JSLHR
Research Note
Face-Viewing Patterns in Young Children With Autism Spectrum Disorders: Speaking up for the Role of Language Comprehension Jakob Åsberg Johnels,a Christopher Gillberg,a Terje Falck-Ytter,b and Carmela Miniscalcoa
Purpose: The aim was to examine whether viewing patterns toward the mouth, eyes, and nonmouth–noneyes areas differed between young children with autism spectrum disorder (ASD) and typically developing (TD) children when viewing a person speaking. The role of language comprehension in such viewing patterns was also examined. Method: Eleven children with ASD (approximately 4.5 years old) and 29 TD toddlers (approximately 2.5 years old) participated. The groups were matched on language comprehension raw scores from the Reynell Developmental Language Scales III. All children viewed short films of a
woman speaking while their eye movements were recorded with eye-tracking equipment. Results: Children with ASD spent proportionally less time viewing the mouth and more time viewing the nonmouth– noneyes areas. Time viewing the eyes did not differ between groups. Increased mouth viewing was associated with lower language comprehension in the group with ASD. Conclusion: Variability in language comprehension is an important factor to monitor when interpreting face-viewing patterns in young children with ASD, particularly with regard to mouth viewing. The results may help explain divergent findings in this field of research.
I
of preference for socially relevant information (Klin, Lin, Gorrindo, Ramsay, & Jones, 2009). However, certain aspects of the original conclusions by Jones et al. (2008) are somewhat controversial. First, it appears as though attention to the mouth and attention to the eyes are both very natural responses for typically developing (TD) individuals at different ages (Falck-Ytter & von Hofsten, 2011; Nakano et al., 2010; Yarbus, 1967). Human beings appear particularly sensitive to the visual qualities of a speaking person. Indeed, audiovisual speech has repeatedly been shown to improve perception of what is said compared with speech that is only heard (Sumby & Pollack, 1954), even in cases in which the auditory speech signal is perfectly clear (Reisberg, McLean, & Goldfield, 1987). It has also been proposed that audiovisual speech provides an important scaffold for language processing by discriminating acoustically confusable but visually distinctive segments (cf. Campbell, 2008, for a review). In fact, recent research has shown that merely seeing the initial articulatory gestures of a word triggers lexical access to this word (Fort et al., 2012), and orientation toward the mouth area can therefore be assumed to support language
n this study, we examined whether face-viewing patterns differ between young children with autism spectrum disorder (ASD) and typically developing (TD) children when shown a short film of a woman talking. In their influential eye-tracking study with young children, Jones, Carr, and Klin (2008) reported that toddlers with ASD tend to look at the mouth rather than the eyes, and they proposed that normative orientation toward the eyes of others is impaired in this group. They later integrated these findings in the context of a more complex account in which increased mouth viewing in ASD reflects a general preference for aspects of the world that are high in physical audiovisual synchrony (i.e., when a person speaks, the articulatory gestures and sounds she or he produces are in synchrony), whereas decreased eye viewing reflects a lack
a
University of Gothenburg, Sweden Uppsala University, Sweden Correspondence to Jakob Åsberg Johnels: [email protected]
b
Editor: Rhea Paul Associate Editor: Linda Watson Received October 1, 2013 Revision received March 31, 2014 Accepted June 17, 2014 DOI: 10.1044/2014_JSLHR-L-13-0268
2246
Disclosure: The authors have declared that no competing interests existed at the time of publication.
Journal of Speech, Language, and Hearing Research • Vol. 57 • 2246–2252 • December 2014 • © American Speech-Language-Hearing Association
comprehension. Research has also suggested adaptive orientation toward the mouth at certain early periods in a child’s normative language development. Interestingly, Lewkowicz and Hansen-Tift (2012) recently showed that typically developing infants increase their viewing of the mouth area during the second half of the first year—a time when they are thought to benefit especially from the audiovisual speech cues available there—whereas proportionally more time is thereafter devoted to eye viewing. Although Lewkowicz and Hansen-Tift did not actually assess language levels (but instead inferred these from the chronological ages of their samples), a study by Young, Merin, Rogers, and Ozonoff (2009) in fact directly showed that increased levels of mouth viewing at age 6 months longitudinally predicted vocabulary levels at age 2 years. Second, a few recent studies have failed to find support for increased mouth viewing and decreased eye viewing in children with ASD. Chawarska and Shic (2009) examined face scanning of static photos and found that young children with ASD showed an atypical pattern of gaze as they attended more to the eyes but less to other key features of faces, including the mouth, than TD children. Furthermore, they found that children with ASD were more inclined to look away from faces and that this pattern was more pronounced in older (4-year-old) than in younger (2-year-old) children. Using films, Nakano et al. (2010) also showed that their group of preschool-age children with ASD devoted less time to viewing faces, including the mouth, than TD toddlers, who in turn spent most of the time viewing either the mouth or the eyes. More recently, Chawarska, Macari, and Shic (2012) showed that toddlers with ASD, compared with both TD toddlers and toddlers with developmental disorders other than ASD, spent less time viewing the mouth when the female actor in a film was involved in bids for dyadic attention, whereas they noted no group difference with regard to eye-viewing time. It is not clear what lies at the root of these differences in relation to the original findings by Jones et al. (2008). Nakano et al. (2010) suggested that the increased mouth viewing reported by Jones et al. might reflect a simple methodological artifact resulting from the fact that the actress in their movie clapped her hands while vocalizing, and the clapping hands might have been responsible for drawing “the attention of the toddlers with ASD to the lower half of the screen, where both the hands and the mouth were located” (p. 2942). Another possibility is cultural differences between participants in the studies. Specifically, Nakano et al. (2010) examined face viewing in Japanese children, and cross-cultural research has shown that face viewing differs somewhat between individuals from the Western world and those from East Asia (e.g., Blais, Jack, Scheepers, Fiset, & Caldara, 2008). However, cross-cultural differences cannot explain the results obtained in the U.S. study by Chawarska et al. (2012), who instead suggested that differences in developmental level or severity of autism-related symptoms in their study samples might be responsible for differences obtained in research on dynamic face viewing in ASD. We believe that it is particularly important to monitor the language
comprehension levels of the participants with and without ASD, given the covariance between language level and faceviewing patterns found in recent research. This was also highlighted in a recent study by Hosozawa, Tanaka, Shimizu, Nakano, and Kitazawa (2012), in which a group of toddlers with specific language impairment (without autism) were included as a comparison group to the participants in the Nakano et al. study. A clear result in the study by Hosozawa et al. was that the group with specific language impairment viewed the mouth significantly more, which was interpreted as a strategy to compensate for their poor auditory language processing abilities. The first aim of our study was to examine whether viewing times for the mouth, eyes, and areas other than the eyes or mouth could discriminate children with ASD from TD toddlers matched for language comprehension. Similar to Jones et al. (2008), we used (dynamic) films rather than (static) pictures as stimuli in our study. Although several previous studies have matched their participants with ASD on verbal mental age with children with other developmental disabilities (Chawarska et al., 2012; Jones et al., 2008), the novel comparison presented here is important because (a) language comprehension is often lower than what would be expected from verbal IQ in young children with ASD (Kjellmer et al., 2012) and (b) it allows us to examine whether children with ASD indeed display a deviant pattern of face viewing relative to TD peers with the same language comprehension. On the basis of the original account proposed by Jones et al., one would predict longer viewing times to the mouth and shorter viewing times to the eyes in the ASD group, but whether the results of Jones et al. will hold when controlling for language comprehension has not previously been examined. The second aim was to examine the associations between face viewing and independently assessed measures of language comprehension. We were particularly interested in examining the association between language comprehension and mouth viewing. However, it was rather difficult to predict a direction of any correlation on the basis of prior research. Given Young et al.’s (2009) research on 6-montholds, one might predict a positive correlation between increased mouth viewing and language level. In contrast, Hosozawa et al.’s (2012) study on toddlers and preschool children suggested that mouth viewing in young children is used to compensate for poor language skills, which possibly suggests a negative correlation between mouth viewing and language comprehension.
