research paper on montessori method

ORIGINAL RESEARCH article

An association between montessori education in childhood and adult wellbeing.

• Department of Psychology, University of Virginia, Charlottesville, VA, United States

Wellbeing, or how people think and feel about their lives, predicts important life outcomes from happiness to health to longevity. Montessori pedagogy has features that enhance wellbeing contemporaneously and predictively, including self-determination, meaningful activities, and social stability. Here, 1905 adults, ages 18–81 ( M = 36), filled out a large set of wellbeing scales followed by demographic information including type of school attended each year from 2 to 17. About half the sample had only attended conventional schools and the rest had attended Montessori for between 2 and 16 years ( M = 8 years). To reduce the variable set, we first developed a measurement model of wellbeing using the survey data with exploratory then confirmatory factor analyses, arriving at four factors: general wellbeing, engagement, social trust, and self-confidence. A structural equation model that accounted for age, gender, race, childhood SES, and years in private school revealed that attending Montessori for at least two childhood years was associated with significantly higher adult wellbeing on all four factors. A second analysis found that the difference in wellbeing between Montessori and conventional schools existed even among the subsample that had exclusively attended private schools. A third analysis found that the more years one attended Montessori, the higher one’s wellbeing as an adult. Unmeasured selection effects could explain the results, in which case research should determine what third variable associated with Montessori schooling causes adult wellbeing. Several other limitations to the study are also discussed. Although some of these limitations need to be addressed, coupled with other research, including studies in which children were randomly assigned to Montessori schools, this study suggests that attending Montessori as a child might plausibly cause higher adult wellbeing.

Introduction

Wellbeing, or the felt experience of health, happiness, and flourishing, predicts several desirable outcomes including better health and work performance, longevity, and more positive social behavior and relations ( Ryff, 2014 ). Low levels of wellbeing predict suicidal behavior even more strongly than does mental illness ( Keyes et al., 2012 ). Even intrinsically, wellbeing could be considered the supreme human outcome ( Diener et al., 2015 ). Although wellbeing is partially determined by genetic inheritance ( Røysamb and Nes, 2019 ), environmental factors are important contributors ( Diener et al., 2016 ). Yet few childhood experiences have been shown to predict adults’ psychological wellbeing. One that does is residential moves: more moves in childhood significantly predicts lower adult wellbeing ( Oishi and Schimmack, 2010 ). Here we explore whether a different childhood experience, Montessori education, might predict higher adult wellbeing. We know of no research examining an association between Montessori specifically and later wellbeing, but one study found that people who had attended various alternative schools including Montessori as children adjusted better to university, controlling for high school baseline ( Shankland et al., 2010 ). Montessori warrants further study, as it is the most common and long-lasting alternative progressive pedagogy in the world ( Debs, 2019 ) and has several features that are endemic to wellbeing-enhancing educational environments ( White and Kern, 2018 ).

A logic model for Montessori education ( Culclasure et al., 2019 ) predicts that Montessori features like choosing one’s activities, using real, hands-on materials, and collaborating with peers would result in a range of positive personal and social outcomes. Summaries of child development research and their implications for educational environments also suggest that attending schools with Montessori features (like collaboration and learning based on interests) should enhance wellbeing ( National Academies of Sciences Engineering and Medicine, 2018 ; Darling-Hammond et al., 2019 ). Actual studies in conventional schools also show that features consistent with Montessori (like low test anxiety: Montessori has no tests) predict higher wellbeing in school ( Baker, 2004 ; Cohen, 2006 ; Felner et al., 2007 ; Seligman et al., 2009 ; Steinmayr et al., 2016 , 2018 ). Furthermore, random lottery studies of Montessori students (discussed later) show differences from waitlisted controls suggesting Montessori lays groundwork that would be expected to lead to higher wellbeing ( Lillard and Else-Quest, 2006 ; Lillard et al., 2017 ). Here we present a series of structural equation models showing that Montessori schooling in childhood is associated with higher adult wellbeing, after accounting for a range of demographic variables. We begin with discussion of the Montessori system and three features that are known to enhance wellbeing in school and other settings: choice or self-determination, meaningful activities, and social stability.

Montessori Schooling

Initiated in 1907, Montessori pedagogy ( Montessori, 1967/1995 ) is the oldest surviving and most prevalent child-centered, constructivist education system in the world ( Debs, 2019 ), practiced in over 500 public and thousands of private American schools ( National Center for Montessori in the Public Sector, 2014 ) and tens of thousands of schools around the world ( 180 Studio and Saunders Eckenhoff, 2020 ). Although some think of it as a preschool model serving mainly White children, Montessori extends through high school, and over half of American public Montessori students today are children of color ( Debs, 2019 ). Three salient Montessori features would be expected, based both on logic and prior research, to lead to certain long term wellbeing outcomes: self-determination, meaningful activity, and stable social relationships (for more discussion of characteristics of Montessori programs, see Lillard and McHugh, 2019a , b ). Although we could not study this directly, we could examine whether there are associations between prior Montessori schooling and adult wellbeing. No study to our knowledge has specifically examined whether attending Montessori is associated with feelings of wellbeing over the long term; school studies tend to look at concurrent or relatively proximate outcomes ( Baker, 2004 ; Cohen, 2006 ; Felner et al., 2007 ; Seligman et al., 2009 ; Steinmayr et al., 2016 , 2018 ), or other long-term outcomes like income ( Chetty et al., 2018 ). Wellbeing in adulthood is a multidetermined outcome, influenced by health, wealth, marital status, and many other features ( Diener et al., 2013 , 2016 ), but childhood school experience could plausibly be another predictor. Here we consider how Montessori embodies three features that other research has shown predict wellbeing. Figure 1 portrays the model we arrived at after conducting the study; below we describe the model we hypothesized prior to analyzing our data.

Figure 1. Final model of hypothesized relation of Montessori features to wellbeing aspects. This is the final model that was arrived at after analyses; the hypothesized model described in the text has a different set of hypothesized results.

Self-Determination

Children in Montessori classrooms choose their own work almost all the time, and in this sense are very much in charge of their own educations. A teacher guides children in individual or small group lessons, but children decide which lessons to follow up on, and even whether to sit in on a lesson: they determine their own activities. The classroom is composed of hundreds of potential activities, laid out on shelves, and children choose among them, or even arrange their own field trips during which they leave the classroom to do something else ( Montessori, 1967/1995 ; Lillard, 2017 ).

Self-determination has been conceptualized as fulfilling a fundamental human need for autonomy ( Deci and Ryan, 2011 ). Choice and its progeny, a sense of control or agency, have been shown to predict stronger intrinsic motivation, self-efficacy, happiness, and sense of competence ( Langer and Rodin, 1976 ; Ryan and Grolnick, 1986 ; Rovee-Collier and Hayne, 2000 ; Patall et al., 2008 , 2010 ); living in a more individualistic country, where self-determination is higher, also predicts wellbeing at a national level ( Rhoads et al., 2021 ). Related to self-determination, Montessori has no grades or tests. Research has shown that external rewards and performance evaluations proffer an external sense of control ( Ryan and Deci, 2000 ), and their absence thus allows for an internal locus of control and a sense of self-determination (see Lillard, 2017 , Chapter 6). In addition, when given external rewards for doing tasks, people tend to opt for easier tasks ( Harter, 1978 ), avoiding challenge.

We hypothesized that self-determination might lead to two sets of outcomes. A first set concerns what one might call general wellbeing: happiness, finding meaning in life, self-confidence, and (relatedly) a sense of personal competence. A second set of potential outcomes is related to intrinsic motivation: people who experience a high degree of self-determination early in life might later be more intrinsically engaged and seek more challenge. Supporting the idea that self-determination in Montessori schools might cause this array of outcomes, an experience sampling study showed that Montessori middle school students rated themselves more highly than did conventionally schooled students on positive affect, experiences of “flow” ( Csikszentmihalyi, 1997 ), potency, and intrinsic motivation ( Rathunde and Csikszentmihalyi, 2005a ). We hypothesized that adults who formerly attended Montessori schools would show higher general wellbeing, including self-confidence, and be more apt to seek challenge and be engaged with their work.

Meaningful Activities

Montessori offers children meaningful activities, by which we mean activities for which the underlying reasons are clear, and thus give people a sense of purpose. Offering meaningful activities is crucial for a school system built on self-determination ( Lillard, 2019 ), because meaning motivates engagement ( Bruner, 1990 ). In prekindergarten, Montessori children learn to prepare, serve, and clean up meals, take care of their classroom, and button their clothes—all meaningful activities for a young child; research shows that children prefer really doing such activities to pretending to do them, because they like having a purpose in their activities ( Taggart et al., 2018 , 2020 ). When learning abstract concepts like number, children in Montessori use concrete materials that make obvious what the abstraction is about; this extends for example to using a cube composed of 27 blocks that make clear why the trinomial formula gives the area of a cube: each block represents an element of the formula. If I see that putting together a set of blocks where three blocks have all sides of length A, B, and C (respectively), 3 correspond to a2b, and so on, then I can understand why the trinomial formula works; the formula then has meaning—it is not merely an abstraction. In addition, Montessori students can pursue activities that interest them, which of course translates to their activities being personally meaningful.

Work that is geared at an optimal level—challenging but possible—engages people, putting them in the positive state referred to as “flow” ( Csikszentmihalyi, 1997 ). Under free conditions in which a range of meaningful options are supplied, and no rewards (like grades) are offered, people tend to choose engagements at an intermediate level of challenge ( Danner and Lonky, 1981 ; Dweck, 1999 ; Simon, 2001 ; Kidd et al., 2012 , 2014 )—not so easy that they learn nothing, but not so difficult that they will collapse in frustration (and therefore not learn). Given this, the fact that children choose their own work should translate to work being meaningful, which would in theory support higher levels of engagement in Montessori schools. The aforementioned research also suggests this is the case ( Rathunde and Csikszentmihalyi, 2005a ): Montessori middle school students reported feeling more engaged than other students. Engagement also increases general wellbeing ( Lewis et al., 2011 ), and so by extension meaningful activities might also increase general wellbeing. As shown in Figure 1 , we hypothesized that the increased engagement that research has shown occurs in Montessori classrooms, perhaps due to one’s activities being meaningful, translates to increased engagement in activities throughout life. We also expected that a pattern of having meaningful activities in the school years could translate to a general sense of meaning in life and happiness, and thus be related to general wellbeing.

Social Stability and Cohesion

Another Montessori education feature that might enhance long term wellbeing is the social environment, which nourishes social-emotional development and sustained relationships. Classrooms span 3 years (for example, 6- to 9-year-olds, 9- to 12-year-olds, and so on through high school) during which children have the same teacher and immediate peer group; children just older and younger are classmates for 1–2 years, and are met again as one moves up classroom levels.

The practice of staying with the same teacher and classmates, called “looping,” supports positive relationships, self-confidence, and academic performance ( Burke, 1996 ; Little and Dacus, 1999 ; Cistone and Shneyderman, 2004 ; Nitecki, 2017 ; Hill and Jones, 2018 ); the one study that showed better academic performance but not better relationships ( Tourigny et al., 2020 ) used only 2 years of looping whereas Montessori and some other studies use 3; 3 years (versus 2) might make a difference to relationship quality. Positive academic performance also predicts later wellbeing ( Ng et al., 2015 ; Steinmayr et al., 2016 ). Being in ungraded multiage classrooms (for example, where “1st grader” is not identified), as Montessori children are, is also particularly beneficial for both academic and social-emotional outcomes, and the benefits increase with more time in such classrooms ( Lloyd, 1999 ).

The Montessori practice of not having grades or tests also might benefit relationships with both teachers and peers. For one, the teacher becomes a guide, rather than one who makes judgments and “grades.” Among students, absence of tests and grades fosters collaboration whereas grades foster competition ( Butler and Ruzany, 1993 ). Collaboration itself is another reason why Montessori schooling might be associated with social aspects of higher adult wellbeing. In Montessori classrooms, particularly at after age 6, students constantly collaborate on schoolwork, which could reasonably be expected to cultivate social skills. As noted, older Montessori students also go on self-arranged small group field trips, and they often create classroom rules themselves (as a group). By middle and high school, Montessori classrooms might go on longer trips together. All these practices could foster greater social cohesion.

We know of no strong studies examining child-teacher relationships in Montessori, but there is evidence (including from lottery-control studies, discussed below) suggesting that peer relationships are stronger in Montessori. This makes sense because, as opposed to conventional schools where students usually work individually, in Montessori schools students often work in small groups. Moreover, research indicates that Montessori student’s social knowledge and skills are more advanced, and the overall school climate is better ( Flynn, 1991 ; Rathunde and Csikszentmihalyi, 2005b ; Lillard and Else-Quest, 2006 ; Lillard et al., 2017 ; Denervaud et al., 2020a ). Random lottery studies also indicate that academic performance is stronger in Montessori ( Lillard and Else-Quest, 2006 ; Lillard et al., 2017 ) and well-controlled matched/growth studies (e.g., Culclasure et al., 2018 ; Denervaud et al., 2019 ) suggest Montessori leads to higher academic performance. Stronger academic performance has been shown to lead to higher well-being ( Ng et al., 2015 ; Steinmayr et al., 2016 ), possibly via improved self-esteem ( Yang et al., 2019 ), which reinforces more positive relationships and sense of community.

Taken together, findings on social stability led us to predict that Montessori students would have more positive social relationships and a stronger sense of community throughout life (see Figure 1 ). Such factors are typically related to general wellbeing, thus the strong social stability in Montessori schools could also predict higher general wellbeing later.

In sum, we hypothesized that Montessori pedagogy in childhood might lead to higher wellbeing later in life. The reasoning behind this hypothesis was that the pedagogy has three features (self-determination, meaningful activities, and social cohesion) that other studies have shown enhance wellbeing along several dimensions (clustering into what we might call general wellbeing, intrinsic motivation/engagement, and social skills/social cohesion), and because studies of Montessori versus conventionally schooled children (including natural experiments) have indicated that Montessori students are different along these dimensions during their school years. In the next section we present those natural experiments.

Natural Experiments Suggesting Montessori Might Cause a Trajectory to Higher Wellbeing

Natural experiments using random lotteries have examined the immediate influence of Montessori on proximal child outcomes, and the results for the two studies which involved highly trained Montessori teachers suggest children have higher wellbeing while in school and are on a trajectory toward higher wellbeing later in their lives. These studies controlled for parent characteristics because admission was determined by a random lottery that parents of the control children had also entered. The studies thus have high internal validity, although the results might not apply to families that do not enter such lotteries, lowering external validity. The first study examined children in Kindergarten and 6th grade ( Lillard and Else-Quest, 2006 ), whereas the second followed across preschool (from ages 3 to 6) children who were equal on all measured outcomes at baseline ( Lillard et al., 2017 ). Both contrasted children in high fidelity public Montessori schools (in that both met the standards for recognition by the Association Montessori Internationale or AMI, which meant all the teachers had AMI’s intensive 9-month training and diploma) with waitlist control children in business-as-usual non-Montessori schools. In terms of self-determination and its benefits, the studies showed better academic performance and mastery orientation. In terms of social skills, they showed better social cognition and behavior, and a stronger sense of community. They also indicated more developed executive function. Academic performance, mastery orientation, social skills, and executive function all predict higher wellbeing ( Moffitt et al., 2011 ; Reynolds et al., 2011 , 2017 ; Sancassiani et al., 2015 ; Steinmayr et al., 2016 , 2018 ; Haimovitz and Dweck, 2017 ; Darling-Hammond et al., 2019 ). Because random assignment designs support causal inference, these lottery control studies suggest attending Montessori might cause higher wellbeing in adults. However, a third natural experiment in France contrasted Montessori children taught by untrained teachers with Ecole Maternale (the highly regarded national preschool program) children; all were randomly assigned at the classroom level ( Courtier, 2020 ). In this study, Montessori children performed unequivocally better on reading, but not on an array of other measures. Further research is needed to determine why these studies had different results; one possibility is that wellbeing-related qualities emerge more reliably when teachers have Montessori training; another is that the Ecole Maternale program has superior outcomes to business-as-usual programs in the United States. Regardless, the U.S. studies lend support to the idea that Montessori causes certain qualities in American children, and other studies show those qualities to be associated with higher wellbeing.

Taking together these findings, as well as the fact that Montessori has conditions that are associated with higher wellbeing, we hypothesized that adults who went to Montessori as children have higher adult wellbeing.

Materials and Methods

Participants.

Participants were recruited through various methods including Facebook ads in cities known to have many Montessori schools, Amazon’s Mechanical Turk, school associations, and snowballing. The final sample consisted of 1905 participants in the US and Canada who had attended Montessori for at least 2 years or who had spent virtually all their school years at conventional schools. Two years of intervention was selected because that duration led to significant outcomes for the Perry Preschool Project ( Heckman, 2006 ), and 2 years of Head Start is significantly more impactful than 1 year ( Wen et al., 2012 ). Although Montessori schooling considers 3 years in a classroom to constitute a full “cycle,” whether specific sets of 3 years are associated with later wellbeing was not addressed here. Participants’ mean age was 37.05 years ( SD = 13.12, range = 18–81 years), 79.2% were female, and 83.0% identified as White, 3.4% as Black or African American, 4.5% as Asian, 3.7% as Hispanic or Latino, and less than 1% as American Indian or Alaska Native, Native Hawaiian or Pacific Islander. Another 3.8% aligned with multiple races/ethnicities, and 1% self-identified in other categories (e.g., Jewish) or preferred not to answer. Because of the small sizes of categories other than White, they were grouped for analyses.

The sample size goal was 2000 participants, 1000 in each group (Montessori and conventional), which is well above the threshold needed for an SEM involving 83 paths which is the size of our largest model ( Wolf et al., 2013 ). Our recruiting cut-off was set at the desired n s or 6 months, whichever came earlier. Based on Wolf et al.’s (2013) analysis, and given our model structure, parameters, and variables, our final sample size of 1905 should have sufficient power to detect the hypothesized effects.

Participants were partitioned into groups using R. Those who had spent no or only 1 year in an unconventional school (like Montessori) were categorized as conventional, n = 1071 (19 had gone to Montessori for just 1 year). Those with at least 2 years of Montessori, n = 834, were classified as such. Although a 2-year cutoff was used, for the Montessori group the average length of attendance was 8 years ( SD = 3.66, range 2–16). An additional group of participants who had attended other alternative schools for two or more years ( n = 506) were excluded from analyses to focus here on Montessori versus conventional schooling. (Their results will be reported elsewhere).

Surveys were administered on the Qualtrics platform with a compensation rate of $0.50/survey. Participants who gave informed consent were then given a series of scales and questions, ending with demographic questions including school types attended. The stated purpose of the research in the informed consent was “to better understand the long-term outcomes of alternative and conventional school education on peoples’ lives.” On the final page, school type history was requested; no specific school system was mentioned until after all other measures had been filled out.

The survey included 18 established scales (see below) that were intended to get at a variety of aspects of wellbeing, and a few ordinal scales and other questions getting at issues of interest. Because the ordinal scales explained little variance in wellbeing, they were not used in our analyses, but those results were consistent with the ones below and will be reported elsewhere. Eleven of our 18 scales are subscales of the Psychological and Social Wellbeing scales (SWB). Below, after considering advantages and disadvantages of online surveys, we describe each scale with its original alpha; in the present study, alphas were the same or exceeded the originals in all cases except one (Social Coherence, for which our α = 0.54 vs. 0.64 in the original).

Online Survey Delivery

Online surveys are an important source of data for psychology research, with pros and cons ( Gosling and Mason, 2015 ). One benefit is the ability to recruit large samples from the general population (versus, for example, undergraduate psychology majors) including samples with relatively rare characteristics of interest. This made them ideal for this study, since Montessori schools are much less common than conventional schools. Other benefits are reducing self-presentation bias and experimenter influence. There are also disadvantages. For example, drop-out rates in online survey research are high, averaging 34% in a meta-analysis that saw ranges from 0 to 87% ( Musch and Reips, 2000 ). Drop-out can be a serious concern when it occurs more in one assigned condition than another, but this does not apply here because conditions were not assigned in this study. Also mitigating this concern, studies have shown that samples completing internet surveys closely resemble the populations from which they are drawn ( Gosling and Mason, 2015 ). Another potential problem is multiple submissions; we guarded against this by ensuring each respondent had a unique computer identifier (IP address). Another issue is that respondents are limited to people who use the internet, and findings might not generalize to the population. As internet usage increases, this is less of a concern—87% of households in the developed world were on the internet in 2018 and these data were collected in 2019 ( ITU Publications, 2019 ). However, it is the case that survey respondents (be they on a telephone or online) tend to be younger, female, White, and affluent ( Curtin et al., 2000 ; Singer et al., 2000 ; Smith, 2008 ). We accounted for these factors in our models, but it is a limitation of the online survey method and therefore of this study.

There are many measures of wellbeing ( Ackerman et al., 2018 ), tapping into outcomes ranging from finding meaning in one’s life to mindful awareness to one’s sense of community support. Our approach was to use a wide range of accepted measures of adult wellbeing, and reduce the measures to a manageable set using exploratory and confirmatory factor analyses, and then examine how the resulting factors align with the three hypothesized outcome clusters. Finally, we conducted a series of Structural Equation Modeling (SEM) analyses to examine whether Montessori was a meaningful predictor of outcomes after accounting for demographic variables.

Psychological Wellbeing Scales (PWB)

Participants filled out six PWB scales ( Ryff and Keyes, 1995 ) of three items each. Participants rated each item using a 7-point scale ranging from strongly agree to strongly disagree . The six PWB scales, with their original reported alphas, are Self-Acceptance (e.g., “When I look at the story of my life, I am pleased with how things have turned out so far”; α = 0.59), Environmental Mastery (“I am good at managing the responsibilities of daily life”; α = 0.52), and Autonomy (“I judge myself by what I think is important”; α = 0.48), all of which seem to tap into General Wellbeing (see Figure 1 ); Personal Growth (“Life is a continuous process of growth”; α = 0.55) and Purpose in Life (“Some people wander aimlessly through life; I am not one of them”; α = 0.36), which seem to tap Engagement; and Positive Relations (“People would describe me as a giving person”; α = 0.58), which seems to tap Social Trust. Higher scores indicate greater levels of wellbeing.

Social Wellbeing Scales

The five SWB Scales ( Keyes, 1998 ) each have three items that use the same 7-point scale as the PWB scales. The SWB scales are Social Coherence (“I can predict/make sense of the world”; α = 0.64) and Social Contribution (“I have something to give”; α = 0.66) which tap into the self-confidence aspect of General Wellbeing and perhaps the meaning aspect of Engagement; and Social Integration (“I feel close to people in my community”; α = 0.73), Social Acceptance (“People are kind”; α = 0.41), and Social Actualization (“Society is getting better”; α = 0.64) which all appear to tap into Social Trust.

Satisfaction With Life Scale

The Satisfaction with Life Scale (SWL) ( Diener et al., 1985 ), one of the most commonly used measures of wellbeing ( Ackerman et al., 2018 ), consists of five items (e.g., “In most ways my life is close to my ideal”; α = 0.87) which participants rate using the same 7-point scale ranging from strongly disagree to strongly agree . Ratings for each of the five items are summed up to calculate a single aggregate score. A high score indicates high satisfaction with one’s own life, and seems to tap into one’s general sense of wellbeing and happiness.

Meaning in Life Questionnaire

This 10-item scale ( Steger et al., 2008 ) measures meaning in life, including its presence and one’s search for meaning; in the current study we used the 5-item MILQ-Presence subscale to assess the presence of meaning in life. Using a 7-point scale ranging from absolutely untrue to absolutely true , participants rate five short statements such as, “My life has a clear sense of purpose”; α = 0.86. An aggregate score is calculated by summing up the five items, and a higher score reflects a higher subjective sense of meaning in one’s life. This scale appears to tap into the meaning aspect of engagement.

Subjective Vitality Scale

This 7-item scale ( Ryan and Frederick, 1997 ) measures the extent to which one feels alive and alert. Using a 7-point scale ranging from not at all true to very true , participants rate seven short statements such as, “I feel alive and vital”; α = 0.83. An aggregate score is calculated by adding ratings from each of the items. As is recommended, the second item (the only one needing to be reverse scored) was dropped ( Bostic et al., 2000 ). This scale also seems to tap general wellbeing and happiness.

Short Need for Cognition Scale

This 18-item scale ( Cacioppo et al., 1984 ) measures the extent to which individuals engage in and enjoy effortful thinking. Using a 5-point scale ranging from extremely uncharacteristic of me to extremely characteristic of me , participants rate statements such as, “I really enjoy a task that involves coming up with new solutions and problems”; α = 0.90. An aggregate score is calculated by adding ratings from the 18 items, with higher scores indicating high interest in thinking, complex problem solving, and intellectual tasks. We expected that having the opportunity to choose difficult and meaningful activities as a child might lead to a lifelong desire to seek challenges, creating Engagement.

Mindful Attention Awareness Scale

This 15-item scale ( Brown and Ryan, 2003 ) measures individuals’ dispositional mindfulness, or awareness and attention to the present moment. Using a 6-point scale ranging from almost always to almost never , participants rate each statement with reference to their day-to-day experiences. For example, one item is, “I find it difficult to stay focused on what is happening in the present;” a high score means that is almost never true; α = 0.81. An aggregate score is calculated by averaging the 15 responses. Higher scores reflect higher dispositional mindfulness, which is thought to be generally related to wellbeing.

As noted we also administered some ordinal scales and a few isolated questions; in addition we administered the Big 5 personality survey ( Brody and Ehrlichman, 1998 ). Because whether personality should be viewed as a wellbeing outcome is controversial, we do not discuss this further here, nor do we discuss the ordinal scales which did not contribute to the variance in our models.

Demographics and School History

After completing the wellbeing measures, participants answered standard demographic questions, reporting factors such as their age, gender, and race. In addition, they were asked how they would describe their family’s social class when they were 3–12 years old, with five options: lower/working, lower middle, middle, upper middle, and upper. Similar scales have been successfully used in other studies to measure adults’ estimates of their childhood SES ( Straughen et al., 2013 ; Listl et al., 2018 ; Lindberg et al., 2021 ). Childhood SES is highly related to child outcomes ( Reardon, 2011 ; Duncan and Murnane, 2014 ), and a meta-analysis showed that one’s own estimate of one’s SES (often called SSS, for subjective social status) is more strongly related to wellbeing than is one’s actual SES ( Tang et al., 2016 ).

Finally, participants were asked, for each year from ages 2 to 17, what type of school they attended, with options including Regular/Traditional, Montessori, Homeschool, Waldorf, Reggio Emilia, Other Alternative, and did not attend. They were also asked how the school they went to each year was funded, with options of Public, Private (non-religious), and Private (religious), and did not attend. The format of this page was as follows: the ages (“2 years old,” and so on) were listed down the left-most column, the next seven columns held circles one could check for the type of school, and the next four columns were for the funding model.

Analytic Approach

There was no missing data. An exploratory factor analysis (EFA) was done on a randomly selected two-thirds of respondents’ data using all variables and maximum likelihood extraction. A four-factor structure was confirmed in a confirmatory factor analysis (CFA) done with the remaining data. These factors were then entered in a series of Structural Equation Models (SEM) that accounted for age, race, childhood SES, gender, and proportion of years attending private schools. All analyses were run using R, OpenMx 2.19.1 ( Neale et al., 2016 ), and SPSS 24.

The purpose of using factor analysis here was to reduce the variable set; the variables were selected because they are frequently used to measure wellbeing, rather than with an eye to expected factors. Adequacy of model fit was determined according to the guidelines set by Plucker (2003) for factor analyses (RMSEA.05-0.10; AGFI 90+, CFI.90+) and using a multi-index approach as suggested by Hu and Bentler (1999) . For this we added the Tucker-Lewis Index (TLI), adopting the common standard for of 0.90. RMSEA was the favored index because of the relatively large sample size and number of indicators, which can lower the values of CFI and TLI indices ( Kenny and McCoach, 2003 ; Shi et al., 2019 ). Chi-square was inappropriate here because of the relatively large sample size ( Hu and Bentler, 1999 ; Russell, 2002 ).

Data Preparation

Several scales, particularly the PWB scales, resulted in data that were on visual inspection negatively skewed, reflecting that as a whole, the sample had high wellbeing. Box–Cox transformations were applied to skewed variables ( Box and Cox, 1964 ). We rounded lambdas to the nearest whole number to decide whether to square or cube the variable. Box–Cox transformations were then applied to check that the new rounded lambda value was close to 1; the one exception to this was Personal Growth, which would need to be raised to the 6th power by this criterion; after cubing had a lambda of 2.07 and was not transformed further. As stated, there were no missing data.

Table 1A shows the raw unadjusted means, SDs, scale alphas in this study, and the correlations among the scales, age, and Montessori status; Table 1B shows the means and SDs or percentages on each demographic variable for the Montessori and conventional samples separately, and Table 1C shows the same for the scales. The vast majority (80%) of participants filled out the surveys in under 30 min; 16 min was the modal time to completion. First we report results from the EFA, then the CFA. We then discuss the resulting factors before turning to the SEM.

Table 1A. Means, SDs, and correlation matrix for observed continuous variables: entire sample.

