Measuring Theory of Mind in Children. Psychometric Properties of the ToM Storybooks

We tested 324 typically developing children that came from preschools, kindergartens and elementary schools. All children had a Dutch linguistic background, and did not have any language acquisition problems that could have hampered their performance on the ToM tasks (for the effect of language on ToM performance see for instance Garfield et al. 2001; Lohmann and Tomasello 2003). Two Dutch language tests were used, depending on the age of the child. For 3–6 year olds, the Reynell was administered (test for receptive language comprehension; Van Eldik et al. 1997); and for 6–9 year olds, the TvK (Taaltest voor Kinderen, Language Test for Children; Van Bon 1982) was used (subtests ‘vocabulary’ and ‘sentence construction’). Language scores were available for 249 children (Reynell: n = 170, TvK: n = 79). Those children who did not receive a language test were older than 6 years and judged as having appropriate language skills by their teachers.

The sample consisted of 157 girls and 167 boys. The ages ranged from three up to and including 11 years (see Table 1 for the age distribution). Because the most rapid changes in ToM occur before the age of 5 years, there is an overrepresentation of young children.¹

The test takes 40–50 min, including a short break. The child sits at the left side of the administrator, so it can see the drawings clearly (the drawings are on the left side of the book, while the accompanying text for the experimenter can be found on the right side). The drawings remain in front of the child during the questioning in order to prevent mistakes due to memory requirements (in agreement with Charman et al. 2001).

The ToM Storybooks have been developed with the aim of providing practitioners and researchers with a comprehensive ToM test assessing different basic ToM components and associated aspects. The test was administered to typically developing children.

The test consists of 34 tasks divided over six storybooks. It holds 74 test questions and 18 justification questions, resulting in a maximum total score of 110. The ToM sumscore can be considered a good estimation of ToM ability, as the OPLM results illustrated. Weighted values are not required. Also, a ToM quotient score can be calculated.

As far as typically developing children are concerned, the test focuses on the age range between three and six, given the rapid developments of ToM that occur in this period. However, the test has standardized norms and is applicable up to the age of 12. As a result, an 11-year-old child with ToM problems can be compared to a typically developing 5-year-old but also to a typically developing 11-year-old. Thus, the test is particularly suited for studies requiring age comparisons, based on the same instrument (e.g. cross-sectional research, assessment of clinical populations at various ages).

The analyses were based on the 324 typically developing children from Study 1. We calculated composite variables for the ToM Storybooks: for the different tasks, means were calculated over theoretically similar items. This resulted in 21 composite variables (between brackets are the number of test questions + number of justification questions) (for example of the tasks see Appendix A): emotion recognition (5 + 0) and emotion naming (5 + 2) (parts from the emotion recognition tasks); desire action (3 + 1), desire emotion recognition (5 + 1), and desire emotion naming (5 + 0) (parts from the desire tasks); standard belief emotion recognition (2 + 0), standard belief emotion naming (2 + 1), standard belief action (3 + 1), changed belief action (1 + 1), not own belief action (1 + 1), not belief action (2 + 1), inferred belief (control) action (4 + 0), and (explicit) FB action (5 + 2) (parts from the belief tasks); mental physical senses (8 + 2), mental physical others (4 + 1), mental physical future (4 + 1), real imaginary (7 + 0), close impostor senses (4 + 0), close impostor others (2 + 1), and close impostor future (2 + 1) (parts from tasks aimed at the distinction between physical and mental entities); and finally, the variable seeing-is-knowing (1 + 1).

The scores of the three age groups (group 1: n = 87, 3–4.5 years old; group 2: n = 119, 4.5–6.5 years old; and group 3: n = 118; 6.5–11 years old) were analyzed using a simultaneous component analysis with equal pattern (SCA-P; Kiers and Ten Berge 1994). SCA-P, which is a variant of principal component analysis, estimates one pattern matrix for all three groups. As a result, the interpretation of the components (or factors) is equal for all groups, but the correlations between components and standard deviations of the component scores can differ across groups. To determine the number of components, the scree-test (Cattell 1966), the eigenvalue-greater-than-one rule (Kaiser 1960), and the substantive meaning of components, was used. Only minimum loadings of 0.400 were considered. Finally, composite variables had to show adequate specificity for their components. Subsequently, internal consistency reliability was calculated for the components found.

