." Includes useful information on language levels, acquisition and 10 things teachers can do today to help ELL students.

Understanding the "Silent Period" with English Language Learners

This article describes some strategies used by two kindergarten teachers to communicate verbally and nonverbally.

More by this author

12 ways classroom teachers can support ells.

• 20 Strategies for School Leaders
• 8 Strategies for Preschool ELLs' Language and Literacy Development
• Addressing ELLs’ Language Learning and Special Education Needs: Questions and Considerations
• An ELL Teacher's Back-to-School Checklist
• Bilingual Family Night for ELL Families
• Communicating Important Information with ELL Families: Strategies for Success
• Communication Strategies for All Classrooms: Focusing on English Language Learners and Students with Learning Disabilities
• Connecting with ELL Families: Strategies for Success
• Creating a Plan of Action for ELL Family Engagement
• Differentiated Instruction for English Language Learners
• Distance Learning for ELLs: Challenges and Opportunities
• Distance Learning for ELLs: Lessons Learned About Family Partnerships
• Distance Learning for ELLs: Planning Instruction
• Distance Learning Strategies for ELLs: What You Need to Know
• Early Literacy Instruction in Dual Language Preschools (Spanish/English)
• Early Literacy Instruction in Spanish: Teaching the Beginning Reader
• Eight English Learner Myths for Administrators
• ELLs and Reading Fluency in English
• ELLs in Early Childhood Education: Recruiting Immigrant Families
• Encouraging and Sustaining ELL Parent Engagement
• Engaging ELL Families Through Community Partnerships
• Engaging ELL Parents as Leaders
• Essential Actions: 15 Research-based Practices to Increase ELL Student Achievement
• Fifteen Things Teachers Can Do to Support ELL Success in the New Year
• Final Thoughts
• Finding Answers for Our Immigrant Students and Families: An ELL District Leader's Perspective
• Fostering Literacy Development in English Language Learners
• Gear Up for a New School Year!
• Getting Ready for College: What ELL Students Need to Know
• How to Connect ELLs' Background Knowledge to Content
• How to Support ELL Students with Interrupted Formal Education (SIFEs)
• How to Support Refugee Students in Your School Community
• Improving Writing Skills: ELLs and the Joy of Writing
• Increase Student Interaction with "Think-Pair-Shares" and "Circle Chats"
• Increasing Academic Language Knowledge for English Language Learner Success
• Increasing ELL Student Reading Comprehension with Non-fiction Text
• Introduction: Strategies for Engaging ELL Families
• Math Instruction for English Language Learners
• Motivating ELL Student Readers
• Music and Language Learning
• No Child Left Behind and the Assessment of English Language Learners
• Phonics Instruction for Middle and High School ELLs
• Preparing an Engaging Social Studies Lesson for English Language Learners
• Preparing ELLs to be 21st-Century Learners
• Reader's Theater: Oral Language Enrichment and Literacy Development for ELLs
• Reading 101 for English Language Learners
• Successful Field Trips with English Language Learners
• Successful Parent-Teacher Conferences with Bilingual Families
• Summer Reading: English Language Learners at the Library
• Tales of a Fourth Grade Slump: How to Help ELL Students Leap to Success
• Teaching and Reading Poetry with English Language Learners
• Technology Review: Ellevation Software
• Ten Things Parents Should Know About the Common Core State Standards
• Tips for Successful Parent-Teacher Conferences with Bilingual Families
• Visual Thinking Strategies for Improved Comprehension
• Writing a Winning Essay
• Writing Poetry with English Language Learners

Major support provided by our founding partner, the American Federation of Teachers, AFL-CIO.

With generous support provided by the National Education Association.

Related Content

• Making a Case for Supporting Students' Home Languages
• The Home Language: An English Language Learner's Most Valuable Resource
• The changing ELL field: Learning a second language
• The changing ELL field: Seeing language and culture as assets

Most Popular

10 Strategies for Building Relationships with ELLs

Culturally Responsive Instruction for Holiday and Religious Celebrations

Language Objectives: The Key to Effective Content Area Instruction for English Learners

edward omar replied on Thu, 2009-10-01 00:05 Permalink

Thank you for having this web page!!!!

Harold Butts replied on Fri, 2009-10-16 03:08 Permalink

Great reference material. I will implement this knowledge in my lesson plans.

Beth replied on Fri, 2009-12-18 01:02 Permalink

These stages and strategies apply to children and youth as well as adult learners of a second language. This is a very concise and useful article.

khalida13 replied on Tue, 2012-01-10 09:26 Permalink

i need your help in first language acqiusition help me pleaze

chimanma replied on Tue, 2012-03-06 07:00 Permalink

please specify categories,as in names for first language acquisition

Ebrahim Zangani replied on Wed, 2013-04-03 02:04 Permalink

This is a preliminary and laconic information regarding SLA. Please specify the areas of further investigation in SLA.

Anonymous replied on Fri, 2013-12-13 19:49 Permalink

Excellent!! It is very useful in my studies about the acquistion of a second language. Thanks a lot!!!

Stephen Kumureba replied on Sat, 2013-12-21 10:31 Permalink

Helped by this page, but cater for 1st language acquisition also.

BENEDICT E. AMADI replied on Sat, 2014-04-05 16:31 Permalink

I love what I have seen. Interesting.

lellingw replied on Tue, 2014-08-19 12:45 Permalink

This page has been up for awhile but how is language acquisition coupled with error correction? And explicit vocabulary instruction, neither of which has been proven to be very effective and NOT part of acquiring a language. Some didn't read their research correctly.

ELIZABETHROA replied on Tue, 2014-10-21 23:23 Permalink

EXCELLENT!!

Kathy replied on Wed, 2014-11-19 17:09 Permalink

Should I be assessing phonics skills with spelling words or sentences? Does it matter?

Kathy replied on Wed, 2014-11-19 17:10 Permalink

peter replied on Mon, 2014-11-24 20:57 Permalink

Can somebody tell me the complete reference of this article? I need the reference for an article that Im writting in the college, but I need to citate very well.

I will thank you a lot.

Julie I. replied on Sat, 2014-11-29 01:00 Permalink

Peter, did you ever figure out how to cite this article, I am planning on using it as well for a powerpoint and would like to reference correctly...

stefanie replied on Fri, 2015-03-20 09:35 Permalink

most of these hot links are no longer good :O( It would be great if someone could update them!

Mary Ellen M. replied on Sun, 2019-04-28 15:37 Permalink

Thank you for the validation on past practices and the continuation of guided , differenciated teaching practices for ELL students at any level of instruction. Being mindful of acquisition of language, for any student, and strategic scaffolding, is key to any language experience in the classroom.

More information about text formats

• No HTML tags allowed.
• Web page addresses and e-mail addresses turn into links automatically.
• Lines and paragraphs break automatically.

24/7 writing help on your phone

To install StudyMoose App tap and then “Add to Home Screen”

The Theories of Language Acquisition

Save to my list

Remove from my list

The Behaviorist Theory

The Nativist Theory

• Language Acquisition: An Overview
• Self-Serving Bias: What It Is, Examples, Negative and Positive Effects

The Theories of Language Acquisition. (2020, Sep 14). Retrieved from https://studymoose.com/the-theories-of-language-acquisition-essay

"The Theories of Language Acquisition." StudyMoose , 14 Sep 2020, https://studymoose.com/the-theories-of-language-acquisition-essay

StudyMoose. (2020). The Theories of Language Acquisition . [Online]. Available at: https://studymoose.com/the-theories-of-language-acquisition-essay [Accessed: 9 Jun. 2024]

"The Theories of Language Acquisition." StudyMoose, Sep 14, 2020. Accessed June 9, 2024. https://studymoose.com/the-theories-of-language-acquisition-essay

"The Theories of Language Acquisition," StudyMoose , 14-Sep-2020. [Online]. Available: https://studymoose.com/the-theories-of-language-acquisition-essay. [Accessed: 9-Jun-2024]

StudyMoose. (2020). The Theories of Language Acquisition . [Online]. Available at: https://studymoose.com/the-theories-of-language-acquisition-essay [Accessed: 9-Jun-2024]

• The Main Theories in Second Language Acquisition Pages: 10 (2764 words)
• Theories of Knowledge Acquisition in Philosophy Pages: 3 (674 words)
• Second Language Acquisition Pages: 8 (2178 words)
• Language Acquisition by Learning Linguistic Gambits Pages: 11 (3125 words)
• Second Language Acquisition: Processs and Hypotheses Pages: 9 (2541 words)
• Language Acquisition According to Interlanguage Theory Pages: 5 (1389 words)
• Language Acquisition Pages: 5 (1321 words)
• Noam Chomsky’s Theory on Language Acquisition Pages: 8 (2166 words)
• Vygotsky on Child Cognitive Development & Language Acquisition Pages: 6 (1509 words)
• Language Acquisition and Literal Translation Pages: 2 (406 words)

👋 Hi! I’m your smart assistant Amy!

Don’t know where to start? Type your requirements and I’ll connect you to an academic expert within 3 minutes.

You are using an outdated browser. Please upgrade your browser or activate Google Chrome Frame to improve your experience.

An Introduction to Language Acquisition

First language or native language acquisition is the process of building the ability to understand a language and use it to communicate with others . It’s how a baby grows from a wordless wonder into somebody who can’t stop talking during class.

And it starts earlier than you might think . Studies have shown that fetuses in the third trimester can hear and learn to distinguish the vowel sounds and rhythms of their native language versus a foreign language. That means they’re born primed to learn their first language!

And babies may begin to understand the meanings of common nouns as early as 6-9 months of age , before they start talking at about 9-15 months.

When you acquired your native tongue, you didn’t need a thick grammar textbook full of highlights or a long list of vocabulary to memorize.

Your learning was instinctive and unconscious . You were just living with your parents, who naturally talked to you about everything from what to eat to how to play and when to sleep. You probably can’t even remember when you started using their words back.

If you are born in Korea to parents who speak Korean with you, you’ll naturally end up speaking Korean. The same goes for whatever native language you grow up hearing.

What Is Foreign Language Acquisition?

Who can acquire a foreign language, theories about how language acquisition occurs.

• Behaviorism (B.F. Skinner)
• Universal Grammar (Noam Chomsky)
• Cognitive Theory (Jean Piaget)

The 5 Components of a Language to Acquire

2. semantics, 3. morphology, 4. phonology, 5. pragmatics, the language skills you’ll need to acquire, and one more thing....

Download: This blog post is available as a convenient and portable PDF that you can take anywhere. Click here to get a copy. (Download)

Foreign language acquisition is the one that happens after you’ve acquired your native tongue. It builds on the existing language framework in your brain .

In contrast with first language acquisition, second language learning usually happens when you’re older , maybe inside a school or university classroom, or nowadays even a virtual one. Maybe you learn a new language because your new job requires you to speak with customers who don’t use your first language. Or maybe you just want to learn how to flirt in a new language.

• Thousands of learner friendly videos (especially beginners)
• Handpicked, organized, and annotated by FluentU's experts
• Integrated into courses for beginners

Whatever the reason, the methods for learning a second language are conscious and with a purpose .

You actively study grammar from textbooks. You have your word lists and flashcards with their corresponding pictures and translations. You have apps, podcasts and YouTube videos.

Most readers of this blog are probably in this same boat, tremendously enriching their lives by learning a second (or third) language.

It’s true that language acquisition is most effective in the “critical period” of early childhood, when our highly elastic brains absorb language like a sponge. Afterward, it’s comparatively more difficult. This has led many to believe that learning a language is the sole province of the young.

But while it’s true that our brains rapidly develop in our early years, it doesn’t lose plasticity over our lifetimes . We can create novel neural connections and learn something new at any age. That means you can embark on a language learning journey at any stage in life , your stabilized brain notwithstanding.

Studies have identified factors that exert a stronger influence than age on an individual’s language performance. For example, one study found that a person’s motivation is a better predictor of linguistic success than age .

• Interactive subtitles: click any word to see detailed examples and explanations
• Slow down or loop the tricky parts
• Show or hide subtitles
• Review words with our powerful learning engine

What is it that drives you to learn the second language? What gets you over the speed bumps? Why do you do it when you could’ve done something else? These are more important than what you write in the blank after “Age.”

Quality of inputs is another factor that does better than age to predict language acquisition. That is, even if you start learning a language later in life, you can still be better off than those who started early, as long as you spend considerable time interacting with native speakers or use authentic materials in your study. The quality of inputs determines your linguistic success.

So really, it’s not that second language acquisition is only for the young or the gifted. It’s just that we need the right tools and the drive to do it.

But whether it’s first or second language acquisition, how do these processes actually take place in the mind of a language learner? Psychologists and linguists have put forth several theories over the decades to explain the phenomenon, and we’re going to look into three of the most influential ones in the next section.

Philosophers have always been fascinated by human linguistic ability, particularly its initial acquisition.

Ever since Socrates intoned “Know thyself,” we have tried to peek behind the curtain and find out how we are actually able to learn language and use it for a myriad of communicative purposes.

Here are some theories on the matter:

• Learn words in the context of sentences
• Swipe left or right to see more examples from other videos
• Go beyond just a superficial understanding

Behaviorism ( B.F. Skinner )

Behaviorism is a school of psychology that had its heyday from the 1900s to the 1950s and still holds some sway in how we think about language acquisition.

In a nutshell, behaviorism attributes animal and human behavior to cause and effect or, in other words, stimulus and response. Some responses are natural, and others are learned, or conditioned .

For example, salivating is a natural response to the stimulus of food, but not to hearing a doorbell. However, if you like to order in a lot, you might learn that the sound of your doorbell means your food has arrived. If you start salivating at the sound of that bell, congratulations! You’ve been conditioned to associate the doorbell with food.

B.F. Skinner, an eminent behaviorist, proposed that language acquisition is really one big and complex case of conditioning .

According to behaviorism, babies first learn the meanings of words via association . If a baby hears the word “milk” often enough right before being fed from the bottle, he’ll soon associate the word “milk” with drinking milk. If he always hear the word “ball” right before being handed a spherical object, he’ll begin to associate “ball” with the object.

A baby will then learn to use words by imitating the adults around them. At first they just babble, but when they make the right sounds, their parents will react by smiling, praising them or giving them what they want . Getting a reward for a correct behavior is a form of conditioning called positive reinforcement.

The child continues to learn the correct form of his language by trial and error , receiving positive reinforcement (a reward) when he uses correct grammar and negative reinforcement (lack of reward) or correction when he gets it wrong.

• FluentU builds you up, so you can build sentences on your own
• Start with multiple-choice questions and advance through sentence building to producing your own output
• Go from understanding to speaking in a natural progression.

In the behaviorist view, language is simply reinforced imitation .

Universal Grammar ( Noam Chomsky )

In the 1960s, the field of behaviorism came under serious attack from the likes of Noam Chomsky, a man recognized as the father of modern linguistics, and about as decorated a scholar as any.

He pointed out that if you look closer, parents really give very little linguistic input for tots to imitate directly. Chomsky argued that parent-child interactions are limited to repeated utterances of things like “Put that back” and “Open your mouth”—not very likely to make significant dents towards the cause of language learning. And besides, when a child says, “I swimmed today,” he didn’t get that from any adult figure in his life. That’s not imitation.

So how does one account for the fact that children learn to speak their native tongues in spite of the “poverty of the stimulus”? One is left with the conclusion, Chomsky argues, that if not from the outside, external input, then the ability must have been there all along .

Chomsky asserts that human beings are biologically wired for language—that we have a “language acquisition device” that allows us to learn any language in the world. Linguistic ability is innate to us.

Proof of this are the emergent abilities that have no external source. For example, how do children make out the individual words in the strings of sounds that they hear? Reading and writing are learned later, so they can’t have worked it out by seeing separate words on a page.