Method Participants The study included a group of 40 Swedish-speaking children—11 with ASD (1 girl, 10 boys; mean age approximately 4.5 years) and 29 who were TD (8 girls, 21 boys; mean age approximately 2.5 years). The ASD and TD groups were matched on raw mean scores on a language comprehension test (Arvidsson & Köröndi, 2011; Edwards et al., 1997;
Johnels et al.: Face-Viewing Patterns in Young Children With ASD
2247
see details below) but differed in chronological age, with the ASD group being significantly older. See Table 1 for participant information on these background measures. ASD group. Participants with ASD were patients at the Child Neuropsychiatric Clinic in Gothenburg, Sweden. The ASD group was systematically and comprehensively neuropsychiatrically assessed by a clinical research team, including a psychiatrist, a psychologist, and a speech-language pathologist at the clinic. Participation in the study required a diagnosis within the autism spectrum (autistic disorder, n = 9; pervasive developmental disorder–not otherwise specified, n = 2) according to consensus clinical judgment. On the Autism Diagnostic Observation Schedule, Module 1 or 2 (Lord et al., 2000), all participants scored above the cutoff for ASD (the lowest obtained total score was 11). Originally, 15 children with ASD were targeted. However, one child was excluded as a result of calibration failure during eye tracking, which resulted in no eye-tracking data at all, and another refused to sit and watch the film. In addition, two children scored 0 on the language comprehension test (see below) and therefore could not be included. TD group. The TD children and their parents were consecutively recruited at either of two time points during 2010 or 2011 while the child was undergoing a general health screening at about age 30 months at a child health care center in a suburb of Gothenburg. According to Magnusson (1997), 97.5% of all Swedish children participate in screening at these centers. The response rate among the contacted parents was 94%. Thirty-five children were originally targeted, but two had parents who declined participation because of a heavy workload. Four of the 33 originally recruited toddlers had to be excluded from the study, three because of a complete lack of eye-tracking data in the face-viewing experiment and one because of the child’s refusal to participate in the experiment. Our TD sample is considered representative of 2.5-year-olds in the region. None were suspected of any major developmental disorder, neither at the time of assessment nor according to previous health assessments. The absence of ASD was also confirmed by the Modified Checklist for Autism in Toddlers (Robins, Fein, Barton & Green, 2001), which was used in an interview with at least one of the child’s parents.
Experimental Setting The experiments were conducted at different locations for the ASD and TD groups. Children with ASD were assessed at the Child Neuropsychiatric Clinic and its research center, and the children in the TD group were assessed at the child health center. Care was taken to ensure that the experimental settings were as similar as possible in terms of sound conditions (i.e., a silent room), lighting, and furnishings (sparsely furnished). The same eye-tracker, laptop computer, and loudspeakers were used for all recordings.
Test Instruments and Stimuli Language comprehension test. ASD and TD children were examined by either of two speech-language pathologists using the Reynell Developmental Language Scales III (RDLS III; Edwards et al., 1997). The original RDLS III is a combined comprehension and language production test. Only the comprehension part has been translated and norm-referenced into Swedish (Arvidsson & Köröndi, 2011), hence only this part was administered. The results of the comprehension part, containing 62 test items, are sorted into 10 different domains, that is, from comprehension of single words to sentences of increasing difficulty. In the TD group, it was possible to calculate a norm-referenced score (Z score) for each participant, which showed that these children performed within the normal range on the RDLS III (z = –0.49, SD = 1.04). Because of their age, it was not possible to calculate norm-referenced scores for all participants in the ASD group. Raw scores were used in the analyses and for group matching. Eye-tracking stimuli and design. Children were seated in a chair and placed about 60 cm from the eye-tracking monitor. Gaze was measured with a Tobii T120 (Tobii Technology Inc., Stockholm, Sweden), which records near infrared reflections of both eyes at 60 Hz as the participant watches an integrated 17-in. monitor. A 5-point infant calibration procedure was used before the experiment in keeping with recommendations in Sasson and Elison (2012). Visual inspection showed that all 40 children had sufficiently high-quality data to be included in the analysis. In addition,
Table 1. Participant information and experimental data with group comparisons for children with autism spectrum disorder and typically developing children. ASD (n = 11) Measure Background measures Chronological age (months) Language comprehension raw score Experimental measures (results) Proportion viewing eyes Proportion viewing mouth Proportion viewing nonmouth and noneyes
TD (n = 29)
M
SD
Range
M
SD
Range
52.5 28.3
9.8 14.8
39–69 4–58
28.7 27.4
2.5 7.7
24–33 11–41
t(10.5) = –7.93, p < .001 t(12.1) = –0.19, p = .85
.04–.55 .01–.85 .04–.61
t(38) = 1.2, p = .20, d = –0.48 t(30.4) = 2.7, p = .012, d = –0.84 t(38) = –5.2, p < .001, d = 1.78
.26 .23 .50
.17 .15 .16
.01–.69 .05–.52 .28–.74
.35 .41 .24
.20 .25 .14
Note. ASD = autism spectrum disorders; TD = typically developing.
2248
Journal of Speech, Language, and Hearing Research • Vol. 57 • 2246–2252 • December 2014
Group comparison
we made an attempt to quantify calibration quality in a separate stimulus set (see details below). Face stimuli included two visually identical films of a woman telling a short story (11 s) in Swedish about a mouse being chased by a cat (see Figure 1). The participants’ total viewing times (duration) for each area of interest (AOI)—that is, the eyes, the mouth, and areas other than the eyes or mouth (nonmouth–noneyes areas)—were recorded. In one of the films, only the woman’s natural, synchronous voice was heard. In the other, her voice plus another unrelated voice (speaking in English) were heard, thereby introducing noise. The reason both of these films were shown was initially to test whether we could block a potential mouth preference in ASD by introducing auditory noise (i.e., if mouth viewing in ASD is driven by a perceptual preference for audiovisual synchrony [Jones et al., 2008], then lowering this synchrony may result in shortened viewing times). However, the proportions of total viewing time for eyes, mouth, and nonmouth–noneyes areas were similar across the two films (i.e., for the one- and two-voice versions) according to paired-samples t tests performed separately for each group (all ts ≤ 1.6, p ≥ .15). Nor did we observe a significant AOI × Film Version × Group interaction effect in a mixed factorial analysis of variance, F(2, 37) ≤ 1.3, p = .29, suggesting that the diagnostic groups did not respond differentially to the manipulation. Because of this result, we chose in the rest of the analyses to examine the viewing times for eyes, mouth, and nonmouth–noneyes areas combined for the two films. This was assumed to (a) ease interpretation and (b) provide a more robust measure of face viewing. The rationale for using total visit duration (or dwell time, including fixations and saccades) for each AOI is that this metric is assumed to more closely reflect real eye movements. Whereas a dwell consists of all samples from entry to exit in an AOI, fixation-based samples do not have to reside in the AOI as
Figure 1. Static representation from video showing a woman telling a short story. Areas of interest (AOIs; colored rectangle and ellipse) for mouth and eyes are superimposed. Viewing times for non-AOI (i.e., viewing times for nonmouth–noneyes areas) are also registered.
long as the central location of the fixation does. It is also known that fixation data are highly influenced by the choice of fixation algorithm and its settings (Holmqvist et al., 2011). Hence, visit duration was considered a more robust measure in the current study. The face stimuli were embedded in a larger package of social stimuli (shown to each participant in a unique pseudorandom order), lasting approximately 4 min in total. Other stimuli (e.g., biological motion from nonhuman animals) were included in the stimuli package, but because they are not of direct relevance to this study, these data are not reported. Each film was preceded by an attentiongrabbing stimulus that included one or several small brightening stars (presented in a randomized order). Although not originally created for this purpose, we took the opportunity to use the viewing patterns for these stars as an additional control of calibration quality. Specifically, an independent staff member in our lab, who was blind to group membership and to the purposes of our study, was asked to rate on a scale ranging from 1 to 4 how close each child’s fixations were to the center of the star or stars of the first attentiongrabber film (with 1 = systematically way off and 4 = spot on). In general, ratings were high (indicating good calibration quality) and not different across the two groups (MASD = 3.36, SD = 0.73, vs. MTD = 3.62, SD = 0.72), t(38) = 0.96, p = .34.
