Journal of the International Neuropsychological Society (2008), 14, 967–979. Copyright © 2008 INS. Published by Cambridge University Press.
Printed in the USA. doi:10.10170S1355617708081216
f MRI activation during a language task in adolescents with ASD
TRACEY A. KNAUS,1 ANDREW M. SILVER,1 KRISTEN A. LINDGREN,1 NOUCHINE HADJIKHANI,2,3 and HELEN TAGER-FLUSBERG 1
1 Department 2 Martinos
of Anatomy and Neurobiology, Boston University School of Medicine, Boston, Massachusetts Center for Biomedical Imaging, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts 3 Brain Mind Institute, EPFL, Lausanne, Switzerland (Received December 7, 2007; Final Revision June 23, 2008; Accepted June 26, 2008)
Abstract Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by language and communication impairments, social impairments, and repetitive behaviors or restricted interests. Previous studies of semantic functions have found differences in semantic processing and differences in the activation of the language network in adults with ASD compared to controls. The goal of this study is to examine semantic functions in adolescents with ASD compared to typically developing adolescents. We utilized f MRI with a reading version of a response-naming task to investigate activation in 12 right-handed adolescent boys with ASD and 12 typically developing boys. Both groups performed the task at ceiling levels. Boys with ASD had significantly stronger activation than controls in Broca’s area, which was less left lateralized in ASD individuals. Controls had a significant correlation between frontal and temporal language area activation in the left hemisphere, whereas ASD adolescents did not. Direct group comparisons revealed additional regions activated in the ASD group relative to the control group. These results suggest differences in semantic organization, approaches to the semantic task, or efficiency in semantic processing in ASD adolescents relative to typically developing adolescents. (JINS, 2008, 14, 967–979.) Keywords: Autistic disorder, Functional MRI, Broca’s area, Wernicke’s area, Semantics, Asymmetry
Autism spectrum disorder (ASD), a term including classic autism, pervasive developmental disorder-not otherwise specified (PDD-NOS), and Asperger syndrome, is a neurodevelopmental disorder characterized by impairments in language and communication, social deficits, and repetitive behaviors or intense interests (American Psychiatric Association, 1994). Although deficits in language and communication are one of the core features of ASD, the specific nature of these impairments remains unclear. Complicating this, the entire range of language abilities occurs in ASD, including individuals who never develop language to those
Correspondence and reprint requests to: Tracey Knaus Ph.D., Department of Anatomy & Neurobiology, Boston University School of Medicine, 715 Albany Street L-814, Boston, MA 02118. E-mail: email@example.com
who perform normal or above on standardized language tests (for review, Tager-Flusberg et al., 2005). Certain aspects of language, such as pragmatics are impaired in all individuals with ASD, regardless of functioning level (see TagerFlusberg et al., 2001). Other aspects of language, however, such as phonology, syntax, and semantics may not show impairment or may be impaired only in a subgroup (see Tager-Flusberg et al., 2001). Semantic processing in ASD has been studied in older adolescents (15 years or older) and adults using several behavioral paradigms. Long-term memory studies have suggested semantic processing differences. Toichi and Kamio (2002) found semantic, compared to phonological or perceptual, processing of verbal items facilitated long-term memory in controls. In ASD, semantic facilitation did not occur, suggesting decreased use of semantic information. Mottron et al. (2001) demonstrated that when oriented to 967
Broca’s and Wernicke’s areas play key roles in semantic functions (for reviews. but found that when pictures. The ASD group compared to controls had increased extrastriate visual cortex activation.. 2004. Knaus et al. Subjects indicated whether a visually presented word was positive or negative compared with a case decision of the same words. There was also decreased functional connectivity between anterior and posterior language regions. 2004).A. Similar to behavioral findings. 2003) and implies different semantic processing in the ASD group. Müller and colleagues studied sentence production compared to sentence repetition. utilizing f MRI.. 2004. but not in ASD. did not involve language generation. when controls were given semantic cues at retrieval. Another group. 1998).. so alternatively. 1998. Thus.. whereas in ASD. 2003. direct group comparisons did not reveal this difference. Müller et al. but not in controls (Kamio & Toichi. ASD individuals also had smaller activation clusters than controls in left inferior frontal regions. However. pants indicated whether high.968 the semantic features. both groups had better recall than when oriented to phonological features or no orientation. A correlation was found between priming performance with performance IQ and Raven’s Progressive Matrices in ASD. Just et al. Harris et al. Foundas. functional neuroimaging studies in ASD have demonstrated differences in these regions during semantic tasks. (2007) had subjects determine whether visually presented words belonged to a given category and contrasted this with subjects indicating whether a specified letter occurred in letter strings. Several studies have examined semantics or syntax using sentences in ASD adults. In one study. For their task. The ASD group also showed similar activation in Broca’s to the semantic and perceptual task. 1997). 