The scree plot of the SCA-P did not give a clear indication for five components. The eigenvalue-greater-than-one rule indicated that seven components should be retained. We established the number of components on the basis of the substantive content of the components, determining whether increasing the number of factors still allowed the items of a factor to measure a clinical concept. A solution consisting of five components provided the best interpretation. This solution accounted for 53.8% of the variance. The pattern matrix was rotated using the oblique Promax rotation criterion. The resulting structure matrix revealed a reasonably simple structure of five components (see Table 5) component 1 = belief action; component 2 = emotion recognition; component 3 = mental physical; component 4 = belief emotion; and component 5 = desire emotion. Two composite variables (from the original 21 formulated) also had loadings on other components, not being entirely specific, namely the composite variables ‘mental physical senses’ and ‘close impostor future’. Two other composite variables did not fit this structure, namely desire action and seeing-is-knowing. The correlations between the components varied from 0.248 to 0.454 (see Table 5).

Cronbach alphas, corrected for age, for these five components were calculated: component 1 (10 items, α = 0.79), component 2 (4 items, α = 0.47), component 3 (9 items, α = 0.80), component 4 (25 items, α = 0.62) and component 5 (14 items, α = 0.61). Since the scores on the justification items depended on the child’s answer on the related dichotomous items, justification items were not included in the calculation of the alphas, in order to avoid artifacts.

To assess the degree of differentiation in the three age groups, inter-factor correlations between the five components were computed within the three age groups. The correlations between the components were largest in the youngest group (average correlations, standard deviations of the correlations: M = 0.47, SD = 0.08), and comparable in the intermediate age group (M = 0.26, SD = 0.013) and in the eldest group (M = 0.26, SD = 0.15).

The component analysis resulted in a structure that largely corresponds with the underlying theoretical constructs from the test. The five components appeal to the five subtypes of abilities named in Appendix A) except for the composite variable ‘seeing leads to knowing’ which did not appear as a separate component. This is not surprising, since the composite variable consists of too few questions (only two). The internal consistency reliability is satisfying (although some Cronbach alphas are not >0.70), since it concerns alphas on subparts of the test each containing a limited number of items, that are also corrected for age. The inter-factor correlations are consistent with the expectations: they are high in the youngest children implying that ToM abilities are not (yet) differentiated.

For the internal test consistency, the data of the 324 typically developing children from Study 1 were used. For the test–retest reliability, a subgroup of 45 typically developing children (age 3–7) was tested again, with the second administration occurring 2–3 weeks later. We presume that ToM ability remains relatively constant when reassessed after such a short period. The test–retest reliability was also measured for children with PDD-NOS (n = 18; age 5–9) (a subgroup from the clinical group that will be presented in Study 4, see Table 7), with the second administration after 1 week. In order to determine the inter-rater reliability of the justifications, the test results of a subsample of 10 children were randomly chosen from both research groups (n = 10 typically developing children and n = 10 children with PDD-NOS). For the analysis of possible diminishing test performance at the end of the test, the data of the 324 typically developing children was used.

The internal consistency was established by means of a Cronbach’s alpha. The test–retest reliability was established by means of a Pearson product-moment correlation coefficient. The inter-rater reliability was calculated on the basis of Cohen’s kappa’s. Five independent raters scored the justifications and the correlations between these five raters were calculated. This was done in two manners: a flexible manner by points awarded to the justifications (2, 1 or 0 points) (compare Appendix B) and a stringent manner by justification category chosen (compare Appendix C). To examine whether the test scores were affected by nuisance factors such as fatigue or boredom, it was checked if the results over the various storybooks showed a significant decline. Books 2–5 were considered, because they have a similar item structure (see Study 1). Since children could choose the order of the books, the average total score of the actual presentation of those books were compared. If nuisance played a part, the last presented book should result in a lower score than the first presented book.

The internal consistency of the ToM Storybooks was good. After correction for the influence of age, Cronbach’s alpha for the dichotomous items was 0.90. The test–retest reliability for the typically developing children was good (M ₁ ToM-total score = 59.91, SD1 = 18.46 versus M ₂ ToM-total score = 66.76, SD2 = 19.73; r = .86, p < 0.001). The children’s scores rose significantly on the second administration (M = 6.84, SD = 10.33; paired samples t-test, p < 0.001). The test–retest reliability for the children with PDD-NOS was also good (M ₁ ToM-total score = 80.22, SD1 = 14.37 versus M ₂ ToM-total score = 79.67, SD2 = 15.67; r = 0.98, no significant difference). The inter-rater reliability was high (Cohen’s Kappa = 0.97–0.99 for the 0–2 points awarded, 0.81–0.97 for the 21 categories). Concerning nuisance effects, no statistically significant decrease in total scores per book were found during the test administration; this applied for the total group as well as for the three separate age groups separately (for test performance from beginning to end of testing, see Table 6).