Table 1B. Means, SDs, and/or percentages for demographic variables for the Montessori and conventional samples.

Table 1C. Untransformed means and standard errors for Montessori and conventional samples.

Factor Analyses

Exploratory factor analysis using maximum likelihood extraction was run with the correlation matrix from data of approximately 2/3 of the participants ( n = 1220). This n allowed covariates (such as gender and childhood SES) to be balanced across conditions (Montessori and conventional). A Promax oblique rotation was used, allowing all factors to intercorrelate. Initially the training set was run using 1–11 factors and all variables. The ordinal items (classified as “Other Questions” in the survey) explained little variance (low communality or h 2 values), which made sense in that they were designed to get at issues a step removed from wellbeing. The ordinal variables were removed, and a new set of EFAs were run using the same methods.

A parallel analysis approach ( Horn, 1965 ) was taken to determining the number of factors. This approach is favored over the Kaiser criterion of eigenvalue > 1, which has been described as “not psychometrically justifiable” ( Reise et al., 2000 , p. 291) and is prone to over and under extraction of factors ( Reise et al., 2000 ; Russell, 2002 ). With parallel analysis, random data sets are generated constituting the same number of items and participants; the scree plot from these data is laid over the scree plot from the actual study data, and where the actual and simulated data lines cross indicates the maximum number of factors (see Figure 2 ); fit statistics are also taken into consideration (see Table 2 ). Examination of the scree plot suggested a maximum of five factors, which explained 54% of the variance. However, a four-factor solution had an adequate fit by the preferred measure (RMSEA of 0.079) and explained 50% of the variance; it was selected because the five-factor solution had ultra-Heywood cases ( Dillon et al., 1987 ). All factor loadings for the four-factor solution are shown in Table 3 . The cut-off value for a variable’s loading on a factor was set to >0.35 because this resulted in the best-fitting model and minimized the number of crossloadings.

Figure 2. Parallel analysis scree plot.

Table 2. Goodness of fit measures and variance explained for analyses 1, 2, and 3.

Table 3. Promax rotated factor loadings: EFA 4-factor solution.

Factor 3 has only two indicators; three or more can improve stability ( Thurstone, 1947 ), but this recommendation is more flexible for SEM models, and Kenny (2012) advises that a two-indicator factor is acceptable if the errors are uncorrelated and their loadings are set to equal each other. Therefore, in the CFA, the loadings of the two indicators were set to equal each other, and these conditions were met.

The factor correlations and variance accounted for are shown in Table 4 . General wellbeing accounted for the largest proportion of variance (38%), and was strongly correlated with all three other factors, as might be expected of General Wellbeing. Engagement accounted for 30% of the variance, and was also strongly related to Social Trust and Self-confidence ( r s or 0.52 and 0.48, respectively). Social Trust and Self confidence each accounted for 16% of the variance; their relation to each other was 0.28.

Table 4. EFA: factor correlations and proportion of variance accounted for.

Next a confirmatory factor analysis using maximum likelihood estimation was conducted on the remaining 36% of respondents’ data; the results are in Table 5 and fit indices are in Table 2 . Figure 3 shows the resulting model. As one might expect from the high correlations, there were mostly very high loadings; the four factors and the item loadings are discussed next.

Table 5. CFA loadings and standard errors (SE).

Figure 3. Diagram of confirmatory factor analytic solution.

The Four Factors

The purpose of the factor analyses was to reduce the set of scales to a smaller number of factors and to suggest latent variables emerging from a set of wellbeing measures administered to all participants. Based on prior research concerning the outcomes or at least associates of self-determination, meaningful activities, and social stability (which can exist to varying degrees in any school experience), we hypothesized three latent clusters would emerge: one concerning happiness, meaning, and self-confidence; one concerning engagement and seeking challenge; and a third concerning social trust and sense of community. The factor analysis actually resulted in four factors that align reasonably well with what was hypothesized, with the first cluster of outcomes splitting into two (self-confidence and general wellbeing; see Figure 1 ). Again, the present study is not directly testing the paths shown in the figure; it only tests their plausibility.

General Wellbeing

The first factor is “General Wellbeing,” with six indicators (see Table 5 ). Self Acceptance from the Psychological Wellbeing Scale was set to 1.0, and one other Psychological Wellbeing scale, Environmental Mastery (e.g., “In general, I feel I am in charge of the situation in which I live”), also loaded highly on this factor (0.90). Meaning in Life (“My life has a clear sense of purpose”) and Satisfaction with Life (“In most ways, my life is close to ideal”) loaded highly on this factor as well, as did Subjective Vitality (“I have energy and spirit”). Mindful Attention Awareness (“I rush through activities without being really attentive to them” [item is reverse scored]) had the lowest loading at 0.62. Cronbach’s alpha for these six items on the CFA test data set was 0.88 (95% confidence interval 0.87–0.89). Dropping any item except Mindful Attention Awareness resulted in a lower alpha; dropping that item slightly increased alpha to 0.89, but this was within the confidence interval, and dropping it from the model resulted in Heywood cases, so it was retained.

We had hypothesized that three Montessori features–self-determination, meaningful activities, and social stability–would lead to happiness, a sense of meaning, and self-confidence. These predictions were upheld, in that the Satisfaction with Life scale, Subjective Vitality, and Self Acceptance all reflect happiness, Meaning in Life is eponymous, and Environmental Mastery reflects self-confidence. However, three other variables that reflect self-confidence in a somewhat different sense loaded instead on a discrete factor we called Self-Confidence (discussed below); in the EFA Environmental Mastery also loaded on that factor (0.23) but not at a level that met the threshold of 0.35.

The second factor reflects investing oneself in one’s activities and social world. Social Contribution (“My daily activities contribute to something worthwhile to my community”) was set to 1.0 and Social Integration (“I feel close to other people in my community”) had a loading of 0.91; both these scales are from the Social Wellbeing scale. Also loading on Engagement were Personal Growth (e.g., “Life is a continuous process of learning,” 0.88), Positive Relations (“People would describe me as a giving person, willing to share my time with others,” 0.93), and Purpose in Life (“Some people wander aimlessly through life, but I am not one of them,” 0.82), all Psychological Wellbeing scales. Cronbach’s alpha for these five items was 0.83 (95% confidence interval 0.81–0.85). Dropping any item resulted in a lower alpha. This factor resembled the second cluster of outcomes we hypothesized would result from the Montessori characteristics of self-determination and meaningful activities, with a stronger social engagement element than was anticipated.

Social Trust

The third factor included two Social Wellbeing subscales, Social Acceptance and Social Actualization, which reflect trust in society—items like, “The world is becoming a better place for everyone” and “I believe that people are kind.” Because there were two items for this factor, their loadings were set to be equal (see above) at 0.71. Cronbach’s alpha for these two items was 0.67 (95% confidence interval 0.63–0.72). This factor reflects outcomes prior research suggested would result from Montessori’s high degree of social stability.

Self-Confidence

Loading on the fourth factor were three variables that reflect confidence in one’s own thinking (as opposed to one’s behaviors, which the Environmental Mastery subscale taps more). Social Coherence from the Social Wellbeing scale, including items like “I find it easy to predict what will happen next” was set to 1.0, and the Need for Cognition scale loaded highly with it (0.99); this scale includes items like, “I like to have the responsibility of handling a situation that requires a lot of thinking” and “I prefer my life to be full of puzzles that I must solve.” Autonomy from the Psychological Wellbeing scale also loaded on this factor (0.86), with items like, “I have confidence in my own opinions” and “I judge myself by what I think is important, not what others think is important.” Cronbach’s alpha for these three items was 0.57 (95% confidence interval 0.52–0.63). Dropping any item resulted in a lower alpha.

In sum, two of the three outcome clusters we had hypothesized, based on prior research, would stem from experiences involving high levels of self-determination, meaningful activities, and social stability were upheld, with Engagement having a social aspect that was not expected. The first hypothesized factor, however, split into two, with outcomes pertinent to general wellbeing (including confidence in one’s abilities) falling into one cluster and outcomes more specifically related to self-confidence about one’s thought processes in a fourth cluster.

Predicting the Structural Model From Montessori Attendance

Having reduced the wellbeing scales to a set of four latent factors, the next step was to examine whether experience with Montessori schooling is associated with participants’ scores on any of those four factors, for the whole sample of 1905 participants, divided into the Montessori and Conventional school groups as explained above. Gender (Male/Female), race (Caucasian/Not), age, childhood SES (lower/working, lower middle, middle, upper middle, upper), and proportion of schooling that was private were accounted for as covariates in the models. There was a significant improvement in model fit when the binary variable of schooling (Montessori for at least 2 years, or Conventional) was added into the model, both overall and ranging across all four factors. Social trust had the largest beta-value (0.32) but even the lowest beta-value, for Self-confidence, showed a highly significant effect of Montessori ( p < 0.001).

This indicates that the means on all four factors are significantly higher for the Montessori group, after accounting for the covariates. The means, SDs, standardized regression coefficients, and corrected p -values are shown in Table 6 , and the covariate values are shown in the Supplementary Table .

Table 6. Means and SDs, standardized regression coefficients, and p -values for SEM.

Montessori attendance significantly predicted higher scores on all four latent variables: General Wellbeing, Engagement, Social Trust, and Self-confidence. This makes theoretical sense, in that Montessori schools have features that are related to these aspects of wellbeing. For example, Montessori gives children free choice and thus a high degree of self-determination, which (as reviewed in the section “Introduction”) has been shown in other research to render happiness and a strong sense of one’s own competence, and which allows one to find and engage in activities that give one a sense of purpose. The second feature we highlighted is that Montessori activities are meaningful, in that they have a clear purpose to which children can relate; this, along with self-determination, allows one to choose work that provides an optimal level of challenge, creating strong engagement. We did not anticipate that higher engagement among people who attended Montessori would include social integration, but it clustered with other variables tapping into engagement in the factor analysis. The third Montessori feature, social stability (including multi-year classrooms), was hypothesized to lead to strong relationships which then predict higher general wellbeing as well as social trust. Classroom looping practices also improve academic performance, which in turn predicts higher wellbeing. Thus, the results are consistent with what we hypothesized, based on prior research.

However, an alternative possible explanation for these results is that they stem not from at least 2 (and in this sample, an average of 8) years of Montessori education or some associate thereof, but instead from a third variable, perhaps something associated with having parents who make the effort to select and in most cases finance a specific school for their child, as opposed to using the default neighborhood public option. In other words, it may be that having parents who go out of their way to find and fund a different school program leads to higher adult wellbeing, or is associated with other factors that lead to higher wellbeing. Of course, many public school parents also are very intentional about their choice, choosing their domicile (and paying property taxes) based on public school district, but nonetheless they do not pay tuition in addition to taxes. Although we had covaried years of private school in the initial analysis, a more focused way to examine whether something associated with parents choosing a private school explains the results is to limit the dataset to participants who always attended a private school, because private schools are never the default option; every child in the United States and Canada lives in a school district where they could attend a tuition-free public school at least from Kindergarten on. A second analysis therefore analyzed data from the subset of participants who attended private schools for all of their schooling.

Robustness Check/Alternative Specification #1

The second analysis involved the subsample of 439 participants who had exclusively attended private schools: a Montessori group of n = 268 for whom at least 2 of those years were in private Montessori (with all or most of the remaining years in conventional private school programs), and Conventional group of n = 171 who went exclusively to conventional private schools. The Montessori group had spent M = 9.22 years ( SD = 3.59) in Montessori schools and M = 6.14 ( SD = 3.71) in conventional private schools, whereas the exclusively private conventional group had spent M = 14.53 years ( SD = 1.30) in school, virtually all of it in conventional private schools. The mean age of participants was 33.99 years ( SD = 11.55, range = 18–71), 93 were male (21.2%), and the rest were female; 87.9% identified as White.

The structural equation model of the initial analysis was conducted using only data from the exclusively privately schooled subset of participants. These results, including mean factor scores and SDs, are shown in Table 7 ; model fit statistics are in Table 2 . Even among the exclusively privately schooled subset—those participants whose parents selected and typically paid tuition at a private school for their entire precollege life, and even after accounting for the effects of age, gender, race, and childhood SES, having attended Montessori for at least 2 years (and an average of 9 years) was significantly associated with higher wellbeing on three of the four factors: General Wellbeing, Engagement, and Social Trust. Self-confidence was not significant ( p = 0.07), suggesting that confidence in one’s own thinking/mind is as strong among those who attended private conventional schools as among those who attended private Montessori schools.

Table 7. Analysis 2, private-only sample: means, SDs, standardized regression coefficients and p -values for SEM.

The standardized Beta values in the SEM were similar to what they were for the whole sample, but slightly stronger for Social Trust and slightly less strong for Self-confidence. Thus, while the initial analysis controlled for the proportion of schooling that was private, this second analysis shows that even among the subsample who only attended private schools, model fit improves when Montessori status is added.

Although wellbeing was still higher for Montessori compared to other participants, it is possible that this is because there is something about parents who choose Montessori (public or private) for their children that differs from other parents, and that it is those differences that lead to the better outcomes. This possibility was addressed in a third analysis by examining duration effects.

Particularly as children get older, duration of Montessori attendance would often reflect availability rather than parent choice. The option to attend Montessori after age 6 is limited, because Montessori elementary, middle, and high schools are far less prevalent than Montessori preschools; even Montessori elementary schools (for children ages 6–12) were extremely rare 30 years ago (when our average participant age was 6); Montessori elementary schools have gradually become more common, whereas Montessori high schools are still rare today. Because duration of Montessori attendance is often constrained by availability, self-selection is less of an issue in such an analysis, raising the odds that (were any significant effects found) the programs caused effects. Although this analysis was conducted because positive results would strengthen the possibility of causality, we caution that it is only a test of association.

Robustness Check/Alternative Specification #2

The subset of 853 respondents from the sample who had attended Montessori for at least 1 year in childhood was examined, omitting those who had never attended it (since such an analysis would in effect virtually repeat the initial analyses). This subset had an average age of 32.07 ( SD = 9.73, range 18–61 years). They had attended Montessori for a mean of 7.88 years ( SD = 3.77, range = 2–16 years) and conventional school for a mean of 7.36 years ( SD = 3.83); 23% were male and 84% were White.

The SEM tested the association between years in Montessori and the latent factors, again controlling for age, gender, race, childhood SES, and proportion of schooling that was private. The model fit statistics are in Table 2 and the regression coefficients and p -values are in Table 8 . In this analysis, the SEM showed the duration of Montessori was significant for two of the four factors (General Wellbeing and Engagement). For all four factors, the direction was positive: being in Montessori school for longer was associated with at least slightly higher scores on all factors.

Table 8. Analysis 3, effect of Montessori duration: means, SDs, standardized regression coefficients and p -values for SEM.

This analysis suggests two possibilities. The first is the causal possibility, that Montessori schooling could cause positive wellbeing outcomes, but that for two of the factors, a threshold number of years delivers those outcomes, and extending beyond those adds little additional benefit to social trust or self-confidence. However, General Wellbeing, composed of variables aimed at life happiness, meaning, and sense of one’s own competence, and Engagement (in one’s activities and social world) might be strengthened by more time in a Montessori environment. Alternatively, it may be that parents who choose to and are able to (because they live in communities where it is available) keep their children in Montessori longer also give their children other experiences that promote even more General Wellbeing and Engagement, but not more experiences that promote more Social Trust and Self-confidence than were indexed by the initial choice.

The present research aimed to determine if Montessori schooling in childhood might plausibly lead to higher wellbeing in adulthood, because many features of Montessori schooling are known to cause higher wellbeing contemporaneously or predictively in other school and non-school situations. Based on this existing research, we developed a hypothesized model of how three Montessori program features—self-determination, meaningful activities, and social stability—might lead to three clusters of wellbeing outcomes—a general cluster including happiness, finding meaning in life, and feeling competent; another including engaging and seeking challenge in one’s activities; and a third around a strong sense of community and social trust. Although we could not test that model directly, we could test its plausibility by seeing whether Montessori schooling was associated with latent factors aligned with those outcomes. Over 1900 individuals filled out a large set of wellbeing surveys and their responses were subjected to exploratory and confirmatory factor analyses, which arranged into four latent factors similar to those originally hypothesized, but with confidence in one’s own thoughts and mind emerging as a distinct fourth factor, and the engagement factor including social engagement as well (see Figure 1 ).

To test the plausibility of the hypothesis about schooling, we conducted a structural equation model analysis to determine if at least 2 years of Montessori schooling in childhood is significantly associated with adult wellbeing; the model accounted for the covariates of gender, race, age, childhood SES, and years in private school. The first analysis showed that Montessori was associated with higher scores on all four latent factors: General Wellbeing, Engagement, Social Trust, and Self-confidence.

This could be due to some feature of the parents; although the initial analysis controlled for SES, a different and unmeasured variable could be operating. The second analysis asked if something associated with selecting a private school for one’s child might be the operational variable, and therefore tested whether Montessori would be significantly associated with higher adult wellbeing even among the subsample who attended private schools at least through age 17. For three of the four latent factors, it was: General Wellbeing, Engagement, and Social Trust; the fourth, Self-confidence (in one’s thinking), showed a trend. Attending Montessori schooling for at least two childhood years was associated with higher wellbeing on these factors even among people who only attended private schools their entire pre-college lives.

Having parents who always had selected a private school was thus not responsible for the generally higher wellbeing associated with having attended Montessori observed in the initial analysis. Perhaps there is something associated with selecting Montessori specifically (whether public or private) that is associated with higher wellbeing. Although three studies that used dozens of measures (like parenting styles measures) to discriminate Montessori from other parents found no significant differences ( Fleege et al., 1967 ; Dreyer and Rigler, 1969 ; Denervaud et al., 2020b ), there must be some different qualities, and the present study rendered no way to examine those directly. However, Montessori enrollment, particularly as one gets older, is constrained by availability, yet the postulated unmeasured parent qualities would be expected to persist regardless of that availability. Therefore the third analysis examined whether duration of Montessori enrollment is associated with higher wellbeing. Duration of Montessori enrollment was associated with the latent factors of General Wellbeing and Engagement, but not Social Trust or Self-confidence, for which the direction of association was positive but non-significant. Either parents who choose Montessori schooling for their children, or something associated with such parents, also engenders these aspects of wellbeing, or very little (Montessori or postulated associate) exposure is needed to engender them.

However, General Wellbeing and Engagement were hypothesized to be influenced by Montessori features (see Figure 1 ), and are significantly and positively associated with duration of Montessori attendance (from 1 to 16 years). The latent variable of General Wellbeing was measured by scales concerning meaning in life and satisfaction with life, self-acceptance, vitality, and environmental mastery. Engagement was measured by variables tapping social contribution, social integration, positive relations, aspiring for personal growth, and a sense of purpose in life.

Although the associations with General Wellbeing and Engagement held across all three analyses, the study design does not allow one to determine if Montessori schooling caused higher wellbeing. An experimental design, in which children were randomly assigned to Montessori and then tested as adults, would be needed.

Lottery-Control Studies

Although we know of no long-term lottery control studies of Montessori education, two short term ones were described in the section “Introduction.” The two natural experiments conducted in the US show that high-fidelity public Montessori causes features associated with higher wellbeing, like stronger mastery orientation, executive function, social knowledge/skills, and academic performance ( Lillard and Else-Quest, 2006 ; Moffitt et al., 2011 ; Reynolds et al., 2011 , 2017 ; Sancassiani et al., 2015 ; Steinmayr et al., 2016 , 2018 ; Haimovitz and Dweck, 2017 ; Lillard et al., 2017 ; Darling-Hammond et al., 2019 ). It is also plausible that having one’s children attend Montessori changes parents, and that the parents’ subsequent behavior led to higher adult wellbeing, but these natural experiment studies lend plausibility to the hypothesized model.

Predictive Features of Montessori for Improved Outcomes

Another support for the plausibility of the hypothesized model is that several of Montessori programs’ features, including the three highlighted here, predict higher wellbeing even when implemented in conventional school settings; these were discussed in the section “Introduction.” For example in classrooms where students are given more agency and opportunities for self-determination, they also have higher sense of their own competence and overall wellbeing ( Ryan and Grolnick, 1986 ), and this is causal: When teachers were trained to increase students’ sense of self-determination, the students’ wellbeing increased ( De Charms, 1976 ). Montessori work is highly engaging ( Rathunde and Csikszentmihalyi, 2005a ), and higher engagement leads to higher wellbeing ( Csikszentmihalyi, 1990 ). Montessori also proffers social stability (3-year age groupings and teacher consistency) and (as indicated in research) stronger relationships ( Rathunde and Csikszentmihalyi, 2005b ; Lillard and Else-Quest, 2006 ). Strong relationships in childhood also predict higher wellbeing in adulthood ( Olsson et al., 2013 ).

Conclusion and Limitations

In sum, although this study only shows an association between Montessori schooling in childhood and higher adult wellbeing, lottery control studies and studies showing that features of Montessori schooling are associated with higher wellbeing in other settings lend weight to the possibility that Montessori might cause higher adult wellbeing. But if this is not the case—if in fact features of Montessori parents or some other third variable associated with Montessori attendance is the cause—then it would be very interesting to determine what the underlying cause for the discovered association is.

This study has several limitations in addition to its being a study of association rather than an experiment. One is that the sample was largely female and White, which is often the case for internet survey samples ( Smith, 2008 ; Peytchev, 2011 ; Boulianne, 2013 ). Other studies that have tested for gender differences in Montessori outcomes typically have not found them ( Lillard and Else-Quest, 2006 ; Culclasure et al., 2018 ), and children of color particularly thrive in Montessori schools ( Ansari and Winsler, 2014 , 2020 ; Brown and Steele, 2015 ; Brown and Lewis, 2017 ; Culclasure et al., 2018 ; Lillard et al., in press ; Snyder et al., 2021 ). Ansari and Winsler (2014 , 2020) found that only Hispanic children thrived, but many other studies show Black children thrive in Montessori as well. Furthermore, race and gender were accounted for in our models. Still, in an ideal sample, gender and race would be representative of the population, and future research should strive for a representative sample.

A second limitation concerns variation in Montessori implementation. Although the core Montessori features we discussed—self-determination, meaningful activities, and social stability—likely characterize all Montessori schools, variations in implementation might accentuate or mitigate them. For example, we have seen Montessori elementary classrooms that require students to fill out checklists of their work activities in ways that likely reduce feelings of self-determination in those classrooms. Studies showing the strongest Montessori outcomes involve high fidelity Montessori implementation ( Lillard, 2019 ). Here, we have no information regarding the fidelity of implementation in the classrooms the adults attended. In future research, it would be useful to gather information on implementation fidelity and examine whether it varies with student wellbeing.

Another limitation is that participants knew the purpose was to consider the impact of alternative schooling on one’s life, and this could have biased people’s responding, although the direction such bias might take is unclear. If one has fond memories of school, whether it was conventional or Montessori, and one knows one is doing a survey about the impact of school on current wellbeing, one might answer more positively; if one has negative memories, one might answer more negatively. Thus, while it is unclear the direction in which knowledge about the survey’s purpose might bias any individual’s responses, in the future one might administer surveys without providing a description of the purpose. This might be difficult to do while obtaining a large sample of alternatively schooled individuals, since alternative schooling is relatively rare (based on other data from our laboratory, we estimate that 5% of American college students attended a non-conventional school at some point). Participants did not answer questions about the types of schools they attended and when until the end of the survey, which might have reduced attention to this aspect of the study until the surveys were complete, but biased responding is a limitation in survey research, particularly when it concerns subjective qualities.

Yet another limitation comes from recruitment itself. In order to get a large enough Montessori sample, Facebook ads were run in communities where Montessori schools are more abundant, such as Washington D.C., Minneapolis/St. Paul, and Milwaukee which have long had Montessori teacher training programs. Although these ads should also have recruited conventionally schooled people in those cities, it is conceivable that the study’s external validity is compromised by the strategy, because there are regional differences in wellbeing ( Lawless and Lucas, 2011 ). For example, at the county level, life satisfaction is highly positively correlated with household income, and negatively correlated with the percentages of persons living in poverty and unemployed. The cities where Facebook ads were run varied in different ways on these metrics. For example, relative to the US average in 2018, there were higher rates of poverty and unemployment and lower median income in Milwaukee, favorable levels of all three metrics in Minneapolis, and varied levels in DC (high poverty coupled with low unemployment and high median income). Such regional variation could in part explain the levels of wellbeing in the adults sampled here, and future research should control for region to measure its contribution to wellbeing.

A further limitation concerns internal validity: We asked people to recall what type of school they attended each year from when they were 2–17, with seven options ranging from Did Not Attend to Homeschool. Some people might not have remembered their school type but guessed a type, which would produce noise in the data, rendering our results less reliable. In general, people’s memories for childhood experiences are thought to be “substantially accurate” ( Brewin et al., 1993 , p. 84), and memories for what type of school one attended, for almost a full year, during each childhood year, is likely to have been rehearsed in the family, lending a degree of confidence to the generally accuracy of the school data. Still, there are likely to be some inaccuracies in the data regarding school history.

In sum, wellbeing is a multidetermined but very important human outcome. If childhood schooling were to influence adult wellbeing, the public health implications would be very important. Pedagogical environments that support children to become adults with high levels of wellbeing are desirable. Although there are important limitations, the research presented here suggests that Montessori schooling might be associated with higher adult wellbeing, and that a causal relation between Montessori schooling in childhood and wellbeing in adulthood is at least plausible.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics Statement

The studies involving human participants were reviewed and approved by the University of Virginia Institutional Review Board for the Social and Behavioral Sciences. The patients/participants provided their written informed consent to participate in this study.

Author Contributions

AL conceived of the study, contributed theory, raised funds, oversaw the entire project, did some data analysis, and wrote the manuscript. MM conducted the data analyses. DV and EF selected stimuli in collaboration with AL, recruited participants, put the surveys on Qualtrics, ran the study, worked with MM to prepare the data file, and contributed to the manuscript. DV ran some analyses. All the authors contributed to the article and approved the submitted version.

This research was supported by grants from the Wildflower and Wend Foundations to AL.

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary Material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyg.2021.721943/full#supplementary-material

180 Studio and Saunders Eckenhoff (2020). Seed + Spark: Using nature as a model to reimagine how we learn and live. Chicago: 180 Press.

Google Scholar

Ackerman, C. E., Warren, M. A., and Donaldson, S. I. (2018). Scaling the heights of positive psychology: A systematic review of measurement scales. Int. J. Wellbeing 8, 1–21. doi: 10.5502/ijw.v8i2.734

PubMed Abstract | CrossRef Full Text | Google Scholar

Ansari, A., and Winsler, A. (2014). Montessori public school pre-k programs and the school readiness of low-income black and latino children. J. Educ. Psychol. 106, 1066–1079. doi: 10.1037/a0036799

Ansari, A., and Winsler, A. (2020). The long-term benefits of Montessori pre-K for Latinx children from low-income families. Appl. Dev. Sci. 2020, 1–15. doi: 10.1080/10888691.2020.1781632

CrossRef Full Text | Google Scholar

Baker, S. R. (2004). Intrinsic, extrinsic, and amotivational orientations: Their role in university adjustment, stress, well-being, and subsequent academic performance. Curr. Psychol. 23, 189–202. doi: 10.1007/s12144-004-1019-9

Bostic, T. J., Rubio, D. M., and Hood, M. (2000). A validation of the subjective vitality scale using structural equation modeling. Soc. Indicat. Res. 52, 313–324.

Boulianne, S. (2013). Examining the gender effects of different incentive amounts in a web survey. Field Methods 25, 91–104. doi: 10.1177/1525822X12453113

Box, G. E., and Cox, D. R. (1964). An analysis of transformations. J. R. Stat. Soc. Ser. B 26, 211–243.

Brewin, C. R., Andrews, B., and Gotlib, I. H. (1993). Psychopathology and early experience: a reappraisal of retrospective reports. Psychol. Bull. 113, 82–98. doi: 10.1037/0033-2909.113.1.82

Brody, N., and Ehrlichman, H. (1998). Personality psychology: The science of individuality. New York City: Pearson College Div.

Brown, K., and Ryan, R. (2003). The benefits of being present: Mindfulness and its role in psychological well-being. J. Personal. Soc. Psychol. 84, 822–848. doi: 10.1037/0022-3514.84.4.822

Brown, K. E., and Lewis, C. W. (2017). A comparison of reading and math achievement for African American third grade students in Montessori and other magnet schools. J. Negro Educ. 86, 439–448. doi: 10.7709/jnegroeducation.86.4.0439

Brown, K. E., and Steele, A. S. (2015). Racial discipline disproportionality in Montessori and traditional public schools: A comparative study using the relative rate index. J. Montessori Res. 1, 14–27. doi: 10.17161/jomr.v1i1.4941

Bruner, J. (1990). Acts of meaning. Cambridge, MA: Harvard University Press.

Burke, D. L. (1996). Multi-year teacher/student relationships are a long-overdue arrangement. Phi Delta Kappan 77:360.