2006. 2000). controls also had activation in the left inferior temporal region (Müller et al. Two f MRI studies focused on semantic processing of words in ASD and suggested atypical semantic processing or organization. 2006).. recall was better than with phonological cues. which relies heavily on Broca’s area. (2007) showed semantic priming effects in controls. pictures may have an advantage over words in accessing meaning. Behavioral and imaging studies have suggested different semantic organization or strategies in semantic processing. whereas controls had activation in this region only during the semantic task. however. Functional neuroimaging has been used to investigate the neurobiological substrates of language in ASD. Based on these findings.. with several studies reporting decreased Broca’s activation (Gaffrey et al. In a study of ASD individuals without early language delay. The first studies were PET and f MRI studies that used passive paradigms to explore auditory language processing. We chose a task that each individual could easily and accurately perform so that differences in activation between the groups could not be attributed to differences in performance levels (Bookheimer. Kamio et al. We used a visually presented response-naming task (Bookheimer et al. 2007. in ASD. suggesting the contribution of non-verbal factors and the possibility of different strategies being used. however. These prior studies. we predicted that
. Harris et al.. 2004. (2006) demonstrated less activation in Broca’s area and increased middle temporal gyrus activation in ASD relative to controls. (2006) demonstrated activation in the ASD group in parietal and occipital imageryrelated regions in low. Another priming study found no group differences. The control group was more accurate than the ASD group for the category task. instead of words. focusing on syntactic processing during a visually presented sentence task.. semantic and phonological cues had the same effect. there was decreased left BA 46 and left thalamus activation in the autistic group relative to controls (Müller et al... whereas controls had more activation in these regions during highrelative to low-imagery sentences. 2006) and some finding increased Wernicke’s area activation (Harris et al.. this is the only study that has used a semantic task involving semantic integration and word generation.. This further supports decreased use of semantic information in ASD. with similar performance in both groups. 2001. The purpose of this study was to investigate semantic processing in adolescents with ASD and typically developing adolescents. which was not found in ASD (Toichi & Kamio. Semantic priming has also been used to examine semantic functioning in ASD. were used as primes there was an increased priming effect in ASD.. Toichi and Kamio (2001) also found no differences in semantic priming in ASD compared to controls. Kana et al. 1999). Gaffrey et al. Kana et al. Almost no previous semantic f MRI tasks in ASD have involved language production and to our knowledge. another study found that controls had better long-term memory and increased verbal associations to concrete than abstract nouns and a correlation between recall and number of associations. this is the first study to examine language activation patterns in ASD adolescents as young as 11 years old. found increased posterior activation (left superior temporal gyrus) and reduced activation in frontal language areas (left inferior frontal gyrus) in autism compared to controls (Just et al. Similarly. 2006. Vigneau et al. The ASD’s task performance was faster and less accurate than controls’.. we might not expect to find group differences in Broca’s activation with this task. In another study. partici-
T. 2002. To our knowledge. 1999). Bookheimer. f MRI semantic studies have been performed mainly in adults. We therefore hypothesized that ASD individuals would have decreased left frontal language activation and increased activation of left temporal language regions. In addition. 2004). Gervais et al. Just et al. both groups had activation in the left inferior0middle frontal gyrus with left lateralization in the perisylvian region. which corresponded to increased errors on the semantic task. These studies found decreased temporal activation and decreased left lateralization of activation in autism compared to controls (Boddaert et al. 2000).and low-imagery sentences were true or false and there were no group differences in performance.and high-imagery sentences.
58) 123. pixel matrix 5 256 3 256. learning. 1995) to assess language abilities. 1994) using the Autism Diagnostic Interview-Revised (ADI-R.
MATERIALS AND METHODS Participants
Participants included 12 adolescents with ASD (autism. flip angle 5 88. an expert clinician confirmed that all individuals met criteria for ASD. parents gave informed written consent and children provided informed written or verbal assent prior to participation.24) ASD (n 5 12) Range 11.