Ethics Parents provided written consent. The study was approved by the Ethics Committees at Uppsala University and the University of Gothenburg.
Statistical Analyses The main outcome measure was the proportion of total viewing time devoted to the defined AOIs (and the non-AOI), calculated as viewing time for the given area divided by the total viewing time anywhere on the screen.1 A 3 (AOI: eyes vs. mouth vs. nonmouth–noneyes) × 2 (group: ASD vs. TD) mixed factorial analysis of variance was conducted to examine whether face-viewing patterns differed between the groups. The most critical outcome from this analysis is an interaction effect between AOI and group. Following such a result, we next used separate t tests to examine group differences on each AOI. Adjusted degrees of freedom are reported when appropriate; therefore, the degrees of freedom can vary even though the n is always the same. To examine whether the group-level patterns held for individual children, receiver operating characteristic curve analysis was used to examine discriminative ability in viewing patterns across children with and without ASD. Pearson correlations were used to analyze associations, and Fisher’s r-to-z transformation was used to compare the size
1 The total viewing time anywhere on the screen did not differ statistically between the groups (M = 15.5 s, SD = 3.0 vs. M = 16.94 s, SD = 4.7, for the ASD and TD groups, respectively), t(28.9) ≤ 1, p = .4.
Johnels et al.: Face-Viewing Patterns in Young Children With ASD
2249
of correlations in the two study groups. SPSS Statistics 19 for Mac (IBM Corp., Armonk, NY) was used for all analyses, except the r-to-z transformation, which was conducted in an interactive resource (http://vassarstats.net/rdiff.html). A two-tailed alpha level of .05 was applied.
Results The analysis of variance did not reveal a main effect of AOI, F(2) < 1, p = .37 (lower bound), but it did reveal a significant Group × AOI interaction, F(1, 38) = 7.73, p = .008, hp 2 = .169. Table 1 reports mean viewing times and standard deviations for each group, along with results from follow-up pairwise independent t tests. The most striking result is that children with ASD spent twice as much time as TD children viewing nonmouth–noneyes areas but less time viewing the mouth. No significant difference was observed between the groups for viewing eyes.2 Because viewing areas other than the mouth and eyes appears to be characteristic of young children with ASD compared with language level–matched TD toddlers, we next examined how this metric discriminates ASD from TD in a receiver operating characteristic curve. The result of this analysis gave an area under the curve of .90 (SE = .05, p < .001), which is considered excellent discriminative ability. Because there is always a trade-off between sensitivity and specificity, the “best” cutoff depends on the purposes of the testing, for example, for screening versus gaining information for diagnostic decision making. Inspection of sensitivity and specificity suggested that two meaningful cutoff values could be identified. A first low cutoff of 30% viewing time for nonmouth–noneyes areas has a sensitivity of .91 and a specificity of .72 for ASD (positive predictive value = 91%; negative predictive value = 28%). An even lower cutoff could raise the sensitivity up to 1.0, but at the cost of greatly lowering the specificity (< .58). A second high cutoff of 44.5% has a sensitivity of .64 and a specificity of .93 for ASD (positive predictive value = 64%; negative predictive value = 96%). An even higher cutoff could raise the specificity but at the cost of disproportionately lowering the sensitivity ( .20). In the TD group, the correlation between language comprehension and mouth-viewing times was not significant (r = .08, p = .67), and the size of this correlation was significantly smaller than that seen in the ASD group (z = –2.12, p = .034). There was no correlation between language comprehension and times viewing eyes (r = –.01, p = .95) or times viewing noneyes–nonmouth areas (r = –14, p = .48). Nor did we find any significant correlations between chronological age and face-viewing patterns in either group (in the ASD group, all ps > .27; in the TD group, all ps > .08).
Discussion In comparison with a group of TD toddlers, children with ASD spent less time viewing the mouth and more time viewing nonmouth–noneyes areas when shown a film of a woman talking. A novel aspect of the analyses presented here was that we closely evaluated the children’s language comprehension level and matched our groups on this variable. This was motivated by recent research showing that language and mouth-viewing patterns are associated (cf. Hosozawa et al., 2012; Young et al., 2009). Arguably, previous research has not been able to determine whether the group differences in viewing patterns between children with and without ASD reflect diagnostic status or the children’s language comprehension level. In this study, we found that even with such control for language comprehension, marked group differences were found; however, these differences only partly mirrored those found in other studies comparing young children with and without ASD. The pattern displayed by the TD group is consistent with a large literature suggesting that the mouth and eyes have a special status when neurotypical individuals view a face (Yarbus, 1967), whereas individuals with ASD show a deviant pattern in this regard, possibly reflecting limited attentional bias to key regions of the face or even to the face in general (e.g., Chawarska, Volkmar, & Klin, 2010; Riby & Hancock, 2009). In fact, our data suggest that viewing areas other than the mouth and eyes may be a strong marker for ASD. A receiver operating characteristic curve analysis further showed that viewing other areas than the mouth and eyes had excellent discriminative ability for ASD also at an individual child level. Hence, this finding may turn out to be of both theoretical and clinical importance. Our finding that children with ASD attended to the mouth less than the TD group is in sharp contrast to that reported by Jones et al. (2008) but is consistent with other recent research (Chawarska & Shic, 2009; Chawarska et al., 2012; Nakano et al., 2010). Of course, it cannot be ruled out that our participants with ASD had viewed the mouth extensively when they were 2.5 years old. More important, though, the influential account forwarded by Jones et al. does not appear to be constrained to a particular age. Furthermore,
Journal of Speech, Language, and Hearing Research • Vol. 57 • 2246–2252 • December 2014
Figure 2. Scatter plot of the negative association between mouth viewing and language comprehension in the group with autism spectrum disorder.
we found no correlations between chronological age and viewing patterns in our group with ASD, suggesting that our group differences in (mouth-) viewing patterns probably cannot be explained by differences in chronological age. That said, an important avenue for future research will be to track and compare the developmental course of face viewing in children with and without ASD, preferably using a longitudinal design. In sum, this article highlights that mouth viewing appears to be a natural response for 30-month-old TD toddlers (cf. Falck-Ytter & von Hofsten, 2011) and suggests that general conclusions regarding extensive mouth viewing in young children with ASD may need to be revised. Interestingly, a few recent studies have looked more deeply into the heterogeneity of face viewing in school-aged children and adolescents with ASD and have described associations between face-viewing patterns and social, communicative, or linguistic proficiency in this group (Falck-Ytter, Fernell, Gillberg, & von Hofsten, 2010; Norbury, Brock, Cragg, Einav, Griffiths, & Nation, 2009; Rice, Moriuchi, Klin, & Jones, 2012). In this study, we were also interested in individual differences within the ASD group. We found that increased levels of mouth viewing in ASD were associated with poor language comprehension. This confirmed our suspicion that variability in language comprehension is an important factor to monitor when interpreting face-viewing patterns in young children with ASD. This result also underscores the importance of matching study groups on this variable; had our groups differed greatly in language comprehension, the outcome from the group comparison might very well have been different. In this sense, the results of the
present study may help explain the many divergent findings obtained in this field of research. Future research is needed to unravel the exact mechanisms involved, but we speculate that viewing the mouth may constitute adaptive (compensatory) adjustments allowing children with ASD and low language comprehension to decode language, whereas children with ASD and better language ability do not need to use visual support from the mouth area to the same extent. This idea is related to the ones proposed in Lewkowicz and Hansen-Tift (2012), although they focused on language production rather than comprehension in TD infants, and Hosozawa et al. (2012), who studied children with specific language impairment. However, if this explanation is correct, it is somewhat difficult to explain why we did not observe the same correlation in the TD group (which was larger and similar in terms of language comprehension). One possible explanation could be that only children with delayed or disordered language perceptually orient to the mouth area for compensatory visual clues (Hosozawa et al., 2012). Support for this account was also obtained in the study of adolescents with ASD by Norbury et al. (2009). These authors—as did we—found that a measure of language skills (i.e., a composite measure of receptive vocabulary and sentence repetition) correlated negatively with mouth viewing. However, this was only the case for the subgroup with impaired language skills; among those with good language skills, the association was not significant and in fact appeared to be possibly reversed. We end by discussing two shortcomings of this study and also describe directions for future research. First, similar
Johnels et al.: Face-Viewing Patterns in Young Children With ASD
2251
to many other studies in this field, our study included a relatively small sample of children. Therefore, future research needs to replicate these findings in larger samples. Second, future research is needed to further explore possible mediating mechanisms and developmental consequences of the results presented here. Why do young children with ASD attend to noneye–nonmouth areas when looking at a person talking? And why, more exactly, is mouth viewing in young children with ASD associated with poor language comprehension? An important question in this context is also whether the nature of the association between mouth viewing and limited language comprehension differs depending on how language comprehension is measured. In a new ongoing study, we are exploring how children’s orientation toward the mouth of a person talking is related not only to test scores of language comprehension, but also to the child’s comprehension of what that specific person in the film is saying.