Butler, R., and Ruzany, N. (1993). Age and socialization effects on the development of social comparison motives and normative ability assessment in kibbutz and urban children. Child Dev. 64, 532–543. doi: 10.2307/1131267

Cacioppo, J. T., Petty, R. E., and Feng Kao, C. (1984). The efficient assessment of need for cognition. J. Personal. Assess. 48, 306–307. doi: 10.1207/s15327752jpa4803_13

Chetty, R., Friedman, J. N., Hendren, N., Jones, M. R., and Porter, S. R. (2018). The Opportunity Atlas: Mapping the Childhood Roots of Social Mobility. Cambridge, MA: National Bureau of Economic Research.

Cistone, P., and Shneyderman, A. (2004). Looping: An empirical evaluation. Int. J. Educ. Policy Res. Pract. Reconceptual. Childhood Stud. 5, 47–61.

Cohen, J. D. (2006). Social, emotional, ethical, and academic education: Creating a climate for learning, participation in democracy, and well-being. Harvard Educ. Rev. 76, 201–237. doi: 10.17763/haer.76.2.j44854x1524644vn

Courtier, P. (2020). L’impact de la pédagogie Montessori sur le développement cognitif, social et académique des enfants en maternelle. France: Université de Lyon.

Csikszentmihalyi, M. (1990). Flow: The psychology of optimal experience. New York City, NJ: Harper Perennial.

Csikszentmihalyi, M. (1997). Finding flow: The psychology of engagement with everyday life. New York: Basic Books, Inc.

Culclasure, B., Fleming, D. J., and Riga, G. (2018). An Evaluation of Montessori Education in South Carolina’s Public Schools. Greenville, SC: The Riley Institute at Furman University.

Culclasure, B. T., Daoust, C. J., Cote, S. M., and Zoll, S. (2019). Designing a logic model to inform Montessori research. J. Montessori Res. 5, 35–74. doi: 10.17161/jomr.v2i2.5067

Curtin, R., Presser, S., and Singer, E. (2000). The effects of response rate changes on the index of consumer sentiment. Public Opin. Q. 64, 413–428. doi: 10.1086/318638

Danner, F. W., and Lonky, E. (1981). A cognitive-developmental approach to the effects of rewards on intrinsic motivation. Child Dev. 52, 1043–1052. doi: 10.1186/s12868-016-0283-6

Darling-Hammond, L., Flook, L., Cook-Harvey, C., Barron, B., and Osher, D. (2019). Implications for educational practice of the science of learning and development. Appl. Dev. Sci. 24, 1–44. doi: 10.1080/10888691.2018.1537791

De Charms, R. (1976). Enhancing Motivation: A Change in the Classroom. New York, NY: Irvington Press.

Debs, M. C. (2019). Diverse Parents, Desirable Schools: Public Montessori in the Era of School Choice. Cambridge, MA: Harvard Education Press.

Deci, E. L., and Ryan, R. M. (2011). “Self-determination theory,” in Handbook of theories of social psychology , Vol. 1, eds P. A. M. V. Lange, A. W. Kruglanski, and E. T. Higgins (Thousand Oaks: Sage), 416–433.

Denervaud, S., Knebel, J.-F., Hagmann, P., and Gentaz, E. (2019). Beyond executive functions, creativity skills benefit academic outcomes: Insights from Montessori education. PloS One 14:e0225319. doi: 10.1371/journal.pone.0225319

Denervaud, S., Mumenthaler, C., Gentaz, E., and Sander, D. (2020a). Emotion recognition development: Preliminary evidence for an effect of school pedagogical practices. Learn. Instruct. 69:101353.

Denervaud, S., Rornari, E., Yang, X. F., Immordino-Yang, M. H., and Sander, D. (2020b). An fMRI study of error monitoring in Montessori and traditionally-schooled children. NPJ Sci. Learn. 5:11. doi: 10.1038/s41539-020-0069-6

Diener, E., Emmons, R. A., Larsen, R. J., and Griffin, S. (1985). The satisfaction with life scale. J. Personal. Assess. 49, 71–75. doi: 10.1207/s15327752jpa4901_13

Diener, E., Heintzelman, S. J., Kushlev, K., Tay, L., Wirtz, D., Lutes, L. D., et al. (2016). Findings all psychologists should know from the new science on subjective well being. Can. Psychol. 58, 87–104. doi: 10.1037/cap0000063

Diener, E., Oishi, S., and Lucas, R. E. (2015). National accounts of subjective well-being. Am. Psychol. 70:234. doi: 10.1037/a0038899

Diener, E., Tay, L., and Oishi, S. (2013). Rising income and the subjective well-being of nations. J. Personal. Soc. Psychol. 104:267.

Dillon, W. R., Kumar, A., and Mulani, N. (1987). Offending estimates in covariance structure analysis: Comments on the causes of and solutions to Heywood cases. Psychol. Bull. 101, 126–135.

Dreyer, A. S., and Rigler, D. (1969). Cognitive performance in Montessori and nursery school children. J. Educ. Res. 62, 411–416. doi: 10.1080/00220671.1969.10883885

Duncan, G. J., and Murnane, R. J. (2014). Restoring Opportunity. Cambridge, MA: Harvard Education Press.

Dweck, C. S. (1999). Self-theories: Their role in motivation, personality, and development. Milton Park: Psychology Press/Taylor & Francis.

Felner, R. D., Seitsinger, A. M., Brand, S., Burns, A., and Bolton, N. (2007). Creating small learning communities: Lessons from the project on high-performing learning communities about “what works” in creating productive, developmentally enhancing, learning contexts. Educ. Psychol. 42, 209–221. doi: 10.1080/00461520701621061

Fleege, U. H., Black, M. S., and Racksauskas, J. A. (1967). Montessori pre-school education. Final report. US Department of Health, Education, and Welfare, Office of Education, Bureau of Research. Eric Database 017320.

Flynn, T. M. (1991). Development of social, personal and cognitive skills of preschool children in Montessori and traditional preschool programs. Early Child Dev. Care 72, 117–124.

Gosling, S. D., and Mason, W. (2015). Internet research in psychology. Annu. Rev. Psychol. 66, 877–902. doi: 10.1146/annurev-psych-010814-015321

Haimovitz, K., and Dweck, C. S. (2017). The origins of children’s growth and fixed mindsets: New research and a new proposal. Child Dev. 88, 1849–1859. doi: 10.1111/cdev.12955

Harter, S. (1978). Pleasure derived from challenge and the effects of receiving grades on children’s difficulty level choices. Child Dev. 49, 788–799. doi: 10.2307/1128249

Heckman, J. J. (2006). Skill formation and the economics of investing in disadvantaged children. Science 312, 1900–1902. doi: 10.1126/science.1128898

Hill, A. J., and Jones, D. B. (2018). A teacher who knows me: The academic benefits of repeat student-teacher matches. Eco. Educ. Rev. 64, 1–12. doi: 10.1016/j.econedurev.2018.03.004

Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis. Psychometrika 30, 179–185. doi: 10.1007/bf02289447

Hu, L. T., and Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct. Equat. Model. 6, 1–55. doi: 10.1080/10705519909540118

ITU Publications (2019). Measuring Digital Development: Facts and Figures. Geneva: International Telecommunication Union.

Kenny, D. A. (2012). Identification. Available Online at: http://davidakenny.net/cm/identify_formal.htm

Kenny, D. A., and McCoach, D. B. (2003). Effect of the number of variables on measures of fit in structural equation modeling. Struct. Equat. Model. 10, 333–351.

Keyes, C. L. M. (1998). Social well-being. Soc. Psychol. Q. 61, 121–140. doi: 10.2307/2787065

Keyes, C. L. M., Eisenberg, D., Perry, G. S., Dube, S. R., Kroenke, K., and Dhingra, S. S. (2012). The relationship of level of positive mental health with current mental disorders in predicting suicidal behavior and academic impairment in college students. J. Am. Coll. Health 60, 126–133. doi: 10.1080/07448481.2011.608393

Kidd, C., Piantadosi, S. T., and Aslin, R. N. (2012). The Goldilocks effect: Human infants allocate attention to visual sequences that are neither too simple nor too complex. PLoS One 7:e36399. doi: 10.1371/journal.pone.0036399

Kidd, C., Piantadosi, S. T., and Aslin, R. N. (2014). The Goldilocks effect in infant auditory attention. Child Dev. 85, 1795–1804. doi: 10.1111/cdev.12263

Langer, E. J., and Rodin, J. (1976). The effects of choice and enhanced personal responsibility for the aged: A field experiment in an institutional setting. J. Personal. Soc. Psychol. 34, 191–198. doi: 10.1037//0022-3514.34.2.191

Lawless, N. M., and Lucas, R. E. (2011). Predictors of regional well-being: A county level analysis. Soc. Indicat. Res. 101, 341–357.

Lewis, A. D., Huebner, E. S., Malone, P. S., and Valois, R. F. (2011). Life satisfaction and student engagement in adolescents. J. Youth Adoles. 40, 249–262. doi: 10.1007/s10964-010-9517-6

Lillard, A. S. (2017). Montessori: The science behind the genius , 3rd Edn. Oxford: Oxford University Press.

Lillard, A. S. (2019). Shunned and admired: montessori, self-determination, and a case for radical school reform. Educ. Psychol. Rev. 31, 939–965. doi: 10.1007/s10648-019-09483-3

Lillard, A. S., and Else-Quest, N. (2006). Evaluating montessori education. Science 313, 1893–1894. doi: 10.1126/science.1132362

Lillard, A. S., Heise, M. J., Richey, E. M., Tong, X., Hart, A., and Bray, P. M. (2017). Montessori preschool elevates and equalizes child outcomes: A longitudinal study. Front. Psychol. 8:1783. doi: 10.3389/fpsyg.2017.01783

Lillard, A. S., and McHugh, V. (2019a). Authentic montessori: The Dottoressa’s view at the end of her life part I: The environment. J. Montessori Res. 5, 1–18. doi: 10.17161/jomr.v5i1.7716

Lillard, A. S., and McHugh, V. (2019b). Authentic montessori: The Dottoressa’s view at the end of her life part II: The teacher and the child. J. Montessori Res. 5, 19–34. doi: 10.17161/jomr.v5i1.9753

Lillard, A. S., Taggart, J., Yonas, D., and Seale, M. N. (in press). An alternative to “no excuses”: Considering Montessori as culturally responsive pedagogy. J. Negro Educ.

Lindberg, M. H., Chen, G., Olsen, J. A., and Abelsen, B. (2021). Explaining subjective social status in two countries: The relative importance of education, occupation, income and childhood circumstances. SSM-Popul. Health 15:100864. doi: 10.1016/j.ssmph.2021.100864

Listl, S., Broadbent, J. M., Thomson, W. M., Stock, C., Shen, J., Steele, J., et al. (2018). Childhood socioeconomic conditions and teeth in older adulthood: evidence from SHARE wave 5. Commun. Dentist. Oral Epidemiol. 46, 78–87. doi: 10.1111/cdoe.12332

Little, T. S., and Dacus, N. B. (1999). Looping: Moving up with the class. Educ. Leadership 57, 42–45.

Lloyd, L. (1999). Multi-age classes and high ability students. Rev. Educ. Res. 69, 187–212. doi: 10.3102/00346543069002187

Moffitt, T. E., Arseneault, L., Belsky, D., Dickson, N., Hancox, R. J., Harrington, H. L., et al. (2011). A gradient of childhood self-control predicts health, wealth, and public safety. Proc. Natl. Acad. Sci. 108, 2693–2698. doi: 10.1073/pnas.1010076108

Montessori, M. (1967/1995). The absorbent mind (C. A. Claremont, Trans.). New York: Henry Holt.

Musch, J., and Reips, U.-D. (2000). “A brief history of Web experimenting,” in Psychological experiments on the Internet , ed. M. Birnbaum (Cambridge: Academic Press), 61–87. doi: 10.1016/b978-012099980-4/50004-6

National Academies of Sciences Engineering and Medicine (2018). How people learn II: Learners, contexts, and cultures. Washington, D.C: The National Academies Press.

National Center for Montessori in the Public Sector (2014). Growth of Public Montessori in the United States: 1975-2014. Available Online at: https://www.public-montessori.org/white-papers/growth-of-public-montessori-in-the-united-states-1975-2014 [Retrieved 14 April 2019]

Neale, M. C., Hunter, M. D., Pritikin, J. N., Zahery, M., Brick, T. R., Kirkpatrick, R. M., et al. (2016). OpenMx 2.0: Extended structural equation and statistical modeling. Psychometrika 81, 535–549. doi: 10.1007/s11336-014-9435-8

Ng, Z. J., Huebner, S. E., and Hills, K. J. (2015). Life satisfaction and academic performance in early adolescents: Evidence for reciprocal association. J. School Psychol. 53, 479–491. doi: 10.1016/j.jsp.2015.09.004

Nitecki, E. (2017). Looping and attachment in early Cchildhood education: How applications of epigenetics demand a change. J. Scholarship Teach. Learn. 17, 85–100.

Oishi, S., and Schimmack, U. (2010). Residential mobility, well-being, and mortality. J. Personal. Soc. Psychol. 98, 980–994. doi: 10.1037/a0019389

Olsson, C. A., McGee, R., Nada-Raja, S., and Williams, S. M. (2013). A 32-year longitudinal study of child and adolescent pathways to well-being in adulthood. J. Happiness Stud. 14, 1069–1083. doi: 10.1007/s10902-012-9369-8

Patall, E. A., Cooper, H., and Robinson, J. C. (2008). The effects of choice on intrinsic motivation and related outcomes: a meta-analysis of research findings. Psychol. Bull. 134, 270–300. doi: 10.1037/0033-2909.134.2.270

Patall, E. A., Cooper, H., and Wynn, S. R. (2010). The effectiveness and relative importance of choice in the classroom. J. Educ. Psychol. 102, 896–915. doi: 10.1037/a0019545

Peytchev, A. (2011). Breakoff and unit nonresponse across web surveys. J. Offic. Stat. 27, 33–47.

Plucker, J. A. (2003). Exploratory and confirmatory factor analysis in gifted education: Examples with self-concept data. J. Educ. Gifted 27, 20–35. doi: 10.1177/016235320302700103

Rathunde, K. R., and Csikszentmihalyi, M. (2005a). Middle school students’ motivation and quality of experience: A comparison of Montessori and traditional school environments. Am. J. Educ. 111, 341–371. doi: 10.1086/428885

Rathunde, K. R., and Csikszentmihalyi, M. (2005b). The social context of middle school: Teachers, friends, and activities in Montessori and traditional school environments. Element. School J. 106, 59–79. doi: 10.1007/978-94-017-9094-9_9

Reardon, S. F. (2011). “The widening academic achievement gap between the rich and the poor: New evidence and possible explanations,” in Whither opportunity; Rising Inequality, Schools, and Children’s Life Chances , eds G. Duncan and R. Murnane (New York: Russell Sage Foundation), 91–116.

Reise, S. P., Waller, N. G., and Comrey, A. L. (2000). Factor analysis and scale revision. Psychol. Assess. 12, 287–297. doi: 10.1037//1040-3590.12.3.287

Reynolds, A. J., Ou, S. R., Mondi, C. F., and Hayakawa, M. (2017). Processes of early childhood interventions to adult well-being. Child Dev. 88, 378–387. doi: 10.1111/cdev.12733

Reynolds, A. J., Temple, J. A., Ou, S.-R., Arteaga, I. A., and White, B. A. B. (2011). School-based early childhood education and age-28 well-being: Effects by timing, dosage, and subgroups. Science 333, 360–364. doi: 10.1126/science.1203618

Rhoads, S. A., Gunter, D., Ryan, R. M., and Marsh, A. A. (2021). Global variation in subjective well-being predicts seven forms of altruism. Psychol. Sci. 32, 1247–1261. doi: 10.1177/0956797621994767

Rovee-Collier, C., and Hayne, H. (2000). “Memory in infancy and early childhood,” in The Oxford handbook of memory , eds E. Tulving and F. I. M. Craik (Oxford: Oxford University Press), 267–282.

Røysamb, E., and Nes, R. B. (2019). The role of genetics in subjective well-being. Nat. Hum. Behav. 3:3. doi: 10.1038/s41562-018-0494-1

Russell, D. W. (2002). In search of underlying dimensions: The use (and abuse) of factor analysis in Personality and Social Psychology Bulletin. Personal. Soc. Psychol. Bull. 28, 1629–1646. doi: 10.1177/014616702237645

Ryan, R. M., and Deci, E. L. (2000). “When rewards compete with nature: The undermining of intrinsic motivation and self-regulation,” in Intrinsic and extrinsic motivation: The search for optimal motivation and performance , ed. C. H. J. M. Sansone (Cambridge: Academic Press, Inc), 13–54. doi: 10.1016/b978-012619070-0/50024-6

Ryan, R. M., and Frederick, C. (1997). On energy, personality, and health: Subjective vitality as a dynamic reflection of well-being. J. Personal. 65, 529–565. doi: 10.1111/j.1467-6494.1997.tb00326.x

Ryan, R. M., and Grolnick, W. S. (1986). Origins and pawns in the classroom: Self-report and projective assessments of individual differences in children’s perceptions. J. Personal. Soc. Psychol. 50, 550–558. doi: 10.1037/0022-3514.50.3.550

Ryff, C. D. (2014). Psychological well-being revisited: Advances in the science and practice of eudaimonia. Psychother. Psychosom. 83, 10–28. doi: 10.1159/000353263

Ryff, C. D., and Keyes, C. L. M. (1995). The structure of psychological well-being revisited. J. Personal. Soc. Psychol. 69, 719–727. doi: 10.1037/0022-3514.69.4.719

Sancassiani, F., Pintus, E., Holte, A., Paulus, P., Moro, M. F., Cossu, G., et al. (2015). Enhancing the emotional and social skills of the youth to promote their wellbeing and positive development: a systematic review of universal school-based randomized controlled trials. Clin. Pract. Epidemiol. Mental Health 11, 21–40. doi: 10.2174/1745017901511010021

Seligman, M. E., Ernst, R. M., Gillham, J., Reivich, K., and Linkins, M. (2009). Positive education: Positive psychology and classroom interventions. Oxford Rev. Educ. 35, 293–311. doi: 10.1080/03054980902934563

Shankland, R., Genolini, C., França, L. R., Guelfi, J.-D., and Ionescu, S. (2010). Student adjustment to higher education: the role of alternative educational pathways in coping with the demands of student life. Higher Educ. 59, 353–366. doi: 10.1007/s10734-009-9252-7

Shi, D., Lee, T., and Maydeu-Olivares, A. (2019). Understanding the model size effect on SEM fit indices. Educ. Psychol. Measur. 79, 310–334. doi: 10.1177/0013164418783530

Simon, H. A. (2001). “Seek and ye shall find”: How curiosity engenders discovery,” in Designing for science: implications from everyday, classroom, and professional settings , eds K. Crowley, C. D. Schunn, and T. Okada (Mahwah: Lawrence Erlbaum).

Singer, E., Van Hoewyk, J., and Maher, M. P. (2000). Experiments with incentives in telephone surveys. Public Opin. Q. 64, 171–188.

Smith, G. (2008). Does Gender Influence Online Survey Participation: A Record-linkage Analysis of University Faculty Online Survey Response Behavior. ERIC Document Reproduction Service No. ED 501717. ERIC. Available online at: https://eric.ed.gov/?id=ED501717 (accessed June, 2008).

Snyder, A., Tong, X., and Lillard, A. S. (2021). Standardized test performance in public Montessori schools. J. School Choice 2021:1958058. doi: 10.1080/15582159.2021.1958058

Steger, M. F., Kashdan, T. B., Sullivan, B. A., and Lorentz, D. (2008). Understanding the search for meaning in life: Personality, cognitive style, and the dynamic between seeking and experiencing meaning. J. Personal. 76, 199–228. doi: 10.1111/j.1467-6494.2007.00484.x

Steinmayr, R., Crede, J., McElvany, N., and Wirthwein, L. (2016). Subjective well-being, test anxiety, academic achievement: Testing for reciprocal effects. Front. Psychol. 6:1994. doi: 10.3389/fpsyg.2015.01994

Steinmayr, R., Heyder, A., Naumburg, C., Michels, J., and Wirthwein, L. (2018). School-related and individual predictors of subjective well-being and academic achievement. Front. Psychol. 9:2631. doi: 10.3389/fpsyg.2018.02631

Straughen, J. K., Caldwell, C. H., Osypuk, T. L., Helmkamp, L., and Misra, D. P. (2013). Direct and proxy recall of childhood socio-economic position and health. Paediat. Perinatal Epidemiol. 27, 294–302. doi: 10.1111/ppe.12045

Taggart, J., Becker, I., Rauen, J., Al Kallas, H., and Lillard, A. S. (2020). What shall we do: Pretend or real? Preschoolers’ choices and parents’ perceptions. J. Cogn. Dev. 21, 261–281. doi: 10.1080/15248372.2019.1709469

Taggart, J., Heise, M. J., and Lillard, A. S. (2018). The real thing: Preschoolers prefer actual activities to pretend one. Dev. Sci. 21:e12582. doi: 10.1111/desc.12582

Tang, K. L., Rashid, R., Godley, J., and Ghali, W. A. (2016). Association between subjective social status and cardiovascular disease and cardiovascular risk factors: a systematic review and meta-analysis. BMJ Open 6:e010137. doi: 10.1136/bmjopen-2015-010137

Thurstone, L. L. (1947). Multiple-factor analysis: A development and expansion of the vectors of mind. Chicago: University of Chicago Press.

Tourigny, R., Plante, I., and Raby, C. (2020). Do students in a looping classroom get higher grades and report a better teacher-student relationship than those in a traditional setting? Educ. Stud. 46, 744–759.

Wen, X., Leow, C., Hahs-Vaughn, D. L., Korfmacher, J., and Marcus, S. M. (2012). Are two years better than one year? A propensity score analysis of the impact of Head Start program duration on children’s school performance in kindergarten. Early Childhood Res. Q. 27, 684–694. doi: 10.1016/j.ecresq.2011.07.006

White, M., and Kern, M. (2018). Positive education: Learning and teaching for wellbeing and academic mastery. Int. J. Wellbeing 8, 1–17. doi: 10.5502/ijw.v8i1.588

Wolf, E. J., Harrington, K. M., Clark, S. L., and Miller, M. W. (2013). Sample size requirements for structural equation models: An evaluation of power, bias, and solution propriety. Educ. Psychol. Measur. 73, 913–934. doi: 10.1177/0013164413495237

Yang, Q., Tian, L., Huebner, E. S., and Zhu, X. (2019). Relations among academic achievement, self-esteem, and subjective well-being in school among elementary school students: A longitudinal mediation model. School Psychol. 34, 328–341. doi: 10.1037/spq0000292

Keywords : wellbeing, human development, education, Montessori, positive psychology

Citation: Lillard AS, Meyer MJ, Vasc D and Fukuda E (2021) An Association Between Montessori Education in Childhood and Adult Wellbeing. Front. Psychol. 12:721943. doi: 10.3389/fpsyg.2021.721943

Received: 07 June 2021; Accepted: 26 October 2021; Published: 25 November 2021.

Reviewed by:

Copyright © 2021 Lillard, Meyer, Vasc and Fukuda. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Angeline S. Lillard, [email protected]

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

Montessori Education at a Distance, Part 1

A survey of montessori educators’ response to a global pandemic.

• Angela K. Murray University of Kansas http://orcid.org/0000-0001-6310-8842
• Katie E. Brown National Center for Montessori in the Public Sector http://orcid.org/0000-0003-4633-6426
• Patricia Barton University of Buffalo

The transition to distance learning in the spring of 2020 caused by COVID-19 was particularly challenging for Montessori educators and students because key elements of the Method were not directly transferable to this new and hastily designed format. Hands-on learning with Montessori materials and learning in a community, as well as careful teacher observation, could not be easily replicated when children were learning from home. To understand how educators applied Montessori principles to serve children and families in these highly unusual circumstances, we surveyed Early Childhood and Elementary Montessori teachers about how they translated core elements of Montessori education to a distance-learning environment. The overall results suggest that Montessori distance-learning arrangements balanced live videoconference experiences for children with offline hands-on activities, while also relying on parents’ and caregivers’ involvement. Teachers reported that they largely designed learning experiences themselves, without significant support or guidance from school leaders. Still, teachers reported that they were able to uphold Montessori principles to only a moderate degree under the circumstances. While teachers understandably hunger for support, professional connections, and a return to the classroom experiences that drew them to the field of Montessori education, this study highlights factors that may affect the transition back to school for teachers, parents and caregivers, and students when face-to-face instruction resumes for all children.

Author Biographies

Angela Murray† is an assistant research professor at the University of Kansas and is the director of the KU Center for Montessori Research. She can be reached at [email protected].

Katie Brown is director of professional learning at the National Center for Montessori in the Public Sector.

Patricia Barton is a graduate student at the University of Buffalo, head of school at Desert Shadows Montessori, and Early Childhood coordinator of the Arizona Montessori Teacher Education Program.

Culclasure, B., Daoust, C., Cote, S., & Zoll, S. (2019). Designing a logic model to inform Montessori research. Journal of Montessori Research, 5(1), 35–49. https://doi.org/10.17161/jomr.v5i1.9788

Debs, M., & Brown, K. E. (2017). Students of color and public Montessori schools: A review of the literature. Journal of Montessori Research, 3(1), 1–15. https://doi.org/10.17161/jomr.v3i1.5859

MacDonald, G. (2016). Technology in the Montessori classroom: Benefits, hazards and preparation for life. NAMTA Journal, 41(2), 99–107. https://files.eric.ed.gov/fulltext/EJ1112230.pdf

Montessori, M. (2012). The 1946 London lectures. Montessori-Pierson Publishing.

National Center for Montessori in the Public Sector. (n.d.). The Montessori census. https://www.montessoricensus.org/

Saldaña, J. (2009). The coding manual for qualitative researcher. SAGE.

Walston, J., Redford, J., & Bhatt, M. P. (2017). Workshop on Survey Methods in Education Research: Facilitator’s guide and resources (REl 2017–214; ED573681). U.S. Department of Education, Institute of Education Sciences, National Center for Education Evaluation and Regional Assistance, Regional Educational Laboratory Midwest. https://files.eric.ed.gov/fulltext/ED573681.pdf

W. K. Kellogg Foundation. (2004). W. K. Kellogg Foundation logic model development guide. https://www.wkkf.org/resource-directory/resources/2004/01/logic-model-development-guide

Copyright (c) 2021 Angela K. Murray, Katie E. Brown, Patricia Barton

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License .

Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution Non-Commercial License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal. Authors can view article download statistics for published articles within their accounts.

Journal of Montessori Research Author Agreement The following is an agreement between the Author (the “Corresponding Author”) acting on behalf of all authors of the work (“Authors”) and the Journal of Montessori Research (the “Journal”) regarding your article (the “Work”) that is being submitted for consideration.   Whereas the parties desire to promote effective scholarly communication that promotes local control of intellectual assets, the parties for valuable consideration agree as follows. A. CORRESPONDING AUTHOR’S GRANT OF RIGHTS After being accepted for publication, the Corresponding Author grants to the Journal, during the full term of copyright and any extensions or renewals of that term, the following: 1.    An irrevocable non-exclusive right to reproduce, republish, transmit, sell, distribute, and otherwise use the Work in electronic and print editions of the Journal and in derivative works throughout the world, in all languages, and in all media now known or later developed. 2.    An irrevocable non-exclusive right to create and store electronic archival copies of theWork, including the right to deposit the Work in open access digital repositories. 3.    An irrevocable non-exclusive right to license others to reproduce, republish, transmit,and distribute the Work under the condition that the Authors are attributed. (Currently this is carried out by publishing the content under a Creative Commons Attribution Non-Commercial 4.0 license (CC BY-NC.) 4.    Copyright in the Work remains with the Authors. B. CORRESPONDING AUTHOR’S DUTIES 1.    When distributing or re-publishing the Work, the Corresponding Author agrees to credit the Journal as the place of first publication. 2.    The Corresponding Author agrees to inform the Journal of any changes in contact information. C. CORRESPONDING AUTHOR’S WARRANTY The Corresponding Author represents and warrants that the Work is the Authors’ original work and that it does not violate or infringe the law or the rights of any third party and, specifically, that the Work contains no matter that is defamatory or that infringes literary or proprietary rights, intellectual property rights, or any rights of privacy. The Corresponding Author also warrants that he or she has the full power to make this agreement, and if the Work was prepared jointly, the Corresponding Author agrees to inform the Authors of the terms of this Agreement and to obtain their written permission to sign on their behalf. The Corresponding Author agrees to hold the Journal harmless from any breach of the aforestated representations. D.  JOURNAL’S DUTIES In consideration of the Author’s grant of rights, the Journal agrees to publish the Work, attributing the Work to the Authors. E. ENTIRE AGREEMENT This agreement reflects the entire understanding of the parties. This agreement may be amended only in writing by an addendum signed by the parties. Amendments are incorporated by reference to this agreement. ACCEPTED AND AGREED BY THE CORRESPONDING AUTHOR ON BEHALF OF ALL AUTHORS CONTRIBUTING TO THIS WORK

How to Cite

• Endnote/Zotero/Mendeley (RIS)

Make a Submission

• Norwegian Bokmål

Developed By

This electronic publication is supported by the

University of Kansas Libraries .

ISSN:  2378-3923

• 2024 Conference
• Montessori Curriculum to Standards Alignment
• Subscribe to Emails

National Center for Montessori in the Public Sector

More montessori to more children.