• Images, examples, video examples, and tips
• Covering all the tricky edge cases, eg.: phrases, idioms, collocations, and separable verbs
• No reliance on volunteers or open source dictionaries
• 100,000+ hours spent by FluentU's team to create and maintain

Chomsky would argue that children use this “language acquisition device” to figure out the rules specific to their native language. He even goes on to assert that there is such a thing as a “Universal Grammar.” For how else did the different languages end up with the same categorization of words (nouns, verbs, adjectives, etc.) when there’s an infinite number of ways words can be categorized? We always have nouns, verbs and adjectives.

Chomsky’s work represented the “nature” side of the “ nature-nurture debate ,” while the behaviorists account for language as part of “nurturing.”

Of course, because of its sweeping and seemingly simplistic assertions, Chomsky’s theory has its own set of strong dissenters. Let’s talk about them next.

Cognitive Theory ( Jean Piaget )

Your churning brain might already be asking any number of questions:

“So what proof do we have for this ‘language acquisition device’? Where in the brain is it located? Can we see it in action?”

“Have we studied all the languages of the world to conclude that there is indeed ‘Universal Grammar’?”

These and other queries prompted a whole different approach to the question of language acquisition. And as is often the case, subsequent theories point out the weaknesses of those that came before them.

Chomsky’s theory did that to behaviorism, and in turn, those that follow will try to fill in the gaps. And instead of taking a side on the nature-nurture debate, the cognitive theory of language acquisition recognizes that both processes have their roles to play.

The psychologist Jean Piaget is a major proponent of this cognitive model, which sees language acquisition in light of developing mental capacities. The idea here is that we’re able to learn language because of our ability to learn . It’s because of our cognitive development. Our brains become more complex, and we learn so many things so fast.

Babies initially don’t talk because their brains and mental capacities still lack the experience and scaffolding necessary for language. But as babies grow, as they interact with adults, as they gain more experience, as they observe more things and as they learn more concepts, language becomes the inevitable result.

Piaget believed that the understanding of concepts must come before language . When a child says, “Ball is red,” he must first understand what a ball and the color red are before he can comment.

So if you notice how language develops, it follows the complexity of our thinking. The more nuanced and layered our thinking, the more textured the language that comes out . That’s why children talk one way, and adults talk a different way.

In this model, language is seen as part of our advancing mental capacities—alongside our ability to reason or to think in the abstract . We are rational beings, information processors that interact and learn from experience.

Those are three of the most influential theories on language acquisition. Each has its merits and each gives a certain view of how we learn language. Needless to say, more research and study is needed on the topic. There’s still so much to discover, and so much to learn in this area of linguistics.

When we say “language acquisition,” what is it exactly that we acquire? Well, we now go to the next section to find out.

Here we get into the nitty-gritty of languages, and look under the hood to see their basic parts.

Languages are governed by rules. Without them, the utterances of one person would be random and meaningless to anyone else.

We need to meet the rules languages follow, behind the scenes, in order to have a proper appreciation of them. I’m talking here about the five components of a language: syntax, semantics, phonology, morphology and pragmatics. Whatever language you’re considering, it has them.

Syntax is how words and phrases are arranged to create a grammatically correct sentence within a language. It also encompasses the parts of speech and other categories words and phrases can occupy.

Because of the specific ways the elements of speech are arranged, we can decipher meaning and understand each other. For example, take the English sentence: “The dog saw the cat.”

English uses subject-verb-object syntax. Thus, we know the dog is the subject of the sentence and the cat is the object. The verb “saw” is what the dog did and what was done to the cat.

If the sentence were written, “Saw the dog the cat” or “The cat the dog saw,” we’d have a heck of a time figuring it out, because it doesn’t follow English syntax rules.

But what is a dog, and what is seeing? The meaning of words is the subject of the next category.

Semantics is all about meaning in a language—what words, phrases and sentences actually signify.

“Shoulder,” for example, is a noun that signifies the part of human anatomy where an arm connects to the body. Its semantic properties include: “connection of forelimb to body,” “outsidemost part,” “burden-bearing part” and more.

We can talk about “a pork shoulder roast,” “the shoulder of the highway” or “shouldering a responsibility” and still be understood, because the word retains its semantic properties across contexts and parts of speech.

But we can’t say shoulder if we mean tree, because they don’t share semantic properties.

In addition to single-word semantics, there are phrase and sentence semantics . These work hand in hand with syntax because different arrangement of words can create different meanings. For example, we have a sentence:

“She tapped him on the shoulder.”

Let’s say we’ll insert the word “ only ” somewhere in the statement. Notice how this changes the whole meaning and complexion of the statement, depending on where exactly we place a single word.

She only tapped him on the shoulder. (She didn’t punch him.)

She tapped only  him on the shoulder. (Nobody else got a similar treatment.)

She tapped him only  on the shoulder. (Not on his head or anywhere else.)

Meaning can change depending on how you arrange specific words. And not only that, meaning can also change depending on the form of individual words. Let’s talk about that next.

Morphology is about the form of words. It’s best observed in the written form of a language. Change in form often brings with it a change in meaning.

Root words —the most basic word forms—can be modified with prefixes and suffixes to form new words, each with a different meaning. A single root word can give birth to many new words, and that’s where the linguistic fun begins.

Take the root word “ drive .”

Add “r” at the end and you have “ driver .” From a verb, your word has become a noun, a person.

Next, add “s” to your newly formed word and you have “ drivers .” You’ve just turned a single person into multiple people by using the plural form of the word.

Change “i” to “o” and you have “ drove .” From a verb in the present tense, you introduced a time change and turned it into a past tense.

You can do many things with the root word “ drive ” and come up with new words like:

• driveability

That’s what morphology is all about. Different meanings come from different word forms. Speaking of forms, when spoken, each of these new words will inevitably sound different. That’s what the next language characteristic is all about.

Phonology is the study of linguistic sounds . And if ever you want to be considered fluent in your target language, you have to be very familiar with the intonations, stresses, pauses, dips and tones of the language.

To sound like a native speaker, you have to pronounce words, phrases and sentences like they do. There are specific sounds and sound patterns that exist in a language. For example, Spanish, Italian and Portuguese have rolling “R’s” that give some English speakers a heck of a time.

In languages like Italian, you oftentimes only need to look at how a word is spelled (morphology) in order to know how it should be pronounced. In other words, in those languages there’s a close correspondence between the language’s written form and its spoken form.

In the case of English, though, if you could guess the correct pronunciations of words like “though,” “rough” and “bough” and  based on spelling alone, you’d stand a good chance of winning the lottery, too.

Pragmatics is concerned with how meaning is negotiated between speaker and listener. This is the part of language that is not spoken, but implied based on context . It’s how we can say one thing and mean another.

When your boss, after reading your submitted proposal, tells you, “This won’t work. Go back to square one,” you begrudgingly know what he means. You don’t take his words literally and look for “square one.” You start again.

Or when you’re hours late for a date with your wife and she asks you, “Do you know what time it is?” you know better than to give her the exact time. You know a rhetorical question when you hear one.

Pragmatics lends languages levity, so we don’t get stuck with being so literal all the time. You know you’re fluent in a language when you understand idiomatic expressions, sarcasm and the like.

Now that we know about the five characteristics of languages, we get to the four modalities in which language acquisition can be judged: listening, speaking, reading and writing.

How do you know if or when you’ve acquired a language?

That’s a difficult question to answer. When you get down to it, language acquisition isn’t an either-or kind of thing, but rather a continuum, and language learners stand at various stages of acquisition.

And to make things a little bit more complicated, there are four basic language modalities or skills involved: listening, speaking, reading and writing. They’re closely related, but still clearly different. You may have thought of “language acquisition” in terms of speaking ability, but it’s just one of four competencies considered.

Let’s look at them.

We know that listening is the first language skill to be developed . Before babies can even talk, they’ve already logged serious hours listening. They listen to how their parents talk, to the intonations and pauses, and take their cues as to the speaker’s emotions.

Babies have this “silent phase” when they simply give you those cute bright eyes. No words are spoken. But you know something is happening inside those brains because one day, they just start babbling—something unintelligible at first, then gradually moving into their first words, like wooden sculptures slowly arising from individual blocks of wood.

Listening has often been mistaken for a passive activity, where you just sit there and orient your ears to the audio. You can even sleep if you want to. But nothing is further from the truth.

To listen effectively, you have to be actively into it. You need to listen for specific things: intonations, motivations, emotions, accents and the natural flow of sound.

A language has a specific musicality unique to it. It’s not just about vocabulary. To be fluent, you need to be aware not only of the words but also of the sounds of those words. And the only way you can hone this skill is by investing the time into listening to both authentic sources and study materials .

You can for example use an audio-based study program like Pimsleur . Listen to it on your commute. For authentic material, you can get podcasts produced by your target language’s native speakers.

And just because you’re listening doesn’t mean you have to limit yourself to audio . There are language learning programs like FluentU that offer authentic videos of all sorts.

FluentU takes authentic videos—like music videos, movie trailers, news and inspiring talks—and turns them into personalized language learning lessons.

You can try FluentU for free for 2 weeks. Check out the website or download the iOS app or Android app.

P.S. Click here to take advantage of our current sale! (Expires at the end of this month.)

Try FluentU for FREE!

At first, you don’t need to go for complete comprehension of what you’re listening to. Heck, you don’t even need to work out the individual words. Close your eyes and consciously notice the dips and rises of the tone. Notice, for example, how the tone evolves from the beginning of a sentence to how it ends.

You have to invest time in this. That is, you do if you want to sound like a native speaker.

Speaking is probably what you think of when we mention “language acquisition.” It is, after all, the most vivid proof of your linguistic chops. There’s nothing like speaking fluent Mandarin to impress a date—never mind that what you actually said was the equivalent of “Where’s the bathroom?”

Ironically, although speaking may be the end goal for many language learners, many devote very little study time to it . Many learners instead dive deep into vocabulary and grammar.

I’m not saying you shouldn’t do that. Vocabulary helps on all fronts—listening, speaking, reading and writing. But it doesn’t score a frontal hit on the main goal of speaking .

The thing that stops language learners is usually embarrassment . Even when we’re totally alone, we’re worried that somebody from far away might hear us butcher the pronunciation of a single word.  You don’t want to mess it up, so you put it off, focusing on word lists, perfect grammar or anything else to avoid opening your mouth.

But the truth is that messing it up is a necessary step on the road to getting good, and it’s nothing to be ashamed of .

Babies don’t have those hangups. They babble away, butchering their mother tongues all day long, while their egos remain intact. Is it any wonder why they acquire their language so easily?

Speaking is a physical phenomenon , so you need to actually practice getting your vocal ensemble—your tongue, mouth, teeth and palate—to move the way native speakers move theirs. You need to feel what it’s like saying those words. You need to hear yourself speak.

To learn to speak, you need to open your mouth . There’s just no way around it.

Being able to read in a second language  opens up a whole new world of literature to you.

Imagine being able to read “The Three Musketeers” in the original French or Dante’s “Divine Comedy” in the original Italian. There’s just nothing like a helping of those works in the language in which they were written, because there are some things that just can’t be adequately translated .

Thankfully, all your time studying vocabulary and grammar rules works in favor of reading comprehension.

In addition, you can gradually build your comprehension prowess by starting off with dual-language books . These are books that give you a line-by-line translation of the story. You can compare and contrast the languages as you go along.

Next in this build-up would be children’s books in the target language only . Children’s books will be easy enough for you to read. Choose stories you’re familiar with so you can do away with plot-guessing and focus on learning.

And remember, just to practice moving your mouth in the target language, try reading aloud the text in front of you. That way, you’re hitting two birds with one stone.

Many consider the ability to write in another language the apex of language acquisition. Maybe they’re thinking about writing in terms of epic volumes, academic in nature, read and revered by one generation and the next.

Here we’re talking about writing in more prosaic terms.

Writing , in many respects, can actually be easier than speaking the target language. With the written form, language learners actually have a visible record in front of them. Written texts are more malleable than spoken words. You can scratch written texts, reorder them and correct their tenses and conjugations.

Again, vocabulary and grammar training help a lot to build this skill.

In addition, you can practice writing by doing short paragraphs on things like :

• My Perfect Day
• My Secret Hobby
• Why I love “Terminator 3”

Your work may not become a fixture in the language classes of the future, but the cool thing about writing is that the more you write, the better you become at expressing yourself in the target language . This inevitably helps in honing the other communication skills, like speaking on the fly, understanding content written by others and listening to native material.

Now you know a lot about language acquisition—from the theories about it, to the differences between native language and second language acquisition, to the five characteristics of languages and the four linguistic skills to hone . I’m hoping that, if anything, this piece has sparked more interest and desire in you to learn the languages of the world .

Happy learning!

If you dig the idea of learning on your own time from the comfort of your smart device with real-life authentic language content, you'll love using FluentU .

With FluentU, you'll learn real languages—as they're spoken by native speakers. FluentU has a wide variety of videos as you can see here:

FluentU App Browse Screen.

FluentU has interactive captions that let you tap on any word to see an image, definition, audio and useful examples. Now native language content is within reach with interactive transcripts.

Didn't catch something? Go back and listen again. Missed a word? Hover your mouse over the subtitles to instantly view definitions.

Interactive, dual-language subtitles.

You can learn all the vocabulary in any video with FluentU's "learn mode." Swipe left or right to see more examples for the word you’re learning.

FluentU Has Quizzes for Every Video

And FluentU always keeps track of vocabulary that you’re learning. It gives you extra practice with difficult words—and reminds you when it’s time to review what you’ve learned. You get a truly personalized experience.

Start using the FluentU website on your computer or tablet or, better yet, download the FluentU app from the iTunes or Google Play store. Click here to take advantage of our current sale! (Expires at the end of this month.)

Related posts:

Enter your e-mail address to get your free pdf.

We hate SPAM and promise to keep your email address safe

Language Acquisition Theories Essays

Exploring language acquisition theories, popular essay topics.

• American Dream
• Artificial Intelligence
• Black Lives Matter
• Bullying Essay
• Career Goals Essay
• Causes of the Civil War
• Child Abusing
• Civil Rights Movement
• Community Service
• Cultural Identity
• Cyber Bullying
• Death Penalty
• Depression Essay
• Domestic Violence
• Freedom of Speech
• Global Warming
• Gun Control
• Human Trafficking
• I Believe Essay
• Immigration
• Importance of Education
• Israel and Palestine Conflict
• Leadership Essay
• Legalizing Marijuanas
• Mental Health
• National Honor Society
• Police Brutality
• Pollution Essay
• Racism Essay
• Romeo and Juliet
• Same Sex Marriages
• Social Media
• The Great Gatsby
• The Yellow Wallpaper
• Time Management
• To Kill a Mockingbird
• Violent Video Games
• What Makes You Unique
• Why I Want to Be a Nurse
• Send us an e-mail

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 03 June 2024

Applying large language models for automated essay scoring for non-native Japanese

• Wenchao Li 1 &
• Haitao Liu 2  

Humanities and Social Sciences Communications volume  11 , Article number:  723 ( 2024 ) Cite this article

254 Accesses

2 Altmetric

Metrics details

• Language and linguistics

Recent advancements in artificial intelligence (AI) have led to an increased use of large language models (LLMs) for language assessment tasks such as automated essay scoring (AES), automated listening tests, and automated oral proficiency assessments. The application of LLMs for AES in the context of non-native Japanese, however, remains limited. This study explores the potential of LLM-based AES by comparing the efficiency of different models, i.e. two conventional machine training technology-based methods (Jess and JWriter), two LLMs (GPT and BERT), and one Japanese local LLM (Open-Calm large model). To conduct the evaluation, a dataset consisting of 1400 story-writing scripts authored by learners with 12 different first languages was used. Statistical analysis revealed that GPT-4 outperforms Jess and JWriter, BERT, and the Japanese language-specific trained Open-Calm large model in terms of annotation accuracy and predicting learning levels. Furthermore, by comparing 18 different models that utilize various prompts, the study emphasized the significance of prompts in achieving accurate and reliable evaluations using LLMs.