Based on the minimum standard for reliability of 0.70 (Nunnally and Bernstein 1994, p. 265), the internal consistency (Cronbach’s α = 0.90) of the total score was good (0.90). This is an adequate value for a test aimed at young children and is consistent with findings from comparable research on standard and complex FB tasks (Hughes et al. 1999, 2000 obtained alphas of 0.83–0.84; Muris et al. 1999 obtained alphas of 0.84–0.92) and suggests that the different tasks measure the same underlying construct. Also, the test–retest reliability is good, both in typically developing children (r = 0.86) and in children with PDD-NOS (r = 0.98). This is consistent with findings from comparable research (r = 0.77: Hughes et al. 2000; r = 0.88: Muris et al. 1999). However, a significant increase in ToM total scores was found at the second measurement in typically developing children. Such a rise is not surprising, since it can be expected that young children learn from being tested (Grigorenko and Sternberg 1998). The average score rise (M = 6.86, SD = 10.33) is of the same magnitude as those obtained with most standard psychometric measures on cognitive skills for young children. For instance, a difference of six IQ points can also be found in test–retest research with intelligence tests (e.g. Tellegen et al. 2003). A similar observation has also been reported in ToM research in typically developing children (Muris et al. 1999). The children with PDD-NOS did not show such a rise. They seemed not to have learned from their former experience. This finding may form an important point of attention in evaluating children with suspected ToM problems.

The inter-rater reliability of scoring the justifications is high (Cohen’s Kappa >0.80, namely 0.81–0.97, even concerning the more stringent scoring criterion) (see also Charman et al. 2001; Muris et al. 1999). There were no differences in difficulty in judging the justifications of typically developing children versus children with PDD-NOS. There was also no evidence for a statistically significant negative effect on the test scores due to increasing fatigue or boredom during the test administration.

Children were referred to an outpatient clinic for child and adolescent psychiatry. After an extensive psycho-diagnostic and psychiatric examination (which included parent interviews and play contacts with the child), the children were diagnosed as having PDD-NOS (pervasive developmental disorder not otherwise specified) according to DSM-IV criteria (APA 1994).

The clinical group consisted of 30 children with PDD-NOS. Their ages ranged from four up to and including 8 years (see Table 7). There were 24 boys and 6 girls, resulting in a sex ratio of 4–1, which is the average sex ratio found in children with autism (compare Yeargin-Allsopp et al. 2003).

In order to check the validity of the clinical diagnosis, two additional tests were administered: the Vineland adaptive behavior scales (VABS) (Sparrow et al. 1984; Dutch version: Researchgroup Developmental Disorders, State University Leiden, 1995) and the Children’s Social Behavior Questionnaire (CSBQ; Luteijn et al. 2000; Dutch version: VISK; Luteijn et al. 2002; Hartman et al. 2007). The VABS is an interview in which parents are questioned about the actual social behavior and skills of their child. We used parts of the expanded form of the VABS. For each child the discrepancy between the Vineland age equivalent (in months) and the chronological age (in months) was computed (VA-CA). The results showed that these children had large and negative discrepancy scores in receptive language, playing skills, interpersonal relationships and coping skills (see also Serra et al. 2002) as can be expected in children with pervasive developmental disorders. Their problems with expressive language and daily living skills (community) were less profound (Paul et al. 2004) (compare Table 7). The parents also filled in the CBSQ. This is a questionnaire in which parents report autism-related behavior. It can be used to facilitate selection of PDD samples for research purposes (Hartman et al. 2006). The CSBQ scores of our group are comparable to those known for children with HFA and PDD-NOS (compare Table V in Hartman et al. 2006) (total score, M = 48, SD = 18; ‘tuned’, M = 10, SD = 5; ‘social’, M = 13, SD = 5; ‘orientation’, M = 8, SD = 3; ‘understanding’, M = 5, SD = 3; ‘stereotyped behavior’, M = 6, SD = 3; ‘change, M = 2, SD = 2).

Next, all children participated in an extensive psychological examination which included the assessment of intelligence (Wechsler Intelligence Scale for Children-Revised: Wechsler 1974; Dutch version, 1986) and the level of language comprehension. Concerning the latter, two Dutch language tests were used, depending on the age of the child. For 3–6 year olds, the Reynell was administered (test for receptive language comprehension; Van Eldik et al. 1997); and for 6–9 year olds, the TvK (Taaltest voor Kinderen, Language Test for Children; Van Bon 1982) was used (subtests ‘vocabulary’ and ‘sentence construction’).