• Montessori Assessment Playbook
• NCMPS Tools in Transparent Classroom
• Essential Elements for Public Montessori
• Developmental Environment Rating Scale (DERS)
• Startup and Expansion
• Program Support
• Essential Elements Review
• Support for Leaders
• Support for Coaches
• Child Study
• Professional Learning
• About Montessori

Featured Research

• What the Research Shows

Webinar Recordings

White Papers

• Mission and Vision
• Our Funders & Supporters
• Annual Report
• Notes from the Executive Director
• Montessori News
• Montessori Schools
• Montessori Jobs

Research and Resources

Research and ResouRces

Research shows the impact of Montessori on outcomes such as school readiness, academic achievement, executive function skills, and more.

Rigorous meta-analysis says montessori works, campbell systematic reviews , 2023.

In a systematic review of Montessori research considering 2000+ articles, researchers found  evidence that Montessori education outperformed traditional education on a wide variety of academic and nonacademic outcomes.

Public Montessori Raises Achievement, Closes Gaps

Frontiers in psychology, 2017.

This three-year study of two public Montessori schools in Hartford  reduced achievement gaps across income levels  and  rated Montessori children higher in academic achievement, social cognition, mastery orientation, and school enjoyment . This work is the basis for an ongoing federally funded national study.

Public Montessori Schools Outperform Districts on Standardized Tests

Journal of school choice, 2021.

In the ten states or regions with the most public Montessori schools, Montessori schools had higher levels of ELA and math proficiency at eighth grade , even when controlling for race and income—in fact, Black, Hispanic, and economically disadvantaged students showed the greatest difference, and so-called “achievement gaps” were smaller in Montessori schools.

Montessori Shows Strengths Across Race and Income

Riley institute at furman, 2018.

This study of 45 public Montessori schools in South Carolina found that  Montessori students met more standards and showed more growth (especially in low-income and Black sub-groups) , showed more creativity, and had fewer absences, discipline events and suspensions.

Montessori Education Associated with Child and Adult Wellbeing

Frontiers in psychology, 2021.

This comprehensive survey of nearly 2,000 adults with and without Montessori experience in childhood found that attending Montessori for at least two years was associated with significantly higher adult wellbeing on four factors: general wellbeing , engagement , social trust , and self-confidence .

Montessori—The Science Behind the Genius by Dr. Angeline Lillard

Oxford university press, 2017.

Lillard’s 2017 book is a comprehensive summary of the developmental psychology underpinning the Montessori method.

Read more Montessori research news at MontessoriPublic »

Research summaries.

• Does it Work? What the Research Says A two-page summary of Montessori outcomes, with citations.
• Annotated Bibliography A selected bibliography of recent research
• Annotated Bibliography by claim Detailed citations for Montessori outcomes

Research Resources

• American Educational Research Association Montessori Special Interest Group
• American Montessori Society Research and Resources
• Center for Montessori Studies at the University of Hartford
• Journal of Montessori Research
• University of Kansas Center for Montessori Research

Research Opportunities

Opportunities to take part in ongoing research projects:

• A-GAME Alternative measures of success
• Montessori Research Pool

What the research shows

Dozens of studies show how Montessori supports a wide range of developmental outcomes.

Executive functions.

Montessori supports executive function skills, including self-regulation, working memory, planning, and inhibitory control. Executive functions predict positive life-long outcomes such as academic achievement, income potential, and marital satisfaction.

Academic Achievement

Montessori supports self-regulation and intrinsic motivation, which correlate with academic achievement. In language arts, children in Montessori classrooms have shown strengths in decoding, letter-word identification, reading assessments, sentence structure, and writing creativity. In mathematics, children have shown higher scores in applied problem solving, understanding of math concepts, and standardized test scores.

School Readiness

While the Montessori approach recognizes that learning begins at birth, well before conventional schooling starts, children in Montessori preschool and kindergarten classrooms have shown strengths in traditional “school readiness” measures.

Social Development

Montessori supports social problem solving ability, a sense of community and social justice, positive perceptions of classmates, and construcive social problem-solving.

An Introduction to the Developmental Environmental Ratings Scale (DERS)

Introduction to the Montessori Curriculum to Standards Alignment

Federal Supports for Public Montessori Schools

Essential Elements Webinar

Making the Case for Public Montessori

Does it work what the research says →.

Scientific research confirms that Montessori children have an advantage not only academically, but also in social and emotional development.

What is Montessori? →

A one-page overview for stakeholders new to Montessori.

Ages Three to Six →

A one-page overview of Montessori education for early childhood.

Montessori Elementary →

A one-page overview of Montessori education for 1st through 6th grade.

Essential Elements for Successful Montessori Schools in the Public School Sector →

Essential guidelines for fully implemented, sustainable public Montessori programs.

Starting a Public Montessori School: A Planning Timeline →

A one-page timeline overview. Download PDF →

Montessori School Design →

Prepared Environments for Learning and Living.

NCMPS produces a series of White Papers to provide support for Montessori implementation and program development for use with administrators, policy makers, legislators, and wherever else they can be helpful.

These papers are free to download and use with attribution to the National Center for Montessori in the Public Sector.

Three-Year Age-Spans in Montessori Classrooms: The Benefits of a Full Upper Elementary cycle

The assistant in a montessori classroom, montessori bilingual education, english language learners and special education students in montessori schools: the case for push-in services, childhood trauma, teacher compensation in montessori charter schools, what is montessori-inspired practice, braided funding for fully implemented montessori early learning, 0–6 early learning initiative, head start, early head start, and montessori, montessori as an intervention for children with dyslexia, growth of public montessori in the united states: 1975-2014, reading and math achievement for african american lower elementary students in public montessori programs, contact ncmps.

• Montessori 101
• DERS App – Developmental Environmental Rating Scale
• Montessori Teaching Jobs
• Montessori Training Centers
• Map of Montessori Schools

Cheyenne Eete Kippenberger

Desmond Blackburn

Username or Email Address

Remember Me

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Front Psychol

Montessori Preschool Elevates and Equalizes Child Outcomes: A Longitudinal Study

Angeline s. lillard.

1 Department of Psychology, University of Virginia, Charlottesville, VA, United States

Megan J. Heise

Eve m. richey, alyssa hart, paige m. bray.

2 Department of Education, University of Hartford, Hartford, CT, United States

Quality preschool programs that develop the whole child through age-appropriate socioemotional and cognitive skill-building hold promise for significantly improving child outcomes. However, preschool programs tend to either be teacher-led and didactic, or else to lack academic content. One preschool model that involves both child-directed, freely chosen activity and academic content is Montessori. Here we report a longitudinal study that took advantage of randomized lottery-based admission to two public Montessori magnet schools in a high-poverty American city. The final sample included 141 children, 70 in Montessori and 71 in other schools, most of whom were tested 4 times over 3 years, from the first semester to the end of preschool (ages 3–6), on a variety of cognitive and socio-emotional measures. Montessori preschool elevated children’s outcomes in several ways. Although not different at the first test point, over time the Montessori children fared better on measures of academic achievement, social understanding, and mastery orientation, and they also reported relatively more liking of scholastic tasks. They also scored higher on executive function when they were 4. In addition to elevating overall performance on these measures, Montessori preschool also equalized outcomes among subgroups that typically have unequal outcomes. First, the difference in academic achievement between lower income Montessori and higher income conventionally schooled children was smaller at each time point, and was not (statistically speaking) significantly different at the end of the study. Second, defying the typical finding that executive function predicts academic achievement, in Montessori classrooms children with lower executive function scored as well on academic achievement as those with higher executive function. This suggests that Montessori preschool has potential to elevate and equalize important outcomes, and a larger study of public Montessori preschools is warranted.

Introduction

Optimizing preschool education is important from both economic and developmental standpoints ( Heckman, 2006 ; Blair and Raver, 2016 ). The human brain undergoes marked development in the first 6 years, and the environment interacts with gene expression producing changes that appear to be permanent ( Zhang and Meaney, 2010 ). Furthermore, neural development proceeds in a hierarchical fashion, with later attainments built on earlier ones ( Merzenich, 2001 ). Economic analyses show that the highest rates of return on educational investments in human capital are derived from preschool programs ( Heckman, 2006 ). Yet the two primary examples of successfull early childhood interventions (Perry Preschool and the Abecedarian Project) are from the 1960s ( Campbell et al., 2002 ; Schweinhart et al., 2005 ) and were small studies with very intensive interventions that would be very expensive (on the order of $20,000/year per child) to implement in today’s dollars ( Minervino and Pianta, 2014 ). Doing such interventions at scale would be exceedingly difficult. However, some alternative public preschool programs can feasibly be widely implemented; one such program is Montessori. Understanding if such programs provide measurable benefit to young children’s development is a prerequisite to determining whether to attempt implementation at scale.

Montessori education aligns with principles and practices that a century of research has shown are more optimal for child development than the principles and practices that undergird conventional schooling ( Lillard, 2017 ). Developed by a physician in the first half of the 20th century, the educational method stemmed from close observation of children in relatively free environments. It provides a complex and interrelated set of hands-on materials and lessons across major topic areas and is designed for children ages 0 to 12+ years ( Montessori, 1994a ). Within a structure created by the materials and teacher oversight, children are free to make constructive choices among activities that they have been taught, to explore personal interests (with the caveat that they also engage broadly), and to decide whether to work alone or with peers in the multi-age classrooms. There are no grades or extrinsic rewards, and learning is situated in real or simulative contexts. Montessori education is aimed at development of the whole child, integrating social and cognitive growth for healthy independent functioning.

The first studies of Montessori outcomes lacked good controls or had small samples and compromises in program quality; for example, they used single-age classrooms, added non-Montessori activities, and/or had teachers with minimal training ( Karnes et al., 1983 ; Miller and Bizzell, 1984 ). Program quality is clearly an important consideration, as children in higher-fidelity Montessori classrooms (where children had only Montessori activities) had larger social and cognitive school-year gains than those in lower-fidelity ones ( Lillard, 2012 ). However, the Lillard () study had serious limitations, including that the children were middle-income and not randomly assigned to the schools, which were private. Such limitations are common in the relatively few existing studies of Montessori education ( Rathunde and Csikszentmihalyi, 2005 ; Peng and Md-Yunus, 2014 ).

Another study avoided these problems by testing 5-year-olds in a high-fidelity public inner-city Montessori school who had gained admission through a computerized district-level random lottery when they were 3 years old, and compared their outcomes to those of 5-year-olds who had lost that lottery and were at non-Montessori schools ( Lillard and Else-Quest, 2006 ). The Montessori children significantly outperformed the control children on an array of measures. In that study, however, the sample of preschoolers was small ( N = 55), and the children were tested just once during the school year. These limitations are also problematic.

In the present study, children in two high-fidelity public Montessori magnet schools (11 classrooms) who had gained admission via a random computerized district-level lottery at 3 years old were compared to a group who had lost the lottery and attended other non-Montessori schools, over half of which were private schools. Children ( N = 141) were tested over the fall semester when they were 3 years old, and then again at the end of the school year for three consecutive years. The tests, described next, assessed a variety of skills known to be important to later success.

Children’s academic ability is considered of primary importance in school assessments. For young children, initial progress in reading, vocabulary, and numerical understanding are valued indicators. Here we measured these with four Woodcock–Johnson IIIR Tests of Achievement: Letter-Word, Picture Vocabulary, Applied Problems, and Calculation ( Woodcock et al., 2001 ). The Woodcock-Johnson tests have good psychometric properties as described in the manual, and are frequently used to measure school outcomes.

Academic benefit might have trade-offs in social learning; indeed, Montessori education has been criticized for being “asocial” since the children rarely participate in whole-class activities ( DeVries and Gonçu, 1987 ). Social cognition was measured with the Theory of Mind scale ( Wellman and Liu, 2004 ), which has good internal and external validity ( Wellman, 2014 ); for example, it predicts later social competence ( Wellman, 2014 ). A central construct in the Theory of Mind scale is understanding of false belief, which has garnered considerable attention in developmental psychology and education in the last 30 years ( Blair and Razza, 2007 ). Understanding that someone can have a false belief entails the crucial understanding that minds represent the world, and that people’s behaviors are based not (necessarily) on the way the world actually is, but on how they represent the world to be ( Dennett, 1987 ). The Theory of Mind scale contextualizes this key understanding with steps leading up to it (understanding of perception and its relation to knowledge, and understanding that people can believe different things) and following it (understanding that the emotions we convey might be different from the emotions we actually feel).

Although theory of mind is related to social competence, they are different constructs. Social competence was measured more directly with stories from the Rubin’s Social Problem-Solving Test - Revised ( Rubin, 1988 ); a different story was used each year, and scoring was modified to home in on the maturity of social competence revealed in children’s responses. In these stories, one child has a coveted resource (like a swing) that another child really wants, and children need to come up with strategies the focal child could use to obtain the resource; responses like “I would ask her to share for 10 min then she could have it for 10 more minutes” are considered highly competent, whereas “I’d tell the teacher” or “I’d say please, please, please” are not. Other studies have shown that children in high-fidelity Montessori preschools show more social competence on this task (as well as better playground interactions) than children in other types of preschools ( Lillard and Else-Quest, 2006 ; Lillard, 2012 ).

Theory of mind is also strongly associated with executive function and involves many of the same neural structures (for example the medial and lateral prefrontal cortex and the temporo-parietal junction) ( Carlson and Moses, 2001 ; Koster-Hale and Saxe, 2013 ; Powell and Carey, 2017 ). Executive function was measured in this study because it undergirds self-regulatory skills that are important to academic and life success ( Blair and Razza, 2007 ; Diamond, 2013 ; Vernon-Feagans et al., 2016 ); in fact, self-regulation at age 4 predicts health, wealth, and criminality outcomes at age 32 ( Moffitt et al., 2011 ). Here executive function was measured with two tasks; a full battery of tests would have been desirable ( Willoughby et al., 2011 ; Lipsey et al., 2017 ), but time constraints only allowed two. One executive function task was Head-Toes-Knees-Shoulders (HTKS), in which a child must do the opposite of a command (for example, touch their toes when asked to touch their head). To do this, a child must keep a command in mind along with the rule to execute its opposite, must inhibit the opposite response, and must executive the required one. This task has good psychometric properties and is related to other tests of executive function as well as concurrent and later academic success ( McClelland et al., 2007 ; Ponitz et al., 2008 , 2009 ; Lipsey et al., 2017 ). The second executive function assessment was the Copy Design subtest from the Visuospatial Processing section of the NEPSY-II ( Korkman et al., 2007 ). For this task, children see a design, and must hold it in mind as they transform the visual image into its motor execution and a new resulting visual copy of that image. Thus working memory, attention, inhibitory control, and execution skills are employed. Design copy is highly related to other tests of executive function ( Grissmer et al., 2010 ; Cameron et al., 2012 ; Fuhs et al., 2014 ; Lipsey et al., 2017 ) and has good test-retest reliability ( r = 0.72 in Lipsey et al., 2017 ). Design copy ability is also related to academic achievement ( Grissmer et al., 2010 ). Although both of these tasks require some similar executive function skills, HTKS involves large motor processes whereas Design Copy involves fine motor skills.

In addition to academic achievement, theory of mind, social competence, and executive function, which have been examined previously, we also used three tasks not previously used in studies of Montessori preschool. The first was the growth of a mastery orientation. Mastery orientation is an important personal quality ( Dweck, 2006 ) indicative of a “growth mindset” ( Dweck, 2017 ): a belief that with effort one can master challenges and increase one’s abilities. People who are mastery oriented want to learn, and take on challenging tasks in order to do so. They are resilient, persisting even in the face of failure. Their implicit theory of intelligence is that it is malleable, such that the harder one works, the better one can be. By contrast, people who are performance oriented seek to look good; their implicit theory of intelligence is that it is fixed, and they tend to give up in the face of failure. About 80% of Americans naturally adopt one orientation or the other, but circumstances can alter those orientations. Clearly if school could increase mastery orientation, this would be positive. Because conventional school practices like extrinsic rewards tend to instead encourage a performance orientation, and Montessori education does not use them, we expected that children might be more mastery oriented by the last 2 years of Montessori preschool. Mastery orientation was measured with a modification of a puzzle task developed by Smiley and Dweck (1994) . Children were given an easy and a very difficult (actually, impossible) puzzle to solve, and then later were offered the opportunity to work on either puzzle again. Convergent evidence suggests that children who choose to continue to work on an unsolvable puzzle are “persisters” with a stronger mastery orientation than children who choose to work again on an easy puzzle ( Smiley and Dweck, 1994 ). Having a mastery-oriented mindset predicts achievement over time ( Dweck, 2006 ). Because it would take time for an orientation like this to develop in a school program, and because it involved a 0–1 response, choices at the first two vs. the last two time points were examined.

The second new construct was feelings about academic tasks. Early academic achievement might occur at the expense of enjoying school tasks, which is undesirable since enjoying kindergarten predicts later school achievement ( Ladd et al., 2000 ). Not liking school tasks could stem from extensive emphasis on academics and could presage burnout, an issue recently raised with regard to a study of Tennessee preschoolers who performed less well by second grade than children who had not gone to preschool ( Lipsey et al., 2015 ; Haskins and Brooks-Gunn, 2016 ). Therefore we assessed children’s liking of academic tasks such as school lessons and reading. However, because preschool-aged children tend to be very positive about many experiences, how much they professed to like leisure activities like playing and watching movies was also taken into account.

Another measure not used in prior studies of Montessori outcomes was the Alternate Uses task, which assesses creativity. Creativity is certainly a desirable construct. Because conventional educational methods often require children to answer questions in specific ways (as on multiple choice tests) but Montessori often encourages independent exploration, Montessori might promote more creativity. On the other hand, there are particular ways that children are instructed to use specific Montessori materials, and this could discourage creativity. Alternate Uses (sometimes called Creative or Unusual Uses) is a commonly used task that asks one to come up with as many uses as one can for common items like paper clips and towels ( Guilford and Christensen, 1973 ). It was administered at each time point after the first fall. Many major current innovators, like both founders of Google (Sergei Brin and Larry Page), the founder of Amazon (Jeff Bezos), the creator of Wikipedia (Jimmy Wales) and the designer of the once-revolutionary video game Sim City (Will Wright) attended Montessori schools ( McAfee, 2011 ; Gaylord, 2012 ), and other studies have shown that Montessori children are more creative in later grades ( Lillard and Else-Quest, 2006 ; Besançon and Lubart, 2008 ), but not in preschool. To our knowledge, no other study has used Alternate Uses with Montessori preschool children.

In sum, the study measured children’s academic achievement, theory of mind and social skills, executive function, mastery orientation, relative enjoyment of school, and creativity at four time points to determine whether Montessori education would have a significant influence on those important constructs.

In addition to examining the overall efficacy of Montessori preschool for these measures, the study (because of its sample size) permitted examination of Montessori’s potential for disrupting the predictive power of certain variables for certain outcomes. One is the predictive power of income for achievement, or the income achievement gap. Childhood poverty is a significant predictor of poor life outcomes ( Brooks-Gunn and Duncan, 1997 ; Yoshikawa et al., 2012 ). Education is widely viewed as a ladder out of poverty, yet socio-economic status (SES) and school achievement are correlated ( National Early Childcare Research Network, 2005 ; Sirin, 2005 ). The income achievement gap, which is larger than the racial achievement gap, is present by kindergarten and persists at that high level throughout school ( Reardon, 2011 ). Here we examined Montessori’s potential to address the income achievement gap in preschool. Second, executive function is known to predict many life outcomes ( Moffitt et al., 2011 ); children with poorer executive function generally do not do as well in school ( Blair and Razza, 2007 ; Duncan et al., 2007 ), and so remedial programs like the Chicago School Readiness Project ( Raver et al., 2011 ) and Tools of the Mind ( Diamond et al., 2007 ) are instituted as costly add-on programs. Montessori is a form of differentiated instruction that can naturally support different levels of executive function. For example, a child who needs more structure can be monitored more closely than a child who needs less structure. This is more difficult to do in conventional schools, since the structure is set up to treat all children in a given class in the same way ( Tomlinson, 2014 ). Because Montessori can more easily and naturally accommodate differences in children, we ask whether executive function might be less predictive in Montessori programs.

The samples were ethnically diverse and equivalent at the first test point in terms of parent education and income (ranging from $0 to $200,000), child age, and Time 1 scores; this lack of pre-existing differences would be expected given the random lottery assignment. Slight (but non-significant) differences in performance at Time 1 could be due school programs already having influenced children at the first test point, which ranged from mid-September to mid-December. Over the subsequent 30 months, significant differences emerged on several measures, all indicating better outcomes for children in the Montessori program.

Materials and Methods

This longitudinal study examined how children in Montessori vs. other preschool environments changed over 3 years. The same basic set of tests were administered to children at each time point. The study was carried out in accordance with the guidelines for human research of the Institutional Review Board for the Social and Behavioral Sciences at the University of Virginia, which approved the protocol.

Participants

Sample characteristics are detailed in Table ​ Table1 1 . In brief, the final sample included 70 children in Montessori and 71 controls who were at other non-Montessori schools. Children were 41.15 months old on average at the first test point, and each sample was ethnically diverse and had slightly more males than females. Household income ranged widely (because the lottery was for a magnet school) as did parent education; the average parent had some college education, but the range was from 9th grade through post-graduate. The two subsamples did not differ on any measured ethnographic variable.

Sample characteristics.

Recruitment

All participants were recruited from Hartford, CT and its outlying suburbs by letters sent home from the school district office following a school choice lottery (see below) in each of 4 years spanning 2010–2013; each participating child was in the study for 3 years, so data collection spanned from fall 2010 through spring 2016. Letters were sent to parents of all 3-year-olds who had been entered in a lottery listing one of two public Montessori magnet schools as their first choice; the letters were accompanied by contact, demographic, and school information forms, a permission letter, and an envelope to return their information to the study coordinator. Parents were sent a $10 gift card as a thank you for returning the information forms. After spring tests each year, children were sent an age-appropriate book and parents were sent a $50 gift card.

The lottery was done by computer at the Connecticut State Department of Education’s Regional School Choice Office in Hartford, CT in May of each year. A child’s parent or guardian had submitted a lottery application during the period spanning October through February, selecting one of the two Montessori schools as their first of five school choices. The lottery selection was random except for neighborhood, sibling, and staff preferences. Staff children were disqualified from the study but 2 study children were admitted to a Montessori via the sibling preference; their siblings had presumably been admitted at random so the latent parent characteristics the lottery was intended to control for were still present. One control child had been admitted to Montessori but did not attend because the parents “did not like the neighborhood the school was in”; all other participants who gained admission to one of the two Montessori schools did become enrolled there. These two siblings and the admitted non-attender were assigned to the school program group they were actually in, but removing the two siblings and placing the cross-over child in the experimental group (“intent-to-treat”) had no meaningfully effect on results. For example, the ANCOVA on Time 4 academic achievement strengthens slightly when these changes are made, from F (2,119) = 7.24, p = 0.008, η p 2 = 0.06 to F (2,117) = 9.58, p = 0.002, η p 2 = 0.08. For philosophical reasons (such as grouping participants according to the treatment actually received) the study’s original group assignment was retained.

Control schools

Forty-three control children attended the same schools for the duration of their time in the study; 26 made one school switch, and 1 switched schools twice. At the beginning of the study, the 71 control children were in 51 schools; most of those schools had 1 child, some had 2–3, and one had 4. Over the course of the entire study (6 school years), control children were at 71 different schools. (Children were tracked at the school, not the classroom level). Thirty of the 71 schools were publicly funded (15 magnet including for example Reggio, Arts, and Environmental Science schools; 8 conventional public schools; and 7 Head Start programs) and 41 were private schools. Thirty-two of the schools attended by control children were in Hartford city (including West Hartford, which is wealthier with an average household income of $120,000) and 39 were in the outlying suburbs. Public early childhood programs in Connecticut must (1) satisfy the NAEYC accreditation standards and (2) be a member of the state’s early childhood professional registry. Connecticut requires an Early Childhood Teaching Credential that entails either (1) being a graduate of an approved higher education program or (2) another higher education degree, teaching experience, and 12 credits in early childhood education.

Montessori schools

One of the Montessori schools was the first public Montessori school in Connecticut, established in 1994. The other one opened in 2008. During the study years both Montessori schools were recognized by the Association Montessori Internationale (AMI) for their strict fidelity to original principles. One school had 5 classrooms and the other had 6 classrooms serving 27 three- to six-year-olds. One school also included students to 6th grade and the other to 8th grade; each had about 350 children in total. The teachers all had AMI training, for which a BA/BS degree is preferred but not required. Three of the teachers originally at one school had previously taught conventionally, and agreed to be retrained when the school converted to Montessori in 2008. There was some teacher turnover during the study but these changes were not tracked at either Montessori or conventional schools.

Missing Data and Exclusions

Over 4 years, 174 children were admitted to the study; 141 were retained in the final sample. Of these 141, 122 children were tested at all 4 time points, and 19 were tested at 3 time points. Of these 19, one joined the study at Time 2, 2 missed one test session, and 16 moved or crossed over between Time 3 and Time 4. 11 of these were in Montessori and 5 were control children. The control children who were lost had all moved; this lost subset of control children had performed significantly lower in academic achievement at earlier time points than the control children who did not move. The Montessori children who were lost at Time 4 did not significantly differ from those who remained in the study. Thus attrition patterns bias Time 4 results toward better outcomes for the control sample. For the variables reported here and the remaining children, 2.6% of data is missing due to experimenter error, child non-compliance, or interruptions in testing.

Of the 33 children who were admitted but excluded from the study, 23 children contributed insufficient data; 4 of these (2 Montessori) were lost between Times 1 and 2 and 19 (9 Montessori) were lost between Times 2 and 3. The children who were lost did not differ from other children in terms of parent education, parent income, ethnicity, or gender. The decision not to include these children was based on a preference for actual over imputed data. The other 10 excluded children (6 Montessori) had insufficient English ( n = 5), speech delay ( n = 3), or other learning disabilities ( n = 2).

All parents provided written informed consent. Testing was conducted one-on-one, usually in the child’s school, but in a few cases in a public library due to lack of school cooperation. Ten trained research assistants tested children over the course of the study (eight graduate students and two project coordinators). Tasks were administered in a fixed order chosen to vary formats for engaging children: Theory of Mind, Letter-Word, Alternate Uses, Design Copy, Puzzle Part 1, Math, Head Toes Knees Shoulders, Social Problem-Solving, Picture Vocabulary, Preference Questionnaire, Puzzle Part 2. Testing was done simultaneously at Montessori and control schools so that test time would not be confounded with school type.

Participants were administered the same tasks at all test points, except the Preferences Questionnaire and the Alternate Uses creativity task, which were added in the spring of 2011, so these tasks are missing at Time 1 from the 29 participants who enrolled in 2010.

On some tasks, having exactly the same items at different test points would threaten validity. For these tasks there were four sets of materials, administered on a rotating basis.

Academic Ability

Children’s academic ability was assessed using the Woodcock–Johnson IIIR Tests of Achievement according to the instructions in the manual ( Woodcock et al., 2001 ). Because there were no age differences across samples, raw scores were used for all Woodcock–Johnson tests. The Picture Vocabulary subtest assessed vocabulary, and the Letter-Word subtest assessed reading. Because the Montessori schools both taught cursive letters, the printed letters in the earlier items on the Letter-Word subscale were overlaid with cursive letters when testing Montessori students. Ordinary print letters were retained from the point when the test changes from letter to word identification. Early mathematical achievement was measured with the Applied Problems subtest, followed by the Calculation subtest if children scored 19 points or higher. These scores were summed for a Math score. The Math, Letter-Word, and Picture Vocabulary score loaded on a common factor (see Appendix ) and were highly correlated ( r s > 0.80), so to reduce the number of comparisons in the study, these scores were combined (by adding Z -scores) for an overall Academic Achievement measure (e.g., Lipsey et al., 2017 ).

Theory of Mind

We used four tasks from the Theory of Mind Scale ( Wellman and Liu, 2004 ) omitting the lowest level (Diverse Desires) for brevity since 3-year-olds typically pass this level. As an example, in the Knowledge Access task, children were shown what was hidden in the drawer of a doll-house-sized bureau, and then shown a doll who they were told had not seen inside the drawer. They were asked if the doll knew what was inside the drawer, and if the doll had seen inside the drawer; both answers had to be correct for a child to be given credit. Children were given Knowledge Access first, followed by Contents False Belief, Diverse Beliefs, and Hidden Emotion, for final scores of 0–4. The contents, dolls, and doll names changed for each test session. For example, for contents false-belief task, one year the child saw a Band-Aid box with crayons inside, another year a raisin box with buttons inside, another year a Crayons box with rubber bands inside, and another year a Cheerios box with beads inside. Since children entered the study for four consecutive years, each material set came first for a portion of the sample.

Social Problem Solving

One object acquisition story from Rubin’s Social Problem-Solving Test - Revised was administered ( Rubin, 1988 ) each year. In these stories, children were shown two other preschoolers, one of whom had a coveted resource like a swing and had had it for a “long, long time” and the other of whom wanted that resource. Children were asked what the second child could do or say to get the resource, what else they could do or say, and what the child him- or herself would do or say. Children’s use of strategies considering fairness and justice for both parties were coded. Although there is no limit to the number of such solutions a child might give, in reality the range was 0–3 at all four test points. Interrater reliability on 20% of all responses across all years was 0.99.