Similar content being viewed by others

Accurate structure prediction of biomolecular interactions with AlphaFold 3

Testing theory of mind in large language models and humans

Scaling neural machine translation to 200 languages

Conventional machine learning technology in aes.

AES has experienced significant growth with the advancement of machine learning technologies in recent decades. In the earlier stages of AES development, conventional machine learning-based approaches were commonly used. These approaches involved the following procedures: a) feeding the machine with a dataset. In this step, a dataset of essays is provided to the machine learning system. The dataset serves as the basis for training the model and establishing patterns and correlations between linguistic features and human ratings. b) the machine learning model is trained using linguistic features that best represent human ratings and can effectively discriminate learners’ writing proficiency. These features include lexical richness (Lu, 2012 ; Kyle and Crossley, 2015 ; Kyle et al. 2021 ), syntactic complexity (Lu, 2010 ; Liu, 2008 ), text cohesion (Crossley and McNamara, 2016 ), and among others. Conventional machine learning approaches in AES require human intervention, such as manual correction and annotation of essays. This human involvement was necessary to create a labeled dataset for training the model. Several AES systems have been developed using conventional machine learning technologies. These include the Intelligent Essay Assessor (Landauer et al. 2003 ), the e-rater engine by Educational Testing Service (Attali and Burstein, 2006 ; Burstein, 2003 ), MyAccess with the InterlliMetric scoring engine by Vantage Learning (Elliot, 2003 ), and the Bayesian Essay Test Scoring system (Rudner and Liang, 2002 ). These systems have played a significant role in automating the essay scoring process and providing quick and consistent feedback to learners. However, as touched upon earlier, conventional machine learning approaches rely on predetermined linguistic features and often require manual intervention, making them less flexible and potentially limiting their generalizability to different contexts.

In the context of the Japanese language, conventional machine learning-incorporated AES tools include Jess (Ishioka and Kameda, 2006 ) and JWriter (Lee and Hasebe, 2017 ). Jess assesses essays by deducting points from the perfect score, utilizing the Mainichi Daily News newspaper as a database. The evaluation criteria employed by Jess encompass various aspects, such as rhetorical elements (e.g., reading comprehension, vocabulary diversity, percentage of complex words, and percentage of passive sentences), organizational structures (e.g., forward and reverse connection structures), and content analysis (e.g., latent semantic indexing). JWriter employs linear regression analysis to assign weights to various measurement indices, such as average sentence length and total number of characters. These weights are then combined to derive the overall score. A pilot study involving the Jess model was conducted on 1320 essays at different proficiency levels, including primary, intermediate, and advanced. However, the results indicated that the Jess model failed to significantly distinguish between these essay levels. Out of the 16 measures used, four measures, namely median sentence length, median clause length, median number of phrases, and maximum number of phrases, did not show statistically significant differences between the levels. Additionally, two measures exhibited between-level differences but lacked linear progression: the number of attributives declined words and the Kanji/kana ratio. On the other hand, the remaining measures, including maximum sentence length, maximum clause length, number of attributive conjugated words, maximum number of consecutive infinitive forms, maximum number of conjunctive-particle clauses, k characteristic value, percentage of big words, and percentage of passive sentences, demonstrated statistically significant between-level differences and displayed linear progression.

Both Jess and JWriter exhibit notable limitations, including the manual selection of feature parameters and weights, which can introduce biases into the scoring process. The reliance on human annotators to label non-native language essays also introduces potential noise and variability in the scoring. Furthermore, an important concern is the possibility of system manipulation and cheating by learners who are aware of the regression equation utilized by the models (Hirao et al. 2020 ). These limitations emphasize the need for further advancements in AES systems to address these challenges.

Deep learning technology in AES

Deep learning has emerged as one of the approaches for improving the accuracy and effectiveness of AES. Deep learning-based AES methods utilize artificial neural networks that mimic the human brain’s functioning through layered algorithms and computational units. Unlike conventional machine learning, deep learning autonomously learns from the environment and past errors without human intervention. This enables deep learning models to establish nonlinear correlations, resulting in higher accuracy. Recent advancements in deep learning have led to the development of transformers, which are particularly effective in learning text representations. Noteworthy examples include bidirectional encoder representations from transformers (BERT) (Devlin et al. 2019 ) and the generative pretrained transformer (GPT) (OpenAI).

BERT is a linguistic representation model that utilizes a transformer architecture and is trained on two tasks: masked linguistic modeling and next-sentence prediction (Hirao et al. 2020 ; Vaswani et al. 2017 ). In the context of AES, BERT follows specific procedures, as illustrated in Fig. 1 : (a) the tokenized prompts and essays are taken as input; (b) special tokens, such as [CLS] and [SEP], are added to mark the beginning and separation of prompts and essays; (c) the transformer encoder processes the prompt and essay sequences, resulting in hidden layer sequences; (d) the hidden layers corresponding to the [CLS] tokens (T[CLS]) represent distributed representations of the prompts and essays; and (e) a multilayer perceptron uses these distributed representations as input to obtain the final score (Hirao et al. 2020 ).

AES system with BERT (Hirao et al. 2020 ).

The training of BERT using a substantial amount of sentence data through the Masked Language Model (MLM) allows it to capture contextual information within the hidden layers. Consequently, BERT is expected to be capable of identifying artificial essays as invalid and assigning them lower scores (Mizumoto and Eguchi, 2023 ). In the context of AES for nonnative Japanese learners, Hirao et al. ( 2020 ) combined the long short-term memory (LSTM) model proposed by Hochreiter and Schmidhuber ( 1997 ) with BERT to develop a tailored automated Essay Scoring System. The findings of their study revealed that the BERT model outperformed both the conventional machine learning approach utilizing character-type features such as “kanji” and “hiragana”, as well as the standalone LSTM model. Takeuchi et al. ( 2021 ) presented an approach to Japanese AES that eliminates the requirement for pre-scored essays by relying solely on reference texts or a model answer for the essay task. They investigated multiple similarity evaluation methods, including frequency of morphemes, idf values calculated on Wikipedia, LSI, LDA, word-embedding vectors, and document vectors produced by BERT. The experimental findings revealed that the method utilizing the frequency of morphemes with idf values exhibited the strongest correlation with human-annotated scores across different essay tasks. The utilization of BERT in AES encounters several limitations. Firstly, essays often exceed the model’s maximum length limit. Second, only score labels are available for training, which restricts access to additional information.

Mizumoto and Eguchi ( 2023 ) were pioneers in employing the GPT model for AES in non-native English writing. Their study focused on evaluating the accuracy and reliability of AES using the GPT-3 text-davinci-003 model, analyzing a dataset of 12,100 essays from the corpus of nonnative written English (TOEFL11). The findings indicated that AES utilizing the GPT-3 model exhibited a certain degree of accuracy and reliability. They suggest that GPT-3-based AES systems hold the potential to provide support for human ratings. However, applying GPT model to AES presents a unique natural language processing (NLP) task that involves considerations such as nonnative language proficiency, the influence of the learner’s first language on the output in the target language, and identifying linguistic features that best indicate writing quality in a specific language. These linguistic features may differ morphologically or syntactically from those present in the learners’ first language, as observed in (1)–(3).

我-送了-他-一本-书

Wǒ-sòngle-tā-yī běn-shū

1 sg .-give. past- him-one .cl- book

“I gave him a book.”

Agglutinative

彼-に-本-を-あげ-まし-た

Kare-ni-hon-o-age-mashi-ta

3 sg .- dat -hon- acc- give.honorification. past

Inflectional

give, give-s, gave, given, giving

Additionally, the morphological agglutination and subject-object-verb (SOV) order in Japanese, along with its idiomatic expressions, pose additional challenges for applying language models in AES tasks (4).

足-が 棒-に なり-ました

Ashi-ga bo-ni nar-mashita

leg- nom stick- dat become- past

“My leg became like a stick (I am extremely tired).”

The example sentence provided demonstrates the morpho-syntactic structure of Japanese and the presence of an idiomatic expression. In this sentence, the verb “なる” (naru), meaning “to become”, appears at the end of the sentence. The verb stem “なり” (nari) is attached with morphemes indicating honorification (“ます” - mashu) and tense (“た” - ta), showcasing agglutination. While the sentence can be literally translated as “my leg became like a stick”, it carries an idiomatic interpretation that implies “I am extremely tired”.

To overcome this issue, CyberAgent Inc. ( 2023 ) has developed the Open-Calm series of language models specifically designed for Japanese. Open-Calm consists of pre-trained models available in various sizes, such as Small, Medium, Large, and 7b. Figure 2 depicts the fundamental structure of the Open-Calm model. A key feature of this architecture is the incorporation of the Lora Adapter and GPT-NeoX frameworks, which can enhance its language processing capabilities.

GPT-NeoX Model Architecture (Okgetheng and Takeuchi 2024 ).

In a recent study conducted by Okgetheng and Takeuchi ( 2024 ), they assessed the efficacy of Open-Calm language models in grading Japanese essays. The research utilized a dataset of approximately 300 essays, which were annotated by native Japanese educators. The findings of the study demonstrate the considerable potential of Open-Calm language models in automated Japanese essay scoring. Specifically, among the Open-Calm family, the Open-Calm Large model (referred to as OCLL) exhibited the highest performance. However, it is important to note that, as of the current date, the Open-Calm Large model does not offer public access to its server. Consequently, users are required to independently deploy and operate the environment for OCLL. In order to utilize OCLL, users must have a PC equipped with an NVIDIA GeForce RTX 3060 (8 or 12 GB VRAM).

In summary, while the potential of LLMs in automated scoring of nonnative Japanese essays has been demonstrated in two studies—BERT-driven AES (Hirao et al. 2020 ) and OCLL-based AES (Okgetheng and Takeuchi, 2024 )—the number of research efforts in this area remains limited.

Another significant challenge in applying LLMs to AES lies in prompt engineering and ensuring its reliability and effectiveness (Brown et al. 2020 ; Rae et al. 2021 ; Zhang et al. 2021 ). Various prompting strategies have been proposed, such as the zero-shot chain of thought (CoT) approach (Kojima et al. 2022 ), which involves manually crafting diverse and effective examples. However, manual efforts can lead to mistakes. To address this, Zhang et al. ( 2021 ) introduced an automatic CoT prompting method called Auto-CoT, which demonstrates matching or superior performance compared to the CoT paradigm. Another prompt framework is trees of thoughts, enabling a model to self-evaluate its progress at intermediate stages of problem-solving through deliberate reasoning (Yao et al. 2023 ).

Beyond linguistic studies, there has been a noticeable increase in the number of foreign workers in Japan and Japanese learners worldwide (Ministry of Health, Labor, and Welfare of Japan, 2022 ; Japan Foundation, 2021 ). However, existing assessment methods, such as the Japanese Language Proficiency Test (JLPT), J-CAT, and TTBJ Footnote 1 , primarily focus on reading, listening, vocabulary, and grammar skills, neglecting the evaluation of writing proficiency. As the number of workers and language learners continues to grow, there is a rising demand for an efficient AES system that can reduce costs and time for raters and be utilized for employment, examinations, and self-study purposes.

This study aims to explore the potential of LLM-based AES by comparing the effectiveness of five models: two LLMs (GPT Footnote 2 and BERT), one Japanese local LLM (OCLL), and two conventional machine learning-based methods (linguistic feature-based scoring tools - Jess and JWriter).

The research questions addressed in this study are as follows:

To what extent do the LLM-driven AES and linguistic feature-based AES, when used as automated tools to support human rating, accurately reflect test takers’ actual performance?

What influence does the prompt have on the accuracy and performance of LLM-based AES methods?

The subsequent sections of the manuscript cover the methodology, including the assessment measures for nonnative Japanese writing proficiency, criteria for prompts, and the dataset. The evaluation section focuses on the analysis of annotations and rating scores generated by LLM-driven and linguistic feature-based AES methods.

Methodology

The dataset utilized in this study was obtained from the International Corpus of Japanese as a Second Language (I-JAS) Footnote 3 . This corpus consisted of 1000 participants who represented 12 different first languages. For the study, the participants were given a story-writing task on a personal computer. They were required to write two stories based on the 4-panel illustrations titled “Picnic” and “The key” (see Appendix A). Background information for the participants was provided by the corpus, including their Japanese language proficiency levels assessed through two online tests: J-CAT and SPOT. These tests evaluated their reading, listening, vocabulary, and grammar abilities. The learners’ proficiency levels were categorized into six levels aligned with the Common European Framework of Reference for Languages (CEFR) and the Reference Framework for Japanese Language Education (RFJLE): A1, A2, B1, B2, C1, and C2. According to Lee et al. ( 2015 ), there is a high level of agreement (r = 0.86) between the J-CAT and SPOT assessments, indicating that the proficiency certifications provided by J-CAT are consistent with those of SPOT. However, it is important to note that the scores of J-CAT and SPOT do not have a one-to-one correspondence. In this study, the J-CAT scores were used as a benchmark to differentiate learners of different proficiency levels. A total of 1400 essays were utilized, representing the beginner (aligned with A1), A2, B1, B2, C1, and C2 levels based on the J-CAT scores. Table 1 provides information about the learners’ proficiency levels and their corresponding J-CAT and SPOT scores.

A dataset comprising a total of 1400 essays from the story writing tasks was collected. Among these, 714 essays were utilized to evaluate the reliability of the LLM-based AES method, while the remaining 686 essays were designated as development data to assess the LLM-based AES’s capability to distinguish participants with varying proficiency levels. The GPT 4 API was used in this study. A detailed explanation of the prompt-assessment criteria is provided in Section Prompt . All essays were sent to the model for measurement and scoring.

Measures of writing proficiency for nonnative Japanese

Japanese exhibits a morphologically agglutinative structure where morphemes are attached to the word stem to convey grammatical functions such as tense, aspect, voice, and honorifics, e.g. (5).

食べ-させ-られ-まし-た-か

tabe-sase-rare-mashi-ta-ka

[eat (stem)-causative-passive voice-honorification-tense. past-question marker]

Japanese employs nine case particles to indicate grammatical functions: the nominative case particle が (ga), the accusative case particle を (o), the genitive case particle の (no), the dative case particle に (ni), the locative/instrumental case particle で (de), the ablative case particle から (kara), the directional case particle へ (e), and the comitative case particle と (to). The agglutinative nature of the language, combined with the case particle system, provides an efficient means of distinguishing between active and passive voice, either through morphemes or case particles, e.g. 食べる taberu “eat concusive . ” (active voice); 食べられる taberareru “eat concusive . ” (passive voice). In the active voice, “パン を 食べる” (pan o taberu) translates to “to eat bread”. On the other hand, in the passive voice, it becomes “パン が 食べられた” (pan ga taberareta), which means “(the) bread was eaten”. Additionally, it is important to note that different conjugations of the same lemma are considered as one type in order to ensure a comprehensive assessment of the language features. For example, e.g., 食べる taberu “eat concusive . ”; 食べている tabeteiru “eat progress .”; 食べた tabeta “eat past . ” as one type.

To incorporate these features, previous research (Suzuki, 1999 ; Watanabe et al. 1988 ; Ishioka, 2001 ; Ishioka and Kameda, 2006 ; Hirao et al. 2020 ) has identified complexity, fluency, and accuracy as crucial factors for evaluating writing quality. These criteria are assessed through various aspects, including lexical richness (lexical density, diversity, and sophistication), syntactic complexity, and cohesion (Kyle et al. 2021 ; Mizumoto and Eguchi, 2023 ; Ure, 1971 ; Halliday, 1985 ; Barkaoui and Hadidi, 2020 ; Zenker and Kyle, 2021 ; Kim et al. 2018 ; Lu, 2017 ; Ortega, 2015 ). Therefore, this study proposes five scoring categories: lexical richness, syntactic complexity, cohesion, content elaboration, and grammatical accuracy. A total of 16 measures were employed to capture these categories. The calculation process and specific details of these measures can be found in Table 2 .