Concerning convergent validity, two additional questionnaires and one test were included. The CSBQ (Luteijn et al. 2000) measures, among other things, ToM related knowledge, namely in the subscale ‘difficulties in understanding social information’. The VABS questionnaire (Vineland adaptive behavior scales questionnaire; Frith et al. 1994; Dutch translation: Hoogewys et al. 1999) consists of 32 theoretically derived items aimed at discriminating between social behaviors for which mentalizing (ToM) is essential (Interactive Sociability Scale, abbreviated as IS scale) or not (Active Sociability Scale, abbreviated as AS scale, concerning social behaviors that can be acquired without mentalizing). Both CSBQ and VABS questionnaire were administered for the clinical group (n = 30 PDD-NOS, 4–8 years). Also a second ToM instrument was administered, namely the Tom-Test, a Dutch test that questions a wide variety of ToM aspects (Steerneman et al. 2002; see also Muris et al. 1999). In contrast with the ToM Storybooks, it also includes second-order-belief tasks. From the 30 children with PDD-NOS, 23 received the Tom-Test (age 4–8).

There were also four groups of typically developing children involved in Study 4. The first group is a subsample of 30 control children drawn from the 324 typically developing children from Study 1. They were matched on age and gender with the PDD-NOS group. This control group was used to make comparisons with the clinical group. The second is a subsample of 249 typically developing children (drawn from the group of typically developing children in Study 1; 3–9 years). For these children, language scores were available (Reynell: n = 170, TvK: n = 79; 59 boys and 48 girls). This control group was used to explore the relationship of ToM scores and language scores in typically developing children. The third is a subsample of 107 typically developing children (drawn from the 324 typically developing children in Study 1; 3–7 years). For these children, intelligence scores were available. They received a nonverbal intelligence test. Depending on the age of the child this consisted of the SON-R 2½-7 years (Snijders-Oomen Nonverbal intelligence scale: Tellegen et al. 1998) or the SON-R 5½-17 years (Snijders-Oomen Nonverbal intelligence scale—Revised: Snijders et al. 1988).³ This control group was used to explore the relationship between ToM scores and IQ scores in typically developing children. The fourth group is a subsample of 106 typically developing children (drawn from the 324 typically developing children in Study 1, 3–8 years; 54 boys and 52 girls). For these children, VABS questionnaire scores were available. This control group was used to explore the relationship between ToM Scores and VABS questionnaire scores in typically developing children.

With regard to convergent validity, we calculated Pearson product-moment correlations between the ToM Total score and the CSBQ subscales, the VABS questionnaire and the Tom-Test. Divergent validity was tested by comparing the ToM quotient scores with language scores and IQ scores by calculating Pearson product-moment correlations.

The ToM scores of children with PDD-NOS are significantly lower than those of the matched control children (ToM total score: M = 67.60, SD = 18.23 versus M = 77.23, SD = 15.24, p = 0.001, one-tailed; ToM-Q score: M = 85.10, SD = 21.28 versus M = 101.09, SD = 13.79, p < 0.001, one-tailed). They had significantly lower scores on the mental physical tasks, the belief-action tasks, the belief-emotion tasks and the desire-action tasks. No significant differences were found for the emotion-recognition tasks and the desire-emotion tasks (see Table 8).

The correlations of the ToM Storybooks with other tests can be found in Table 9. The correlations of the ToM total score with the CSBQ subscales in children with PDD-NOS were negative and significant (p = 0.01, one-tailed): subscale 1 ‘not optimally tuned to the social situation’ (r = −0.26), subscale 2 ‘reduced contact and social interest’(r = −0.26), subscale 3 ‘orientation problems in time, place, or activity’ (r = −0.60), subscale 4 ‘difficulties in understanding social information’ (r = −0.47), subscale 5 ‘stereotyped behavior’ (r = −0.39), and subscale 6 ‘fear of and resistance to changes’ (r = −0.41). The lower children with PDD-NOS scored on the ToM Storybooks, the more problems they exhibited on the CSBQ-subscales.

The correlations of the ToM Storybooks with the VABS questionnaire subscores are significant for the IS scale (r = 0.19 and r = 0.35, for, respectively, typically developing children and children with PDD-NOS, p = 0.01, one-tailed) and show a trend for the AS scale (p = 0.06, one-tailed, for both typically developing children and children with PDD-NOS). Thus, a higher ToM score implies higher sociability.