References Arvidsson, E., & Köröndi, J. (2011). Svensk normering av språkföstålsedelen i Reynell Developmental Language Scales III 4:6-4:11 år—Sambandet mellan gemensam läsning och språkförståelse [Swedish norms on Reynell Developmental Language Scales III]. Gothenburg, Sweden: University of Gothenburg. Blais, C., Jack, R. E., Scheepers, C., Fiset, D., & Caldara, R. (2008). Culture shapes how we look at faces. PLoS ONE, 3, e3022. Campbell, R. (2008). The processing of audio-visual speech: Empirical and neural bases. Philosophical Transactions of the Royal Society B: Biological Sciences, 363, 1001–1010. Chawarska, K., Macari, S., & Shic, F. (2012). Context modulates attention to social scenes in toddlers with autism. Journal of Child Psychology and Psychiatry, 53, 903–913. Chawarska, K., & Shic, F. (2009). Looking but not seeing: Abnormal visual scanning and recognition of faces in 2- and 4-year old children with autism spectrum disorder. Journal of Autism and Developmental Disorders, 39, 1663–1672. Chawarska, K., Volkmar, F., & Klin, A. (2010). Limited attentional bias for faces in toddlers with autism spectrum disorders. Archives of General Psychiatry, 67, 178–185. Edwards, S., Fletcher, P., Garman, M., Hughes, A., Letts, C., & Sinka, I. (1997). The Reynell Developmental Language Scales III: The University of Reading Edition. Windsor, England: NferNelson. Falck-Ytter, T., Fernell, E., Gillberg, C., & von Hofsten, C. (2010). Face scanning distinguishes social from communication impairments in autism. Developmental Science, 13, 864–865. Falck-Ytter, T., & von Hofsten, C. (2011). How special is social looking in ASD—A review. Progress in Brain Research, 198, 209–222. Fort, M., Kandel, S., Chipot, J., Savariaux, C., Granjon, L., & Spinelli, E. (2012). Language and Cognitive Processes, 28, 1207–1223. Holmqvist, K., Nyström, M., Andersson, R., Dewhurst, R., Halszka, J., & van de Weijer, J. (2011). Eye tracking: A comprehensive guide to methods and measures. Oxford, England: Oxford University Press. Hosozawa, M., Tanaka, K., Shimizu, T., Nakano, T., & Kitazawa, S. (2012). How children with specific language impairment view social situations. Pediatrics, 129, 1453–1460. Jones, W., Carr, K., & Klin, A. (2008). Absence of preferential looking to the eyes of approaching adults predicts level of social
2252
disability in 2-year-old toddlers with autism spectrum disorder. Archives of General Psychiatry, 65, 946–954. Kjellmer, L., Hedvall, Å., Holm, A., Fernell, E., Gillberg, C., & Norrelgen, F. (2012). Language comprehension in preschoolers with autism spectrum disorders without intellectual disability: Use of the Reynell Developmental Language Scales. Research in Autism Spectrum Disorders, 6, 1119–1125. Klin, A., Lin, D. J., Gorrindo, P., Ramsay, G., & Jones, W. (2009). Two-year-olds with autism orient to non-social contingencies rather than biological motion. Nature, 459, 257–261. Lewkowicz, D. J., & Hansen-Tift, A. M. (2012). Infants deploy selective attention to the mouth of a talking face when learning speech. Proceedings of the National Academy of Sciences, USA, 109, 1431–1436. Lord, C., Risi, S., Lambrecht, L., Cock, E. H., Jr., Leventhal, B. L., DiLavore, P. C., Pickles, A., & Rutter, M. (2000). The Autism Diagnostic Schedule—Generic: A standard measure of social and communication deficits associated with the spectrum of autism. Journal of Autism and Developmental Disorders, 30, 205–223. Magnusson, M. (1997). Rationality of routine health examinations by physicians of 18-month-old children: Experiences based on data from a Swedish county. Acta Paediatrica, 86, 881–887. Nakano, T., Tanaka, K., Endo, Y., Yamane, Y., Yamamoto, T., Nakano, Y., . . . Kitazawa, S. (2010). Atypical gaze patterns in children and adults with autism spectrum disorders dissociated from developmental changes in gaze behaviour. Proceedings of the Royal Society B: Biological Sciences, 277, 2935–2943. Norbury, C. F., Brock, J., Cragg, L., Einav, S., Griffiths, H., & Nation, K. (2009). Eye-movement patterns are associated with communicative competence in autistic spectrum disorders. Journal of Child Psychology and Psychiatry, 50, 834–842. Reisberg, D., McLean, J., & Goldfield, A. (1987). Easy to hear but hard to understand: A lip-reading advantage with intact auditory stimuli. In B. Dodd & R. Campbell (Eds.), Hearing by eye: The psychology of lip-reading (pp. 97–113). Hillsdale, NJ: Erlbaum. Riby, D. M., & Hancock, P. J. B. (2009). Looking at movies and cartoons: Eye-tracking evidence from Williams syndrome and autism. Journal of Intellectual Disability Research, 53, 169–181. Rice, K., Moriuchi, J. M., Klin, A., & Jones, W. (2012). Parsing heterogeneity in autism spectrum disorders: Visual scanning of dynamic social scenes in school-aged children. Journal of the American Academy of Child & Adolescent Psychiatry, 53, 238–248. Robins, D. L., Fein, D., Barton, M. L., & Green, J. A. (2001). The modified checklist for autism in toddlers: An initial study investigating the early detection of autism and pervasive developmental disorders. Journal of Autism and Developmental Disorders, 31, 131–144. Sasson, N. J., & Elison, J. T. (2012). Eye tracking young children with autism. Journal of Visualized Experiments, 61, e3675. Sumby, W. H., & Pollack, I. (1954). Visual contribution to speech intelligibility in noise. Journal of the Acoustical Society of America, 26, 212–215. Yarbus, A. L. (1967). Eye movements and vision. New York, NY: Plenum Press. Young, G. S., Merin, N., Rogers, S. J., & Ozonoff, S. (2009). Gaze behavior and affect at 6 months: Predicting clinical outcomes and language development in typically developing infants and infants at-risk for autism. Developmental Science, 12, 798–814.