T-unit, first introduced by Hunt ( 1966 ), is a measure used for evaluating speech and composition. It serves as an indicator of syntactic development and represents the shortest units into which a piece of discourse can be divided without leaving any sentence fragments. In the context of Japanese language assessment, Sakoda and Hosoi ( 2020 ) utilized T-unit as the basic unit to assess the accuracy and complexity of Japanese learners’ speaking and storytelling. The calculation of T-units in Japanese follows the following principles:

A single main clause constitutes 1 T-unit, regardless of the presence or absence of dependent clauses, e.g. (6).

ケンとマリはピクニックに行きました (main clause): 1 T-unit.

If a sentence contains a main clause along with subclauses, each subclause is considered part of the same T-unit, e.g. (7).

天気が良かった の で (subclause)、ケンとマリはピクニックに行きました (main clause): 1 T-unit.

In the case of coordinate clauses, where multiple clauses are connected, each coordinated clause is counted separately. Thus, a sentence with coordinate clauses may have 2 T-units or more, e.g. (8).

ケンは地図で場所を探して (coordinate clause)、マリはサンドイッチを作りました (coordinate clause): 2 T-units.

Lexical diversity refers to the range of words used within a text (Engber, 1995 ; Kyle et al. 2021 ) and is considered a useful measure of the breadth of vocabulary in L n production (Jarvis, 2013a , 2013b ).

The type/token ratio (TTR) is widely recognized as a straightforward measure for calculating lexical diversity and has been employed in numerous studies. These studies have demonstrated a strong correlation between TTR and other methods of measuring lexical diversity (e.g., Bentz et al. 2016 ; Čech and Miroslav, 2018 ; Çöltekin and Taraka, 2018 ). TTR is computed by considering both the number of unique words (types) and the total number of words (tokens) in a given text. Given that the length of learners’ writing texts can vary, this study employs the moving average type-token ratio (MATTR) to mitigate the influence of text length. MATTR is calculated using a 50-word moving window. Initially, a TTR is determined for words 1–50 in an essay, followed by words 2–51, 3–52, and so on until the end of the essay is reached (Díez-Ortega and Kyle, 2023 ). The final MATTR scores were obtained by averaging the TTR scores for all 50-word windows. The following formula was employed to derive MATTR:

\({\rm{MATTR}}({\rm{W}})=\frac{{\sum }_{{\rm{i}}=1}^{{\rm{N}}-{\rm{W}}+1}{{\rm{F}}}_{{\rm{i}}}}{{\rm{W}}({\rm{N}}-{\rm{W}}+1)}\)

Here, N refers to the number of tokens in the corpus. W is the randomly selected token size (W < N). \({F}_{i}\) is the number of types in each window. The \({\rm{MATTR}}({\rm{W}})\) is the mean of a series of type-token ratios (TTRs) based on the word form for all windows. It is expected that individuals with higher language proficiency will produce texts with greater lexical diversity, as indicated by higher MATTR scores.

Lexical density was captured by the ratio of the number of lexical words to the total number of words (Lu, 2012 ). Lexical sophistication refers to the utilization of advanced vocabulary, often evaluated through word frequency indices (Crossley et al. 2013 ; Haberman, 2008 ; Kyle and Crossley, 2015 ; Laufer and Nation, 1995 ; Lu, 2012 ; Read, 2000 ). In line of writing, lexical sophistication can be interpreted as vocabulary breadth, which entails the appropriate usage of vocabulary items across various lexicon-grammatical contexts and registers (Garner et al. 2019 ; Kim et al. 2018 ; Kyle et al. 2018 ). In Japanese specifically, words are considered lexically sophisticated if they are not included in the “Japanese Education Vocabulary List Ver 1.0”. Footnote 4 Consequently, lexical sophistication was calculated by determining the number of sophisticated word types relative to the total number of words per essay. Furthermore, it has been suggested that, in Japanese writing, sentences should ideally have a length of no more than 40 to 50 characters, as this promotes readability. Therefore, the median and maximum sentence length can be considered as useful indices for assessment (Ishioka and Kameda, 2006 ).

Syntactic complexity was assessed based on several measures, including the mean length of clauses, verb phrases per T-unit, clauses per T-unit, dependent clauses per T-unit, complex nominals per clause, adverbial clauses per clause, coordinate phrases per clause, and mean dependency distance (MDD). The MDD reflects the distance between the governor and dependent positions in a sentence. A larger dependency distance indicates a higher cognitive load and greater complexity in syntactic processing (Liu, 2008 ; Liu et al. 2017 ). The MDD has been established as an efficient metric for measuring syntactic complexity (Jiang, Quyang, and Liu, 2019 ; Li and Yan, 2021 ). To calculate the MDD, the position numbers of the governor and dependent are subtracted, assuming that words in a sentence are assigned in a linear order, such as W1 … Wi … Wn. In any dependency relationship between words Wa and Wb, Wa is the governor and Wb is the dependent. The MDD of the entire sentence was obtained by taking the absolute value of governor – dependent:

MDD = \(\frac{1}{n}{\sum }_{i=1}^{n}|{\rm{D}}{{\rm{D}}}_{i}|\)

In this formula, \(n\) represents the number of words in the sentence, and \({DD}i\) is the dependency distance of the \({i}^{{th}}\) dependency relationship of a sentence. Building on this, the annotation of sentence ‘Mary-ga-John-ni-keshigomu-o-watashita was [Mary- top -John- dat -eraser- acc -give- past] ’. The sentence’s MDD would be 2. Table 3 provides the CSV file as a prompt for GPT 4.

Cohesion (semantic similarity) and content elaboration aim to capture the ideas presented in test taker’s essays. Cohesion was assessed using three measures: Synonym overlap/paragraph (topic), Synonym overlap/paragraph (keywords), and word2vec cosine similarity. Content elaboration and development were measured as the number of metadiscourse markers (type)/number of words. To capture content closely, this study proposed a novel-distance based representation, by encoding the cosine distance between the essay (by learner) and essay task’s (topic and keyword) i -vectors. The learner’s essay is decoded into a word sequence, and aligned to the essay task’ topic and keyword for log-likelihood measurement. The cosine distance reveals the content elaboration score in the leaners’ essay. The mathematical equation of cosine similarity between target-reference vectors is shown in (11), assuming there are i essays and ( L i , …. L n ) and ( N i , …. N n ) are the vectors representing the learner and task’s topic and keyword respectively. The content elaboration distance between L i and N i was calculated as follows:

\(\cos \left(\theta \right)=\frac{{\rm{L}}\,\cdot\, {\rm{N}}}{\left|{\rm{L}}\right|{\rm{|N|}}}=\frac{\mathop{\sum }\nolimits_{i=1}^{n}{L}_{i}{N}_{i}}{\sqrt{\mathop{\sum }\nolimits_{i=1}^{n}{L}_{i}^{2}}\sqrt{\mathop{\sum }\nolimits_{i=1}^{n}{N}_{i}^{2}}}\)

A high similarity value indicates a low difference between the two recognition outcomes, which in turn suggests a high level of proficiency in content elaboration.

To evaluate the effectiveness of the proposed measures in distinguishing different proficiency levels among nonnative Japanese speakers’ writing, we conducted a multi-faceted Rasch measurement analysis (Linacre, 1994 ). This approach applies measurement models to thoroughly analyze various factors that can influence test outcomes, including test takers’ proficiency, item difficulty, and rater severity, among others. The underlying principles and functionality of multi-faceted Rasch measurement are illustrated in (12).

\(\log \left(\frac{{P}_{{nijk}}}{{P}_{{nij}(k-1)}}\right)={B}_{n}-{D}_{i}-{C}_{j}-{F}_{k}\)

(12) defines the logarithmic transformation of the probability ratio ( P nijk /P nij(k-1) )) as a function of multiple parameters. Here, n represents the test taker, i denotes a writing proficiency measure, j corresponds to the human rater, and k represents the proficiency score. The parameter B n signifies the proficiency level of test taker n (where n ranges from 1 to N). D j represents the difficulty parameter of test item i (where i ranges from 1 to L), while C j represents the severity of rater j (where j ranges from 1 to J). Additionally, F k represents the step difficulty for a test taker to move from score ‘k-1’ to k . P nijk refers to the probability of rater j assigning score k to test taker n for test item i . P nij(k-1) represents the likelihood of test taker n being assigned score ‘k-1’ by rater j for test item i . Each facet within the test is treated as an independent parameter and estimated within the same reference framework. To evaluate the consistency of scores obtained through both human and computer analysis, we utilized the Infit mean-square statistic. This statistic is a chi-square measure divided by the degrees of freedom and is weighted with information. It demonstrates higher sensitivity to unexpected patterns in responses to items near a person’s proficiency level (Linacre, 2002 ). Fit statistics are assessed based on predefined thresholds for acceptable fit. For the Infit MNSQ, which has a mean of 1.00, different thresholds have been suggested. Some propose stricter thresholds ranging from 0.7 to 1.3 (Bond et al. 2021 ), while others suggest more lenient thresholds ranging from 0.5 to 1.5 (Eckes, 2009 ). In this study, we adopted the criterion of 0.70–1.30 for the Infit MNSQ.

Moving forward, we can now proceed to assess the effectiveness of the 16 proposed measures based on five criteria for accurately distinguishing various levels of writing proficiency among non-native Japanese speakers. To conduct this evaluation, we utilized the development dataset from the I-JAS corpus, as described in Section Dataset . Table 4 provides a measurement report that presents the performance details of the 14 metrics under consideration. The measure separation was found to be 4.02, indicating a clear differentiation among the measures. The reliability index for the measure separation was 0.891, suggesting consistency in the measurement. Similarly, the person separation reliability index was 0.802, indicating the accuracy of the assessment in distinguishing between individuals. All 16 measures demonstrated Infit mean squares within a reasonable range, ranging from 0.76 to 1.28. The Synonym overlap/paragraph (topic) measure exhibited a relatively high outfit mean square of 1.46, although the Infit mean square falls within an acceptable range. The standard error for the measures ranged from 0.13 to 0.28, indicating the precision of the estimates.

Table 5 further illustrated the weights assigned to different linguistic measures for score prediction, with higher weights indicating stronger correlations between those measures and higher scores. Specifically, the following measures exhibited higher weights compared to others: moving average type token ratio per essay has a weight of 0.0391. Mean dependency distance had a weight of 0.0388. Mean length of clause, calculated by dividing the number of words by the number of clauses, had a weight of 0.0374. Complex nominals per T-unit, calculated by dividing the number of complex nominals by the number of T-units, had a weight of 0.0379. Coordinate phrases rate, calculated by dividing the number of coordinate phrases by the number of clauses, had a weight of 0.0325. Grammatical error rate, representing the number of errors per essay, had a weight of 0.0322.

Criteria (output indicator)

The criteria used to evaluate the writing ability in this study were based on CEFR, which follows a six-point scale ranging from A1 to C2. To assess the quality of Japanese writing, the scoring criteria from Table 6 were utilized. These criteria were derived from the IELTS writing standards and served as assessment guidelines and prompts for the written output.

A prompt is a question or detailed instruction that is provided to the model to obtain a proper response. After several pilot experiments, we decided to provide the measures (Section Measures of writing proficiency for nonnative Japanese ) as the input prompt and use the criteria (Section Criteria (output indicator) ) as the output indicator. Regarding the prompt language, considering that the LLM was tasked with rating Japanese essays, would prompt in Japanese works better Footnote 5 ? We conducted experiments comparing the performance of GPT-4 using both English and Japanese prompts. Additionally, we utilized the Japanese local model OCLL with Japanese prompts. Multiple trials were conducted using the same sample. Regardless of the prompt language used, we consistently obtained the same grading results with GPT-4, which assigned a grade of B1 to the writing sample. This suggested that GPT-4 is reliable and capable of producing consistent ratings regardless of the prompt language. On the other hand, when we used Japanese prompts with the Japanese local model “OCLL”, we encountered inconsistent grading results. Out of 10 attempts with OCLL, only 6 yielded consistent grading results (B1), while the remaining 4 showed different outcomes, including A1 and B2 grades. These findings indicated that the language of the prompt was not the determining factor for reliable AES. Instead, the size of the training data and the model parameters played crucial roles in achieving consistent and reliable AES results for the language model.

The following is the utilized prompt, which details all measures and requires the LLM to score the essays using holistic and trait scores.

Please evaluate Japanese essays written by Japanese learners and assign a score to each essay on a six-point scale, ranging from A1, A2, B1, B2, C1 to C2. Additionally, please provide trait scores and display the calculation process for each trait score. The scoring should be based on the following criteria:

Moving average type-token ratio.

Number of lexical words (token) divided by the total number of words per essay.

Number of sophisticated word types divided by the total number of words per essay.

Mean length of clause.

Verb phrases per T-unit.

Clauses per T-unit.

Dependent clauses per T-unit.

Complex nominals per clause.

Adverbial clauses per clause.

Coordinate phrases per clause.

Mean dependency distance.

Synonym overlap paragraph (topic and keywords).

Word2vec cosine similarity.

Connectives per essay.

Conjunctions per essay.

Number of metadiscourse markers (types) divided by the total number of words.

Number of errors per essay.

Japanese essay text

出かける前に二人が地図を見ている間に、サンドイッチを入れたバスケットに犬が入ってしまいました。それに気づかずに二人は楽しそうに出かけて行きました。やがて突然犬がバスケットから飛び出し、二人は驚きました。バスケット の 中を見ると、食べ物はすべて犬に食べられていて、二人は困ってしまいました。(ID_JJJ01_SW1)

The score of the example above was B1. Figure 3 provides an example of holistic and trait scores provided by GPT-4 (with a prompt indicating all measures) via Bing Footnote 6 .

Example of GPT-4 AES and feedback (with a prompt indicating all measures).

Statistical analysis

The aim of this study is to investigate the potential use of LLM for nonnative Japanese AES. It seeks to compare the scoring outcomes obtained from feature-based AES tools, which rely on conventional machine learning technology (i.e. Jess, JWriter), with those generated by AI-driven AES tools utilizing deep learning technology (BERT, GPT, OCLL). To assess the reliability of a computer-assisted annotation tool, the study initially established human-human agreement as the benchmark measure. Subsequently, the performance of the LLM-based method was evaluated by comparing it to human-human agreement.

To assess annotation agreement, the study employed standard measures such as precision, recall, and F-score (Brants 2000 ; Lu 2010 ), along with the quadratically weighted kappa (QWK) to evaluate the consistency and agreement in the annotation process. Assume A and B represent human annotators. When comparing the annotations of the two annotators, the following results are obtained. The evaluation of precision, recall, and F-score metrics was illustrated in equations (13) to (15).

\({\rm{Recall}}(A,B)=\frac{{\rm{Number}}\,{\rm{of}}\,{\rm{identical}}\,{\rm{nodes}}\,{\rm{in}}\,A\,{\rm{and}}\,B}{{\rm{Number}}\,{\rm{of}}\,{\rm{nodes}}\,{\rm{in}}\,A}\)

\({\rm{Precision}}(A,\,B)=\frac{{\rm{Number}}\,{\rm{of}}\,{\rm{identical}}\,{\rm{nodes}}\,{\rm{in}}\,A\,{\rm{and}}\,B}{{\rm{Number}}\,{\rm{of}}\,{\rm{nodes}}\,{\rm{in}}\,B}\)

The F-score is the harmonic mean of recall and precision:

\({\rm{F}}-{\rm{score}}=\frac{2* ({\rm{Precision}}* {\rm{Recall}})}{{\rm{Precision}}+{\rm{Recall}}}\)

The highest possible value of an F-score is 1.0, indicating perfect precision and recall, and the lowest possible value is 0, if either precision or recall are zero.