The correlation of the ToM Storybooks with the Tom-Test is high (M = 47.09, SD = 9.74 versus M = 87.39, SD = 11.36, scores of, respectively, Tom-Test and ToM Storybooks, r = 0.79, p < 0.001, tested two-tailed). Children with PDD-NOS evidence ToM problems on both the ToM Storybooks and the Tom test.

The correlation with language comprehension in typically developing children varies for the different language tests from 0.43 to 0.47 (p ≤ 0.001, tested two tailed; a common variance of 18–22%) (see Table 9). Concerning IQ, in typically developing children only a performance IQ was obtained. The correlation with ToM-Q was 0.47 (p = 0.001, tested two-tailed; a common variance of 22%); while in children with PDD-NOS, there was no significant correlation with performance IQ (r = 0.07). However, the correlation of their verbal IQ with ToM-Q was 0.41 (p < 0.05, tested one-tailed; a common variance of 17%).

The results show that children with PDD-NOS evidence ToM problems. Children with PDD-NOS have problems with beliefs, both in predicting behaviors and emotions. In addition, they have problems on emotion recognition, real-imaginary, real-mental, close impostor, and desire-action tasks. These findings largely agree with the findings from Serra et al. (2002). Despite differences in p-values, the findings from both studies coincide. The only contrary finding is that beliefs used to predict actions were significantly more difficult for children with PDD-NOS than for typically developing children, whereas Serra and colleagues found the opposite. The finding from the present study, however, is more consistent with clinical expectations.

The construct validity of the ToM Storybooks is good, both for the convergent and the divergent validity. Concerning the convergent validity, substantial correlations with ToM-related tests were found. The correlations of the ToM total score with the CSBQ subscales were good. Average negative correlations were found with all subscales. The highest correlations were found for the subscale ‘difficulties in understanding social information’, which can be perceived of as related to ToM skills, and for the subscale ‘orientation problems in time, place, or activity’, which can be perceived of as related to executive functions. It is known that executive functions are somehow linked with ToM development (e.g. Carlson et al. 2002).

The results from the ToM Storybooks also correlated with the VABS supplementary items from Frith et al. (1994). We found significant correlations with the IS Scale (requiring ToM) and a trend for the AS Scale (not requiring ToM) with the ToM Storybooks, for both the typically developing group and the PDD-NOS group. Our results agree to a large extent with the results of Frith et al. (1994), except that the latter found significant differences for the AS only in the normal control group and for the IS only in the autistic group. Purely speculatively, the differences in results can be due to the restricted use of FB measurements (Smarties test and Three Boxes test instead of a comprehensive ToM test) in a more seriously affected group (children with an autistic disorder compared to children with PDD-NOS in our research).

With regard to the convergent validity, the correlation of the ToM Storybooks with the Tom-Test of Steerneman et al. (2002) is as expected. The ToM Storybooks test also has adequate discriminant validity. It can distinguish children with a normal ToM development from children with ToM problems, such as children with PDD-NOS. For future research, examining the applicability and discriminatory power of the ToM Storybooks, it is recommended to include children with an autistic disorder and other clinical groups, like for instance children with ADHD.

Finally, the correlations between children’s scores on the ToM Storybooks and language acquisition tests are high (>0.40). Also, correlations with IQ scores were inspected. The verbal IQ results of the children with PDD-NOS were somewhat lower than the normal sample, which is often seen in subsamples of children with autism (compare Joseph et al. 2002; Kraijer 2004; Siegel et al. 1996). As regards the correlations with IQ scores, our research showed a significant correlation with PIQ for the typically developing group (compare Muris et al. 1997, 1999; Carlson et al. 2002), but not for the PDD-NOS group. The latter group showed significant correlations with VIQ, consistent with findings of other researchers (r(230) = 0.43 in typically developing children, p < 0.001 in Hughes et al. 1999; r(52) = 0.61 in typically developing children and children with PDD-NOS, p < 0.001 in Muris et al. 1999). Final conclusions on the differences between these two groups cannot be drawn since different IQ tests were used. Due to the age limitations of IQ tests, different tests were used for the children with PDD-NOS in comparison with typically developing children. Moreover the PDD-NOS group was much smaller. As concerns future research on the relationship between IQ and ToM, the present authors recommend the use of a comprehensive ToM instrument, as was done in the present study and in studies of Hughes and colleagues, and Muris and colleagues.