Journal of Speech, Language, and Hearing Research • Vol. 57 • 2246–2252 • December 2014
Copyright of Journal of Speech, Language & Hearing Research is the property of American Speech-Language-Hearing Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
Research Note
Face-Viewing Patterns in Young Children With Autism Spectrum Disorders: Speaking up for the Role of Language Comprehension Jakob Åsberg Johnels,a Christopher Gillberg,a Terje Falck-Ytter,b and Carmela Miniscalcoa
Purpose: The aim was to examine whether viewing patterns toward the mouth, eyes, and nonmouth–noneyes areas differed between young children with autism spectrum disorder (ASD) and typically developing (TD) children when viewing a person speaking. The role of language comprehension in such viewing patterns was also examined. Method: Eleven children with ASD (approximately 4.5 years old) and 29 TD toddlers (approximately 2.5 years old) participated. The groups were matched on language comprehension raw scores from the Reynell Developmental Language Scales III. All children viewed short films of a
woman speaking while their eye movements were recorded with eye-tracking equipment. Results: Children with ASD spent proportionally less time viewing the mouth and more time viewing the nonmouth– noneyes areas. Time viewing the eyes did not differ between groups. Increased mouth viewing was associated with lower language comprehension in the group with ASD. Conclusion: Variability in language comprehension is an important factor to monitor when interpreting face-viewing patterns in young children with ASD, particularly with regard to mouth viewing. The results may help explain divergent findings in this field of research.
I
of preference for socially relevant information (Klin, Lin, Gorrindo, Ramsay, & Jones, 2009). However, certain aspects of the original conclusions by Jones et al. (2008) are somewhat controversial. First, it appears as though attention to the mouth and attention to the eyes are both very natural responses for typically developing (TD) individuals at different ages (Falck-Ytter & von Hofsten, 2011; Nakano et al., 2010; Yarbus, 1967). Human beings appear particularly sensitive to the visual qualities of a speaking person. Indeed, audiovisual speech has repeatedly been shown to improve perception of what is said compared with speech that is only heard (Sumby & Pollack, 1954), even in cases in which the auditory speech signal is perfectly clear (Reisberg, McLean, & Goldfield, 1987). It has also been proposed that audiovisual speech provides an important scaffold for language processing by discriminating acoustically confusable but visually distinctive segments (cf. Campbell, 2008, for a review). In fact, recent research has shown that merely seeing the initial articulatory gestures of a word triggers lexical access to this word (Fort et al., 2012), and orientation toward the mouth area can therefore be assumed to support language
n this study, we examined whether face-viewing patterns differ between young children with autism spectrum disorder (ASD) and typically developing (TD) children when shown a short film of a woman talking. In their influential eye-tracking study with young children, Jones, Carr, and Klin (2008) reported that toddlers with ASD tend to look at the mouth rather than the eyes, and they proposed that normative orientation toward the eyes of others is impaired in this group. They later integrated these findings in the context of a more complex account in which increased mouth viewing in ASD reflects a general preference for aspects of the world that are high in physical audiovisual synchrony (i.e., when a person speaks, the articulatory gestures and sounds she or he produces are in synchrony), whereas decreased eye viewing reflects a lack
a
University of Gothenburg, Sweden Uppsala University, Sweden Correspondence to Jakob Åsberg Johnels: [email protected]
b
Editor: Rhea Paul Associate Editor: Linda Watson Received October 1, 2013 Revision received March 31, 2014 Accepted June 17, 2014 DOI: 10.1044/2014_JSLHR-L-13-0268
2246
Disclosure: The authors have declared that no competing interests existed at the time of publication.
Journal of Speech, Language, and Hearing Research • Vol. 57 • 2246–2252 • December 2014 • © American Speech-Language-Hearing Association
comprehension. Research has also suggested adaptive orientation toward the mouth at certain early periods in a child’s normative language development. Interestingly, Lewkowicz and Hansen-Tift (2012) recently showed that typically developing infants increase their viewing of the mouth area during the second half of the first year—a time when they are thought to benefit especially from the audiovisual speech cues available there—whereas proportionally more time is thereafter devoted to eye viewing. Although Lewkowicz and Hansen-Tift did not actually assess language levels (but instead inferred these from the chronological ages of their samples), a study by Young, Merin, Rogers, and Ozonoff (2009) in fact directly showed that increased levels of mouth viewing at age 6 months longitudinally predicted vocabulary levels at age 2 years. Second, a few recent studies have failed to find support for increased mouth viewing and decreased eye viewing in children with ASD. Chawarska and Shic (2009) examined face scanning of static photos and found that young children with ASD showed an atypical pattern of gaze as they attended more to the eyes but less to other key features of faces, including the mouth, than TD children. Furthermore, they found that children with ASD were more inclined to look away from faces and that this pattern was more pronounced in older (4-year-old) than in younger (2-year-old) children. Using films, Nakano et al. (2010) also showed that their group of preschool-age children with ASD devoted less time to viewing faces, including the mouth, than TD toddlers, who in turn spent most of the time viewing either the mouth or the eyes. More recently, Chawarska, Macari, and Shic (2012) showed that toddlers with ASD, compared with both TD toddlers and toddlers with developmental disorders other than ASD, spent less time viewing the mouth when the female actor in a film was involved in bids for dyadic attention, whereas they noted no group difference with regard to eye-viewing time. It is not clear what lies at the root of these differences in relation to the original findings by Jones et al. (2008). Nakano et al. (2010) suggested that the increased mouth viewing reported by Jones et al. might reflect a simple methodological artifact resulting from the fact that the actress in their movie clapped her hands while vocalizing, and the clapping hands might have been responsible for drawing “the attention of the toddlers with ASD to the lower half of the screen, where both the hands and the mouth were located” (p. 2942). Another possibility is cultural differences between participants in the studies. Specifically, Nakano et al. (2010) examined face viewing in Japanese children, and cross-cultural research has shown that face viewing differs somewhat between individuals from the Western world and those from East Asia (e.g., Blais, Jack, Scheepers, Fiset, & Caldara, 2008). However, cross-cultural differences cannot explain the results obtained in the U.S. study by Chawarska et al. (2012), who instead suggested that differences in developmental level or severity of autism-related symptoms in their study samples might be responsible for differences obtained in research on dynamic face viewing in ASD. We believe that it is particularly important to monitor the language
comprehension levels of the participants with and without ASD, given the covariance between language level and faceviewing patterns found in recent research. This was also highlighted in a recent study by Hosozawa, Tanaka, Shimizu, Nakano, and Kitazawa (2012), in which a group of toddlers with specific language impairment (without autism) were included as a comparison group to the participants in the Nakano et al. study. A clear result in the study by Hosozawa et al. was that the group with specific language impairment viewed the mouth significantly more, which was interpreted as a strategy to compensate for their poor auditory language processing abilities. The first aim of our study was to examine whether viewing times for the mouth, eyes, and areas other than the eyes or mouth could discriminate children with ASD from TD toddlers matched for language comprehension. Similar to Jones et al. (2008), we used (dynamic) films rather than (static) pictures as stimuli in our study. Although several previous studies have matched their participants with ASD on verbal mental age with children with other developmental disabilities (Chawarska et al., 2012; Jones et al., 2008), the novel comparison presented here is important because (a) language comprehension is often lower than what would be expected from verbal IQ in young children with ASD (Kjellmer et al., 2012) and (b) it allows us to examine whether children with ASD indeed display a deviant pattern of face viewing relative to TD peers with the same language comprehension. On the basis of the original account proposed by Jones et al., one would predict longer viewing times to the mouth and shorter viewing times to the eyes in the ASD group, but whether the results of Jones et al. will hold when controlling for language comprehension has not previously been examined. The second aim was to examine the associations between face viewing and independently assessed measures of language comprehension. We were particularly interested in examining the association between language comprehension and mouth viewing. However, it was rather difficult to predict a direction of any correlation on the basis of prior research. Given Young et al.’s (2009) research on 6-montholds, one might predict a positive correlation between increased mouth viewing and language level. In contrast, Hosozawa et al.’s (2012) study on toddlers and preschool children suggested that mouth viewing in young children is used to compensate for poor language skills, which possibly suggests a negative correlation between mouth viewing and language comprehension.