In accordance with Taghipour and Ng ( 2016 ), the calculation of QWK involves two steps:

Step 1: Construct a weight matrix W as follows:

\({W}_{{ij}}=\frac{{(i-j)}^{2}}{{(N-1)}^{2}}\)

i represents the annotation made by the tool, while j represents the annotation made by a human rater. N denotes the total number of possible annotations. Matrix O is subsequently computed, where O_( i, j ) represents the count of data annotated by the tool ( i ) and the human annotator ( j ). On the other hand, E refers to the expected count matrix, which undergoes normalization to ensure that the sum of elements in E matches the sum of elements in O.

Step 2: With matrices O and E, the QWK is obtained as follows:

K = 1- \(\frac{\sum i,j{W}_{i,j}\,{O}_{i,j}}{\sum i,j{W}_{i,j}\,{E}_{i,j}}\)

The value of the quadratic weighted kappa increases as the level of agreement improves. Further, to assess the accuracy of LLM scoring, the proportional reductive mean square error (PRMSE) was employed. The PRMSE approach takes into account the variability observed in human ratings to estimate the rater error, which is then subtracted from the variance of the human labels. This calculation provides an overall measure of agreement between the automated scores and true scores (Haberman et al. 2015 ; Loukina et al. 2020 ; Taghipour and Ng, 2016 ). The computation of PRMSE involves the following steps:

Step 1: Calculate the mean squared errors (MSEs) for the scoring outcomes of the computer-assisted tool (MSE tool) and the human scoring outcomes (MSE human).

Step 2: Determine the PRMSE by comparing the MSE of the computer-assisted tool (MSE tool) with the MSE from human raters (MSE human), using the following formula:

\({\rm{PRMSE}}=1-\frac{({\rm{MSE}}\,{\rm{tool}})\,}{({\rm{MSE}}\,{\rm{human}})\,}=1-\,\frac{{\sum }_{i}^{n}=1{({{\rm{y}}}_{i}-{\hat{{\rm{y}}}}_{{\rm{i}}})}^{2}}{{\sum }_{i}^{n}=1{({{\rm{y}}}_{i}-\hat{{\rm{y}}})}^{2}}\)

In the numerator, ŷi represents the scoring outcome predicted by a specific LLM-driven AES system for a given sample. The term y i − ŷ i represents the difference between this predicted outcome and the mean value of all LLM-driven AES systems’ scoring outcomes. It quantifies the deviation of the specific LLM-driven AES system’s prediction from the average prediction of all LLM-driven AES systems. In the denominator, y i − ŷ represents the difference between the scoring outcome provided by a specific human rater for a given sample and the mean value of all human raters’ scoring outcomes. It measures the discrepancy between the specific human rater’s score and the average score given by all human raters. The PRMSE is then calculated by subtracting the ratio of the MSE tool to the MSE human from 1. PRMSE falls within the range of 0 to 1, with larger values indicating reduced errors in LLM’s scoring compared to those of human raters. In other words, a higher PRMSE implies that LLM’s scoring demonstrates greater accuracy in predicting the true scores (Loukina et al. 2020 ). The interpretation of kappa values, ranging from 0 to 1, is based on the work of Landis and Koch ( 1977 ). Specifically, the following categories are assigned to different ranges of kappa values: −1 indicates complete inconsistency, 0 indicates random agreement, 0.0 ~ 0.20 indicates extremely low level of agreement (slight), 0.21 ~ 0.40 indicates moderate level of agreement (fair), 0.41 ~ 0.60 indicates medium level of agreement (moderate), 0.61 ~ 0.80 indicates high level of agreement (substantial), 0.81 ~ 1 indicates almost perfect level of agreement. All statistical analyses were executed using Python script.

Results and discussion

Annotation reliability of the llm.

This section focuses on assessing the reliability of the LLM’s annotation and scoring capabilities. To evaluate the reliability, several tests were conducted simultaneously, aiming to achieve the following objectives:

Assess the LLM’s ability to differentiate between test takers with varying levels of oral proficiency.

Determine the level of agreement between the annotations and scoring performed by the LLM and those done by human raters.

The evaluation of the results encompassed several metrics, including: precision, recall, F-Score, quadratically-weighted kappa, proportional reduction of mean squared error, Pearson correlation, and multi-faceted Rasch measurement.

Inter-annotator agreement (human–human annotator agreement)

We started with an agreement test of the two human annotators. Two trained annotators were recruited to determine the writing task data measures. A total of 714 scripts, as the test data, was utilized. Each analysis lasted 300–360 min. Inter-annotator agreement was evaluated using the standard measures of precision, recall, and F-score and QWK. Table 7 presents the inter-annotator agreement for the various indicators. As shown, the inter-annotator agreement was fairly high, with F-scores ranging from 1.0 for sentence and word number to 0.666 for grammatical errors.

The findings from the QWK analysis provided further confirmation of the inter-annotator agreement. The QWK values covered a range from 0.950 ( p  = 0.000) for sentence and word number to 0.695 for synonym overlap number (keyword) and grammatical errors ( p  = 0.001).

Agreement of annotation outcomes between human and LLM

To evaluate the consistency between human annotators and LLM annotators (BERT, GPT, OCLL) across the indices, the same test was conducted. The results of the inter-annotator agreement (F-score) between LLM and human annotation are provided in Appendix B-D. The F-scores ranged from 0.706 for Grammatical error # for OCLL-human to a perfect 1.000 for GPT-human, for sentences, clauses, T-units, and words. These findings were further supported by the QWK analysis, which showed agreement levels ranging from 0.807 ( p  = 0.001) for metadiscourse markers for OCLL-human to 0.962 for words ( p  = 0.000) for GPT-human. The findings demonstrated that the LLM annotation achieved a significant level of accuracy in identifying measurement units and counts.

Reliability of LLM-driven AES’s scoring and discriminating proficiency levels

This section examines the reliability of the LLM-driven AES scoring through a comparison of the scoring outcomes produced by human raters and the LLM ( Reliability of LLM-driven AES scoring ). It also assesses the effectiveness of the LLM-based AES system in differentiating participants with varying proficiency levels ( Reliability of LLM-driven AES discriminating proficiency levels ).

Reliability of LLM-driven AES scoring

Table 8 summarizes the QWK coefficient analysis between the scores computed by the human raters and the GPT-4 for the individual essays from I-JAS Footnote 7 . As shown, the QWK of all measures ranged from k  = 0.819 for lexical density (number of lexical words (tokens)/number of words per essay) to k  = 0.644 for word2vec cosine similarity. Table 9 further presents the Pearson correlations between the 16 writing proficiency measures scored by human raters and GPT 4 for the individual essays. The correlations ranged from 0.672 for syntactic complexity to 0.734 for grammatical accuracy. The correlations between the writing proficiency scores assigned by human raters and the BERT-based AES system were found to range from 0.661 for syntactic complexity to 0.713 for grammatical accuracy. The correlations between the writing proficiency scores given by human raters and the OCLL-based AES system ranged from 0.654 for cohesion to 0.721 for grammatical accuracy. These findings indicated an alignment between the assessments made by human raters and both the BERT-based and OCLL-based AES systems in terms of various aspects of writing proficiency.

Reliability of LLM-driven AES discriminating proficiency levels

After validating the reliability of the LLM’s annotation and scoring, the subsequent objective was to evaluate its ability to distinguish between various proficiency levels. For this analysis, a dataset of 686 individual essays was utilized. Table 10 presents a sample of the results, summarizing the means, standard deviations, and the outcomes of the one-way ANOVAs based on the measures assessed by the GPT-4 model. A post hoc multiple comparison test, specifically the Bonferroni test, was conducted to identify any potential differences between pairs of levels.

As the results reveal, seven measures presented linear upward or downward progress across the three proficiency levels. These were marked in bold in Table 10 and comprise one measure of lexical richness, i.e. MATTR (lexical diversity); four measures of syntactic complexity, i.e. MDD (mean dependency distance), MLC (mean length of clause), CNT (complex nominals per T-unit), CPC (coordinate phrases rate); one cohesion measure, i.e. word2vec cosine similarity and GER (grammatical error rate). Regarding the ability of the sixteen measures to distinguish adjacent proficiency levels, the Bonferroni tests indicated that statistically significant differences exist between the primary level and the intermediate level for MLC and GER. One measure of lexical richness, namely LD, along with three measures of syntactic complexity (VPT, CT, DCT, ACC), two measures of cohesion (SOPT, SOPK), and one measure of content elaboration (IMM), exhibited statistically significant differences between proficiency levels. However, these differences did not demonstrate a linear progression between adjacent proficiency levels. No significant difference was observed in lexical sophistication between proficiency levels.

To summarize, our study aimed to evaluate the reliability and differentiation capabilities of the LLM-driven AES method. For the first objective, we assessed the LLM’s ability to differentiate between test takers with varying levels of oral proficiency using precision, recall, F-Score, and quadratically-weighted kappa. Regarding the second objective, we compared the scoring outcomes generated by human raters and the LLM to determine the level of agreement. We employed quadratically-weighted kappa and Pearson correlations to compare the 16 writing proficiency measures for the individual essays. The results confirmed the feasibility of using the LLM for annotation and scoring in AES for nonnative Japanese. As a result, Research Question 1 has been addressed.

Comparison of BERT-, GPT-, OCLL-based AES, and linguistic-feature-based computation methods

This section aims to compare the effectiveness of five AES methods for nonnative Japanese writing, i.e. LLM-driven approaches utilizing BERT, GPT, and OCLL, linguistic feature-based approaches using Jess and JWriter. The comparison was conducted by comparing the ratings obtained from each approach with human ratings. All ratings were derived from the dataset introduced in Dataset . To facilitate the comparison, the agreement between the automated methods and human ratings was assessed using QWK and PRMSE. The performance of each approach was summarized in Table 11 .

The QWK coefficient values indicate that LLMs (GPT, BERT, OCLL) and human rating outcomes demonstrated higher agreement compared to feature-based AES methods (Jess and JWriter) in assessing writing proficiency criteria, including lexical richness, syntactic complexity, content, and grammatical accuracy. Among the LLMs, the GPT-4 driven AES and human rating outcomes showed the highest agreement in all criteria, except for syntactic complexity. The PRMSE values suggest that the GPT-based method outperformed linguistic feature-based methods and other LLM-based approaches. Moreover, an interesting finding emerged during the study: the agreement coefficient between GPT-4 and human scoring was even higher than the agreement between different human raters themselves. This discovery highlights the advantage of GPT-based AES over human rating. Ratings involve a series of processes, including reading the learners’ writing, evaluating the content and language, and assigning scores. Within this chain of processes, various biases can be introduced, stemming from factors such as rater biases, test design, and rating scales. These biases can impact the consistency and objectivity of human ratings. GPT-based AES may benefit from its ability to apply consistent and objective evaluation criteria. By prompting the GPT model with detailed writing scoring rubrics and linguistic features, potential biases in human ratings can be mitigated. The model follows a predefined set of guidelines and does not possess the same subjective biases that human raters may exhibit. This standardization in the evaluation process contributes to the higher agreement observed between GPT-4 and human scoring. Section Prompt strategy of the study delves further into the role of prompts in the application of LLMs to AES. It explores how the choice and implementation of prompts can impact the performance and reliability of LLM-based AES methods. Furthermore, it is important to acknowledge the strengths of the local model, i.e. the Japanese local model OCLL, which excels in processing certain idiomatic expressions. Nevertheless, our analysis indicated that GPT-4 surpasses local models in AES. This superior performance can be attributed to the larger parameter size of GPT-4, estimated to be between 500 billion and 1 trillion, which exceeds the sizes of both BERT and the local model OCLL.

Prompt strategy

In the context of prompt strategy, Mizumoto and Eguchi ( 2023 ) conducted a study where they applied the GPT-3 model to automatically score English essays in the TOEFL test. They found that the accuracy of the GPT model alone was moderate to fair. However, when they incorporated linguistic measures such as cohesion, syntactic complexity, and lexical features alongside the GPT model, the accuracy significantly improved. This highlights the importance of prompt engineering and providing the model with specific instructions to enhance its performance. In this study, a similar approach was taken to optimize the performance of LLMs. GPT-4, which outperformed BERT and OCLL, was selected as the candidate model. Model 1 was used as the baseline, representing GPT-4 without any additional prompting. Model 2, on the other hand, involved GPT-4 prompted with 16 measures that included scoring criteria, efficient linguistic features for writing assessment, and detailed measurement units and calculation formulas. The remaining models (Models 3 to 18) utilized GPT-4 prompted with individual measures. The performance of these 18 different models was assessed using the output indicators described in Section Criteria (output indicator) . By comparing the performances of these models, the study aimed to understand the impact of prompt engineering on the accuracy and effectiveness of GPT-4 in AES tasks.

Based on the PRMSE scores presented in Fig. 4 , it was observed that Model 1, representing GPT-4 without any additional prompting, achieved a fair level of performance. However, Model 2, which utilized GPT-4 prompted with all measures, outperformed all other models in terms of PRMSE score, achieving a score of 0.681. These results indicate that the inclusion of specific measures and prompts significantly enhanced the performance of GPT-4 in AES. Among the measures, syntactic complexity was found to play a particularly significant role in improving the accuracy of GPT-4 in assessing writing quality. Following that, lexical diversity emerged as another important factor contributing to the model’s effectiveness. The study suggests that a well-prompted GPT-4 can serve as a valuable tool to support human assessors in evaluating writing quality. By utilizing GPT-4 as an automated scoring tool, the evaluation biases associated with human raters can be minimized. This has the potential to empower teachers by allowing them to focus on designing writing tasks and guiding writing strategies, while leveraging the capabilities of GPT-4 for efficient and reliable scoring.

PRMSE scores of the 18 AES models.

This study aimed to investigate two main research questions: the feasibility of utilizing LLMs for AES and the impact of prompt engineering on the application of LLMs in AES.

To address the first objective, the study compared the effectiveness of five different models: GPT, BERT, the Japanese local LLM (OCLL), and two conventional machine learning-based AES tools (Jess and JWriter). The PRMSE values indicated that the GPT-4-based method outperformed other LLMs (BERT, OCLL) and linguistic feature-based computational methods (Jess and JWriter) across various writing proficiency criteria. Furthermore, the agreement coefficient between GPT-4 and human scoring surpassed the agreement among human raters themselves, highlighting the potential of using the GPT-4 tool to enhance AES by reducing biases and subjectivity, saving time, labor, and cost, and providing valuable feedback for self-study. Regarding the second goal, the role of prompt design was investigated by comparing 18 models, including a baseline model, a model prompted with all measures, and 16 models prompted with one measure at a time. GPT-4, which outperformed BERT and OCLL, was selected as the candidate model. The PRMSE scores of the models showed that GPT-4 prompted with all measures achieved the best performance, surpassing the baseline and other models.

In conclusion, this study has demonstrated the potential of LLMs in supporting human rating in assessments. By incorporating automation, we can save time and resources while reducing biases and subjectivity inherent in human rating processes. Automated language assessments offer the advantage of accessibility, providing equal opportunities and economic feasibility for individuals who lack access to traditional assessment centers or necessary resources. LLM-based language assessments provide valuable feedback and support to learners, aiding in the enhancement of their language proficiency and the achievement of their goals. This personalized feedback can cater to individual learner needs, facilitating a more tailored and effective language-learning experience.

There are three important areas that merit further exploration. First, prompt engineering requires attention to ensure optimal performance of LLM-based AES across different language types. This study revealed that GPT-4, when prompted with all measures, outperformed models prompted with fewer measures. Therefore, investigating and refining prompt strategies can enhance the effectiveness of LLMs in automated language assessments. Second, it is crucial to explore the application of LLMs in second-language assessment and learning for oral proficiency, as well as their potential in under-resourced languages. Recent advancements in self-supervised machine learning techniques have significantly improved automatic speech recognition (ASR) systems, opening up new possibilities for creating reliable ASR systems, particularly for under-resourced languages with limited data. However, challenges persist in the field of ASR. First, ASR assumes correct word pronunciation for automatic pronunciation evaluation, which proves challenging for learners in the early stages of language acquisition due to diverse accents influenced by their native languages. Accurately segmenting short words becomes problematic in such cases. Second, developing precise audio-text transcriptions for languages with non-native accented speech poses a formidable task. Last, assessing oral proficiency levels involves capturing various linguistic features, including fluency, pronunciation, accuracy, and complexity, which are not easily captured by current NLP technology.