Executive Function

Executive Function was assessed with two tasks. For Head-Toes-Knees-Shoulders ( Ponitz et al., 2009 ), children were first asked to touch their head, then to touch their toes. Children were then told that they were playing an “opposite game” in which they must touch the opposite part of the body than the experimenter said. Children were then administered 10 items, each scored 0–2, with 0 indicating the child followed the command literally, 1 meaning the child touched the incorrect body part first and then corrected themselves without prompting, and 2 meaning the child touched the correct (opposite) body part. If a child scored 10 points or more on the first 10 items, a second series of 10 items was administered which included knees and shoulders; the maximum points a child could earn was 40.

Second, the Design Copy subtest from the Visuospatial Processing section of the NEPSY-II was administered and scored according to the manual ( Korkman et al., 2007 ). Children were shown a paper with a 4 × 4 grid with four figures across the top and third rows. The first figure was a vertical line; the experimenter showed children how to copy the line in the box below it (first box, second row), saying (for 3- and 4-year-olds), “See this line? I will draw one here. Now you draw one here,” handing the child the pencil and pointing to the second figure (a horizontal line) and the box below it. For 5-year-olds, and for the remaining items, the experimenter simply pointed to the top figure then the blank box below it, saying, “Copy this one here.” This continued for up to 16 figures until a child failed to successfully copy three figures consecutively. An independent coder coded a randomly selected subset of children at each test period, and interrater reliabilities across the two coders were excellent: r s = 0.98 (32 children at Time 1); 0.96 (22 children at Time 2); 0.95 (14 children at Time 3); 0.90 (22 children at Time 4).

Mastery Orientation

The puzzle task (modified from Smiley and Dweck, 1994 ) designed to test mastery orientation was given in two parts. First, children were given a fairly easy puzzle for their age, along with a picture of what the completed puzzle should look like. The picture was turned over while children solved the puzzle. After 2 min or when children completed the puzzle (whichever occurred first), they were given a much more difficult puzzle to solve and its completed picture which was then turned over. However, in this puzzle there were also pieces that had been switched with a similar puzzle, rendering the puzzle unsolvable. Children were again given 2 min to work on the puzzle. Then they completed several other tasks, and finally the experimenter brought out both puzzles again, told children that they had some extra time, and asked which one they wanted to work on and why; children could opt for neither or the easier puzzle (scored 0), or the more difficult puzzle (scored 1).

School Enjoyment: Preference Questionnaire

A questionnaire was developed to assess children’s enjoyment of academic (school and reading) and leisure (media and play) tasks; four filler questions were included as well. There were four questions about each of the focal topics, and children rated their enjoyment by pointing to a sad, neutral, or happy face. These responses were coded as 0, 1, or 2, and added together. Since young children often give the highest possible ratings on such scales ( Ladd et al., 2000 ), to get variability, responses at the end of each school year (so they had experience with the school tasks) were summed, and liking for academic tasks was subtracted from liking of recreational tasks, reflecting how much more each child liked recreational than scholastic activities across preschool.

Alternative Uses was used to assess creativity ( Guilford and Christensen, 1973 ). First, as a warm-up, children were shown a photograph of an object (e.g., a pencil) and the experimenter said, “See this? This is a pencil. Can you tell me as many different things that you can think of that you can do, play or make with this?” If children made no reply in 10 s, the experimenter prompted with one use. The first of two test items was presented in the same way (“See this? This is a bucket…”). Responses were recorded for 1 min, with the experimenter prompting “What else?” If a child was producing responses and then appeared to run out of ideas (did not respond for a few seconds), the second item was shown and the same process repeated. For both test items the total time during which responses counted was 2 min; responses given after 2 min were not included.

Each intelligible response was scored as standard or non-standard. Categories were exclusive. For example, a standard use for a towel would be to wipe one’s body, and a non-standard use would be to place it over one’s head to pretend that one is a ghost. Analyses were conducted on the number of non-standard uses each child gave, collapsed across both items at each assessment. The actual range of responses was 0 to 5 total non-standard uses. Two coders independently coded a randomly selected subset of the data ( n s below). Reliability was r = 0.80 on 16 children who were double-coded at Time 1; 0.73 (45 children at Time 2); 0.79 (46 children at Time 3); 0.82 (40 children at Time 4).

Statistical Analyses

Some analyses reported here employed growth curve modeling, one of the most frequently used analytic techniques for longitudinal data analysis with repeated measures. Growth curve modeling can directly analyze intraindividual change over time and interindividual differences in intraindividual change ( McArdle and Nesselroade, 2014 ). Growth curve analysis obtains a description of the mean growth in a population over a specific period of time. Individual variations around the mean growth curve are due to random effects and intraindividual measurement errors.

A typical growth curve model can be expressed as

where y i = (y i1 ,y i2 ,...,y iT )′ is a T × 1 vector and y ij is an observation for individual i at time j ( i = 1, ..., T; j = 1, ..., T where N is the sample size and T is the total number of measurement occasions); Λ is a T × q factor loading matrix determining the shape of growth trajectories, b i is a q × 1 vector of random effects, and e i is a vector of intraindividual measurement errors. The vector of random effects b i varies for each individual, and its mean, representing the fixed effects, can be interpreted by a function of covariates X i with parameters β. The residual vector u i represents the random component of b i .

We use maximum likelihood estimation methods to fit the model. Missing values are believed to be missing completely at random (MCAR) or missing at random (MAR). Thus, Full Information Maximum Likelihood method (FIML) is applied to deal with missing data.

Data were not nested in control classrooms for the obvious reason that most control schools had only one child, and children’s classrooms and teachers were not tracked because they were not the focus of this study. Data were also not nested within Montessori classrooms, and the reason for this might be less obvious: Every year the 11 Montessori classrooms were differently constituted. First, peers changed: Always, at least 33% of children turned over as the oldest group of 9 moved on and a new group of 9 three-year-olds entered. In addition, several teachers and assistants turned over at some point during the study (although this was not closely tracked, at least three teachers at one school turned over), rendering different teacher experiences for each wave of children entering a given physical class (some had teacher A for 3 years, others for 2, others for 1, and others did not have teacher A at all). For this reason, treating children who entered a given classroom in 2010 and those who entered that classroom in 2013 as being in the same class (as a nested design would do) would not make sense; they had no overlap in peers, and many had different teachers as well. If we treated each entering year as different classrooms, we would have many tiny groups (1.6 children per nested group on average, given the average of 6.36 children per classroom entering over 4 years). Nesting Montessori children in classrooms therefore did not make sense. Analyses comparing results at the two Montessori schools revealed no school differences.

Time 1 Equivalence

T -tests were done on all results to determine whether the samples differed already at their initial test (Time 1), conducted at some point during their first 3 months of school. The p -values exceeded 0.05 for all tests, indicating that the samples were equivalent at the start of the study.

The groups were slightly (although not significantly) different in academic achievement at the first test point. Since the children were randomly assigned to Montessori or the waitlist, it seems most likely that these small differences were due to their respective school programs beginning to have an effect between the time of school entry and the initial test point (which was mid-December for some children, 3 months into the school year). This is further supported by lack of group differences in all the demographic variables.

Here we first explain how data were reduced, then discuss the results showing that Montessori preschool elevated performance overall for the whole sample. We next discuss results showing that Montessori equalized performance of subgroups by raising the typically lower-performing subgroups towards the level of the higher-performing subgroups. We end with a comparison of public Montessori with public and private non-Montessori schools.

Data Reduction

The Woodcock-Johnson scores loaded on a single factor and were significantly intercorrelated within each time point ( r s > 0.80), so were converted to Z -scores and summed for an Academic Achievement score at each test point. The Copy Design and Head-Toes-Knees-Shoulders task also loaded on a single factor and were also significantly correlated ( r = 0.66) so were converted to Z -scores and summed for each test point. Figure ​ Figure1 1 shows the correlations across the composite variables and Theory of Mind across time points, and the Appendix describes the factor analysis.

Correlation Table for Academic Achievement, Theory of Mind, and Executive Function across four time points. These variables were selected because their interrelations are of significant interest in preschool research. In this graphic representation, all squares are red because all correlations were positive. The shading legend is on the right. Darker colors (as well as larger squares) represent stronger correlations.

Overall Findings: Montessori vs. Business-As-Usual

Academic achievement.

Although equal at the start of school, the Montessori group advanced at a higher rate across the study years, as illustrated in Figure ​ Figure2 2 ; Δ B = 0.13 ( SE = 0.067), p < 0.05. This initial analysis did not control for demographic variables because there were no differences, as would be expected given random assignment, but to confirm this a second growth model was created controlling for gender, household income, and Time 1 executive function. This confirmed that while both groups were equal at intercept in academic achievement, Montessori predicted a steeper slope of growth, whereas none of the control variables predicted a steeper slope in the overall sample. The result from the growth curve analysis was confirmed by an ANCOVA on Time 4 academic achievement, controlling for academic achievement at Time 1, F (2,119) = 7.24, p = 0.008, η p 2 = 0.06. Independent samples t -tests showed that the groups were not yet different at Time 1 or Time 2, and that significant differences in academic achievement had emerged by the last two time points (approximately 4 and 5 years of age): t (136) = 2.10, p = 0.04, Cohen’s d = 0.36, and t (122) = 2.26, p = 0.03, Cohen’s d = 0.41, respectively.

Academic achievement across preschool by school type. The figure shows significantly greater growth in academic achievement across preschool for children enrolled in Montessori preschool (dashed blue lines, n = 70) than waitlisted controls (dotted black lines, n = 71). Groups were statistically equivalent at Time 1 (the non-significant difference at Time 1 is likely due the Time 1 tests occurring into mid-December, thus school programs could already have made a difference) and Time 2 (late in the spring of their 1st year in preschool) and significantly different by the end of their 2nd and 3rd years in preschool (Times 3 and 4). Dashed/dotted lines represent actual data and solid lines represent fitted linear growth curves. Standard error bars are shown.

Although children’s scores were equal at the initial test, a linear growth curve model showed that Montessori children had a significantly steeper rate of growth across the preschool years, Δ B = 0.10 ( SE = 0.04), p < 0.05. This result remained in a second growth curve model that controlled for age, household income, and Time 1 Executive Function. Using a different analytic approach, an ANCOVA on Time 4 Theory of Mind scores controlling for Time 1 scores also showed a significant difference favoring the Montessori group, F (2,115) = 4.47 p = 0.04, η p 2 = 0.04. Scores were examined at each time point. For Times 1 and 2 the two groups were not different. At Time 3, the difference was significant, t (135) = 2.09, p = 0.04, Cohen’s d = 0.36, and at the end of kindergarten (Time 4), the difference was a trend, t (122) = 1.74, p = 0.08, Cohen’s d = 0.32. These results show that social cognition developed more rapidly in children attending Montessori schools.

Children in the two samples were equivalent throughout the study with respect to their social problem-solving skills; the average number of justice-related responses ranged from 0.24 to 0.97 across the 4 time points. An ANCOVA on Time 4 Social Problem Solving controlling for Time 1 comparing Montessori and control samples was non-significant F (1,117) = 0.20 p = 0.66, η p 2 = 0.002, nor was the group difference significant at any time point with independent samples t -tests.

Linear growth curve analyses did not indicate differences in the growth of executive function. An ANCOVA on Time 4 executive function controlling for Time 1 only showed a trend toward a difference, with Montessori children scoring more highly: F (2,118) = 3.00, p = 0.09, η p 2 = 0.03. Only at Time 3 was the difference significant, t (135) = 2.09, p = 0.04, Cohen’s d = 0.35. Evidence that Montessori magnet preschools lead to better executive function as compared to that developed by control children attending other preschools is not strong here.

At the first two time points, there were no group differences: 37 of 70 Montessori (53%) and 35 of 71 control children (49%) chose to try a difficult puzzle again on one or both occasions (Fisher’s Exact test, p = 0.74). By the time children were 4 and 5, at Times 3 and 4, school program effects were significant, with Fisher’s Exact test showing more Montessori children made the mastery choice (45 of 69 or 65%) than did control children (33 of 71 or 47%), p = 0.03, two-tailed. Thus, children who were randomly assigned to a Montessori program were more likely to have a growth mindset by the latter half of their preschool years. Children’s explanations for their choices were consistent with the underlying orientation. Easy puzzle choosers said things like, “Because it’s easier,” whereas difficult puzzle choosers said things like, “Because I think I can do it.”

School Enjoyment

An ANOVA showed that the Montessori children were relatively more positive about school-related activities than were the control children, F (1,116) = 5.69 p = 0.02, η p 2 = 0.05 (see Figure ​ Figure3 3 ). This suggests that the Montessori children’s achievement gains were not at the expense of their enjoying school.

Enjoyment of recreational (left panel) and academic (right panel) activities across preschool. Montessori children ( n = 55, blue beans, on right side of each panel) were relatively more favorable to academic tasks than control children ( n = 63, gray beans). Dots represent children, bars represent means, and shaded areas represent 95% confidence intervals.

Children in the two samples were equivalent throughout the study with respect to their creativity; average non-standard uses scores ranged from 0.31 to 1.55 across the 4 time points. An ANCOVA on Time 4 Creativity controlling for Time 1 Creativity comparing Montessori and control samples was non-significant F (1,94) = 0.96 p = 0.33, η p 2 = 0.01, nor was the group difference significant at any time point with independent samples t -tests.

Comparison of Subgroups in Montessori vs. Business-As-Usual Schools

We examined two sets of subgroups. First, we looked at the association of achievement with household income in Montessori vs. control schools. Because this achievement gap has been of considerable interest in the country historically, we present several analyses of this issue, before examining the influence of different levels of executive function in each sample.

Levels of Achievement for Children of Different Income Levels

Income is typically associated with school achievement. This was the case in the control sample, as shown in the right hand side of Figure ​ Figure4 4 using data from the final test point (Time 4). The left hand side shows this relation for the Montessori sample. Among children in Montessori, the correlation between academic achievement and household income across the entire study was 0.23, whereas in the control sample it was twice that: 0.46. Using the Fisher transformation, this difference in correlations was significant, Z = 2.46, p = 0.01. To further examine this, 1000 bootstrapped samples were generated; the 95% bootstrap confidence intervals of Δ r was (0.04, 0.39), supporting that the correlations between income and academic achievement in the two samples are significantly different. The smaller correlation among Montessori children might be a simple function of their being in magnet schools, since this is in essence the point of magnet schools [although their success at this is mixed ( Ballou, 2009 )]. However, for the subgroup of 15 control children who were at other magnet schools, the correlation between academic achievement and household income was even stronger, suggesting the mitigated income-achievement correlation for Montessori children is not merely due to their being in magnet schools.

Relation between academic achievement and household income in Montessori and control children at the end of the kindergarten year. The relation is significantly smaller in Montessori children ( n = 58, left panel ) than in control children ( n = 66, right panel ).

How strong the gains in academic achievement were among just the lower income children is also of interest, because of the income achievement gap. Although the income range was very broad, there was not a sufficiently large subsample to only examine those living below the poverty line, so instead we examined the study subsample with a household income below the median split. For this lower income half of the sample ( n = 67), mean household income was $32,627; SD = 18,443; the federal poverty line for a family of 4 in Connecticut was $24,600. At Time 1, an ANCOVA on academic achievement controlling for age (because there was a slight age difference in the subsamples), showed no difference between the Montessori and control lower income subsamples, whereas by Time 4 the lower income Montessori subsample had significantly higher academic achievement than the lower income control subsample, F (1,62) = 6.86, p = 0.01, η p 2 = 0.10; see Figure ​ Figure5 5 . This result also held when controlling for Time 1 academic achievement: F (1,61) = 7.25, p = 0.009, η p 2 = 0.11.

Academic achievement across four time points by school condition and income group. Although equal to the lower income control children at Time 1, by Time 4 the lower income children in Montessori showed a strong positive trajectory towards closing the achievement gap with the higher income children in control and Montessori schools. Standard error bars are shown.

Furthermore, Montessori education greatly reduced the achievement gap across the preschool years. A series of four t -tests compared the lower income Montessori children with the higher income control children at each time point. For the higher income half of the sample ( n = 74, including 7 at the median income of 70,000), mean household income was $105,804; SD = 33,123. The higher income control children outperformed lower income Montessori children at Times 1 and 2, t (64) = 2.47, p = 0.02, Cohen’s d = 0.61 and t (61) = 2.43, p = 0.02, Cohen’s d = 0.61, respectively. At Time 3, the difference was reduced by a third in terms of effect size and was no longer significant, t (62) = 1.59, p = 0.12, Cohen’s d = 0.40, and by the end of kindergarten (Time 4), the difference was reduced by yet another third, t (62) = 1.59, p = 0.41, Cohen’s d = 0.21. Thus, the effect size of the income achievement gap went from 3/5 of a standard deviation at age 3, to 2/5 at age 4, and finally to 1/5 at the end of the 3rd year in Montessori. Within the Montessori sample, the same series of tests showed trending ( p = 0.06 at Time 1) or significant income-group differences in academic achievement at the first three time points but not at the last one, t (56) = 1.41, p = 0.16, although the difference was still a third of a standard deviation in size, Cohen’s d = 0.37. By contrast, within the control sample, the higher income subgroup performed a full standard deviation better than the lower income subgroup, Cohen’s d = 0.98. The higher income Montessori children were the highest performers in the study by the end of kindergarten (Time 4, see Figure ​ Figure5 5 ), but the lower-income children were doing much better in Montessori classrooms than in control schools by this last time point.

Outcomes for Children with Different Levels of Executive Function

Second, we examined the predictive power of executive function for achievement. For both Montessori and control children, higher executive function predicted academic achievement at Time 1 (the intercept). In the control sample, as expected from many studies, executive function also predicted the slope of academic achievement in the latent growth curve model, Δ B = -0.067, SE = 0.03, p = 0.05. By contrast, initial levels of executive function had no influence on the slope of academic achievement for children in the Montessori programs, Δ B = 0.009, SE = 0.03, p = 0.76. Thus, in terms of academic outcomes, in Montessori classrooms children with low executive function do as well as children with high executive function. In other words, special supplementary curricula targeting executive function are not needed to equalize achievement outcomes for children in Montessori programs; academic achievement was higher overall, and children with lower executive function were not at a disadvantage.

Montessori vs. Public or Private Business-As-Usual

Because control children were at both private and publically funded schools, we examined how Montessori children compared to both groups on academic achievement, theory of mind, and executive function. Controlling for academic achievement at the first time point, there was a significant school type effect on academic achievement at the final time point, F (2,122) = 3.94, p = 0.022, η p 2 = 0.06. Post hoc tests showed a significant mean difference (favoring Montessori, for all results described here) between public Montessori and public control schools ( p = 0.012) and a trend between public Montessori and private control schools ( p = 0.055). There was no difference between public and private control schools ( p = 0.42). For theory of mind, the same analyses indicated a group difference, F (2,114) = 4.30, p = 0.016, η p 2 = 0.07, which post hoc tests revealed was both between public Montessori and public control schools ( p = 0.004) and public and private control schools (favoring private, p = 0.048), but not between public Montessori and private control schools ( p = 0.40). Executive function at the final time point controlling for the first time point approached a trend on the omnibus test F (2,117) = 2.27, p = 0.11, η p 2 = 0.04 attributable to a significant difference in growth of children in public Montessori vs. in public control schools ( p = 0.04).

Assisting young children’s development is an essential societal task; the human brain undergoes tremendous development in the early school years, setting in place patterns that predict life trajectories ( Moffitt et al., 2011 ). Yet in the United States, the methods by which we try to help young children oscillate between didactic academic and pure discovery learning approaches, neither of which supports whole-child development in optimal ways ( Fisher et al., 2011 ). Montessori education takes a different, whole-child approach and could feasibly be implemented at scale, but there have been no strong studies of its outcomes.

Taking advantage of a computerized random lottery for placement in two Montessori magnet preschools, this study compared 70 preschool-aged children who attended Montessori with 71 who did not. This is to our knowledge the first study spanning three years of Montessori education, and the second Montessori study to use a lottery-loser control design; the present study had a much larger sample size, and used new measures.

Montessori education elevated all children’s performance on several measures, and made the performance of groups that typically do less well more equal. First, academic performance of children in Montessori programs was significantly stronger over time. They performed slightly (but not significantly) better at the first time point, perhaps because children had on average almost 2 months of school program experience at the first test, with some children having a full 3.5 months. By the third and fourth time point, the differences in academic achievement were significant.

Furthermore, Montessori education made substantial headway in reducing the income gap in achievement across the preschool years. Whereas lower income control children were performing a full standard deviation lower than higher income control children by the end of preschool, the difference in income groups in Montessori was just a third of a standard deviation. Statistically, the lower income Montessori children did not differ from the higher income children in either school group by the fourth time point. In keeping with this, the income-achievement correlation was significantly smaller for children in Montessori than for children in the control group. This is a very important and impressive finding in our national search for ways to better help children born at an economic disadvantage.

Importantly, the higher achievement in Montessori was not at the expense of social skills or of liking school. Children who had by lottery ended up in Montessori programs performed better on tests of social cognition, were more mastery oriented, and expressed more liking of academic tasks relative to how much they liked recreational tasks. All these variables have predicted better outcomes in other studies, cited earlier. Montessori children fared equally well on tests of social problem solving and creativity, and had better executive function at age 4.

Finally, many studies have shown better academic and life outcomes for children with higher executive function or self-control. While for the control children in this study as well, executive function predicted academic achievement, this was not the case for children in Montessori. In Montessori classrooms, having lower or higher executive function did not matter for achievement; children with lower executive function performed as well as children with higher executive function in Montessori on academic achievement, which is impressive given that academic achievement in the Montessori sample was higher overall. Next we speculate on some possible reasons for these results, considering first intrinsic program differences in outcomes, followed by the possibility that Montessori teachers are superior.

Children in Montessori programs excelled in academic achievement. The Montessori materials and presentations are one possible reason. The materials capitalize on the embodiment of cognition, for example having children trace letters as they say the letter sounds, and match cards with words to small objects. Ample research suggests that this is a more effective way to learn than sitting and listening ( Lillard, 2017 ) as children often do in conventional preschool environments ( Bassok et al., 2016 ). Furthermore, the content via which educational topics are approached in Montessori might be helpful. For example, in Montessori environments, children approach math through spatial learning, when Red Rods that systematically vary in length are transformed into Number Rods that name alternately colored segments with unit numbers ( Montessori, 1914/1965 , 1994b ). The purpose of mathematics is to measure the physical world, and spatial and math skills are correlated ( Verdine et al., 2017 ). Conventional education typically begins math education with counting discrete objects; perhaps starting with spatial relations as is done in Montessori is more helpful. In addition, the Montessori curricula and materials are very logical and very interesting (e.g., Montessori, 2016 ), and this could also be a reason for the difference. Another intrinsic program difference that could result in better learning outcomes is order. The Montessori environment and materials are also highly ordered, and more orderly environments are also associated with better cognitive and academic outcomes ( Fisher et al., 2014 ). These are just a few of many possible reasons for the stronger academic outcomes for children in Montessori classrooms.

This study aligns with two prior studies in showing that children in authentic (in this case, AMI-recognized) Montessori environments perform better on theory of mind than other children ( Lillard and Else-Quest, 2006 ; Lillard, 2012 ). One possible reason for this is that Montessori classrooms combine children of three ages. In China, under the one-child policy, children in multi-age classrooms did better on theory of mind tests than children in single-age classrooms ( Wang and Su, 2009 ). Other studies have shown that children with more older siblings also do better on theory of mind ( Ruffman et al., 1998 ; Peterson, 2000 ). These advantages are believed to stem from the need to consider others’ mental states during conflicts that arise more often with similar-aged siblings or peers ( Lillard and Eisen, 2017 ). A Montessori environment might present even more conflict than a typical preschool classroom, because there is only one of each type of Montessori material—one set of “Pink Tower” blocks, and one set of Musical Bells, for example. This scarcity in the context of 3-year age groupings might create challenges that lead to faster development in theory of mind. Alternatively, Dr. Montessori noted personality changes that accompanied deep concentration on work in preschool classrooms; one of these changes was to become more socially competent ( Montessori, 1917/1965 ), which is associated with theory of mind; note, however, that the more direct measure of social competence (Social Problem Solving) did not show differences in this study.

Children in Montessori programs were more mastery oriented by ages 4 and 5 than were children in the control sample. One possible reason for this is the lack of extrinsic rewards in Montessori programs. The reward systems used in conventional school programs tend to lead to ability-oriented theories about oneself ( Ames, 1992 ), which tend to go along with performance goals. People with performance goals tend to choose easier tasks that will make them look good ( Dweck, 1999 ). Montessori programs encourage repetition of exercises to the point of mastery, and feedback comes from the materials rather than a teacher. These differences might explain the findings obtained here with regard to mastery orientation.

Liking School Enjoyment

Although the children in this study all really liked recreational activities like watching television and movies and playing, children in Montessori showed relatively more liking of academic tasks like reading and getting lessons from a teacher. One possible reason for this is that children have choices about how they spent their time in Montessori; such choice is increasingly rare in preschool programs generally ( Bassok et al., 2016 ). People are generally happier when they have choices, which provide a sense of self-determination ( Deci and Ryan, 2011 ). Other possible reasons for more school liking dovetail with those given for achievement and mastery orientation.

Unlike some other studies ( Lillard and Else-Quest, 2006 ; Lillard, 2012 ; Kayılı, 2016 ), this study did not show significantly stronger development of executive function overall for children in Montessori; their executive function was significantly higher only at age 4. It might be that children whose parents enroll them in lottery magnets are different; this is the first study of magnet Montessori preschools. Alternatively, it might be that conventional preschools are improving in these areas because of social-emotional learning programs ( Ursache et al., 2012 ). Further research is needed to tease apart these possibilities.

The finding concerning executive function and prediction of academic achievement is notable. Many studies have shown that executive function in the early school years predicts academic achievement ( Blair and Razza, 2007 ; Duncan et al., 2007 ; Fuhs et al., 2014 ; Cameron et al., 2015 ), likely because in order for children to learn in conventional school they need to behave in ways that exercise executive function: They need to sit still, listen, follow directions, and inhibit engaging in other activities. But children across the full range of executive function who were in Montessori classrooms grew equally in academic achievement, and overall the Montessori children’s level of academic achievement was higher than that of controls. This suggests that having low executive function is not a disadvantage for children in this type of school program. Whether this translates to executive function being less predictive of later (such as Elementary school) outcomes for children who attended Montessori preschool is topic for further research.

One possible reason why executive function was not predictive of outcomes within the Montessori preschool program is that Montessori is a form of differentiated instruction. Children are not all treated alike; a child who needs more structure can be given that by the teacher. For example, a child who has not developed an ability to make constructive choices can be given limited, or even no choice, by the teacher, whereas a child who makes good choices (for example, chooses challenging work) is allowed to make their own choices. Closer examination of in-classroom processes, noting whether teachers do in fact scaffold lower executive function more effectively in Montessori programs, would shed light on this.

One might ask whether executive function near the time of school entry not predicting academic achievement is problematic. It does not seem so, since executive function still developed similarly in both groups and academic achievement was higher overall in the non-predictive group (Montessori).

Montessori Teachers

In addition to intrinsic program differences, another possible reason for better Montessori outcomes is that Montessori teachers might be better teachers; if so, perhaps children in their classrooms would excel regardless of what educational program the teachers implemented. The teachers were not the focus of study here, but future research should consider this possibility. It is notable that at one of the two schools, three of six teachers had been teaching in a conventional way prior to 2008, and opted for retraining when the school adopted a Montessori program.

Considering the possibility that the study is revealing teacher rather than program effects, we note two points at which the Montessori teachers might have become better teachers: prior to their teacher training, or during (and as a result of) the teacher training.

Possible Pre-existing Differences in Teachers

First, one might ask whether the standards for entering a program to be a Montessori teacher are higher. Most of the Montessori teachers in this study trained at the AMI teacher training center in Hartford. Up to the time of this study, the training center courses were usually undersubscribed, so the center took virtually all applicants (Hall, personal communication, June, 2017). In addition, virtually all those who take the 9-month course are awarded a diploma. However, it is feasible that people who are attracted to Montessori teacher training interact differently with children, and this difference could be responsible for the results obtained. Other studies have shown non-trivial teacher effects at preschool. For example, a large study of prekindergarten classrooms in states that support pre-K (as does Connecticut) indicated two teacher variables that are most predictive of child achievement ( Mashburn et al., 2008 ): (1) teacher emotional support, which predicts social outcomes and (2) teacher instructional support (asking high-level questions, scaffolding children’s thinking), which supports academic outcomes. It is possible that the Montessori teachers were higher on these variables even prior to their Montessori teacher training. Further research should examine this, perhaps through questionnaires given to people commencing Montessori vs. conventional teacher education programs.