Method Participants The study included a group of 40 Swedish-speaking children—11 with ASD (1 girl, 10 boys; mean age approximately 4.5 years) and 29 who were TD (8 girls, 21 boys; mean age approximately 2.5 years). The ASD and TD groups were matched on raw mean scores on a language comprehension test (Arvidsson & Köröndi, 2011; Edwards et al., 1997;
Johnels et al.: Face-Viewing Patterns in Young Children With ASD
2247
see details below) but differed in chronological age, with the ASD group being significantly older. See Table 1 for participant information on these background measures. ASD group. Participants with ASD were patients at the Child Neuropsychiatric Clinic in Gothenburg, Sweden. The ASD group was systematically and comprehensively neuropsychiatrically assessed by a clinical research team, including a psychiatrist, a psychologist, and a speech-language pathologist at the clinic. Participation in the study required a diagnosis within the autism spectrum (autistic disorder, n = 9; pervasive developmental disorder–not otherwise specified, n = 2) according to consensus clinical judgment. On the Autism Diagnostic Observation Schedule, Module 1 or 2 (Lord et al., 2000), all participants scored above the cutoff for ASD (the lowest obtained total score was 11). Originally, 15 children with ASD were targeted. However, one child was excluded as a result of calibration failure during eye tracking, which resulted in no eye-tracking data at all, and another refused to sit and watch the film. In addition, two children scored 0 on the language comprehension test (see below) and therefore could not be included. TD group. The TD children and their parents were consecutively recruited at either of two time points during 2010 or 2011 while the child was undergoing a general health screening at about age 30 months at a child health care center in a suburb of Gothenburg. According to Magnusson (1997), 97.5% of all Swedish children participate in screening at these centers. The response rate among the contacted parents was 94%. Thirty-five children were originally targeted, but two had parents who declined participation because of a heavy workload. Four of the 33 originally recruited toddlers had to be excluded from the study, three because of a complete lack of eye-tracking data in the face-viewing experiment and one because of the child’s refusal to participate in the experiment. Our TD sample is considered representative of 2.5-year-olds in the region. None were suspected of any major developmental disorder, neither at the time of assessment nor according to previous health assessments. The absence of ASD was also confirmed by the Modified Checklist for Autism in Toddlers (Robins, Fein, Barton & Green, 2001), which was used in an interview with at least one of the child’s parents.
Experimental Setting The experiments were conducted at different locations for the ASD and TD groups. Children with ASD were assessed at the Child Neuropsychiatric Clinic and its research center, and the children in the TD group were assessed at the child health center. Care was taken to ensure that the experimental settings were as similar as possible in terms of sound conditions (i.e., a silent room), lighting, and furnishings (sparsely furnished). The same eye-tracker, laptop computer, and loudspeakers were used for all recordings.
Test Instruments and Stimuli Language comprehension test. ASD and TD children were examined by either of two speech-language pathologists using the Reynell Developmental Language Scales III (RDLS III; Edwards et al., 1997). The original RDLS III is a combined comprehension and language production test. Only the comprehension part has been translated and norm-referenced into Swedish (Arvidsson & Köröndi, 2011), hence only this part was administered. The results of the comprehension part, containing 62 test items, are sorted into 10 different domains, that is, from comprehension of single words to sentences of increasing difficulty. In the TD group, it was possible to calculate a norm-referenced score (Z score) for each participant, which showed that these children performed within the normal range on the RDLS III (z = –0.49, SD = 1.04). Because of their age, it was not possible to calculate norm-referenced scores for all participants in the ASD group. Raw scores were used in the analyses and for group matching. Eye-tracking stimuli and design. Children were seated in a chair and placed about 60 cm from the eye-tracking monitor. Gaze was measured with a Tobii T120 (Tobii Technology Inc., Stockholm, Sweden), which records near infrared reflections of both eyes at 60 Hz as the participant watches an integrated 17-in. monitor. A 5-point infant calibration procedure was used before the experiment in keeping with recommendations in Sasson and Elison (2012). Visual inspection showed that all 40 children had sufficiently high-quality data to be included in the analysis. In addition,
Table 1. Participant information and experimental data with group comparisons for children with autism spectrum disorder and typically developing children. ASD (n = 11) Measure Background measures Chronological age (months) Language comprehension raw score Experimental measures (results) Proportion viewing eyes Proportion viewing mouth Proportion viewing nonmouth and noneyes
TD (n = 29)
M
SD
Range
M
SD
Range
52.5 28.3
9.8 14.8
39–69 4–58
28.7 27.4
2.5 7.7
24–33 11–41
t(10.5) = –7.93, p < .001 t(12.1) = –0.19, p = .85
.04–.55 .01–.85 .04–.61
t(38) = 1.2, p = .20, d = –0.48 t(30.4) = 2.7, p = .012, d = –0.84 t(38) = –5.2, p < .001, d = 1.78
.26 .23 .50
.17 .15 .16
.01–.69 .05–.52 .28–.74
.35 .41 .24
.20 .25 .14
Note. ASD = autism spectrum disorders; TD = typically developing.
2248
Journal of Speech, Language, and Hearing Research • Vol. 57 • 2246–2252 • December 2014
Group comparison
we made an attempt to quantify calibration quality in a separate stimulus set (see details below). Face stimuli included two visually identical films of a woman telling a short story (11 s) in Swedish about a mouse being chased by a cat (see Figure 1). The participants’ total viewing times (duration) for each area of interest (AOI)—that is, the eyes, the mouth, and areas other than the eyes or mouth (nonmouth–noneyes areas)—were recorded. In one of the films, only the woman’s natural, synchronous voice was heard. In the other, her voice plus another unrelated voice (speaking in English) were heard, thereby introducing noise. The reason both of these films were shown was initially to test whether we could block a potential mouth preference in ASD by introducing auditory noise (i.e., if mouth viewing in ASD is driven by a perceptual preference for audiovisual synchrony [Jones et al., 2008], then lowering this synchrony may result in shortened viewing times). However, the proportions of total viewing time for eyes, mouth, and nonmouth–noneyes areas were similar across the two films (i.e., for the one- and two-voice versions) according to paired-samples t tests performed separately for each group (all ts ≤ 1.6, p ≥ .15). Nor did we observe a significant AOI × Film Version × Group interaction effect in a mixed factorial analysis of variance, F(2, 37) ≤ 1.3, p = .29, suggesting that the diagnostic groups did not respond differentially to the manipulation. Because of this result, we chose in the rest of the analyses to examine the viewing times for eyes, mouth, and nonmouth–noneyes areas combined for the two films. This was assumed to (a) ease interpretation and (b) provide a more robust measure of face viewing. The rationale for using total visit duration (or dwell time, including fixations and saccades) for each AOI is that this metric is assumed to more closely reflect real eye movements. Whereas a dwell consists of all samples from entry to exit in an AOI, fixation-based samples do not have to reside in the AOI as
Figure 1. Static representation from video showing a woman telling a short story. Areas of interest (AOIs; colored rectangle and ellipse) for mouth and eyes are superimposed. Viewing times for non-AOI (i.e., viewing times for nonmouth–noneyes areas) are also registered.
long as the central location of the fixation does. It is also known that fixation data are highly influenced by the choice of fixation algorithm and its settings (Holmqvist et al., 2011). Hence, visit duration was considered a more robust measure in the current study. The face stimuli were embedded in a larger package of social stimuli (shown to each participant in a unique pseudorandom order), lasting approximately 4 min in total. Other stimuli (e.g., biological motion from nonhuman animals) were included in the stimuli package, but because they are not of direct relevance to this study, these data are not reported. Each film was preceded by an attentiongrabbing stimulus that included one or several small brightening stars (presented in a randomized order). Although not originally created for this purpose, we took the opportunity to use the viewing patterns for these stars as an additional control of calibration quality. Specifically, an independent staff member in our lab, who was blind to group membership and to the purposes of our study, was asked to rate on a scale ranging from 1 to 4 how close each child’s fixations were to the center of the star or stars of the first attentiongrabber film (with 1 = systematically way off and 4 = spot on). In general, ratings were high (indicating good calibration quality) and not different across the two groups (MASD = 3.36, SD = 0.73, vs. MTD = 3.62, SD = 0.72), t(38) = 0.96, p = .34.