Data availability

The dataset utilized was obtained from the International Corpus of Japanese as a Second Language (I-JAS). The data URLs: [ https://www2.ninjal.ac.jp/jll/lsaj/ihome2.html ].

J-CAT and TTBJ are two computerized adaptive tests used to assess Japanese language proficiency.

SPOT is a specific component of the TTBJ test.

J-CAT: https://www.j-cat2.org/html/ja/pages/interpret.html

SPOT: https://ttbj.cegloc.tsukuba.ac.jp/p1.html#SPOT .

The study utilized a prompt-based GPT-4 model, developed by OpenAI, which has an impressive architecture with 1.8 trillion parameters across 120 layers. GPT-4 was trained on a vast dataset of 13 trillion tokens, using two stages: initial training on internet text datasets to predict the next token, and subsequent fine-tuning through reinforcement learning from human feedback.

https://www2.ninjal.ac.jp/jll/lsaj/ihome2-en.html .

http://jhlee.sakura.ne.jp/JEV/ by Japanese Learning Dictionary Support Group 2015.

We express our sincere gratitude to the reviewer for bringing this matter to our attention.

On February 7, 2023, Microsoft began rolling out a major overhaul to Bing that included a new chatbot feature based on OpenAI’s GPT-4 (Bing.com).

Appendix E-F present the analysis results of the QWK coefficient between the scores computed by the human raters and the BERT, OCLL models.

Attali Y, Burstein J (2006) Automated essay scoring with e-rater® V.2. J. Technol., Learn. Assess., 4

Barkaoui K, Hadidi A (2020) Assessing Change in English Second Language Writing Performance (1st ed.). Routledge, New York. https://doi.org/10.4324/9781003092346

Bentz C, Tatyana R, Koplenig A, Tanja S (2016) A comparison between morphological complexity. measures: Typological data vs. language corpora. In Proceedings of the workshop on computational linguistics for linguistic complexity (CL4LC), 142–153. Osaka, Japan: The COLING 2016 Organizing Committee

Bond TG, Yan Z, Heene M (2021) Applying the Rasch model: Fundamental measurement in the human sciences (4th ed). Routledge

Brants T (2000) Inter-annotator agreement for a German newspaper corpus. Proceedings of the Second International Conference on Language Resources and Evaluation (LREC’00), Athens, Greece, 31 May-2 June, European Language Resources Association

Brown TB, Mann B, Ryder N, et al. (2020) Language models are few-shot learners. Advances in Neural Information Processing Systems, Online, 6–12 December, Curran Associates, Inc., Red Hook, NY

Burstein J (2003) The E-rater scoring engine: Automated essay scoring with natural language processing. In Shermis MD and Burstein JC (ed) Automated Essay Scoring: A Cross-Disciplinary Perspective. Lawrence Erlbaum Associates, Mahwah, NJ

Čech R, Miroslav K (2018) Morphological richness of text. In Masako F, Václav C (ed) Taming the corpus: From inflection and lexis to interpretation, 63–77. Cham, Switzerland: Springer Nature

Çöltekin Ç, Taraka, R (2018) Exploiting Universal Dependencies treebanks for measuring morphosyntactic complexity. In Aleksandrs B, Christian B (ed), Proceedings of first workshop on measuring language complexity, 1–7. Torun, Poland

Crossley SA, Cobb T, McNamara DS (2013) Comparing count-based and band-based indices of word frequency: Implications for active vocabulary research and pedagogical applications. System 41:965–981. https://doi.org/10.1016/j.system.2013.08.002

Article   Google Scholar  

Crossley SA, McNamara DS (2016) Say more and be more coherent: How text elaboration and cohesion can increase writing quality. J. Writ. Res. 7:351–370

CyberAgent Inc (2023) Open-Calm series of Japanese language models. Retrieved from: https://www.cyberagent.co.jp/news/detail/id=28817

Devlin J, Chang MW, Lee K, Toutanova K (2019) BERT: Pre-training of deep bidirectional transformers for language understanding. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics, Minneapolis, Minnesota, 2–7 June, pp. 4171–4186. Association for Computational Linguistics

Diez-Ortega M, Kyle K (2023) Measuring the development of lexical richness of L2 Spanish: a longitudinal learner corpus study. Studies in Second Language Acquisition 1-31

Eckes T (2009) On common ground? How raters perceive scoring criteria in oral proficiency testing. In Brown A, Hill K (ed) Language testing and evaluation 13: Tasks and criteria in performance assessment (pp. 43–73). Peter Lang Publishing

Elliot S (2003) IntelliMetric: from here to validity. In: Shermis MD, Burstein JC (ed) Automated Essay Scoring: A Cross-Disciplinary Perspective. Lawrence Erlbaum Associates, Mahwah, NJ

Google Scholar  

Engber CA (1995) The relationship of lexical proficiency to the quality of ESL compositions. J. Second Lang. Writ. 4:139–155

Garner J, Crossley SA, Kyle K (2019) N-gram measures and L2 writing proficiency. System 80:176–187. https://doi.org/10.1016/j.system.2018.12.001

Haberman SJ (2008) When can subscores have value? J. Educat. Behav. Stat., 33:204–229

Haberman SJ, Yao L, Sinharay S (2015) Prediction of true test scores from observed item scores and ancillary data. Brit. J. Math. Stat. Psychol. 68:363–385

Halliday MAK (1985) Spoken and Written Language. Deakin University Press, Melbourne, Australia

Hirao R, Arai M, Shimanaka H et al. (2020) Automated essay scoring system for nonnative Japanese learners. Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pp. 1250–1257. European Language Resources Association

Hunt KW (1966) Recent Measures in Syntactic Development. Elementary English, 43(7), 732–739. http://www.jstor.org/stable/41386067

Ishioka T (2001) About e-rater, a computer-based automatic scoring system for essays [Konpyūta ni yoru essei no jidō saiten shisutemu e − rater ni tsuite]. University Entrance Examination. Forum [Daigaku nyūshi fōramu] 24:71–76

Hochreiter S, Schmidhuber J (1997) Long short- term memory. Neural Comput. 9(8):1735–1780

Article   CAS   PubMed   Google Scholar  

Ishioka T, Kameda M (2006) Automated Japanese essay scoring system based on articles written by experts. Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Sydney, Australia, 17–18 July 2006, pp. 233-240. Association for Computational Linguistics, USA

Japan Foundation (2021) Retrieved from: https://www.jpf.gp.jp/j/project/japanese/survey/result/dl/survey2021/all.pdf

Jarvis S (2013a) Defining and measuring lexical diversity. In Jarvis S, Daller M (ed) Vocabulary knowledge: Human ratings and automated measures (Vol. 47, pp. 13–44). John Benjamins. https://doi.org/10.1075/sibil.47.03ch1

Jarvis S (2013b) Capturing the diversity in lexical diversity. Lang. Learn. 63:87–106. https://doi.org/10.1111/j.1467-9922.2012.00739.x

Jiang J, Quyang J, Liu H (2019) Interlanguage: A perspective of quantitative linguistic typology. Lang. Sci. 74:85–97

Kim M, Crossley SA, Kyle K (2018) Lexical sophistication as a multidimensional phenomenon: Relations to second language lexical proficiency, development, and writing quality. Mod. Lang. J. 102(1):120–141. https://doi.org/10.1111/modl.12447

Kojima T, Gu S, Reid M et al. (2022) Large language models are zero-shot reasoners. Advances in Neural Information Processing Systems, New Orleans, LA, 29 November-1 December, Curran Associates, Inc., Red Hook, NY

Kyle K, Crossley SA (2015) Automatically assessing lexical sophistication: Indices, tools, findings, and application. TESOL Q 49:757–786

Kyle K, Crossley SA, Berger CM (2018) The tool for the automatic analysis of lexical sophistication (TAALES): Version 2.0. Behav. Res. Methods 50:1030–1046. https://doi.org/10.3758/s13428-017-0924-4

Article   PubMed   Google Scholar  

Kyle K, Crossley SA, Jarvis S (2021) Assessing the validity of lexical diversity using direct judgements. Lang. Assess. Q. 18:154–170. https://doi.org/10.1080/15434303.2020.1844205

Landauer TK, Laham D, Foltz PW (2003) Automated essay scoring and annotation of essays with the Intelligent Essay Assessor. In Shermis MD, Burstein JC (ed), Automated Essay Scoring: A Cross-Disciplinary Perspective. Lawrence Erlbaum Associates, Mahwah, NJ

Landis JR, Koch GG (1977) The measurement of observer agreement for categorical data. Biometrics 159–174

Laufer B, Nation P (1995) Vocabulary size and use: Lexical richness in L2 written production. Appl. Linguist. 16:307–322. https://doi.org/10.1093/applin/16.3.307

Lee J, Hasebe Y (2017) jWriter Learner Text Evaluator, URL: https://jreadability.net/jwriter/

Lee J, Kobayashi N, Sakai T, Sakota K (2015) A Comparison of SPOT and J-CAT Based on Test Analysis [Tesuto bunseki ni motozuku ‘SPOT’ to ‘J-CAT’ no hikaku]. Research on the Acquisition of Second Language Japanese [Dainigengo to shite no nihongo no shūtoku kenkyū] (18) 53–69

Li W, Yan J (2021) Probability distribution of dependency distance based on a Treebank of. Japanese EFL Learners’ Interlanguage. J. Quant. Linguist. 28(2):172–186. https://doi.org/10.1080/09296174.2020.1754611

Article   MathSciNet   Google Scholar  

Linacre JM (2002) Optimizing rating scale category effectiveness. J. Appl. Meas. 3(1):85–106

PubMed   Google Scholar  

Linacre JM (1994) Constructing measurement with a Many-Facet Rasch Model. In Wilson M (ed) Objective measurement: Theory into practice, Volume 2 (pp. 129–144). Norwood, NJ: Ablex

Liu H (2008) Dependency distance as a metric of language comprehension difficulty. J. Cognitive Sci. 9:159–191

Liu H, Xu C, Liang J (2017) Dependency distance: A new perspective on syntactic patterns in natural languages. Phys. Life Rev. 21. https://doi.org/10.1016/j.plrev.2017.03.002

Loukina A, Madnani N, Cahill A, et al. (2020) Using PRMSE to evaluate automated scoring systems in the presence of label noise. Proceedings of the Fifteenth Workshop on Innovative Use of NLP for Building Educational Applications, Seattle, WA, USA → Online, 10 July, pp. 18–29. Association for Computational Linguistics

Lu X (2010) Automatic analysis of syntactic complexity in second language writing. Int. J. Corpus Linguist. 15:474–496

Lu X (2012) The relationship of lexical richness to the quality of ESL learners’ oral narratives. Mod. Lang. J. 96:190–208

Lu X (2017) Automated measurement of syntactic complexity in corpus-based L2 writing research and implications for writing assessment. Lang. Test. 34:493–511

Lu X, Hu R (2022) Sense-aware lexical sophistication indices and their relationship to second language writing quality. Behav. Res. Method. 54:1444–1460. https://doi.org/10.3758/s13428-021-01675-6

Ministry of Health, Labor, and Welfare of Japan (2022) Retrieved from: https://www.mhlw.go.jp/stf/newpage_30367.html

Mizumoto A, Eguchi M (2023) Exploring the potential of using an AI language model for automated essay scoring. Res. Methods Appl. Linguist. 3:100050

Okgetheng B, Takeuchi K (2024) Estimating Japanese Essay Grading Scores with Large Language Models. Proceedings of 30th Annual Conference of the Language Processing Society in Japan, March 2024

Ortega L (2015) Second language learning explained? SLA across 10 contemporary theories. In VanPatten B, Williams J (ed) Theories in Second Language Acquisition: An Introduction

Rae JW, Borgeaud S, Cai T, et al. (2021) Scaling Language Models: Methods, Analysis & Insights from Training Gopher. ArXiv, abs/2112.11446

Read J (2000) Assessing vocabulary. Cambridge University Press. https://doi.org/10.1017/CBO9780511732942

Rudner LM, Liang T (2002) Automated Essay Scoring Using Bayes’ Theorem. J. Technol., Learning and Assessment, 1 (2)

Sakoda K, Hosoi Y (2020) Accuracy and complexity of Japanese Language usage by SLA learners in different learning environments based on the analysis of I-JAS, a learners’ corpus of Japanese as L2. Math. Linguist. 32(7):403–418. https://doi.org/10.24701/mathling.32.7_403

Suzuki N (1999) Summary of survey results regarding comprehensive essay questions. Final report of “Joint Research on Comprehensive Examinations for the Aim of Evaluating Applicability to Each Specialized Field of Universities” for 1996-2000 [shōronbun sōgō mondai ni kansuru chōsa kekka no gaiyō. Heisei 8 - Heisei 12-nendo daigaku no kaku senmon bun’ya e no tekisei no hyōka o mokuteki to suru sōgō shiken no arikata ni kansuru kyōdō kenkyū’ saishū hōkoku-sho]. University Entrance Examination Section Center Research and Development Department [Daigaku nyūshi sentā kenkyū kaihatsubu], 21–32

Taghipour K, Ng HT (2016) A neural approach to automated essay scoring. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, Austin, Texas, 1–5 November, pp. 1882–1891. Association for Computational Linguistics

Takeuchi K, Ohno M, Motojin K, Taguchi M, Inada Y, Iizuka M, Abo T, Ueda H (2021) Development of essay scoring methods based on reference texts with construction of research-available Japanese essay data. In IPSJ J 62(9):1586–1604

Ure J (1971) Lexical density: A computational technique and some findings. In Coultard M (ed) Talking about Text. English Language Research, University of Birmingham, Birmingham, England

Vaswani A, Shazeer N, Parmar N, et al. (2017) Attention is all you need. In Advances in Neural Information Processing Systems, Long Beach, CA, 4–7 December, pp. 5998–6008, Curran Associates, Inc., Red Hook, NY

Watanabe H, Taira Y, Inoue Y (1988) Analysis of essay evaluation data [Shōronbun hyōka dēta no kaiseki]. Bulletin of the Faculty of Education, University of Tokyo [Tōkyōdaigaku kyōiku gakubu kiyō], Vol. 28, 143–164

Yao S, Yu D, Zhao J, et al. (2023) Tree of thoughts: Deliberate problem solving with large language models. Advances in Neural Information Processing Systems, 36

Zenker F, Kyle K (2021) Investigating minimum text lengths for lexical diversity indices. Assess. Writ. 47:100505. https://doi.org/10.1016/j.asw.2020.100505

Zhang Y, Warstadt A, Li X, et al. (2021) When do you need billions of words of pretraining data? Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, Online, pp. 1112-1125. Association for Computational Linguistics. https://doi.org/10.18653/v1/2021.acl-long.90

Download references

This research was funded by National Foundation of Social Sciences (22BYY186) to Wenchao Li.

Author information

Authors and affiliations.

Department of Japanese Studies, Zhejiang University, Hangzhou, China

Department of Linguistics and Applied Linguistics, Zhejiang University, Hangzhou, China

You can also search for this author in PubMed   Google Scholar

Contributions

Wenchao Li is in charge of conceptualization, validation, formal analysis, investigation, data curation, visualization and writing the draft. Haitao Liu is in charge of supervision.

Corresponding author

Correspondence to Wenchao Li .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Ethical approval

Ethical approval was not required as the study did not involve human participants.