Teacher Training Causing Teacher Differences

Second, the teacher training for Montessori might create better teachers. In terms of time and course intensity, the AMI training seems comparable to the training required for an early childhood teaching certificate. It involves 9 months of lectures and practice teaching, creation of a set of notes explaining Montessori theory and curriculum, and a final examination. The AMI “professors”—the people who teach the teacher-trainees—typically had at least 5 years as an AMI-certified classroom teacher followed by about 7 years of apprenticeship to another teacher trainer, so they are also highly trained. However, one difference to early childhood education is that in Montessori teacher training courses, one focuses on just one system and theory ( Cossentino, 2005 ). By contrast, teachers in conventional teacher education programs typically learn many theories and methods. Whether learning a single theory or multiple ones creates better teachers is an empirical question.

Another possibility, which also needs to be studied, is that Montessori teacher training changes teachers, perhaps by making them more sensitively responsive or higher in instructional support. If this is the case, then Montessori teachers are different but for a reason that is generic to Montessori education. Throughout Dr. Montessori’s books, a warm and loving attitude to children is expressed, and Montessori teachers are expected to come to embody this attitude ( Lillard, 2017 ). In addition, Montessori teachers adopt high expectations of children, for example expecting them to achieve independence in ways that people rarely expect at least in American culture today. Even before age 3, Montessori children are expected to set the table, prepare a meal, and clean up, for example. Five-year-olds multiply and divide 4-digit numbers [see Figure ​ Figure6 6 ; Montessori (2016) describes how this is achieved in high-fidelity Montessori classrooms], and carry out other complex tasks on their own. The combination of warmth, trust, and high expectations that is imparted to teachers during the Montessori teacher training might change them in ways that would make their students have better outcomes even if the teachers did not go on to implement a Montessori curriculum.

Two children working with Montessori decimal materials, with which preschool children perform multiplication and division of 4-digit numbers. Photograph by Laura Joyce-Hubbard, provided by courtesy of Forest Bluff School.

Various means should be used in future studies to look at the degree to which teachers might be responsible for better outcomes in Montessori education. First, one could examine attitudes toward and interactions with children prior to, during, and following teacher education courses, comparing those in Montessori and conventional training, to see how each type of teacher training changes people. Second, measures of teacher–child interaction could be used in studies like this, and entered as separate predictors in regression models, to see whether teacher interaction style in Montessori loads as or more strongly on outcomes than it does in studies of conventional teachers, for example using the CLASS ( Pianta et al., 2012 ).

Value-Added of Montessori Materials and Methods

Even if Montessori teachers differ in some ways from other teachers that cause better child outcomes, the Montessori materials and the methods with which the materials are used probably also add value. Two studies speak to this issue, both capitalizing on the fact that many Montessori classrooms do not offer exclusively Montessori materials. In one study, among 14 Montessori classrooms, children advanced more across a school year in classrooms that offered only Montessori materials than in “Montessori” classrooms that mixed in conventional materials like commercial puzzles ( Lillard, 2012 ). In another study, conventional materials were removed midyear from two of three Montessori classrooms, and children in those two classrooms experienced significantly greater gains in the subsequent 4 months than children in the third classroom ( Lillard and Heise, 2016 ). Because all the Montessori teachers in these studies were Montessori-trained, these studies suggest there might be something in the Montessori materials and the methods with which they are used that allow for steeper growth.

Limitations

A major strength of this study is also a major limitation: It is based on a lottery for admission to two oversubscribed schools. Not all lottery entrants could be located (some had moved and left no forwarding address) and not all who were contacted agreed to enroll. School lottery entrants are not representative of all children, and oversubscribed schools differ from undersubscribed ones. In the real world, lottery designs are often the best available; longitudinal lottery studies are supreme. However, a lottery study is not as good as a true randomized control trial, where everyone is randomly assigned and is made to stay in their assigned group.

Another major strength that is also a limitation is that the study used high fidelity Montessori schools. Montessori outcomes appear to depend on the quality of the Montessori program ( Lillard, 2012 ); outcomes at lower fidelity Montessori schools might not be the same. The Montessori programs in this study were recognized by the AMI, and we do not know if unrecognized Montessori schools, or ones associated with other Montessori organizations and teacher trainings, or even other AMI Montessori schools, would have similar outcomes. Another limitation is that the Montessori and control schools vary on many dimensions, and it is unclear whether specific dimensions might have contributed to outcomes, or whether Montessori programs must be fully implemented to have benefits. This study does suggest that very rigorous Montessori preschool programs significantly affect outcomes relative to business as usual, but less rigorous Montessori programs might not. Another limitation is that people who choose to become Montessori teachers might be different, and might teach more effectively regardless of program type. Ideally one could randomly assign future teachers to Montessori or conventional teacher training, but in lieu of that, other research strategies should be undertaken.

Conclusions and Future Directions

Bearing these limitations in mind, the present study offers evidence that high fidelity Montessori preschool programs are more effective than other business-as-usual school programs at elevating the performance of all children, while also equalizing outcomes for subgroups of children who typically have worse outcomes. First, Montessori programs reduced the income achievement gap, raising achievement of lower income children well beyond the levels achieved by the lower income waitlisted controls. In addition, Montessori programs appeared to work as well for children who were lower in executive function at the outset as for children who were higher in executive function at the outset. Since preschool achievement predicts later achievement ( Duncan et al., 2007 ), these benefits could feasibly extend upward, but whether they do so remains to be tested. Importantly these gains at preschool were not at the expense of “soft skills” that are the most important predictors of life outcomes ( Heckman and Kautz, 2012 ).

Widespread implementation of Montessori programs would be premature prior to further research to examine the external validity of this study. There are over 450 public schools in the United States that offer Montessori education ( National Center for Montessori in the Public Sector, 2014 ), and many of these admit by lottery. (There are also over 4000 private Montessori schools, but random lottery admission in those is unlikely). A large-scale study should examine outcomes in many more public Montessori schools, with an eye to Montessori implementation fidelity, as well as teachers and their training. The present study supports the legitimacy of such a study to determine more definitively if Montessori education should be implemented at scale.

Ethics Statement

The study was carried out in accordance with the recommendations in the guidelines for human research of the Institutional Review Board for the Social and Behavioral Sciences at the University of Virginia, which approved the study protocol. Parents or guardians provided written consent for all children’s participation in accordance with the Declaration of Helsinki.

Author Contributions

AL conceived of and obtained funding for the study, arranged with the sites, submitted initial IRBs, chose stimuli, oversaw all aspects of running, led effort in writing and statistical analyses and submissions. MH arranged for and did data collection in final study year, entered and cleaned data, maintained family contacts, assisted with analyses and writing. ER arranged for and did data collection in 5th year, entered data and maintained family contacts for 4 years. XT conducted growth curve and bootstrapping analyses as well as conceptualization of data, assisted with manuscript. AH created procedure manuals and materials sets, and maintained family contacts and data base, trained and maintained contacts with on-site RAs, and arranged for data collection visits in first several years of study. PB supervised RAs on site in Hartford, stored material sets, facilitated local contacts, provided Hartford school information, and assisted with manuscript.

The following factor model was fitted separately at each time point:

Table ​ TableA1 A1 below shows the factor loadings and fit indices with factor loadings freely estimated. All models show excellent fit (from Kenney, 2015 : for CFI, values over 0.9 are considered good; for RMSEA, 0.10 is the cut-off; for SRMR, less than 0.08 indicates good fit).

Factor loadings and fit indices for academic achievement and executive function.

A further analysis was done to determine fit with factors constrained to be equal; these results are shown in Table ​ TableA2 A2 .

Factor loadings and fit indices for academic achievement and executive function: constrained.

In this analysis, for Time 1, when factors are constrained to be equal, model fit is more than adequate by two indices (CFI and SRMR) but by the RMSEA model fit is not good initially, when children are younger and there is more error (some very young children might not understand test instructions, for example); it becomes acceptable by Times 3 and 4.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

The authors thank the children, parents, and school administrators as well as the Regional School Choice Office in Hartford for their participation; Tim Nee for facilitating the project; and Hedy L. Azarhooshang, Samantha Cusak, Theresa Heinz, Erin Kenney, Sheila Morely, Ariel Rodriguez, Carmen Trainer, and Ashley Wodzicki for collecting data.

Funding. Funding for this project was provided by the Brady Education Foundation.

• Ames C. (1992). Classrooms: goals, structures, and student motivation. J. Educ. Psychol. 84 261–271. 10.1037/0022-0663.84.3.261 [ CrossRef ] [ Google Scholar ]
• Ballou D. (2009). “Magnet school outcomes,” in Handbook of Research on School Choice , eds Berends M., Springer M. G., Ballou D. D., Walberg H. J. (New York, NY: Routledge; ), 409–426. [ Google Scholar ]
• Bassok D., Latham S., Rorem A. (2016). Is kindergarten the new first grade? AERA Open 1 1–31. 10.1177/2332858415616358 [ CrossRef ] [ Google Scholar ]
• Besançon M., Lubart T. (2008). Differences in the development of creative competencies in children schooled in diverse learning environments. Learn. Individ. Dif. 18 381–389. 10.1016/j.lindif.2007.11.009 [ CrossRef ] [ Google Scholar ]
• Blair C., Raver C. C. (2016). Poverty, stress, and brain development: new directions for prevention and intervention. Acad. Pediatr. 16 S30–S36. 10.1016/j.acap.2016.01.010 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Blair C., Razza R. P. (2007). Relating effortful control, executive function, and false belief understanding to emerging math and literacy ability in kindergarten. Child Dev. 78 647–663. 10.1111/j.1467-8624.2007.01019.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Brooks-Gunn J., Duncan G. J. (1997). The effects of poverty on children. Future Child 7 55–71. 10.2307/1602387 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Cameron C. E., Brock L. L., Hatfield B. E., Cottone E. A., Rubinstein E., LoCasale-Crouch J., et al. (2015). Visuomotor integration and inhibitory control compensate for each other in school readiness. Dev. Psychol. 51 1529–1543. 10.1037/a0039740 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Cameron C. E., Brock L. L., Murrah W. M., Bell L. H., Worzalla S. L., Grissmer D., et al. (2012). Fine motor skills and executive function both contribute to kindergarten achievement. Child Dev. 83 1229–1244. 10.1111/j.1467-8624.2012.01768.x [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Campbell F. A., Ramey C. T., Pungello E., Sparling J., Miller-Johnson S. (2002). Early childhood education: young adult outcomes from the abecedarian project. Appl. Dev. Sci. 6 42–57. 10.1037/a0026644 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Carlson S. M., Moses L. J. (2001). Individual differences in inhibitory control and children’s theory of mind. Child Dev. 72 1032–1053. 10.1037/0012-1649.40.6.1105 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Cossentino J. (2005). Ritualizing expertise: a non-montessorian view of the Montessori method. Am. J. Educ. 111 211–244. 10.1086/426838 [ CrossRef ] [ Google Scholar ]
• Deci E. L., Ryan R. M. (2011). “Self-determination theory,” in Handbook of Theories of Social Psychology Vol. 1 eds Lange P. A. M. V., Kruglanski A. W., Higgins E. T. (London: Sage; ), 416–433. [ Google Scholar ]
• Dennett D. (1987). The Intentional Stance. Cambridge, MA: MIT Press. [ Google Scholar ]
• DeVries R., Gonçu A. (1987). Interpersonal relations in four-year dyads from constructivist and Montessori programs. J. Appl. Dev. Psychol. 8 481–501. 10.1016/0193-3973(87)90035-9 [ CrossRef ] [ Google Scholar ]
• Diamond A. (2013). Executive functions. Annu. Rev. Psychol. 64 135–168. 10.1146/annurev-psych-113011-143750 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Diamond A., Barnett W., Thomas J., Munro S. (2007). Preschool program improves cognitive control. Science 318 1387–1388. 10.1126/science.1151148 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Duncan G., Dowsett C., Claessens A., Magnuson K., Huston A., Klebanov P., et al. (2007). School readiness and later achievement. Dev. Psychol. 43 1428–1446. 10.1037/0012-1649.43.6.1428 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Dweck C. S. (1999). Self-Theories: Their Role in Motivation, Personality, and Development. Philadelphia, PA: Psychology Press. [ Google Scholar ]
• Dweck C. S. (2006). Mindset: The New Psychology of Success. New York, NY: Random House. [ Google Scholar ]
• Dweck C. S. (2017). The journey to children’s mindsets—and beyond. Child Dev. Perspect. 11 139–144. 10.1111/cdep.12225 [ CrossRef ] [ Google Scholar ]
• Fisher A. V., Godwin K. E., Seltman H. (2014). Visual environment, attention allocation, and learning in young children when too much of a good thing may be bad. Psychol. Sci. 25 1362–1370. 10.1177/0956797614533801 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Fisher K., Hirsh-Pasek K., Golinkoff R., Singer D., Berk L. W. (2011). “ Playing around in school: implications for learning and education policy ,” in Oxford Handbook of the Development of Play , ed. Peligrini A. (New York, NY: Oxford University Press; ), 341–362. [ Google Scholar ]
• Fuhs M. W., Nesbitt K. T., Farran D. C., Dong N. (2014). Longitudinal associations between executive functioning and academic skills across content areas. Dev. Psychol. 50 1698–1709. 10.1037/a0036633 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Gaylord C. (2012). Maria Montessori and 10 Famous Graduates from Her Schools. The Christian Science Monitor. Avaialble at: http://www.csmonitor.com/Technology/Tech-Culture/2012/0831/Maria-Montessori-and-10-famous-graduates-from-her-schools/Google-founders-Larry-Page-and-Sergey-Brin [accessed November 10 2015] [ Google Scholar ]
• Grissmer D., Grimm K. J., Aiyer S. M., Murrah W. M., Steele J. S. (2010). Fine motor skills and early comprehension of the world: two new school readiness indicators. Dev. Psychol. 46 1008–1017. 10.1037/a0020104 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Guilford J., Christensen P. R. (1973). The one-way relation between creative potential and IQ. J. Creat. Behav. 7 247–252. 10.1002/j.2162-6057.1973.tb01096.x [ CrossRef ] [ Google Scholar ]
• Haskins R., Brooks-Gunn J. (2016). Trouble in the land of early childhood education? The Future of Children. Princeton, NJ: Brookings. [ Google Scholar ]
• Heckman J. J. (2006). Skill formation and the economics of investing in disadvantaged children. Science 312 1900–1902. 10.1126/science.1128898 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Heckman J. J., Kautz T. (2012). Hard evidence on soft skills. Labour Econ. 19 451–464. 10.1016/j.labeco.2012.05.014 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Karnes M., Shewedel A., Williams M. (1983). “A comparison of five approaches for educating young children from low-income homes,” in As the Twig is Bent: Lasting Effects of Preschool Programs , ed. Consortium for Longitudinal Studies. (Hillsdale, NJ: Lawrence Elbaum; ), 133–171. [ Google Scholar ]
• Kayılı G. (2016). The effect of Montessori method on cognitive tempo of kindergarten children. Early Child Dev. Care. 10.1080/03004430.2016.1217849 [ CrossRef ] [ Google Scholar ]
• Kenney D. A. (2015). Measuring Model Fit. Available at: http://davidakenny.net/cm/fit.htm [accessed September 13 2017]. [ Google Scholar ]
• Korkman M., Kirk U., Kemp S. (2007). Nepsy-II: Clinical and Interpretive Manual. San Antonio, TX: The Psychological Corporation. [ Google Scholar ]
• Koster-Hale J., Saxe R. (2013). “Functional neuroimaging of theory of mind,” in Understanding Other Minds: Perspectives from Developmental Social Neuroscience , 3rd Edn, eds Baron-Cohen S., Lombardo M., Tager-Flusberg H. (New York, NY: Oxford University Press; ), 132–163. [ Google Scholar ]
• Ladd G. W., Buhs E. S., Seid M. (2000). Children’s initial sentiments about kindergarten: Is school liking an antecedent of early classroom participation and achievement? Merrill Palmer Q. 46 255–279. [ Google Scholar ]
• Lillard A. S. (2012). Preschool children’s development in classic Montessori, supplemented Montessori, and conventional programs. J. Sch. Psychol. 50 379–401. 10.1016/j.jsp.2012.01.001 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Lillard A. S. (2017). Montessori: The Science behind the Genius , 3rd Edn. New York, NY: Oxford University Press. [ Google Scholar ]
• Lillard A. S., Eisen S. (2017). “Why Montessori is a facilitative environment for theory of mind: three speculations,” in Theory of Mind Development in Context , eds Slaughter V., Rosnay M. de. (London: Routledge; ), 57–70. [ Google Scholar ]
• Lillard A. S., Else-Quest N. (2006). Evaluating Montessori education. Science 313 1893–1894. 10.1126/science.1132362 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Lillard A. S., Heise M. J. (2016). Removing supplementary materials from Montessori classrooms changed child outcomes. J. Montessori Res. 2 17–27. 10.17161/jomr.v2i1.5678 [ CrossRef ] [ Google Scholar ]
• Lipsey M. W., Farran D. C., Hofer K. G. (2015). A Randomized Control Trial of a Statewide Voluntary Prekindergarten Program on Children’s Skills and Behaviors through Third Grade. Research Report Nashville, TN: Peabody Research Institute. [ Google Scholar ]
• Lipsey M. W., Nesbitt K. T., Farran D. C., Dong N., Fuhs M. W., Wilson S. J. (2017). Learning-related cognitive self-regulation measures for prekindergarten children: a comparative evaluation of the educational relevance of selected measures. J. Educ. Psychol. 10.1037/edu0000203 [ CrossRef ] [ Google Scholar ]
• Mashburn A., Pianta R., Hamre B., Downer J., Barbarin O., Bryant D., et al. (2008). Measures of classroom quality in prekindergarten and children’s development of academic, language, and social skills. Child Dev. 79 732–749. 10.1111/j.1467-8624.2008.01154.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• McAfee A. (2011). Montessori builds innovators. Harvard Bus. Rev. Available at: https://hbr.org/2011/07/montessori-builds-innovators [ Google Scholar ]
• McArdle J. J., Nesselroade J. R. (2014). Longitudinal Data Analysis Using Structural Equation Models. Washington, DC: American Psychological Association. [ Google Scholar ]
• McClelland M. M., Cameron C. E., Connor C. M., Farris C. L., Jewkes A. M., Morrison F. J. (2007). Links between behavioral regulation and preschoolers’ literacy, vocabulary, and math skills. Dev. Psychol. 43 947–959. 10.1037/0012-1649.43.4.947 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Merzenich M. M. (2001). “Cortical plasticity contributing to child development,” in Mechanisms of Cognitive Development: Behavioral and Neural Perspectives. Carnegie Mellon Symposia on Cognition , eds McClelland J. L., Siegler R. S. (Mahwah, NJ: Lawrence Erlbaum; ), 67–95. [ Google Scholar ]
• Miller L. B., Bizzell R. P. (1984). Long-term effects of four preschool programs: ninth- and tenth-grade results. Child Dev. 55 1570–1587. 10.2307/1130027 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Minervino J., Pianta R. (2014). “Early learning: the new fact base and cost sustainability,” in Lessons from Research and the Classroom , ed. Minervino J. (Washington, DC: Bill & Melinda Gates Foundation; ). [ Google Scholar ]
• Moffitt T. E., Arseneault L., Belsky D., Dickson N., Hancox R. J., Harrington H. L., et al. (2011). A gradient of childhood self-control predicts health, wealth, and public safety. Proc. Natl. Acad. Sci. U.S.A. 108 2693–2698. 10.1073/pnas.1010076108 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Montessori M. (1914/1965). Dr. Montessori’s Own Handbook. New York, NY: Schocken. [ Google Scholar ]
• Montessori M. (1917/1965). Spontaneous Activity in Education: The Advanced Montessori Method , trans. F. Simmonds New York, NY: Schocken. [ Google Scholar ]
• Montessori M. (1994a). Creative Development in the Child I , trans. R. Ramachandran Madras: Kalakshetra Press. [ Google Scholar ]
• Montessori M. (1994b). Creative Development in the Child II , trans. R. Ramachandran Madras: Kalakshetra Press. [ Google Scholar ]
• Montessori M. (2016). Psychoarithmetic , Vol. 20 Amsterdam: Montessori-Pierson Publishing. [ Google Scholar ]
• National Center for Montessori in the Public Sector (2014). 2014 Census Data Snapshot. Available at: http://www.public-montessori.org/public-montessori-censussnapshot-2014 [ Google Scholar ]
• National Early Childcare Research Network (2005). Early child care and children’s development in the primary grades: follow-up results from the NICHD study of early child care. Am. Educ. Res. J. 42 537–570. 10.3102/00028312042003537 [ CrossRef ] [ Google Scholar ]
• Peng H.-H., Md-Yunus S. (2014). Do children in Montessori schools perform better in the achievement test? A Taiwanese perspective. Int. J. Early Child. 46 299–311. 10.1007/s13158-014-0108-7 [ CrossRef ] [ Google Scholar ]
• Peterson C. C. (2000). Kindred spirits: influences of siblings’ perspectives on theory of mind. Cogn. Dev. 15 435–455. 10.1016/S0885-2014(01)00040-5 [ CrossRef ] [ Google Scholar ]
• Pianta R. C., Hamre B. K., Allen J. P. (2012). “Teacher-student relationships and engagement: conceptualizing, measuring, and improving the capacity of classroom interactions,” in Handbook of Research on Student Engagement , ed. Christensen C. M. (New York, NY: Springer; ), 365–386. [ Google Scholar ]
• Ponitz C. C., McClelland M. M., Jewkes A. M., Connor C. M., Farris C. L., Morrison F. J. (2008). Touch your toes! Developing a direct measure of behavioral regulation in early childhood. Early Child. Res. Q. 23 141–158. 10.1016/j.ecresq.2007.01.004 [ CrossRef ] [ Google Scholar ]
• Ponitz C. C., McClelland M. M., Matthews J. S., Morrison F. J. (2009). A structured observation of behavioral self-regulation and its contribution to kindergarten outcomes. Dev. Psychol. 45 605–619. 10.1037/a0015365 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Powell L. J., Carey S. (2017). Executive function depletion in children and its impact on theory of mind. Cognition 164 150–162. 10.1016/j.cognition.2017.03.022 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Rathunde K. R., Csikszentmihalyi M. (2005). Middle school students’ motivation and quality of experience: a comparison of Montessori and traditional school environments. Am. J. Educ. 111 341–371. 10.1086/428885 [ CrossRef ] [ Google Scholar ]
• Raver C. C., Jones S. M., Li-Grining C., Zhai F., Bub K., Pressler E. (2011). CSRP’s impact on low-income preschoolers’ preacademic skills: self-regulation as a mediating mechanism. Child Dev. 82 362–378. 10.1111/j.1467-8624.2010.01561.x [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Reardon S. F. (2011). “The widening academic achievement gap between the rich and the poor: new evidence and possible explanations,” in Whither Opportunity; Rising Inequality, Schools, and Children’s Life Chances , eds Duncan G., Murnane R. (New York, NY: Russell Sage Foundation; ), 91–116. [ Google Scholar ]
• Rubin K. H. (1988). The Social Problem-Solving Test-Revised. Waterloo, ON: University of Waterloo. [ Google Scholar ]
• Ruffman T., Perner J., Naito M., Parkin L., Clements W. (1998). Older (but not younger) siblings facilitate false belief understanding. Dev. Psychol. 34 161–174. 10.1037//0012-1649.34.1.161 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Schweinhart L. J., Montie J., Xiang Z., Barnett W. S., Belfield C. R., Nores M. (2005). Lifetime Effects: The High/Scope Perry Preschool Study through Age 40. Ypsilanti, MI: High/Scope. [ Google Scholar ]
• Sirin S. R. (2005). Socioeconomic status and academic achievement: a meta-analytic review of research. Rev. Educ. Res. 75 417–453. 10.3102/00346543075003417 [ CrossRef ] [ Google Scholar ]
• Smiley P. A., Dweck C. S. (1994). Individual differences in achievement goals among young children. Child Dev. 65 1723–1743. 10.2307/1131290 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Tomlinson C. A. (2014). Differentiated Classroom: Responding to the Needs of All Learners , 2nd Edn Alexandria, VA: ASCD. [ Google Scholar ]
• Ursache A., Blair C., Raver C. C. (2012). The promotion of self-regulation as a means of enhancing school readiness and early achievement in children at risk for school failure. Child Dev. Perspect. 6 122–128. 10.1111/j.1750-8606.2011.00209.x [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Verdine B., Golinkoff R. M., Hirsh-Pasek K., Newcombe N. (2017). Links between spatial and mathematical skills across the preschool years. Soc. Res. Child Dev. Monogr. 82 1–150. [ PubMed ] [ Google Scholar ]
• Vernon-Feagans L., Willoughby M., Garrett-Peters P., Project T. F. L. (2016). Predictors of behavioral regulation in kindergarten: household chaos, parenting, and early executive functions. Dev. Psychol. 52 430–441. 10.1037/dev0000087 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Wang Y., Su Y. (2009). False belief understanding: children catch it from classmates of different ages. Int. J. Behav. Dev. 33 331–337. 10.1177/0165025409104525 [ CrossRef ] [ Google Scholar ]
• Wellman H. M. (2014). Making Minds: How Theory of Mind Develops. New York, NY: Oxford University Press. [ Google Scholar ]
• Wellman H. M., Liu D. (2004). Scaling of theory-of-mind tasks. Child Dev. 75 523–541. 10.1111/j.1467-8624.2004.00691.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Willoughby M. T., Wirth R., Blair C. B. (2011). Contributions of modern measurement theory to measuring executive function in early childhood: an empirical demonstration. J. Exp. Child Psychol. 108 414–435. 10.1016/j.jecp.2010.04.007 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Woodcock R. W., McGrew K. S., Mather N. (2001). Woodcock-Johnson III Tests of Achievement. Rolling Meadows, IL: Riverside Publishing. [ Google Scholar ]
• Yoshikawa H., Aber J. L., Beardslee W. R. (2012). The effects of poverty on the mental, emotional, and behavioral health of children and youth: implications for prevention. Am. Psychol. 67 272–284. 10.1037/a0028015 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Zhang T., Meaney M. (2010). Epigenetics and the environmental regulation of the genome and its function. Annu. Rev. Psychol. 61 439–466. 10.1146/annurev.psych.60.110707.163625 [ PubMed ] [ CrossRef ] [ Google Scholar ]

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 .

•  We're Hiring!
•  Help Center

Montessori Method

• Most Cited Papers
• Most Downloaded Papers
• Newest Papers
• Save to Library
• Last »
• Montessori Education Follow Following
• Psycoanalisys Follow Following
• Children Follow Following
• Montessori Follow Following
• Queer Studies Follow Following
• Montessori system of education Follow Following
• Readiness to Primary Education Follow Following
• Preschool Education Follow Following
• Features of Authentic Montessori Follow Following
• Primary Education Follow Following

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

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

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
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 27 May 2024

Research on domain ontology construction based on the content features of online rumors

• Jianbo Zhao 1 ,
• Huailiang Liu 1 ,
• Weili Zhang 1 ,
• Tong Sun 1 ,
• Qiuyi Chen 1 ,
• Yuehai Wang 2 ,
• Jiale Cheng 2 ,
• Yan Zhuang 1 ,
• Xiaojin Zhang 1 ,
• Shanzhuang Zhang 1 ,
• Bowei Li 3 &
• Ruiyu Ding 2  

Scientific Reports volume  14 , Article number:  12134 ( 2024 ) Cite this article

175 Accesses

1 Altmetric

Metrics details

• Computational neuroscience
• Computer science
• Data acquisition
• Data integration
• Data mining
• Data processing
• Human behaviour
• Information technology
• Literature mining
• Machine learning
• Scientific data

Online rumors are widespread and difficult to identify, which bring serious harm to society and individuals. To effectively detect and govern online rumors, it is necessary to conduct in-depth semantic analysis and understand the content features of rumors. This paper proposes a TFI domain ontology construction method, which aims to achieve semantic parsing and reasoning of the rumor text content. This paper starts from the term layer, the frame layer, and the instance layer, and based on the reuse of the top-level ontology, the extraction of core literature content features, and the discovery of new concepts in the real corpus, obtains the core classes (five parent classes and 88 subclasses) of the rumor domain ontology and defines their concept hierarchy. Object properties and data properties are designed to describe relationships between entities or their features, and the instance layer is created according to the real rumor datasets. OWL language is used to encode the ontology, Protégé is used to visualize it, and SWRL rules and pellet reasoner are used to mine and verify implicit knowledge of the ontology, and judge the category of rumor text. This paper constructs a rumor domain ontology with high consistency and reliability.

Similar content being viewed by others

From rumor to genetic mutation detection with explanations: a GAN approach

Improving long COVID-related text classification: a novel end-to-end domain-adaptive paraphrasing framework

How do we study misogyny in the digital age? A systematic literature review using a computational linguistic approach

Introduction.

Online rumors are false information spread through online media, which have the characteristics of wide content 1 , hard to identify 2 , 3 . Online rumors can mislead the public, disrupt social order, damage personal and collective reputations, and pose a great challenge to the governance of internet information content. Therefore, in order to effectively detect and govern online rumors, it is necessary to conduct an in-depth semantic analysis and understanding of the rumor text content features.