Ethics Parents provided written consent. The study was approved by the Ethics Committees at Uppsala University and the University of Gothenburg.
Statistical Analyses The main outcome measure was the proportion of total viewing time devoted to the defined AOIs (and the non-AOI), calculated as viewing time for the given area divided by the total viewing time anywhere on the screen.1 A 3 (AOI: eyes vs. mouth vs. nonmouth–noneyes) × 2 (group: ASD vs. TD) mixed factorial analysis of variance was conducted to examine whether face-viewing patterns differed between the groups. The most critical outcome from this analysis is an interaction effect between AOI and group. Following such a result, we next used separate t tests to examine group differences on each AOI. Adjusted degrees of freedom are reported when appropriate; therefore, the degrees of freedom can vary even though the n is always the same. To examine whether the group-level patterns held for individual children, receiver operating characteristic curve analysis was used to examine discriminative ability in viewing patterns across children with and without ASD. Pearson correlations were used to analyze associations, and Fisher’s r-to-z transformation was used to compare the size
1 The total viewing time anywhere on the screen did not differ statistically between the groups (M = 15.5 s, SD = 3.0 vs. M = 16.94 s, SD = 4.7, for the ASD and TD groups, respectively), t(28.9) ≤ 1, p = .4.
Johnels et al.: Face-Viewing Patterns in Young Children With ASD
2249
of correlations in the two study groups. SPSS Statistics 19 for Mac (IBM Corp., Armonk, NY) was used for all analyses, except the r-to-z transformation, which was conducted in an interactive resource (http://vassarstats.net/rdiff.html). A two-tailed alpha level of .05 was applied.
Results The analysis of variance did not reveal a main effect of AOI, F(2) < 1, p = .37 (lower bound), but it did reveal a significant Group × AOI interaction, F(1, 38) = 7.73, p = .008, hp 2 = .169. Table 1 reports mean viewing times and standard deviations for each group, along with results from follow-up pairwise independent t tests. The most striking result is that children with ASD spent twice as much time as TD children viewing nonmouth–noneyes areas but less time viewing the mouth. No significant difference was observed between the groups for viewing eyes.2 Because viewing areas other than the mouth and eyes appears to be characteristic of young children with ASD compared with language level–matched TD toddlers, we next examined how this metric discriminates ASD from TD in a receiver operating characteristic curve. The result of this analysis gave an area under the curve of .90 (SE = .05, p < .001), which is considered excellent discriminative ability. Because there is always a trade-off between sensitivity and specificity, the “best” cutoff depends on the purposes of the testing, for example, for screening versus gaining information for diagnostic decision making. Inspection of sensitivity and specificity suggested that two meaningful cutoff values could be identified. A first low cutoff of 30% viewing time for nonmouth–noneyes areas has a sensitivity of .91 and a specificity of .72 for ASD (positive predictive value = 91%; negative predictive value = 28%). An even lower cutoff could raise the sensitivity up to 1.0, but at the cost of greatly lowering the specificity (< .58). A second high cutoff of 44.5% has a sensitivity of .64 and a specificity of .93 for ASD (positive predictive value = 64%; negative predictive value = 96%). An even higher cutoff could raise the specificity but at the cost of disproportionately lowering the sensitivity ( .20). In the TD group, the correlation between language comprehension and mouth-viewing times was not significant (r = .08, p = .67), and the size of this correlation was significantly smaller than that seen in the ASD group (z = –2.12, p = .034). There was no correlation between language comprehension and times viewing eyes (r = –.01, p = .95) or times viewing noneyes–nonmouth areas (r = –14, p = .48). Nor did we find any significant correlations between chronological age and face-viewing patterns in either group (in the ASD group, all ps > .27; in the TD group, all ps > .08).
Discussion In comparison with a group of TD toddlers, children with ASD spent less time viewing the mouth and more time viewing nonmouth–noneyes areas when shown a film of a woman talking. A novel aspect of the analyses presented here was that we closely evaluated the children’s language comprehension level and matched our groups on this variable. This was motivated by recent research showing that language and mouth-viewing patterns are associated (cf. Hosozawa et al., 2012; Young et al., 2009). Arguably, previous research has not been able to determine whether the group differences in viewing patterns between children with and without ASD reflect diagnostic status or the children’s language comprehension level. In this study, we found that even with such control for language comprehension, marked group differences were found; however, these differences only partly mirrored those found in other studies comparing young children with and without ASD. The pattern displayed by the TD group is consistent with a large literature suggesting that the mouth and eyes have a special status when neurotypical individuals view a face (Yarbus, 1967), whereas individuals with ASD show a deviant pattern in this regard, possibly reflecting limited attentional bias to key regions of the face or even to the face in general (e.g., Chawarska, Volkmar, & Klin, 2010; Riby & Hancock, 2009). In fact, our data suggest that viewing areas other than the mouth and eyes may be a strong marker for ASD. A receiver operating characteristic curve analysis further showed that viewing other areas than the mouth and eyes had excellent discriminative ability for ASD also at an individual child level. Hence, this finding may turn out to be of both theoretical and clinical importance. Our finding that children with ASD attended to the mouth less than the TD group is in sharp contrast to that reported by Jones et al. (2008) but is consistent with other recent research (Chawarska & Shic, 2009; Chawarska et al., 2012; Nakano et al., 2010). Of course, it cannot be ruled out that our participants with ASD had viewed the mouth extensively when they were 2.5 years old. More important, though, the influential account forwarded by Jones et al. does not appear to be constrained to a particular age. Furthermore,
Journal of Speech, Language, and Hearing Research • Vol. 57 • 2246–2252 • December 2014
Figure 2. Scatter plot of the negative association between mouth viewing and language comprehension in the group with autism spectrum disorder.
we found no correlations between chronological age and viewing patterns in our group with ASD, suggesting that our group differences in (mouth-) viewing patterns probably cannot be explained by differences in chronological age. That said, an important avenue for future research will be to track and compare the developmental course of face viewing in children with and without ASD, preferably using a longitudinal design. In sum, this article highlights that mouth viewing appears to be a natural response for 30-month-old TD toddlers (cf. Falck-Ytter & von Hofsten, 2011) and suggests that general conclusions regarding extensive mouth viewing in young children with ASD may need to be revised. Interestingly, a few recent studies have looked more deeply into the heterogeneity of face viewing in school-aged children and adolescents with ASD and have described associations between face-viewing patterns and social, communicative, or linguistic proficiency in this group (Falck-Ytter, Fernell, Gillberg, & von Hofsten, 2010; Norbury, Brock, Cragg, Einav, Griffiths, & Nation, 2009; Rice, Moriuchi, Klin, & Jones, 2012). In this study, we were also interested in individual differences within the ASD group. We found that increased levels of mouth viewing in ASD were associated with poor language comprehension. This confirmed our suspicion that variability in language comprehension is an important factor to monitor when interpreting face-viewing patterns in young children with ASD. This result also underscores the importance of matching study groups on this variable; had our groups differed greatly in language comprehension, the outcome from the group comparison might very well have been different. In this sense, the results of the
present study may help explain the many divergent findings obtained in this field of research. Future research is needed to unravel the exact mechanisms involved, but we speculate that viewing the mouth may constitute adaptive (compensatory) adjustments allowing children with ASD and low language comprehension to decode language, whereas children with ASD and better language ability do not need to use visual support from the mouth area to the same extent. This idea is related to the ones proposed in Lewkowicz and Hansen-Tift (2012), although they focused on language production rather than comprehension in TD infants, and Hosozawa et al. (2012), who studied children with specific language impairment. However, if this explanation is correct, it is somewhat difficult to explain why we did not observe the same correlation in the TD group (which was larger and similar in terms of language comprehension). One possible explanation could be that only children with delayed or disordered language perceptually orient to the mouth area for compensatory visual clues (Hosozawa et al., 2012). Support for this account was also obtained in the study of adolescents with ASD by Norbury et al. (2009). These authors—as did we—found that a measure of language skills (i.e., a composite measure of receptive vocabulary and sentence repetition) correlated negatively with mouth viewing. However, this was only the case for the subgroup with impaired language skills; among those with good language skills, the association was not significant and in fact appeared to be possibly reversed. We end by discussing two shortcomings of this study and also describe directions for future research. First, similar
Johnels et al.: Face-Viewing Patterns in Young Children With ASD
2251
to many other studies in this field, our study included a relatively small sample of children. Therefore, future research needs to replicate these findings in larger samples. Second, future research is needed to further explore possible mediating mechanisms and developmental consequences of the results presented here. Why do young children with ASD attend to noneye–nonmouth areas when looking at a person talking? And why, more exactly, is mouth viewing in young children with ASD associated with poor language comprehension? An important question in this context is also whether the nature of the association between mouth viewing and limited language comprehension differs depending on how language comprehension is measured. In a new ongoing study, we are exploring how children’s orientation toward the mouth of a person talking is related not only to test scores of language comprehension, but also to the child’s comprehension of what that specific person in the film is saying.