Informed consent

This article does not contain any studies with human participants performed by any of the authors.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplemental material file #1, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Li, W., Liu, H. Applying large language models for automated essay scoring for non-native Japanese. Humanit Soc Sci Commun 11 , 723 (2024). https://doi.org/10.1057/s41599-024-03209-9

Download citation

Received : 02 February 2024

Accepted : 16 May 2024

Published : 03 June 2024

DOI : https://doi.org/10.1057/s41599-024-03209-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Albert Bandura’s Social Cognitive Theory

Charlotte Nickerson

Research Assistant at Harvard University

Undergraduate at Harvard University

Charlotte Nickerson is a student at Harvard University obsessed with the intersection of mental health, productivity, and design.

Learn about our Editorial Process

Saul Mcleod, PhD

Editor-in-Chief for Simply Psychology

BSc (Hons) Psychology, MRes, PhD, University of Manchester

Saul Mcleod, PhD., is a qualified psychology teacher with over 18 years of experience in further and higher education. He has been published in peer-reviewed journals, including the Journal of Clinical Psychology.

Olivia Guy-Evans, MSc

Associate Editor for Simply Psychology

BSc (Hons) Psychology, MSc Psychology of Education

Olivia Guy-Evans is a writer and associate editor for Simply Psychology. She has previously worked in healthcare and educational sectors.

On This Page:

Key Takeaways

• Social cognitive theory emphasizes the learning that occurs within a social context. In this view, people are active agents who can both influence and are influenced by their environment.
• The theory was founded most prominently by Albert Bandura, who is also known for his work on observational learning, self-efficacy, and reciprocal determinism.
• One assumption of social learning is that we learn new behaviors by observing the behavior of others and the consequences of their behavior.
• If the behavior is rewarded (positive or negative reinforcement), we are likely to imitate it; however, if the behavior is punished, imitation is less likely. For example, in Bandura and Walters’ experiment,  the children imitated more the aggressive behavior of the model who was praised for being aggressive to the Bobo doll.
• Social cognitive theory has been used to explain a wide range of human behavior, ranging from positive to negative social behaviors such as aggression, substance abuse, and mental health problems.

How We Learn From the Behavior of Others

Social cognitive theory views people as active agents who can both influence and are influenced by their environment.

The theory is an extension of social learning that includes the effects of cognitive processes — such as conceptions, judgment, and motivation — on an individual’s behavior and on the environment that influences them.

Rather than passively absorbing knowledge from environmental inputs, social cognitive theory argues that people actively influence their learning by interpreting the outcomes of their actions, which, in turn, affects their environments and personal factors, informing and altering subsequent behavior (Schunk, 2012).

By including thought processes in human psychology, social cognitive theory is able to avoid the assumption made by radical behaviorism that all human behavior is learned through trial and error. Instead, Bandura highlights the role of observational learning and imitation in human behavior.

Numerous psychologists, such as Julian Rotter and the American personality psychologist Walter Mischel, have proposed different social-cognitive perspectives.

Albert Bandura (1989) introduced the most prominent perspective on social cognitive theory.

Bandura’s perspective has been applied to a wide range of topics, such as personality development and functioning, the understanding and treatment of psychological disorders, organizational training programs, education, health promotion strategies, advertising and marketing, and more.

The central tenet of Bandura’s social-cognitive theory is that people seek to develop a sense of agency and exert control over the important events in their lives.

This sense of agency and control is affected by factors such as self-efficacy, outcome expectations, goals, and self-evaluation (Schunk, 2012).

Origins: The Bobo Doll Experiments

Social cognitive theory can trace its origins to Bandura and his colleagues, in particular, a series of well-known studies on observational learning known as the Bobo Doll experiments .

In these experiments, researchers exposed young, preschool-aged children to videos of an adult acting violently toward a large, inflatable doll.

This aggressive behavior included verbal insults and physical violence, such as slapping and punching. At the end of the video, the children either witnessed the aggressor being rewarded, or punished or received no consequences for his behavior (Schunk, 2012).

After being exposed to this model, the children were placed in a room where they were given the same inflatable Bobo doll.

The researchers found that those who had watched the model either received positive reinforcement or no consequences for attacking the doll were more likely to show aggressive behavior toward the doll (Schunk, 2012).

This experiment was notable for being one that introduced the concept of observational learning to humans.

Bandura’s ideas about observational learning were in stark contrast to those of previous behaviorists, such as B.F. Skinner.

According to Skinner (1950), learning can only be achieved through individual action.

However, Bandura claimed that people and animals can also learn by watching and imitating the models they encounter in their environment, enabling them to acquire information more quickly.

Observational Learning

Bandura agreed with the behaviorists that behavior is learned through experience. However, he proposed a different mechanism than conditioning.

He argued that we learn through observation and imitation of others’ behavior.

This theory focuses not only on the behavior itself but also on the mental processes involved in learning, so it is not a pure behaviorist theory.

Stages of the Social Learning Theory (SLT)

Not all observed behaviors are learned effectively. There are several factors involving both the model and the observer that determine whether or not a behavior is learned. These include attention, retention, motor reproduction, and motivation (Bandura & Walters, 1963).

The individual needs to pay attention to the behavior and its consequences and form a mental representation of the behavior. Some of the things that influence attention involve characteristics of the model.

This means that the model must be salient or noticeable. If the model is attractive, prestigious, or appears to be particularly competent, you will pay more attention. And if the model seems more like yourself, you pay more attention.

Storing the observed behavior in LTM where it can stay for a long period of time. Imitation is not always immediate. This process is often mediated by symbols. Symbols are “anything that stands for something else” (Bandura, 1998).

They can be words, pictures, or even gestures. For symbols to be effective, they must be related to the behavior being learned and must be understood by the observer.

Motor Reproduction

The individual must be able (have the ability and skills) to physically reproduce the observed behavior. This means that the behavior must be within their capability. If it is not, they will not be able to learn it (Bandura, 1998).

The observer must be motivated to perform the behavior. This motivation can come from a variety of sources, such as a desire to achieve a goal or avoid punishment.

Bandura (1977) proposed that motivation has three main components: expectancy, value, and affective reaction. Firstly, expectancy refers to the belief that one can successfully perform the behavior. Secondly, value refers to the importance of the goal that the behavior is meant to achieve.

The last of these, Affective reaction, refers to the emotions associated with the behavior.

If behavior is associated with positive emotions, it is more likely to be learned than a behavior associated with negative emotions. Reinforcement and punishment each play an important role in motivation.

Individuals must expect to receive the same positive reinforcement (vicarious reinforcement) for imitating the observed behavior that they have seen the model receiving.

Imitation is more likely to occur if the model (the person who performs the behavior) is positively reinforced. This is called vicarious reinforcement.

Imitation is also more likely if we identify with the model. We see them as sharing some characteristics with us, i.e., similar age, gender, and social status, as we identify with them.

Features of Social Cognitive Theory

The goal of social cognitive theory is to explain how people regulate their behavior through control and reinforcement in order to achieve goal-directed behavior that can be maintained over time.

Bandura, in his original formulation of the related social learning theory, included five constructs, adding self-efficacy to his final social cognitive theory (Bandura, 1986).

Reciprocal Determinism

Reciprocal determinism is the central concept of social cognitive theory and refers to the dynamic and reciprocal interaction of people — individuals with a set of learned experiences — the environment, external social context, and behavior — the response to stimuli to achieve goals.

Its main tenet is that people seek to develop a sense of agency and exert control over the important events in their lives.

This sense of agency and control is affected by factors such as self-efficacy, outcome expectations, goals, and self-evaluation (Bandura, 1989).

To illustrate the concept of reciprocal determinism, Consider A student who believes they have the ability to succeed on an exam (self-efficacy) is more likely to put forth the necessary effort to study (behavior).

If they do not believe they can pass the exam, they are less likely to study. As a result, their beliefs about their abilities (self-efficacy) will be affirmed or disconfirmed by their actual performance on the exam (outcome).

This, in turn, will affect future beliefs and behavior. If the student passes the exam, they are likely to believe they can do well on future exams and put forth the effort to study.

If they fail, they may doubt their abilities (Bandura, 1989).

Behavioral Capability

Behavioral capability, meanwhile, refers to a person’s ability to perform a behavior by means of using their own knowledge and skills.

That is to say, in order to carry out any behavior, a person must know what to do and how to do it. People learn from the consequences of their behavior, further affecting the environment in which they live (Bandura, 1989).

Reinforcements

Reinforcements refer to the internal or external responses to a person’s behavior that affect the likelihood of continuing or discontinuing the behavior.

These reinforcements can be self-initiated or in one’s environment either positive or negative. Positive reinforcements increase the likelihood of a behavior being repeated, while negative reinforcers decrease the likelihood of a behavior being repeated.

Reinforcements can also be either direct or indirect. Direct reinforcements are an immediate consequence of a behavior that affects its likelihood, such as getting a paycheck for working (positive reinforcement).

Indirect reinforcements are not immediate consequences of behavior but may affect its likelihood in the future, such as studying hard in school to get into a good college (positive reinforcement) (Bandura, 1989).

Expectations

Expectations, meanwhile, refer to the anticipated consequences that a person has of their behavior.

Outcome expectations, for example, could relate to the consequences that someone foresees an action having on their health.

As people anticipate the consequences of their actions before engaging in a behavior, these expectations can influence whether or not someone completes the behavior successfully (Bandura, 1989).

Expectations largely come from someone’s previous experience. Nonetheless, expectancies also focus on the value that is placed on the outcome, something that is subjective from individual to individual.

For example, a student who may not be motivated to achieve high grades may place a lower value on taking the steps necessary to achieve them than someone who strives to be a high performer.

Self-Efficacy

Self-efficacy refers to the level of a person’s confidence in their ability to successfully perform a behavior.

Self-efficacy is influenced by a person’s own capabilities as well as other individual and environmental factors.

These factors are called barriers and facilitators (Bandura, 1989). Self-efficacy is often said to be task-specific, meaning that people can feel confident in their ability to perform one task but not another.

For example, a student may feel confident in their ability to do well on an exam but not feel as confident in their ability to make friends.

This is because self-efficacy is based on past experience and beliefs. If a student has never made friends before, they are less likely to believe that they will do so in the future.

Modeling Media and Social Cognitive Theory

Learning would be both laborious and hazardous in a world that relied exclusively on direct experience.

Social modeling provides a way for people to observe the successes and failures of others with little or no risk.

This modeling can take place on a massive scale. Modeling media is defined as “any type of mass communication—television, movies, magazines, music, etc.—that serves as a model for observing and imitating behavior” (Bandura, 1998).

In other words, it is a means by which people can learn new behaviors. Modeling media is often used in the fashion and taste industries to influence the behavior of consumers.

This is because modeling provides a reference point for observers to imitate. When done effectively, modeling can prompt individuals to adopt certain behaviors that they may not have otherwise engaged in.

Additionally, modeling media can provide reinforcement for desired behaviors.

For example, if someone sees a model wearing a certain type of clothing and receives compliments for doing so themselves, they may be more likely to purchase clothing like that of the model.

Observational Learning Examples

There are numerous examples of observational learning in everyday life for people of all ages.

Nonetheless, observational learning is especially prevalent in the socialization of children. For example:

• A newer employee avoids being late to work after seeing a colleague be fired for being late.
• A new store customer learns the process of lining up and checking out by watching other customers.
• A traveler to a foreign country learning how to buy a ticket for a train and enter the gates by witnessing others do the same.
• A customer in a clothing store learns the procedure for trying on clothes by watching others.
• A person in a coffee shop learns where to find cream and sugar by watching other coffee drinkers locate the area.
•  A new car salesperson learning how to approach potential customers by watching others.
•  Someone moving to a new climate and learning how to properly remove snow from his car and driveway by seeing his neighbors do the same.
•  A tenant learning to pay rent on time as a result of seeing a neighbor evicted for late payment.
•  An inexperienced salesperson becomes successful at a sales meeting or in giving a presentation after observing the behaviors and statements of other salespeople.
•  A viewer watches an online video to learn how to contour and shape their eyebrows and then goes to the store to do so themselves.
•  Drivers slow down after seeing that another driver has been pulled over by a police officer.
•  A bank teller watches their more efficient colleague in order to learn a more efficient way of counting money.
•  A shy party guest watching someone more popular talk to different people in the crowd, later allowing them to do the same thing.
• Adult children behave in the same way that their parents did when they were young.
• A lost student navigating a school campus after seeing others do it on their own.

Social Learning vs. Social Cognitive Theory

Social learning theory and Social Cognitive Theory are both theories of learning that place an emphasis on the role of observational learning.

However, there are several key differences between the two theories. Social learning theory focuses on the idea of reinforcement, while Social Cognitive Theory emphasizes the role of cognitive processes.

Additionally, social learning theory posits that all behavior is learned through observation, while Social Cognitive Theory allows for the possibility of learning through other means, such as direct experience.

Finally, social learning theory focuses on individualistic learning, while Social Cognitive Theory takes a more holistic view, acknowledging the importance of environmental factors.

Though they are similar in many ways, the differences between social learning theory and Social Cognitive Theory are important to understand. These theories provide different frameworks for understanding how learning takes place.

As such, they have different implications in all facets of their applications (Reed et al., 2010).

Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory . Prentice-Hall, Inc.

Bandura, A. (1977). Social learning theory . Englewood Cliffs, NJ: Prentice Hall.

Bandura, A. (1977). Self-efficacy: toward a unifying theory of behavioral change. Psychological Review, 84 (2), 191.

 Bandura, A. (1986). Fearful expectations and avoidant actions as coeffects of perceived self-inefficacy.

Bandura, A. (1989). Human agency in social cognitive theory. American psychologist, 44 (9), 1175.

Bandura, A. (1998). Health promotion from the perspective of social cognitive theory. Psychology and health, 13 (4), 623-649.

Bandura, A. (2003). Social cognitive theory for personal and social change by enabling media. In Entertainment-education and social change (pp. 97-118). Routledge.

Bandura, A. Ross, D., & Ross, S. A. (1961). Transmission of aggression through the imitation of aggressive models. Journal of Abnormal and Social Psychology , 63, 575-582.

LaMort, W. (2019). The Social Cognitive Theory. Boston University.

Reed, M. S., Evely, A. C., Cundill, G., Fazey, I., Glass, J., Laing, A., … & Stringer, L. C. (2010). What is social learning?. Ecology and society, 15 (4).

Schunk, D. H. (2012). Social cognitive theory .

Skinner, B. F. (1950). Are theories of learning necessary?. Psychological Review, 57 (4), 193.

Related Articles

Learning Theories

Aversion Therapy & Examples of Aversive Conditioning

Learning Theories , Psychology , Social Science

Albert Bandura’s Social Learning Theory

Learning Theories , Psychology

Behaviorism In Psychology

Famous Experiments , Learning Theories

Bandura’s Bobo Doll Experiment on Social Learning

Bloom’s Taxonomy of Learning

Child Psychology , Learning Theories

Jerome Bruner’s Theory Of Learning And Cognitive Development

Purdue Online Writing Lab Purdue OWL® College of Liberal Arts

Welcome to the Purdue Online Writing Lab

Welcome to the Purdue OWL

This page is brought to you by the OWL at Purdue University. When printing this page, you must include the entire legal notice.

Copyright ©1995-2018 by The Writing Lab & The OWL at Purdue and Purdue University. All rights reserved. This material may not be published, reproduced, broadcast, rewritten, or redistributed without permission. Use of this site constitutes acceptance of our terms and conditions of fair use.

The Online Writing Lab at Purdue University houses writing resources and instructional material, and we provide these as a free service of the Writing Lab at Purdue. Students, members of the community, and users worldwide will find information to assist with many writing projects. Teachers and trainers may use this material for in-class and out-of-class instruction.

The Purdue On-Campus Writing Lab and Purdue Online Writing Lab assist clients in their development as writers—no matter what their skill level—with on-campus consultations, online participation, and community engagement. The Purdue Writing Lab serves the Purdue, West Lafayette, campus and coordinates with local literacy initiatives. The Purdue OWL offers global support through online reference materials and services.