The research on the content features of online rumors focuses on the lexical, syntactic and semantic features of the rumor text, including lexical, syntactic and semantic features 4 , syntactic structure and functional features 5 , source features 5 , 6 , rhetorical methods 7 , narrative structure 6 , 7 , 8 , language style 6 , 9 , 10 , corroborative means 10 , 11 and emotional features 10 , 12 , 13 , 14 , 15 , 16 , 17 , 18 . Most of the existing researches on rumor content features are feature mining under a single domain topic type, and lack of mining the influence relationship between multiple features. Therefore, this paper proposes to build an online rumor domain ontology to realize fine-grained hierarchical modeling of the relationship between rumor content features and credible verification of its effectiveness. Domain ontology is a systematic description of the objective existence in a specific discipline 19 . The construction methods mainly include TOVE method 20 , skeleton method 21 , IDEF-5 method 22 , 23 , methontology method 24 , 25 and seven-step method 26 , 27 , among which seven-step method is the most mature and widely used method at present 28 , which has strong systematicness and applicability 29 , but it does not provide quantitative indicators and methods about the quality and effect of ontology. The construction technology can be divided into the construction technology based on thesaurus conversion, the construction technology based on existing ontology reuse and the semi-automatic and automatic construction technology based on ontology engineering method 30 . The construction technology based on thesaurus conversion and the construction technology based on existing ontology reuse can save construction time and cost, and improve ontology reusability and interoperability, but there are often differences in structure, semantics and scene. Semi-automatic and automatic construction technology based on ontology engineering method The application of artificial intelligence technology can automatically extract ontology elements and structures from data sources with high efficiency and low cost, but the quality and accuracy are difficult to guarantee. Traditional domain ontology construction methods lack effective quality evaluation support, and construction technology lacks effective integration application. Therefore, this paper proposes an improved TFI network rumor domain ontology construction method based on the seven-step method. Starting from the terminology layer, the framework layer and the instance layer, it integrates the top-level ontology and core document content feature reuse technology, the bottom-up semi-automatic construction technology based on N-gram new word discovery algorithm and RoBERTa-Kmeans clustering algorithm, defines the fine-grained features of network rumor content and carries out hierarchical modeling. Using SWRL rules and pellet inference machine, the tacit knowledge of ontology is mined, and the quality of ontology validity and consistency is evaluated and verified.

The structure of this paper is as follows: Sect “ Related work ” introduces the characteristics of rumor content and the related work of domain ontology construction.; Sect “ Research method ” constructs the term layer, the frame layer and the instance layer of the domain ontology; Sect “ Domain ontology construction ” mines and verifies the implicit knowledge of the ontology based on SWRL rules and Pellet reasoner; Sect “ Ontology reasoning and validation ” points out the research limitations and future research directions; Sect “ Discussion ” summarizes the research content and contribution; Sect “ Conclusion ” summarizes the research content and contribution of this paper.

Related Work

Content features of online rumors.

The content features of online rumors refer to the adaptive description of vocabulary, syntax and semantics in rumor texts. Fu et al. 5 have made a linguistic analysis of COVID-19’s online rumors from the perspectives of pragmatics, discourse analysis and syntax, and concluded that the source of information, the specific place and time of the event, the length of the title and statement, and the emotions aroused are the important characteristics to judge the authenticity of the rumors; Zhang et al. 6 summarized the narrative theme, narrative characteristics, topic characteristics, language style and source characteristics of new media rumors; Li et al. 7 found that rumors have authoritative blessing and fear appeal in headline rhetoric, and they use news and digital headlines extensively, and the topic construction mostly uses programmed fixed structure; Yu et al. 8 analyzed and summarized the content distribution, narrative structure, topic scene construction and title characteristics of rumors in detail; Mourao et al. 9 found that the language style of rumors is significantly different from that of real texts, and rumors tend to use simpler, more emotional and more radical discourse strategies; Zhou et al. 10 analyzed the rumor text based on six analysis categories, such as content type, focus object and corroboration means, and found that the epidemic rumors were mostly “infectious” topics, with narrative expression being the most common, strong fear, and preference for exaggerated and polarized discourse style. Huang et al. 11 conducted an empirical study based on WeChat rumors, and found that the “confirmation” means of rumors include data corroboration and specific information, hot events and authoritative release; Butt et al. 12 analyzed the psycholinguistic features of rumors, and extracted four features from the rumor data set: LIWC, readability, senticnet and emotions. Zhou et al. 13 analyzed the semantic features of fake news content in theme and emotion, and found that the distribution of fake news and real news is different in theme features, and the overall mood, negative mood and anger of fake news are higher; Tan et al. 14 divided the content characteristics of rumors into content characteristics with certain emotional tendency and social characteristics that affect credibility; Damstra et al. 15 identified the elements as a consistent indicator of intentionally deceptive news content, including negative emotions causing anger or fear, lengthy sensational headlines, using informal language or swearing, etc. Lai et al. 16 put forward that emotional rumors can make the rumor audience have similar positive and negative emotions through emotional contagion; Yuan et al. 17 found that multimedia evidence form and topic shaping are important means to create rumors, which mostly convey negative emotions of fear and anger, and the provision of information sources is related to the popularity and duration of rumors; Ruan et al. 18 analyzed the content types, emotional types and discourse focus of Weibo’s rumor samples, and found that the proportion of social life rumors was the highest, and the emotional types were mainly hostile and fearful, with the focus on the general public and the personnel of the party, government and military institutions.

The forms and contents of online rumors tend to be diversified and complicated. The existing research on the content features of rumors is mostly aimed at the mining of content characteristics under specific topics, which cannot cover various types of rumor topics, and lacks fine-grained hierarchical modeling of the relationship between features and credible verification of their effectiveness.

Domain ontology construction

Domain ontology is a unified definition, standardized organization and visual representation of the concepts of knowledge in a specific domain 31 , 32 , and it is an important source of information for knowledge-based systems 19 , 33 . Theoretical methods include TOVE method 20 , skeleton method 21 , IDEF-5 method 22 , 23 , methontology method 24 , 25 and seven-step method 26 , 27 . TOVE method transforms informal description into formal ontology, which is suitable for fields that need accurate knowledge, but it is complex and time-consuming, requires high-level domain knowledge and is not easy to expand and maintain. Skeleton method forms an ontology skeleton by defining the concepts and relationships of goals, activities, resources, organizations and environment, which can be adjusted according to needs and is suitable for fields that need multi-perspective and multi-level knowledge, but it lacks formal semantics and reasoning ability. Based on this method, Ran et al. 34 constructed the ontology of idioms and allusions. IDEF5 method uses chart language and detailed description language to construct ontology, formalizes and visualizes objective knowledge, and is suitable for fields that need multi-source data and multi-participation, but it lacks a unified ontology representation language. Based on this method, Li et al. 35 constructed the business process activity ontology of military equipment maintenance support, and Song et al. 36 established the air defense and anti-missile operation process ontology. Methontology is a method close to software engineering. It systematically develops ontologies through the processes of specification, knowledge acquisition, conceptualization, integration, implementation, evaluation and document arrangement, which is suitable for fields that need multi-technology and multi-ontology integration, but it is too complicated and tedious, and requires a lot of resources and time 37 . Based on this method, Yang et al. 38 completed the ontology of emergency plan, Duan et al. 39 established the ontology of high-resolution images of rural residents, and Chen et al. 40 constructed the corpus ontology of Jiangui. Seven-step method is the most mature and widely used method at present 28 . It is systematic and applicable to construct ontology by determining its purpose, scope, terms, structure, attributes, limitations and examples 29 , but it does not provide quantitative indicators and methods about the quality and effect of ontology. Based on this method, Zhu et al. 41 constructed the disease ontology of asthma, Li et al. 42 constructed the ontology of military events, the ontology of weapons and equipment and the ontology model of battlefield environment, and Zhang et al. 43 constructed the ontology of stroke nursing field, and verified the construction results by expert consultation.

Domain ontology construction technology includes thesaurus conversion, existing ontology reuse and semi-automatic and automatic construction technology based on ontology engineering method 30 . The construction technology based on thesaurus transformation takes the existing thesaurus as the knowledge source, and transforms the concepts, terms and relationships in the thesaurus into the entities and relationships of domain ontology through certain rules and methods, which saves the time and cost of ontology construction and improves the quality and reusability of ontology. However, it is necessary to solve the structural and semantic differences between thesaurus and ontology and adjust and optimize them according to the characteristics of different fields and application scenarios. Wu et al. 44 constructed the ontology of the natural gas market according to the thesaurus of the natural gas market and the mapping of subject words to ontology, and Li et al. 45 constructed the ontology of the medical field according to the Chinese medical thesaurus. The construction technology based on existing ontology reuse uses existing ontologies or knowledge resources to generate new domain ontologies through modification, expansion, merger and mapping, which saves time and cost and improves the consistency and interoperability of ontologies, but it also needs to solve semantic differences and conflicts between ontologies. Chen et al. 46 reuse the top-level framework of scientific evidence source information ontology (SEPIO) and traditional Chinese medicine language system (TCMLS) to construct the ontology of clinical trials of traditional Chinese medicine, and Xiao et al. 47 construct the domain ontology of COVID-19 by extracting the existing ontology and the knowledge related to COVID-19 in the diagnosis and treatment guide. Semi-automatic and automatic construction technology based on ontology engineering method semi-automatically or automatically extracts the elements and structures of ontology from data sources by using natural language processing, machine learning and other technologies to realize large-scale, fast and low-cost domain ontology construction 48 , but there are technical difficulties, the quality and accuracy of knowledge extraction can not be well guaranteed, and the quality and consistency of different knowledge sources need to be considered. Suet al. 48 used regular templates and clustering algorithm to construct the ontology of port machinery, Zheng et al. 49 realized the automatic construction of mobile phone ontology through LDA and other models, Dong et al. 50 realized the automatic construction of ontology for human–machine ternary data fusion in manufacturing field, Linli et al. 51 proposed an ontology learning algorithm based on hypergraph, and Zhai et al. 52 learned from it through part-of-speech tagging, dependency syntax analysis and pattern matching.

At present, domain ontology construction methods are not easy to expand, lack of effective quality evaluation support, lack of effective integration and application of construction technology, construction divorced from reality can not guide subsequent practice, subjective ontology verification and so on. Aiming at the problems existing in the research of content characteristics and domain ontology construction of online rumors, this paper proposes an improved TFI network rumor domain ontology construction method based on seven-step method, which combines top-down existing ontology reuse technology with bottom-up semi-automatic construction technology, and establishes rumor domain ontology based on top-level ontology reuse, core document content feature extraction and new concept discovery in the real corpus from the terminology layer, framework layer and instance layer. Using Protégé as a visualization tool, the implicit knowledge mining of ontology is carried out by constructing SWRL rules to verify the semantic parsing ability and consistency of domain ontology.

Research method

This paper proposes a TFI online rumor domain ontology construction method based on the improvement of the seven-step method, which includes the term layer, the frame layer and the instance layer construction.

Term layer construction

Determine the domain and scope: the purpose of constructing the rumor domain ontology is to support the credible detection and governance of online rumors, and the domain and scope of the ontology are determined by answering questions.

Three-dimensional term set construction: investigate the top-level ontology and related core literature, complete the mapping of reusable top-level ontology and rumor content feature concept extraction semi-automatically from top to bottom; establish authoritative real rumor datasets, and complete the domain new concept discovery automatically from bottom to top; based on this, determine the term set of the domain ontology.

Frame layer construction

Define core classes and hierarchical relationships: combine the concepts of the three-dimensional rumor term set, based on the data distribution of the rumor dataset, define the parent class, summarize the subclasses, design hierarchical relationships and explain the content of each class.

Define core properties and facets of properties: in order to achieve deep semantic parsing of rumor text contents, define object properties, data properties and property facets for each category in the ontology.

Instance layer construction

Create instances: analyze the real rumor dataset, extract instance data, and add them to the corresponding concepts in the ontology.

Encode and visualize ontology: use OWL language to encode ontology, and use Protégé to visualize ontology, so that ontology can be understood and operated by computer.

Ontology verification: use SWRL rules and pellet reasoner to mine implicit knowledge of ontology, and verify its semantic parsing ability and consistency.

Ethical statements

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

Determine the professional domain and scope of the ontology description

This paper determines the domain and scope of the online rumor domain ontology by answering the following four questions:

(1) What is the domain covered by the ontology?

The “Rumor Domain Ontology” constructed in this paper only considers content features, not user features and propagation features; the data covers six rumor types of politics and military, disease prevention and treatment, social life, science and technology, nutrition and health, and others involved in China’s mainstream internet rumor-refuting websites.

(2) What is the purpose of the ontology?

To perform fine-grained hierarchical modeling of the relationships among the features of multi-domain online rumor contents, realize semantic parsing and credibility reasoning verification of rumor texts, and guide fine-grained rumor detection and governance. It can also be used as a guiding framework and constraint condition for online rumor knowledge graph construction.

(3) What kind of questions should the information in the ontology provide answers for?

To provide answers for questions such as the fine-grained rumor types of rumor instances, the valid features of rumor types, etc.

(4) Who will use the ontology in the future?

Users of online rumor detection and governance, users of online rumor knowledge graphs construction.

Three-dimensional term set construction

Domain concepts reused by top-level ontology.

As a mature and authoritative common ontology, top-level ontology can be shared and reused in a large range, providing reference and support for the construction of domain ontology. The domain ontology of online rumors established in this paper focuses on the content characteristics, mainly including the content theme, events and emotions of rumor texts. By reusing the terminology concepts in the existing top-level ontology, the terminology in the terminology set can be unified and standardized. At the same time, the top-level concept and its subclass structure can guide the framework construction of domain ontology and reduce the difficulty and cost of ontology construction. Reusable top-level ontologies include: SUMO, senticnet and ERE after screening.

SUMO ontology: a public upper-level knowledge ontology containing some general concepts and relations for describing knowledge in different domains. The partial reusable SUMO top-level concepts and subclasses selected in this paper are shown in Table 1 , which provides support for the sub-concept design of text topics in rumor domain ontology.

Senticnet: a knowledge base for concept-based sentiment analysis, which contains semantic, emotional, and polarity information related to natural language concepts. The partial reusable SenticNet top-level concepts and subclasses selected in this paper are shown in Table 2 , which provides support for the sub-concept design of text topics in rumor domain ontology.

Entities, relations, and events (ERE): a knowledge base of events and entity relations. The partial reusable ERE top-level concepts and subclasses selected in this paper are shown in Table 3 , which provides support for the sub-concept design of text elements in the rumor domain ontology.

Extracting domain concepts based on core literature content features

Domain core literature is an important source for extracting feature concepts. This paper uses ‘rumor detection’ as the search term to retrieve 274 WOS papers and 257 CNKI papers from the WOS and CNKI core literature databases. The content features of rumor texts involved in the literature samples are extracted, the repetition content features are eliminated, the core content features are screened, and the canonical naming of synonymous concepts from different literatures yields the domain concepts as shown in Table 4 . Among them, text theme, text element, text style, text feature and text rhetoric are classified as text features; emotional category, emotional appeal and rumor motive are classified as emotional characteristics; source credibility, evidence credibility and testimony method are classified as information credibility characteristics; social context is implicit.

Extracting domain concepts based on new concept discovery

This paper builds a general rumor dataset based on China’s mainstream rumor-refuting websites as data sources, and proposes a domain new concept discovery algorithm to discover domain new words in the dataset, add them to the word segmentation dictionary to improve the accuracy of word segmentation, and cluster them according to rumor type, resulting in a concept subclass dictionary based on the real rumor dataset, which provided realistic basis and data support for the conceptual design of each subclass in domain ontology.

Building a general rumor dataset

The rumor dataset constructed in this paper contains 12,472 texts, with 6236 rumors and 6236 non-rumors; the data sources are China’s mainstream internet rumor-refuting websites: 1032 from the internet rumor exposure platform of China internet joint rumor-refuting platform, 270 from today’s rumor-refuting of China internet joint rumor-refuting platform, 1852 from Tencent news Jiaozhen platform, 1744 from Baidu rumor-refuting platform, 7036 from science rumor-refuting platform, and 538 from Weibo community management center. This paper invited eight researchers to annotate the labels (rumor, non-rumor), categories (politics and military, disease prevention and treatment, social life, science and technology, nutrition and health, others) of the rumor dataset. Because data annotation is artificial and subjective, in order to ensure the effectiveness and consistency of annotation, before inviting researchers to annotate, this paper formulates annotation standards, including the screening method, trigger words and sentence break identification of rumor information and corresponding rumor information, and clearly explains and exemplifies the screening method and trigger words of rumor categories, so as to reduce the understanding differences among researchers; in view of this standard, researchers are trained in labeling to familiarize them with labeling specifications, so as to improve their labeling ability and efficiency. The method of multi-person cross-labeling is adopted when labeling, and each piece of data is independently labeled by at least two researchers. In case of conflicting labeling results, the labeling results are jointly decided by the data annotators to increase the reliability and accuracy of labeling. After labeling, multi-person cross-validation method is used to evaluate the labeling results. Each piece of data is independently verified by at least two researchers who did not participate in labeling, and conflicting labeling results are jointly decided by at least five researchers to ensure the consistency of evaluation results. Examples of the results are shown in Table 5 .

N-gram word granularity rumor text new word discovery algorithm

Existing neologism discovery algorithms are mostly based on the granularity of Chinese characters, and the time complexity of long word discovery is high and the accuracy rate is low. The algorithm’s usefulness is low, and the newly discovered words are mostly already found in general domain dictionaries. To solve these problems, this paper proposes an online rumor new word discovery algorithm based on N-gram word granularity, as shown in Fig.  1 .

Flowchart of domain new word discovery algorithm.

First, obtain the corpus to be processed \({\varvec{c}}=\{{{\varvec{s}}}_{1},{{\varvec{s}}}_{2},...,{{\varvec{s}}}_{{{\varvec{n}}}_{{\varvec{c}}}}\}\) , and perform the first preprocessing on the corpus to be processed, which includes: sentence segmentation, Chinese word segmentation and punctuation removal for the corpus to be processed. Obtain the first corpus \({{\varvec{c}}}^{{\varvec{p}}}=\{{{\varvec{s}}}_{1}^{{\varvec{p}}},{{\varvec{s}}}_{2}^{{\varvec{p}}},...,{{\varvec{s}}}_{{{\varvec{n}}}_{{\varvec{c}}}}^{{\varvec{p}}}\}\) ; where \({s}_{i}\) represents the \(i\) -th sentence in the corpus to be processed, \({n}_{c}\) represents the number of sentences in the corpus to be processed, and \({s}_{i}^{p}\) is the i-th sentence in the first corpus; perform N-gram operation on each sentence in the first corpus separately, and obtain multiple candidate words \(n=2\sim 5\) ; count the word frequency of each candidate word in the first corpus, and remove the candidate words with word frequency less than the first threshold, and obtain the first class of candidate word set;calculate the cohesion of each candidate word in the first class of candidate word set according to the following formula:

In the formula, \(P(\cdot )\) represents word frequency.Then filter according to the second threshold corresponding to N-gram operation, and obtain the second class of candidate word set; after loading the new words in the second class of candidate word set into LTP dictionary, perform the second preprocessing on the corpus to be processed \({\varvec{c}}=\{{{\varvec{s}}}_{1},{{\varvec{s}}}_{2},...,{{\varvec{s}}}_{{{\varvec{n}}}_{{\varvec{c}}}}\}\) ; and obtain the second corpus \({{\varvec{c}}}^{{\varvec{p}}\boldsymbol{^{\prime}}}=\{{{\varvec{s}}}_{1}^{{\varvec{p}}\boldsymbol{^{\prime}}},{{\varvec{s}}}_{2}^{{\varvec{p}}\boldsymbol{^{\prime}}},...,{{\varvec{s}}}_{{{\varvec{n}}}_{{\varvec{c}}}}^{{\varvec{p}}\boldsymbol{^{\prime}}}\}\) ; where the second preprocessing includes: sentence segmentation, Chinese word segmentation and stop word removal for the corpus to be processed; after obtaining the vector representation of each word in the second corpus, determine the vector representation of each new word in the second class of candidate word set; according to the vector representation of each new word, use K-means algorithm for clustering; according to the clustering results and preset classification rules, classify each new word to the corresponding domain. The examples of new words discovered are shown in Table 6 :

RoBERTa-Kmeans rumor text concepts extraction algorithm

After adding the new words obtained by the new word discovery to the LTP dictionary, the accuracy of LTP word segmentation is improved. The five types of rumor texts established in this paper are segmented by using the new LTP dictionary, and the word vectors are obtained by inputting them into the RoBERTa word embedding layer after removing the stop words. The word vectors are clustered by k-means according to rumor type to obtain the concept subclass dictionary. The main process is as follows:

(1) Word embedding layer

The RoBERTa model uses Transformer-Encode for computation, and each module contains multi-head attention mechanism, residual connection and layer normalization, feed-forward neural network. The word vectors are obtained by representing the rumor texts after accurate word segmentation through one-hot encoding, and the position encoding represents the relative or absolute position of the word in the sequence. The word embedding vectors generated by superimposing the two are used as input X. The multi-head attention mechanism uses multiple independent Attention modules to perform parallel operations on the input information, as shown in formula ( 2 ):

where \(\left\{{\varvec{Q}},{\varvec{K}},{\varvec{V}}\right\}\) is the input matrix, \({{\varvec{d}}}_{{\varvec{k}}}\) is the dimension of the input matrix. After calculation, the hidden vectors obtained after computation are residual concatenated with layer normalization, and then calculated by two fully connected layers of feed-forward neural network for input, as shown in formula ( 3 ):

where \(\left\{{{\varvec{W}}}_{{\varvec{e}}},{{\varvec{W}}}_{0}\boldsymbol{^{\prime}}\right\}\) are the weight matrices of two connected layers, \(\left\{{{\varvec{b}}}_{{\varvec{e}}},{{\varvec{b}}}_{0}\boldsymbol{^{\prime}}\right\}\) are the bias terms of two connected layers.

After calculation, a bidirectional association between word embedding vectors is established, which enables the model to learn the semantic features contained in each word embedding vector in different contexts. Through fine-tuning, the learned knowledge is transferred to the downstream clustering task.

(2) K-means clustering

Randomly select k initial points to obtain k classes, and iterate until the loss function of the clustering result is minimized. The loss function can be defined as the sum of squared errors of each sample point from its cluster center point, as shown in formula ( 4 ).

where \({x}_{i}\) represents the \(i\) sample, \({a}_{i}\) is the cluster that \({x}_{i}\) belongs to, \({u}_{{a}_{i}}\) represents the corresponding center point, \(N\) is the total number of samples.

After RoBERTa-kmeans calculation, the concept subclasses obtained are manually screened, merged repetition items, deleted invalid items, and finally obtained 79 rumor concept subclasses, including 14 politics and military subclasses, 23 disease prevention and treatment subclasses, 15 social life subclasses, 13 science and technology subclasses, and 14 nutrition and health subclasses. Some statistics are shown in Table 7 .

Each concept subclass is obtained by clustering several topic words. For example, the topic words that constitute the subclasses of body part, epidemic prevention and control, chemical drugs, etc. under the disease prevention and treatment topic are shown in Table 8 .

(3) Determining the terminology set

This paper constructs a three-dimensional rumor domain ontology terminology set based on the above three methods, and unifies the naming of the terms. Some of the terms are shown in Table 9 .

Framework layer construction

Define core classes and hierarchy, define parent classes.

This paper aims at fine-grained hierarchical modeling of the relationship between the content characteristics of multi-domain network rumors. Therefore, the top-level parent class needs to include the rumor category and the main content characteristics of a sub-category rumor design. The main content characteristics are the clustering results of domain concepts extracted based on the content characteristics of core documents, that is, rumor text feature, rumor emotional characteristic, rumor credibility and social context. The specific contents of the five top parent classes are as follows:

Rumor type: the specific classification of rumors under different subject categories; Rumor text feature, the common features of rumor texts in terms of theme, style, rhetoric, etc. Rumor emotional characteristic: the emotional elements of rumor texts, the Rumor motive of the publisher, and the emotional changes they hope to trigger in the receiver. Rumor credibility: the authority of the information source, the credibility of the evidence material provided by the publisher, and the effectiveness of the testimony method. Social context: the relevant issues and events in the society when the rumor is published.

Induce subclasses and design hierarchical relationships

In this paper, under the top-level parent class, according to the top-level concepts of top-level ontologies such as SUMO, senticnet and ERE and their subclass structures, and the rumor text features of each category extracted from the real rumor text dataset, we summarize its 88 subclasses and design the hierarchical relationships, as shown in Fig.  2 , which include:

(1) Rumor text feature

Diagram of the core classes and hierarchy of the rumor domain ontology.

① Text theme 6 , 8 , 13 , 18 , 53 : the theme or topic that the rumor text content involves. Based on the self-built rumor dataset, it is divided into politics and military 54 , involving information such as political figures, political policies, political relations, political activities, military actions, military events, strategic objectives, politics and military reviews, etc.; nutrition and health 55 , involving information such as the relationship between human health and nutrition, the nutritional components and value of food, the plan and advice for healthy eating, health problems and habits, etc.; disease prevention and treatment 10 , involving information such as the definition of disease, vaccine, treatment, prevention, data, etc.; social life 56 , involving information such as social issues, social environment, social values, cultural activities, social media, education system, etc.; science and technology 57 , involving information such as scientific research, scientific discovery, technological innovation, technological application, technological enterprise, etc.; other categories.

② Text element 15 : the structured information of the rumor text contents. It is divided into character, political character, public character, etc.; geographical position, city, region, area, etc.; event, historical event, current event, crisis event, policy event, etc.; action, protection, prevention and control, exercise, fighting, crime, eating, breeding, health preservation, rest, exercise, education, sports, social, cultural, ideological, business, economic, transportation, etc.; material, food, products (food, medicine, health products, cosmetics, etc.) and the materials they contain and their relationship with human health. effect, nutrition, health, harm, natural disaster, man-made disaster, guarantee, prevention, treatment, etc.; institution, government, enterprise, school, hospital, army, police, social group, etc.; nature, weather, astronomy, environment, agriculture, disease, etc.

③ Text style 7 , 10 : the discourse style of the rumor text contents, preferring exaggerated and emotional expression. It is divided into gossip style, creating conflict or entertainment effect; curious style, satisfying people’s curiosity and stimulation; critical style, using receivers’ stereotypes or preconceptions; lyrical style, creating resonance and influencing emotion; didactic style influencing receivers’ thought and behavior from an authoritative perspective; plain style concise objective arousing resonance etc.

④ Text feature 7 , 58 : special language means in the rumor text contents that can increase the transmission and influence of the rumor. It is divided into extensive punctuation reminding or attracting receivers’ attention; many mood words enhancing emotional color and persuasiveness; many emoji conveying attitude; induce forwarding using @ symbol etc. to induce receivers to forward etc.

⑤ Text rhetoric 15 : common rhetorical devices in rumor contents. It is divided into metaphor hyperbole repetition personification etc.

(2) Rumor emotional characteristic

① Emotion category 17 , 59 , 60 : the emotional tendency and intensity expressed in the rumor texts. It is divided into positive emotion happy praise etc.; negative emotion fear 10 anger sadness anxiety 61 dissatisfaction depression etc.; neutral emotion no preference plain objective etc.

② Emotional appeal 16 , 62 , 63 : the online rumor disseminator hopes that the rumor they disseminate can trigger some emotional changes in the receiver. It is divided into “joy” happy pleasant satisfied emotions that prompt receivers to spread or believe some rumors that are conducive to social harmony; “love” love appreciation admiration emotions that prompt receivers to spread or believe some rumors that are conducive to some people or group interests; “anger” angry annoyed dissatisfied emotions that prompt receivers to spread or believe some rumors that are anti-social or intensify conflicts; “fear” fearful afraid nervous emotions that prompt receivers to spread or believe some rumors that have bad effects deliberately exaggerated; “repugnance” disgusted nauseous emotions that prompt receivers to spread or believe some rumors that are detrimental to social harmony; “surprise” surprised shocked amazed emotions that prompt receivers to spread or believe some rumors that deliberately attract traffic exaggerated fabricated etc.

③ Rumor motive 17 , 64 , 65 , 66 : the purpose and need of the rumor publisher to publish rumors and the receiver to forward rumors. Such as profit-driven seeking fame and fortune deceiving receivers; emotional catharsis relieving dissatisfaction emotions by venting; creating panic creating social unrest and riots disrupting social order; entertainment fooling receivers seeking stimulation; information verification digging out the truth of events etc.

(3) Rumor credibility

① source credibility 7 , 17 : the degree of trustworthiness that the information source has. Such as official institutions and authoritative experts and scholars in the field with high credibility; well-known encyclopedias and large-scale civil organizations with medium credibility; small-scale civil organizations and personal hearsay personal experience with low credibility etc.

② evidence credibility 61 : the credibility of the information proof material provided by the publisher. Data support such as scientific basis based on scientific theory or method; related feature with definite research or investigation result in data support; temporal background with clear time place character event and other elements which related to the information content; the common sense of life in line with the facts and scientific common sense that are widely recognized.

③ testimony method 10 , 11 , 17 : the method to support or refute a certain point of view. Such as multimedia material expressing or fabricating content details through pictures videos audio; authority endorsement policy documents research papers etc. of authorized institutions or persons; social identity identity of social relation groups.

(4) Social context

① social issue 67 : some bad phenomena or difficulties in society such as poverty pollution corruption crime government credibility decline 68 etc.