References Arvidsson, E., & Köröndi, J. (2011). Svensk normering av språkföstålsedelen i Reynell Developmental Language Scales III 4:6-4:11 år—Sambandet mellan gemensam läsning och språkförståelse [Swedish norms on Reynell Developmental Language Scales III]. Gothenburg, Sweden: University of Gothenburg. Blais, C., Jack, R. E., Scheepers, C., Fiset, D., & Caldara, R. (2008). Culture shapes how we look at faces. PLoS ONE, 3, e3022. Campbell, R. (2008). The processing of audio-visual speech: Empirical and neural bases. Philosophical Transactions of the Royal Society B: Biological Sciences, 363, 1001–1010. Chawarska, K., Macari, S., & Shic, F. (2012). Context modulates attention to social scenes in toddlers with autism. Journal of Child Psychology and Psychiatry, 53, 903–913. Chawarska, K., & Shic, F. (2009). Looking but not seeing: Abnormal visual scanning and recognition of faces in 2- and 4-year old children with autism spectrum disorder. Journal of Autism and Developmental Disorders, 39, 1663–1672. Chawarska, K., Volkmar, F., & Klin, A. (2010). Limited attentional bias for faces in toddlers with autism spectrum disorders. Archives of General Psychiatry, 67, 178–185. Edwards, S., Fletcher, P., Garman, M., Hughes, A., Letts, C., & Sinka, I. (1997). The Reynell Developmental Language Scales III: The University of Reading Edition. Windsor, England: NferNelson. Falck-Ytter, T., Fernell, E., Gillberg, C., & von Hofsten, C. (2010). Face scanning distinguishes social from communication impairments in autism. Developmental Science, 13, 864–865. Falck-Ytter, T., & von Hofsten, C. (2011). How special is social looking in ASD—A review. Progress in Brain Research, 198, 209–222. Fort, M., Kandel, S., Chipot, J., Savariaux, C., Granjon, L., & Spinelli, E. (2012). Language and Cognitive Processes, 28, 1207–1223. Holmqvist, K., Nyström, M., Andersson, R., Dewhurst, R., Halszka, J., & van de Weijer, J. (2011). Eye tracking: A comprehensive guide to methods and measures. Oxford, England: Oxford University Press. Hosozawa, M., Tanaka, K., Shimizu, T., Nakano, T., & Kitazawa, S. (2012). How children with specific language impairment view social situations. Pediatrics, 129, 1453–1460. Jones, W., Carr, K., & Klin, A. (2008). Absence of preferential looking to the eyes of approaching adults predicts level of social
2252
disability in 2-year-old toddlers with autism spectrum disorder. Archives of General Psychiatry, 65, 946–954. Kjellmer, L., Hedvall, Å., Holm, A., Fernell, E., Gillberg, C., & Norrelgen, F. (2012). Language comprehension in preschoolers with autism spectrum disorders without intellectual disability: Use of the Reynell Developmental Language Scales. Research in Autism Spectrum Disorders, 6, 1119–1125. Klin, A., Lin, D. J., Gorrindo, P., Ramsay, G., & Jones, W. (2009). Two-year-olds with autism orient to non-social contingencies rather than biological motion. Nature, 459, 257–261. Lewkowicz, D. J., & Hansen-Tift, A. M. (2012). Infants deploy selective attention to the mouth of a talking face when learning speech. Proceedings of the National Academy of Sciences, USA, 109, 1431–1436. Lord, C., Risi, S., Lambrecht, L., Cock, E. H., Jr., Leventhal, B. L., DiLavore, P. C., Pickles, A., & Rutter, M. (2000). The Autism Diagnostic Schedule—Generic: A standard measure of social and communication deficits associated with the spectrum of autism. Journal of Autism and Developmental Disorders, 30, 205–223. Magnusson, M. (1997). Rationality of routine health examinations by physicians of 18-month-old children: Experiences based on data from a Swedish county. Acta Paediatrica, 86, 881–887. Nakano, T., Tanaka, K., Endo, Y., Yamane, Y., Yamamoto, T., Nakano, Y., . . . Kitazawa, S. (2010). Atypical gaze patterns in children and adults with autism spectrum disorders dissociated from developmental changes in gaze behaviour. Proceedings of the Royal Society B: Biological Sciences, 277, 2935–2943. Norbury, C. F., Brock, J., Cragg, L., Einav, S., Griffiths, H., & Nation, K. (2009). Eye-movement patterns are associated with communicative competence in autistic spectrum disorders. Journal of Child Psychology and Psychiatry, 50, 834–842. Reisberg, D., McLean, J., & Goldfield, A. (1987). Easy to hear but hard to understand: A lip-reading advantage with intact auditory stimuli. In B. Dodd & R. Campbell (Eds.), Hearing by eye: The psychology of lip-reading (pp. 97–113). Hillsdale, NJ: Erlbaum. Riby, D. M., & Hancock, P. J. B. (2009). Looking at movies and cartoons: Eye-tracking evidence from Williams syndrome and autism. Journal of Intellectual Disability Research, 53, 169–181. Rice, K., Moriuchi, J. M., Klin, A., & Jones, W. (2012). Parsing heterogeneity in autism spectrum disorders: Visual scanning of dynamic social scenes in school-aged children. Journal of the American Academy of Child & Adolescent Psychiatry, 53, 238–248. Robins, D. L., Fein, D., Barton, M. L., & Green, J. A. (2001). The modified checklist for autism in toddlers: An initial study investigating the early detection of autism and pervasive developmental disorders. Journal of Autism and Developmental Disorders, 31, 131–144. Sasson, N. J., & Elison, J. T. (2012). Eye tracking young children with autism. Journal of Visualized Experiments, 61, e3675. Sumby, W. H., & Pollack, I. (1954). Visual contribution to speech intelligibility in noise. Journal of the Acoustical Society of America, 26, 212–215. Yarbus, A. L. (1967). Eye movements and vision. New York, NY: Plenum Press. Young, G. S., Merin, N., Rogers, S. J., & Ozonoff, S. (2009). Gaze behavior and affect at 6 months: Predicting clinical outcomes and language development in typically developing infants and infants at-risk for autism. Developmental Science, 12, 798–814.
Journal of Speech, Language, and Hearing Research • Vol. 57 • 2246–2252 • December 2014
Copyright of Journal of Speech, Language & Hearing Research is the property of American Speech-Language-Hearing Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