A Message From the Assistant Director of Content Development 

The Purdue OWL® is committed to supporting  students, instructors, and writers by offering a wide range of resources that are developed and revised with them in mind. To do this, the OWL team is always exploring possibilties for a better design, allowing accessibility and user experience to guide our process. As the OWL undergoes some changes, we welcome your feedback and suggestions by email at any time.

Please don't hesitate to contact us via our contact page  if you have any questions or comments.

All the best,

Social Media

Facebook twitter.

Reference Examples

More than 100 reference examples and their corresponding in-text citations are presented in the seventh edition Publication Manual . Examples of the most common works that writers cite are provided on this page; additional examples are available in the Publication Manual .

To find the reference example you need, first select a category (e.g., periodicals) and then choose the appropriate type of work (e.g., journal article ) and follow the relevant example.

When selecting a category, use the webpages and websites category only when a work does not fit better within another category. For example, a report from a government website would use the reports category, whereas a page on a government website that is not a report or other work would use the webpages and websites category.

Also note that print and electronic references are largely the same. For example, to cite both print books and ebooks, use the books and reference works category and then choose the appropriate type of work (i.e., book ) and follow the relevant example (e.g., whole authored book ).

Examples on these pages illustrate the details of reference formats. We make every attempt to show examples that are in keeping with APA Style’s guiding principles of inclusivity and bias-free language. These examples are presented out of context only to demonstrate formatting issues (e.g., which elements to italicize, where punctuation is needed, placement of parentheses). References, including these examples, are not inherently endorsements for the ideas or content of the works themselves. An author may cite a work to support a statement or an idea, to critique that work, or for many other reasons. For more examples, see our sample papers .

Reference examples are covered in the seventh edition APA Style manuals in the Publication Manual Chapter 10 and the Concise Guide Chapter 10

Related handouts

• Common Reference Examples Guide (PDF, 147KB)
• Reference Quick Guide (PDF, 225KB)

Textual Works

Textual works are covered in Sections 10.1–10.8 of the Publication Manual . The most common categories and examples are presented here. For the reviews of other works category, see Section 10.7.

• Journal Article References
• Magazine Article References
• Newspaper Article References
• Blog Post and Blog Comment References
• UpToDate Article References
• Book/Ebook References
• Diagnostic Manual References
• Children’s Book or Other Illustrated Book References
• Classroom Course Pack Material References
• Religious Work References
• Chapter in an Edited Book/Ebook References
• Dictionary Entry References
• Wikipedia Entry References
• Report by a Government Agency References
• Report with Individual Authors References
• Brochure References
• Ethics Code References
• Fact Sheet References
• ISO Standard References
• Press Release References
• White Paper References
• Conference Presentation References
• Conference Proceeding References
• Published Dissertation or Thesis References
• Unpublished Dissertation or Thesis References
• ERIC Database References
• Preprint Article References

Data and Assessments

Data sets are covered in Section 10.9 of the Publication Manual . For the software and tests categories, see Sections 10.10 and 10.11.

• Data Set References
• Toolbox References

Audiovisual Media

Audiovisual media are covered in Sections 10.12–10.14 of the Publication Manual . The most common examples are presented together here. In the manual, these examples and more are separated into categories for audiovisual, audio, and visual media.

• Artwork References
• Clip Art or Stock Image References
• Film and Television References
• Musical Score References
• Online Course or MOOC References
• Podcast References
• PowerPoint Slide or Lecture Note References
• Radio Broadcast References
• TED Talk References
• Transcript of an Audiovisual Work References
• YouTube Video References

Online Media

Online media are covered in Sections 10.15 and 10.16 of the Publication Manual . Please note that blog posts are part of the periodicals category.

• Facebook References
• Instagram References
• LinkedIn References
• Online Forum (e.g., Reddit) References
• TikTok References
• X References
• Webpage on a Website References
• Clinical Practice References
• Open Educational Resource References
• Whole Website References

What is generative AI?

In the months and years since ChatGPT burst on the scene in November 2022, generative AI (gen AI) has come a long way. Every month sees the launch of new tools, rules, or iterative technological advancements. While many have reacted to ChatGPT (and AI and machine learning more broadly) with fear, machine learning clearly has the potential for good. In the years since its wide deployment, machine learning has demonstrated impact in a number of industries, accomplishing things like medical imaging analysis  and high-resolution weather forecasts. A 2022 McKinsey survey shows that AI adoption has more than doubled  over the past five years, and investment in AI is increasing apace. It’s clear that generative AI tools like ChatGPT (the GPT stands for generative pretrained transformer) and image generator DALL-E (its name a mashup of the surrealist artist Salvador Dalí and the lovable Pixar robot WALL-E) have the potential to change how a range of jobs are performed. The full scope of that impact, though, is still unknown—as are the risks.

Get to know and directly engage with McKinsey's senior experts on generative AI

Aamer Baig is a senior partner in McKinsey’s Chicago office;  Lareina Yee  is a senior partner in the Bay Area office; and senior partners  Alex Singla  and Alexander Sukharevsky , global leaders of QuantumBlack, AI by McKinsey, are based in the Chicago and London offices, respectively.

Still, organizations of all stripes have raced to incorporate gen AI tools into their business models, looking to capture a piece of a sizable prize. McKinsey research indicates that gen AI applications stand to add up to $4.4 trillion  to the global economy—annually. Indeed, it seems possible that within the next three years, anything in the technology, media, and telecommunications space not connected to AI will be considered obsolete or ineffective .

But before all that value can be raked in, we need to get a few things straight: What is gen AI, how was it developed, and what does it mean for people and organizations? Read on to get the download.

To stay up to date on this critical topic, sign up for email alerts on “artificial intelligence” here .

Learn more about QuantumBlack , AI by McKinsey.

What every CEO should know about generative AI

What’s the difference between machine learning and artificial intelligence, about quantumblack, ai by mckinsey.

QuantumBlack, McKinsey’s AI arm, helps companies transform using the power of technology, technical expertise, and industry experts. With thousands of practitioners at QuantumBlack (data engineers, data scientists, product managers, designers, and software engineers) and McKinsey (industry and domain experts), we are working to solve the world’s most important AI challenges. QuantumBlack Labs is our center of technology development and client innovation, which has been driving cutting-edge advancements and developments in AI through locations across the globe.

Artificial intelligence is pretty much just what it sounds like—the practice of getting machines to mimic human intelligence to perform tasks. You’ve probably interacted with AI even if you don’t realize it—voice assistants like Siri and Alexa are founded on AI technology, as are customer service chatbots that pop up to help you navigate websites.

Machine learning is a type of artificial intelligence. Through machine learning, practitioners develop artificial intelligence through models that can “learn” from data patterns without human direction. The unmanageably huge volume and complexity of data (unmanageable by humans, anyway) that is now being generated has increased machine learning’s potential , as well as the need for it.

What are the main types of machine learning models?

Machine learning is founded on a number of building blocks, starting with classical statistical techniques  developed between the 18th and 20th centuries for small data sets. In the 1930s and 1940s, the pioneers of computing—including theoretical mathematician Alan Turing—began working on the basic techniques for machine learning. But these techniques were limited to laboratories until the late 1970s, when scientists first developed computers powerful enough to mount them.

Until recently, machine learning was largely limited to predictive models, used to observe and classify patterns in content. For example, a classic machine learning problem is to start with an image or several images of, say, adorable cats. The program would then identify patterns among the images, and then scrutinize random images for ones that would match the adorable cat pattern. Generative AI was a breakthrough. Rather than simply perceive and classify a photo of a cat, machine learning is now able to create an image or text description of a cat on demand.

Introducing McKinsey Explainers : Direct answers to complex questions

How do text-based machine learning models work how are they trained.

ChatGPT may be getting all the headlines now, but it’s not the first text-based machine learning model to make a splash. OpenAI’s GPT-3 and Google’s BERT both launched in recent years to some fanfare. But before ChatGPT, which by most accounts works pretty well most of the time (though it’s still being evaluated), AI chatbots didn’t always get the best reviews. GPT-3 is “by turns super impressive and super disappointing,” said New York Times tech reporter Cade Metz in a video where he and food writer Priya Krishna asked GPT-3 to write recipes for a (rather disastrous) Thanksgiving dinner .

The first machine learning models to work with text were trained by humans to classify various inputs according to labels set by researchers. One example would be a model trained to label social media  posts as either positive or negative. This type of training is known as supervised learning because a human is in charge of “teaching” the model what to do.

The next generation of text-based machine learning models rely on what’s known as self-supervised learning. This type of training involves feeding a model a massive amount of text so it becomes able to generate predictions. For example, some models can predict, based on a few words, how a sentence will end. With the right amount of sample text—say, a broad swath of the internet—these text models become quite accurate. We’re seeing just how accurate with the success of tools like ChatGPT.

What does it take to build a generative AI model?

Building a generative AI model has for the most part been a major undertaking, to the extent that only a few well-resourced tech heavyweights have made an attempt . OpenAI, the company behind ChatGPT, former GPT models, and DALL-E, has billions in funding from bold-face-name donors. DeepMind is a subsidiary of Alphabet, the parent company of Google, and even Meta has dipped a toe into the generative AI model pool with its Make-A-Video product. These companies employ some of the world’s best computer scientists and engineers.

But it’s not just talent. When you’re asking a model to train using nearly the entire internet, it’s going to cost you. OpenAI hasn’t released exact costs, but estimates indicate that GPT-3 was trained on around 45 terabytes of text data—that’s about one million feet of bookshelf space, or a quarter of the entire Library of Congress—at an estimated cost of several million dollars. These aren’t resources your garden-variety start-up can access.

What kinds of output can a generative AI model produce?

As you may have noticed above, outputs from generative AI models can be indistinguishable from human-generated content, or they can seem a little uncanny. The results depend on the quality of the model—as we’ve seen, ChatGPT’s outputs so far appear superior to those of its predecessors—and the match between the model and the use case, or input.

ChatGPT can produce what one commentator called a “ solid A- ” essay comparing theories of nationalism from Benedict Anderson and Ernest Gellner—in ten seconds. It also produced an already famous passage describing how to remove a peanut butter sandwich from a VCR in the style of the King James Bible. Image-generating AI models like DALL-E 2 can create strange, beautiful images on demand, like a Raphael painting of a Madonna and child, eating pizza . Other generative AI models can produce code, video, audio, or business simulations .

But the outputs aren’t always accurate—or appropriate. When Priya Krishna asked DALL-E 2 to come up with an image for Thanksgiving dinner, it produced a scene where the turkey was garnished with whole limes, set next to a bowl of what appeared to be guacamole. For its part, ChatGPT seems to have trouble counting, or solving basic algebra problems—or, indeed, overcoming the sexist and racist bias that lurks in the undercurrents of the internet and society more broadly.

Generative AI outputs are carefully calibrated combinations of the data used to train the algorithms. Because the amount of data used to train these algorithms is so incredibly massive—as noted, GPT-3 was trained on 45 terabytes of text data—the models can appear to be “creative” when producing outputs. What’s more, the models usually have random elements, which means they can produce a variety of outputs from one input request—making them seem even more lifelike.

What kinds of problems can a generative AI model solve?

The opportunity for businesses is clear. Generative AI tools can produce a wide variety of credible writing in seconds, then respond to criticism to make the writing more fit for purpose. This has implications for a wide variety of industries, from IT and software organizations that can benefit from the instantaneous, largely correct code generated by AI models to organizations in need of marketing copy. In short, any organization that needs to produce clear written materials potentially stands to benefit. Organizations can also use generative AI to create more technical materials, such as higher-resolution versions of medical images. And with the time and resources saved here, organizations can pursue new business opportunities and the chance to create more value.

We’ve seen that developing a generative AI model is so resource intensive that it is out of the question for all but the biggest and best-resourced companies. Companies looking to put generative AI to work have the option to either use generative AI out of the box or fine-tune them to perform a specific task. If you need to prepare slides according to a specific style, for example, you could ask the model to “learn” how headlines are normally written based on the data in the slides, then feed it slide data and ask it to write appropriate headlines.

What are the limitations of AI models? How can these potentially be overcome?

Because they are so new, we have yet to see the long tail effect of generative AI models. This means there are some inherent risks  involved in using them—some known and some unknown.

The outputs generative AI models produce may often sound extremely convincing. This is by design. But sometimes the information they generate is just plain wrong. Worse, sometimes it’s biased (because it’s built on the gender, racial, and myriad other biases of the internet and society more generally) and can be manipulated to enable unethical or criminal activity. For example, ChatGPT won’t give you instructions on how to hotwire a car, but if you say you need to hotwire a car to save a baby, the algorithm is happy to comply. Organizations that rely on generative AI models should reckon with reputational and legal risks involved in unintentionally publishing biased, offensive, or copyrighted content.

These risks can be mitigated, however, in a few ways. For one, it’s crucial to carefully select the initial data used to train these models to avoid including toxic or biased content. Next, rather than employing an off-the-shelf generative AI model, organizations could consider using smaller, specialized models. Organizations with more resources could also customize a general model based on their own data to fit their needs and minimize biases. Organizations should also keep a human in the loop (that is, to make sure a real human checks the output of a generative AI model before it is published or used) and avoid using generative AI models for critical decisions, such as those involving significant resources or human welfare.

It can’t be emphasized enough that this is a new field. The landscape of risks and opportunities  is likely to change rapidly in coming weeks, months, and years. New use cases are being tested monthly, and new models are likely to be developed in the coming years. As generative AI becomes increasingly, and seamlessly, incorporated into business, society, and our personal lives, we can also expect a new regulatory climate  to take shape. As organizations begin experimenting—and creating value—with these tools, leaders will do well to keep a finger on the pulse of regulation and risk.

Articles referenced include:

• " Implementing generative AI with speed and safety ,” March 13, 2024, Oliver Bevan, Michael Chui , Ida Kristensen , Brittany Presten, and Lareina Yee
• “ Beyond the hype: Capturing the potential of AI and gen AI in tech, media, and telecom ,” February 22, 2024, Venkat Atluri , Peter Dahlström , Brendan Gaffey , Víctor García de la Torre, Noshir Kaka , Tomás Lajous , Alex Singla , Alex Sukharevsky , Andrea Travasoni , and Benjamim Vieira
• “ As gen AI advances, regulators—and risk functions—rush to keep pace ,” December 21, 2023, Andreas Kremer, Angela Luget, Daniel Mikkelsen , Henning Soller , Malin Strandell-Jansson, and Sheila Zingg
• “ The economic potential of generative AI: The next productivity frontier ,” June 14, 2023, Michael Chui , Eric Hazan , Roger Roberts , Alex Singla , Kate Smaje , Alex Sukharevsky , Lareina Yee , and Rodney Zemmel
• “ What every CEO should know about generative AI ,” May 12, 2023, Michael Chui , Roger Roberts , Tanya Rodchenko, Alex Singla , Alex Sukharevsky , Lareina Yee , and Delphine Zurkiya
• “ Exploring opportunities in the generative AI value chain ,” April 26, 2023, Tobias Härlin, Gardar Björnsson Rova , Alex Singla , Oleg Sokolov, and Alex Sukharevsky
• “ The state of AI in 2022—and a half decade in review ,” December 6, 2022,  Michael Chui ,  Bryce Hall ,  Helen Mayhew , Alex Singla , and Alex Sukharevsky
• “ McKinsey Technology Trends Outlook 2023 ,” July 20, 2023,  Michael Chui , Mena Issler,  Roger Roberts , and  Lareina Yee  
• “ An executive’s guide to AI ,” Michael Chui , Vishnu Kamalnath, and Brian McCarthy
• “ What AI can and can’t do (yet) for your business ,” January 11, 2018,  Michael Chui , James Manyika , and Mehdi Miremadi

This article was updated in April 2024; it was originally published in January 2023.

Want to know more about generative AI?

Related articles.

The data dividend: Fueling generative AI

Don’t wait—create, with generative AI