② public attention 63 : events or topics that arouse widespread attention or discussion in the society such as sports events technological innovation food safety religious beliefs Myanmar fraud nuclear wastewater discharge etc.

③ emergency(public sentiment) 69 : some major or urgent events that suddenly occur in society such as earthquake flood public safety malignant infectious disease outbreaks etc.

(5) Rumor type

① Political and military rumor:

Political image rumor: rumors related to images closely connected to politics and military, such as countries, political figures, institutions, symbols, etc. These include positive political image smear rumor, negative political image whitewash rumor, political image fabrication and distortion rumor, etc.

Political event rumor: rumors about military and political events, such as international relations, security cooperation, military strategy, judicial trial, etc. These include positive political event smear rumor, negative political event whitewash rumor, political event fabrication and distortion rumor, etc.

② Nutrition and health rumor:

Food product rumor: rumors related to food, products (food, medicine, health products, cosmetics, etc.), the materials they contain and their association with human health. These include positive effect of food product rumor, negative effect of food product rumor, food product knowledge rumor, etc.

Living habit rumor: rumors related to habitual actions in life and their association with human health. These include positive effect of living habit rumor, negative effect of living habit rumor, living habit knowledge rumor, etc.

③ Disease prevention and treatment rumor:

Disease management rumor: rumors related to disease management and control methods that maintain and promote individual and group health. These include positive prevention and treatment rumor, negative aggravating disease rumor, disease management knowledge rumor, etc.

Disease confirmed transmission rumor: rumors about the confirmation, transmission, and immunity of epidemic diseases at the social level in terms of causes, processes, results, etc. These include local confirmed cases rumor, celebrity confirmed cases rumor, transmission mechanism rumor, etc.

Disease notification and advice rumor: rumors that fabricate or distort the statements of authorized institutions or experts in the field, and provide false policies or suggestions related to diseases. These include institutional notification rumor, expert advice rumor, etc.

④ Social life rumor:

Public figure public opinion rumor: rumors related to public figures’ opinions, actions, private lives, etc. These include positive public figure smear rumor, negative public figure whitewash rumor, public figure life exposure rumor, etc.

Social life event rumor: rumors related to events, actions, and impacts on people's social life. These include positive event sharing rumor, negative event exposure rumor, neutral event knowledge rumor, etc.

Disaster occurrence rumor: rumors related to natural disasters or man-made disasters and their subsequent developments. These include natural disaster occurrence rumor, man-made disaster occurrence rumor, etc.

⑤ Science and technology rumor:

Scientific knowledge rumor: rumors related to natural science or social science theories and knowledge. These include scientific theory rumor, scientific concept rumor, etc.

Science and technology application rumor: rumors related to the research and development and practical application of science and technology and related products. These include scientific and technological product rumor, scientific and technological information rumor, etc.

⑥ Other rumor: rumors that do not contain elements from the above categories.

Definition of core properties and facets of properties

Properties in the ontology are used to describe the relationships between entities or the characteristics of entities. Object properties are relationships that connect two entities, describing the interactions between entities; data properties represent the characteristics of entities, usually in the form of some data type. Based on the self-built rumor dataset, this paper designs object properties, data properties and facets of properties for the parent classes and subclasses of the rumor domain ontology.

Object properties

A partial set of object properties is shown in Table 10 .

Data attributes

The partial data attribute set is shown in Table 11 .

Creating instances

Based on the defined core classes and properties, this paper creates instances according to the real rumor dataset. An example is shown in Table 12 .

This paper selects the online rumor that “Lin Chi-ling was abused by her husband Kuroki Meisa, the tears of betrayal, the shadow of gambling, all shrouded her head. Even if she tried to divorce, she could not get a solution…..” as an example, and draws a structure diagram of the rumor domain ontology instance, as shown in Fig.  3 . This instance shows the seven major text features of the rumor text: text theme, text element, text style, emotion category, emotional appeal, rumor motivation, and rumor credibility, as well as the related subclass instances, laying a foundation for building a multi-source rumor domain knowledge graph.

Schematic example of the rumor domain ontology.

Encoding ontology and visualization

Encoding ontology.

This paper uses OWL language to encode the rumor domain ontology, to accurately describe the entities, concepts and their relationships, and to facilitate knowledge reasoning and semantic understanding. Classes in the rumor domain ontology are represented by the class “Class” in OWL and the hierarchical relationship is represented by subclassof. For example, in the creation of the rumor emotional characteristic class and its subclasses, the OWL code is shown in Fig.  4 :

Partial OWL codes of the rumor domain ontology.

The ontology is formalized and stored as a code file using the above OWL language, providing support for reasoning.

Ontology visualization

This paper uses protégé5.5 to visualize the rumor domain ontology, showing the hierarchical structure and relationship of the ontology parent class and its subclasses. Due to space limitations, this paper only shows the ontology parent class “RumorEmotionalFeatures” and its subclasses, as shown in Fig.  5 .

Ontology parent class “RumorEmotionalFeatures” and its subclasses.

Ontology reasoning and validation

Swrl reasoning rule construction.

SWRL reasoning rule is an ontology-based rule language that can be used to define Horn-like rules to enhance the reasoning and expressive ability of the ontology. This paper uses SWRL reasoning rules to deal with the conflict relationships between classes and between classes and instances in the rumor domain ontology, and uses pellet reasoner to deeply mine the implicit semantic relationships between classes and instances, to verify the semantic parsing ability and consistency of the rumor domain ontology.

This paper summarizes the object property features of various types of online rumors based on the self-built rumor dataset, maps the real rumor texts with the rumor domain ontology, constructs typical SWRL reasoning rules for judging 32 typical rumor types, as shown in Table 13 , and imports them into the protégé rule library, as shown in Fig.  6 . In which x, n, e, z, i, t, v, l, etc. are instances of rumor types, text theme, emotion category, effect, institution, event, action, geographical position, etc. in the ontology. HasTheme, HasEmotion, HasElement, HasSource, HasMood and HasSupport are object property relationships. Polarity value is a data property relationship.

Partial SWRL rules for the rumor domain ontology.

Implicit knowledge mining and verification based on pellet reasoner

This paper extracts corresponding instances from the rumor dataset, imports the rumor domain ontology and SWRL rule description into the pellet reasoner in the protégé software, performs implicit knowledge mining of the rumor domain ontology, judges the rumor type of the instance, and verifies the semantic parsing ability and consistency of the ontology.

Positive prevention and treatment of disease rumors are mainly based on the theme of disease prevention and treatment, usually containing products to be sold (including drugs, vaccines, equipment, etc.) and effect of disease names, claiming to have positive effects (such as prevention, cure, relief, etc.) on certain diseases or symptoms, causing positive emotions such as surprise and happiness among patients and their families, thereby achieving the purpose of selling products. The text features and emotional features of this kind of rumors are relatively clear, so this paper takes the rumor text “Hong Kong MDX Medical Group released the ‘DCV Cancer Vaccine’, which can prevent more than 12 kinds of cancers, including prostate cancer, breast cancer and lung cancer.” as an example to verify the semantic parsing ability of the rumor domain ontology. The analysis result of this instance is shown in Fig.  7 . The text theme is cancer prevention in disease prevention and treatment, the text style is plain narrative style, and the text element includes product-DCV cancer vaccine, positive effect-prevention, disease name-prostate cancer, disease name-breast cancer, disease name-lung cancer; the emotion category of this instance is a positive emotion, emotional appeal is joy, love, surprise; The motive for releasing rumors is profit-driven in selling products, the information source is Hong Kong MDX medical group, and pictures and celebrity endorsements are used as testimony method. This paper uses a pellet reasoner to reason on the parsed instance based on SWRL rules, and mines out the specific rumor type of this instance as positive prevention and treatment of disease rumor. This paper also conducted similar instance analysis and reasoning verification for other types of rumor texts, and the results show that the ontology has high consistency and reliability.

Implicit relationship between rumor instance parsing results and pellet reasoner mining.

Comparison and evaluation of ontology performance

In this paper, the constructed ontology is compared with the representative rumor index system in the field. By inviting four experts to make a comprehensive evaluation based on the self-built index system 70 , 71 , 72 , their performance in the indicators of reliability, coverage and operability is evaluated. According to the ranking order given by experts, they are given 1–4 points, and the first place in each indicator item gets four points. The average value given by three experts is taken as the single indicator score of each subject, and the total score of each indicator item is taken as the final score of the subject.

As can be seen from Table 14 , the rumor domain ontology constructed in this paper constructs a term set through three ways: reusing the existing ontology, extracting the content features of core documents and discovering new concepts based on real rumor data sets, and the ontology structure has been verified by SWRL rule reasoning of pellet inference machine, which has high reliability; ontology covers six kinds of Chinese online rumors, including the grammatical, semantic, pragmatic and social characteristics of rumor text characteristics, emotional characteristics, rumor credibility and social background, which has a high coverage; ontology is coded by OWL language specification and displayed visually on protege, which is convenient for further expansion and reuse of scholars and has high operability.

The construction method of TFI domain ontology proposed in this paper includes terminology layer, framework layer and instance layer. Compared with the traditional methods, this paper adopts three-dimensional data set construction method in terminology layer construction, investigates top-level ontology and related core documents, and completes the mapping of reusable top-level ontology from top to bottom and the concept extraction of rumor content features in existing literature research. Based on the mainstream internet rumor websites in China, the authoritative real rumor data set is established, and the new word discovery algorithm of N-gram combined with RoBERTa-Kmeans clustering algorithm is used to automatically discover new concepts in the field from bottom to top; determine the terminology set of domain ontology more comprehensively and efficiently. This paper extracts the clustering results of domain concepts based on the content characteristics of core documents in the selection of parent rumors content characteristics in the framework layer construction, that is, rumors text characteristics, rumors emotional characteristics, rumors credibility characteristics and social background characteristics; based on the emotional characteristics and the entity categories of real rumor data sets, the characteristics of rumor categories are defined. Sub-category rumor content features combine the concept of three-dimensional rumor term set and the concept distribution based on real rumor data set, define the sub-category concept and hierarchical relationship close to the real needs, and realize the fine-grained hierarchical modeling of the relationship between multi-domain network rumor content features. In this paper, OWL language is used to encode the rumor domain ontology in the instance layer construction, and SWRL rule language and Pellet inference machine are used to deal with the conflict and mine tacit knowledge, judge the fine-grained categories of rumor texts, and realize the effective quality evaluation of rumor ontology. This makes the rumor domain ontology constructed in this paper have high consistency and reliability, and can effectively analyze and reason different types of rumor texts, which enriches the knowledge system in this field and provides a solid foundation for subsequent credible rumor detection and governance.

However, the study of the text has the following limitations and deficiencies:

(1) The rumor domain ontology constructed in this paper only considers the content characteristics, but does not consider the user characteristics and communication characteristics. User characteristics and communication characteristics are important factors affecting the emergence and spread of online rumors, and the motivation and influence of rumors can be analyzed. In this paper, these factors are not included in the rumor feature system, which may limit the expressive ability and reasoning ability of the rumor ontology and fail to fully reflect the complexity and multidimensional nature of online rumors.

(2) In this paper, the mainstream Internet rumor-dispelling websites in China are taken as the data source of ontology instantiation. The data covers five rumor categories: political and military, disease prevention, social life, science and technology, and nutrition and health, and the data range is limited. And these data sources are mainly official or authoritative rumor websites, and their data volume and update frequency may not be enough to reflect the diversity and variability of online rumors, and can not fully guarantee the timeliness and comprehensiveness of rumor data.

(3) The SWRL reasoning rules used in this paper are based on manual writing, which may not cover all reasoning scenarios, and the degree of automation needs to be improved. The pellet inference engine used in this paper is an ontology inference engine based on OWL-DL, which may have some computational complexity problems and lack of advanced reasoning ability.

The following aspects can be considered for optimization and improvement in the future:

(1) This paper will introduce user characteristics into the rumor ontology, and analyze the factors that cause and accept rumors, such as social attributes, psychological state, knowledge level, beliefs and attitudes, behavioral intentions and so on. This paper will introduce the characteristics of communication, and analyze the propagation dynamic factors of various types of rumors, such as propagation path, propagation speed, propagation range, propagation period, propagation effect, etc. This paper hopes to introduce these factors into the rumor feature system, increase the breadth and depth of the rumor domain ontology, and provide more credible clues and basis for the detection, intervention and prevention of rumors.

(2) This paper will expand the data sources, collect the original rumor data directly from social media, news media, authoritative rumor dispelling institutions and other channels, and build a rumor data set with comprehensive types, diverse expressions and rich characteristics; regularly grab the latest rumor data from these data sources and update and improve the rumor data set in time; strengthen the expressive ability of rumor ontology instance layer, and provide full data support and verification for the effective application of ontology.

(3) The text will introduce GPT, LLaMA, ChantGLM and other language models, and explore the automatic generation algorithm and technology of ontology inference rules based on rumor ontology and dynamic Prompt, so as to realize more effective and intelligent rumor ontology evaluation and complex reasoning.

This paper proposed a method of constructing TFI network rumor domain ontology. Based on the concept distribution of three-dimensional term set and real rumor data set, the main features of network rumors are defined, including text features, emotional features, credibility features, social background features and category features, and the relationships among these multi-domain features are modeled in a fine-grained hierarchy, including five parent classes and 88 subcategories. At the instance level, 32 types of typical rumor category judgment and reasoning rules are constructed, and the ontology is processed by using SWRL rule language and pellet inference machine for conflict processing and tacit knowledge mining, so that the semantic analysis and reasoning of rumor text content are realized, which proves its effectiveness in dealing with complex, fuzzy and uncertain information in online rumors and provides a new perspective and tool for the interpretable analysis and processing of online rumors.

Data availability

The datasets generated during the current study are available from the corresponding author upon reasonable request.

Jiang, S. The production scene and content characteristics of scientific rumors. Youth J. https://doi.org/10.15997/j.cnki.qnjz.2020.33.011 (2020).

Article   Google Scholar  

Jin, X. & Zhao, Y. Analysis of internet rumors from the perspective of co-governance—Practice of rumor governance on wechat platform. News and Writing. 6 , 41–44 (2017).

Bai, S. Research on the causes and countermeasures of internet rumors. Press https://doi.org/10.15897/j.cnki.cn51-1046/g2.2010.04.035 (2010).

Garg, S. & Sharma, D. K. Linguistic features based framework for automatic fake news detection. Comput. Ind. Eng. 172 , 108432 (2022).

Zhao, J., Fu, C. & Kang, X. Content characteristics predict the putative authenticity of COVID-19 rumors. Front. Public Health 10 , 920103 (2022).

Article   PubMed   PubMed Central   Google Scholar  

Zhang, Z., Shu, K. & He, L. The theme and characteristics of wechat rumors. News and Writing. 1 , 60–64 (2016).

Li, B. & Yu, G. Research on the discourse space and communication field of internet rumors in the post-truth era—Based on the analysis of 4160 rumors in wechat circle of friends. Journalism Research. 2 , 103–112 (2018).

Yu, G. Text structure and expression characteristics of internet rumors—Analysis of 6000+ rumors based on tencent big data screening and identification. News and Writing. 2 , 53–59 (2018).

Mourão, R. R. & Robertson, C. T. Fake news as discursive integration: An analysis of sites that publish false, misleading, hyperpartisan and sensational information. J. Stud. 20 , 2077–2095 (2019).

Google Scholar  

Zhou, G. Analysis on the content characteristics and strategies of epidemic rumors—Based on Sina’s “novel coronavirus epidemic rumors list”. Sci. Popul. https://doi.org/10.19293/j.cnki.1673-8357.2021.05.002 (2021).

Huang, Y. An analysis of the internal logic and methods of rumor “confirmation”—An empirical study based on 60 rumors spread on wechat. J. Party Sch. Tianjin Munic. Comm. CPC 20 , 7 (2018).

Butt, S. et al . What goes on inside rumour and non-rumour tweets and their reactions: A psycholinguistic analyses. Comput. Hum. Behav. 135 , 107345 (2022).

Zhou, L., Tao, J. & Zhang, D. Does fake news in different languages tell the same story? An analysis of multi-level thematic and emotional characteristics of news about COVID-19. Inf. Syst. Front. 25 , 493–512. https://doi.org/10.1007/s10796-022-10329-7 (2023).

Article   PubMed   Google Scholar  

Tan, L. et al . Research status of deep learning methods for rumor detection. Multimed. Tools Appl. 82 , 2941–2982 (2023).

Damstra, A. et al. What does fake look like? A review of the literature on intentional deception in the news and on social media. J. Stud. 22 , 1947–1963. https://doi.org/10.1080/1461670X.2021.1979423 (2021).

Lai, S. & Tang, X. Research on the influence of information emotionality on the spread of online rumors. J. Inf. 35 , 116–121 (2016).

ADS   Google Scholar  

Yuan, H. & Xie, Y. Research on the rumor maker of internet rumors about public events—Based on the content analysis of 118 influential Internet rumors about public events. Journalist https://doi.org/10.16057/j.cnki.31-1171/g2.2015.05.008 (2015).

Ruan, Z. & Yin, L. Types and discourse focus of weibo rumors—Based on the content analysis of 307 weibo rumors. Contemporary Communication. 4 , 77–78+84 (2014).

Zhang, W. & Zhu, Q. Research on the Construction Method of Domain Ontology. Books and Information. 5 , 16–19+40 (2011).

Tham, K.D., Fox, M.S. & Gruninger, M. A cost ontology for enterprise modelling. In Proceedings of 3rd IEEE Workshop on Enabling Technologies: Infrastructure for Collaborative Enterprises. IEEE , 197–210. https://doi.org/10.1109/ENABL.1994.330502 (1994).

Uschold, M. & Gruninger, M. Ontologies: Principles, methods and applications. Knowl. Eng. Rev. 11 , 93–136 (1996).

Menzel, C. P., Mayer, R. J. & Painter, M. K. IDEF5 ontology description capture method: Concepts and formal foundations (Armstrong Laboratory, Air Force Materiel Command, Wright-Patterson Air Force, 1992).

Book   Google Scholar  

Song, Z., Zhu, F. & ZHANG, D. Research on air and missile defense domain ontology development based on IDEF5 and OWL. Journal of Projectiles, Rockets, Missiles and Guidance. 30 , 176–178 (2010).

Fernández-López, M., Gómez-Pérez, A. & Juristo, N. Methontology: From ontological art towards ontological engineering. AAAI-97 Spring Symposium Series . https://oa.upm.es/5484/ (1997).

Sawsaa, A. & Lu, J. Building information science ontology (OIS) with methontology and protégé. J. Internet Technol. Secur. Trans. 1 , 100–109 (2012).

Yue, L. & Liu, W. Comparative study on the construction methods of domain ontology at home and abroad. Inf. Stud. Theory Appl. 39 , 119–125. https://doi.org/10.16353/j.cnki.1000-7490.2016.08.024 (2016).

Noy, N.F. & McGuinness, D.L. Ontology development 101: A guide to creating your first ontology. Stanford knowledge systems laboratory technical report. KSL-01–05 (2001).

Luo, Y. et al . vim: Research on OWL-based vocabulary ontology construction method for units of measurement. Electronics 12 , 3783 (2023).

Al-Aswadi, F. N., Chan, H. Y. & Gan, K. H. Automatic ontology construction from text: A review from shallow to deep learning trend. Artif. Intell. Rev. 53 , 3901–3928 (2020).

Chen, X. & Mao, T. Ontology construction of documentary heritage—Taking China archives documentary heritage list as an example. Libr. Trib. 43 , 120–131 (2023).

CAS   Google Scholar  

Zhao, X. & Li, T. Research on the ontology construction of archives oriented to digital humanism—Taking Wanli tea ceremony archives as an example. Inf. Stud. Theory Appl. 45 , 154–161. https://doi.org/10.16353/j.cnki.1000-7490.2022.08.021 (2022).

Huang, X. et al . Construction of special knowledge base of government website pages based on domain ontology—Taking “COVID-19 vaccine science popularization” as an example. Libr. Inf. Serv. 66 , 35–46. https://doi.org/10.13266/j.issn.0252-3116.2022.17.004 (2022).

Jindal, R., Seeja, K. & Jain, S. Construction of domain ontology utilizing formal concept analysis and social media analytics. Int. J. Cogn. Comput. Eng. 1 , 62–69 (2020).

Ran, J. et al . Research on ontology construction of idioms and allusions based on OWL. Comput. Technol. Dev. 20 , 63–66 (2010).

Li, L. et al . Research on business process modeling of army equipment maintenance support based on IDEF5. Technol. Innov. Appl. 11 , 80–82 (2021).

Song, Z. et al . Ontology modeling of air defense and anti-missile operation process based on IDEF5/OWL. J. Missiles Guid. 30 , 176–178 (2010).

Li, A., Xu, Y. & Chi, Y. Summary of ontology construction and application. Inf. Stud. Theory Appl 46 , 189–195. https://doi.org/10.16353/j.cnki.1000-7490.2023.11.024 (2023).

Yang, J., Song, C. & Jin, L. Ontology construction of emergency plan based on methontology method. J. Saf. Environ. 18 , 1427–1431. https://doi.org/10.13637/j.issn.1009-6094.2018.04.033 (2018).

Duan, L. & Li, H. Ontology modeling method of high-resolution image rural residential area supported by OIA technology. Modern Agricultural Science and Technology. 2 , 338–340 (2016).

Chen, Y. & Jiang, H. Construction of fire inspection knowledge map based on GIS geospatial relationship. J. Subtrop. Resour. Environ. 18 , 109–118. https://doi.org/10.19687/j.cnki.1673-7105.2023.03.014 (2023).

Zhu, L. et al. Construction of TCM asthma domain ontology. Chin. J. Exp. Tradit. Med. Formulae 23 , 222–226. https://doi.org/10.13422/j.cnki.syfjx.2017150222 (2017).

Li, H. et al . Domain ontology construction and relational reasoning. J. Inf. Eng. Univ. 24 , 321–327 (2023).

Zhang, Y. et al. Construction of ontology of stroke nursing field based on corpus. Chin. Nurs. Res. 36 , 4186–4190 (2022).

Wu, M. et al. Ontology construction of natural gas market knowledge map. Pet. New Energy 34 , 71–76 (2022).

Li, X. et al . Research on ontology construction based on thesaurus and its semantic relationship. Inf. Sci. 36 , 83–87 (2018).

Article   ADS   CAS   Google Scholar  

Chen, Q. et al . Construction of knowledge ontology of clinical trial literature of traditional Chinese medicine. Chin. J. Exp. Tradit. Med. Formulae 29 , 190–197. https://doi.org/10.13422/j.cnki.syfjx.20231115 (2023).

Xiao, Y. et al. Construction and application of novel coronavirus domain ontology. Mil. Med. 46 , 263–268 (2022).

Su, N. et al . Automatic construction method of domain-limited ontology. Lifting the Transport Machinery. 8 , 49–57 (2023).

Zheng, S. et al . Ontology construction method for user-generated content. Inf. Sci. 37 , 43–47. https://doi.org/10.13833/j.issn.1007-7634.2019.11.007 (2019).

Dong, J., Wang, J. & Wang, Z. Ontology automatic construction method for human-machine-object ternary data fusion in manufacturing field. Control Decis. 37 , 1251–1257. https://doi.org/10.13195/j.kzyjc.2020.1298 (2022).

Zhu, L., Hua, G. & Gao, W. Mapping ontology vertices to a line using hypergraph framework. Int. J. Cogn. Comput. Eng. 1 , 1–8 (2020).

Zhai, Y. & Wang, F. Research on the construction method of Chinese domain ontology based on text mining. Inf. Sci. 33 , 3–10. https://doi.org/10.13833/j.cnki.is.2015.06.001 (2015).

Duan, Z. Generation mechanism of internet rumors and countermeasures. Guizhou Soc. Sci. https://doi.org/10.13713/j.cnki.cssci.2016.04.014 (2016).

Du, Z. & Zhi, S. The harm and governance of network political rumors. Academic Journal of Zhongzhou. 4 , 161–165 (2019).

Song, X. et al . Research on influencing factors of health rumor sharing willingness based on MOA theory. J. China Soc. Sci. Tech. Inf. 39 , 511–520 (2020).

Jiang, S. Research on the characteristics, causes and countermeasures of social rumors dissemination in china in recent years. Red Flag Manuscript . 16 , 4 (2011).

Huang, J., Wang, G. & Zhong, S. Research on the propagation law and function mode of sci-tech rumors. Journal of Information. 34 , 156–160 (2015).

Liu, Y. et al . A survey of rumor recognition in social media. Chin. J. Comput. 41 , 1536–1558 (2018).

Wei, D. et al. Public emotions and rumors spread during the covid-19 epidemic in China: Web-based correlation study. J. Med. Internet Res. 22 , e21933 (2020).

Runxi, Z. & Di, Z. A model and simulation of the emotional contagion of netizens in the process of rumor refutation. Sci. Rep. https://doi.org/10.1038/s41598-019-50770-4 (2019).

Tang, X. & Lai, S. Research on the forwarding of network health rumors in public health security incidents—Interaction between perceived risk and information credibility. J. Inf. 40 , 101–107 (2021).

Nicolas, P., Dominik, B. & Stefan, F. Emotions in online rumor diffusion. EPJ Data Sci. https://doi.org/10.1140/epjds/s13688-021-00307-5 (2021).

Deng, G. & Tang, G. Research on the spread of network rumors and its social impact. Seeker https://doi.org/10.16059/j.cnki.cn43-1008/c.2005.10.031 (2005).

Ji, Y. Research on the communication motivation of wechat rumors. Youth J. https://doi.org/10.15997/j.cnki.qnjz.2019.17.006 (2019).

Yuan, G. Analysis on the causes and motives of internet rumors in emergencies—Taking social media as an example. Media. 21 , 80–83 (2016).

Zhao, N., Li, Y. & Zhang, J. A review of the research on influencing factors and motivation mechanism of rumor spread. J. Psychol. Sci. 36 , 965–970. https://doi.org/10.16719/j.cnki.1671-6981.2013.04.015 (2013).

Article   CAS   Google Scholar  

Hu, H. On the formation mechanism of social rumors from the perspective of “rumors and salt storm”. J. Henan Univ. 52 , 63–68 (2012).

Yue, Y. et al. Trust in government buffers the negative effect of rumor exposure on people’s emotions. Curr. Psychol. 42 , 23917–23930 (2023).

Wang, C. & Hou, X. Analysis of rumor discourse in major emergencies. J. Commun. 19 , 34–38 (2012).

Xu, L. Research progress of ontology evaluation. J. China Soc. Scie. Tech. Inf. 35 , 772–784 (2016).

Lantow, B. & Sandkuhl, K. An analysis of applicability using quality metrics for ontologies on ontology design patterns. Intell. Syst. Acc. Financ. Manag. 22 , 81–99 (2015).

Pak, J. & Zhou, L. A framework for ontology evaluationIn. Exploring the Grand Challenges for Next Generation E-Business: 8th Workshop on E-Business, WEB 2009, Phoenix, AZ, USA, December 15, 2009, Revised Selected Papers 8. , 10–18. https://doi.org/10.1007/978-3-642-17449-0_2 (Springer Berlin Heidelberg, 2011).

Download references

Acknowledgements

This study was financially supported by Xi'an Major Scientific and Technological Achievements Transformation and Industrialization Project (20KYPT0003-10).

This work was supported by Xi’an Municipal Bureau of Science and Technology, 20KYPT0003-10.

Author information

Authors and affiliations.

School of Economics and Management, Xidian University, 266 Xifeng Road, Xi’an, 710071, China

Jianbo Zhao, Huailiang Liu, Weili Zhang, Tong Sun, Qiuyi Chen, Yan Zhuang, Xiaojin Zhang & Shanzhuang Zhang

School of Artificial Intelligence, Xidian University, 266 Xifeng Road, Xi’an, 710071, China

Yuehai Wang, Jiale Cheng & Ruiyu Ding

School of Telecommunications Engineering, Xidian University, 266 Xifeng Road, Xi’an, 710071, China

You can also search for this author in PubMed   Google Scholar

Contributions

H.L. formulated the overall research strategy and guided the work. J.Z kept the original data on which the paper was based and verified whether the charts and conclusions accurately reflected the collected data. J.Z. W.Z. and T.S. wrote the main manuscript text. W.Z. Y.W. and Q.C. finished collecting and sorting out the data. J.C. Y.Z. and X.Z. prepared Figs.  1 – 7 , S.Z. B.L. and R.D. prepared Tables 1 – 14 . All authors reviewed the manuscript.

Corresponding author

Correspondence to Jianbo Zhao .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher's note.

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

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.

Zhao, J., Liu, H., Zhang, W. et al. Research on domain ontology construction based on the content features of online rumors. Sci Rep 14 , 12134 (2024). https://doi.org/10.1038/s41598-024-62459-4

Download citation

Received : 07 December 2023

Accepted : 16 May 2024

Published : 27 May 2024

DOI : https://doi.org/10.1038/s41598-024-62459-4

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

• Rumor content features
• Domain ontology
• Top-level ontology reuse
• New concept discovery

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

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

Sign up for the Nature Briefing: AI and Robotics newsletter — what matters in AI and robotics research, free to your inbox weekly.