Correlations between ToM Storybooks and IQ scores were notably high. However, this is not surprising. One could say that, if we look at intelligence in a broad way, comprehensive ToM instruments measure a specific aspect of intelligence, namely a kind of social intelligence. After all, these tests look into the logical reasoning of people and correlations of one type of intelligence with another are highly common. In addition, intelligence contributes to acquiring ToM skills, making it possible for children to understand connections between causes and results. In that view, comparison with IQ should perhaps not be considered as a test for divergent validity.

The reliance of this kind of task on language comprehension with this kind of population, may lead to potential complications. Children with weak language comprehension undoubtedly will have more problems with successfully completing the test. The literature shows that there are strong relationships between language and ToM (Astington and Baird 2004; Astington and Jenkins 1999; de Villiers 2000; Tager-Flusberg 2000). In addition, many children with autism have language problems. In people with autism, ToM results are correlated to verbal mental age (Frith et al. 1991; Prior et al. 1990) and verbal skills (Happé 1995). However, early research has shown that language problems do not contribute to mental state impairment, because children with for instance semantic language impairment do not show such problems (Leslie and Frith 1988; Perner et al. 1989). On the other hand, the influence of language on ToM development should not be underestimated (Ruffman et al. 2003; Sparrevohn and Howie 1995), also in testing. Language is a medium through which children learn about beliefs (Astington 2001). Reading storybooks, for instance, form a rich source of mentalizing information for children (Dyer et al. 2000).

The test includes a wider range of ToM aspects commonly tested. It includes not only tasks on first-order beliefs and desires, but also tasks on associated aspects such as the distinction between mental and physical entities. It is a comprehensive test consisting of tasks with different developmental challenges. The primary advantage of this test over existing batteries is that it targets skills that develop in typically developing children prior to the age of five, and further refine and increase during the early school years. The test, however, is applicable beyond the age of five; it has norm scores up to the age of 12 years and thus allows for comparisons between children of widely varying age, which makes it particularly appropriate for comparison with clinical groups in which ToM development is delayed. As a consequence, this test may have potential for a range of applications to both fundamental and applied work. Moreover, since this study covers a wider age range than is normally included in ToM research, valid comparisons between older children with ToM problems and their age mates with normal ToM functioning can be made. We like to remark that the older age group included in this study is not intended for discrimination between typically developing children, but between older children with clinical diagnosis. Since older typically developing children have, as a group, a smaller range in ToM total scores, a lower ToM score on these simple ToM tasks is very informative. Because of the use of simple ToM tasks and a motivating storyline, the test might also be useful in the field of intellectual disability, where autism spectrum disorders and related ToM problems are common. However, for future research it is advisable to include more complicated ToM tasks, such as a second-order belief task (see for instance Hughes et al. 2000) and a ‘faux pas’ task (Baron-Cohen et al. 1999), so that older children with more subtle problems can also be detected.

The test–retest correlations of the typically developing children suggested a small learning effect. As stated before, this is consistent with findings from Muris et al. (1999). Grigorenko and Sternberg (1998) recommended that this effect—the learning potential of individual children—be included in normal diagnostics. In that case, the pretest–posttest difference can eventually be considered an estimation of learning abilities that are, at least in part, ToM specific. The absence of a comparable learning effect in specific groups of children, like we have found in children with PDD-NOS, could provide interesting information about the nature of ToM abilities in such children. In this line, further research on ToM might profit from dynamic testing—as opposed to static testing—where the learning potential of a child is quantified on the basis of his or her understanding and use of feedback given during testing (Grigorenko and Sternberg 1998). Dynamic indexes can represent a quality step-up compared with static indexes (Fabio 2005).

To conclude, one of the methodological strengths of the current test is that it has extended the limitations common in the majority of the researches done in the field of ToM. Most research has been undertaken in young children only (mostly up to 6 years, with a major focus on 3–4 year olds), has used only a few tasks (FB tasks, mainly single tasks) and considered small research groups (exceptions in the latter can be found in Charman et al. 2002; Hughes et al. 1999). The present research, aimed at constructing a new ToM Storybooks, used a wide range of tasks (not only FB tasks) and consisted of a substantial number of children over a wide age range. The test not only allows for comparisons on the basis of raw scores but standardized norms and norm scores are also available (Blijd-Hoogewys et al. submitted). In our opinion, the ToM Storybooks provide a comprehensive, valid and reliable instrument for researchers and clinicians who wish to measure Theory-of-Mind in young typically developing children, as well as children with an autism spectrum disorder from a broader age range.