A block-design was used with a reading version of a response-naming task (Bookheimer et al. All participants were male.17) 122. Descriptive characteristics of participants.5–19. TE 5 3. with a known genetic disorder.4–19. and Recalling Sentences (Expressive language subtests). For ASD subjects. Subjects 18 years and older provided informed written consent prior to participation. diagnosis was based on DSM-IV criteria (American Psychiatric Association. TE 5 35 ms.4 mm gapless axial images.33 (4.55) 108.25 (23. Asperger syndrome. Images were acquired on a Philips 3 Tesla Intera scanner.25 (5. f MRI scans were acquired using Blood Oxygen Level Dependent (BOLD) contrast with the following parameters: TR 5 2000 ms.69) Range 11.15) 98. indicating complete left-handedness.71) 18. During the response-naming task. 1998).g.8 13–124 67–136 70–121 64–128 66–137 62–120 62–132 4–13 1–5
the ASD group compared to typically developing adolescents would rely on different cortical areas during the semantic task. by pressing a button. 1997) and a control letter-judgment task.3 ms. In addition. the word that best matched what
. subjects were shown a three-word phrase (e. or who had seizures within the last three years were excluded. from two displayed options. indicating complete right-handedness. pixel matrix 5 128 3 128. and a modified version of the Dean handedness inventory (Piro. The modified Dean handedness consists of 12 unimanual tasks and scores range from 224.. See Table 1 for participant details. ages 11–19 years.10) 116.25 (9.5 mm.92 (1. Individuals with frank neurological damage.17 (14. monolingual (English). or PDD-NOS) and 12 typically developing adolescents.
All participants practiced in a mock scanner prior to the actual MR scanning.92 (9. and right-handed based on writing hand.92 (15. 36 contiguous slices.23) 105.92 (8. flip angle 5 908.14) 2. 1994) and the Autism Diagnostic Observation Schedule (ADOS.8 110–123 101–149 99–132 100–136 94–131 91–123 95–130 Mean 15.33 (23. They then chose. Mean. self-report.46 (2. gloves). keeps hands warm) and asked to think of what word was being described (e.4 ms.47) 102.67) 106.. FOV 5 230 mm. Lord et al.. psychiat-
ric. All data in this manuscript were collected in compliance with the Boston University School of Medicine Institutional Review Board. Kaufman & Kaufman. Lord et al. Formulated Sentences.94 (2. and the Clinical Evaluation of Language Fundamentals (CELF-III. who were born prematurely (less than 35 weeks).88) 93. to 124.Language activation in ASD
Table 1.33 (11. 2004) to assess IQ.52) 113.g. Word Classes (Receptive language subtests). slice thickness 5 3. or neurologic disorders.86) 8. FOV 5 230 mm.08 (21. Typically developing individuals had no history or current diagnosis of developmental. 1. Volumetric T1-weighted images were obtained as a series of 95–110. standard deviation.42 (19. and range of demographic variables for each group Subject Demographics Controls (n 5 12) Mean Age Handedness Verbal IQ Non-Verbal IQ Full-Scale IQ CELF receptive CELF expressive CELF total ADOS social ADOS communication 14.48) 17.. Four subtests of the CELF-III were administered: Concepts and Directions.25 (11. Participants were administered the Kaufman Brief Intelligence Test (K-BIT-II. aligned parallel to the intercommisural plane... 2000).35) 103.50 (21. Two FE-EPI axial sequences aligned parallel to the intercommisural plane were acquired. Threedimensional MPRage was used with technical factors of: TR 5 7.67 (3. For subjects under 18 years old. Semel et al.57) 115.
and superiorly by the inferior frontal sulcus. and non-verbal IQ (NVIQ) measures from the KBIT-II were computed. To control for multiple comparisons. Temporal language areas (pSTG. a blank screen for 0. each consisting of three blocks of response-naming and three blocks of the control task. To examine differences in percent signal change. A MANOVA was performed with the mean translation and rotation in each of the 3 directions for each run as dependent variables and group as the independent variable. . Because frontal and temporal regions are critical for semantic processing and to better account for individual differences in anatomy.2 seconds. three strings of letters were presented and subjects indicated. Percent signal change was calculated in both regions in the right and left hemisphere as (mean of the modeled effect0mean of the baseline effect) *100. a general linear model (GLM) fitting the task block’s time vector convolved with a gamma variate estimate of hemodynamic response was performed for each run. Each block was 28 seconds and consisted of 4 trials.org). correlations between percent signal change of each ROI with ADOS communication and social scores were examined. For the control task. To test for group differences in movement. A trial was presented every six seconds. ROIs were defined in the sagittal plane. with the three-word phrase or letter strings presented for 3. and a blank screen for 0. The relationship between percent signal change of each ROI with CELF-III receptive and expressive standard scores was examined. For each run. For individual analyses. Regions of interest (ROIs) were anatomically defined using well-established anatomical landmarks and all measurements were done by 1 rater (TAK) experienced in anatomically defining these regions (see Knaus et al. 1b). resulting in 48 trials across the two runs. verbal IQ (VIQ). resulting in an activation map (the 2log probability map which corresponds to the t-statistic). with a button press. The group analysis function (Worsley et al. posteriorly by the pre-central sulcus. At the beginning of each block a crosshair was presented for 4 seconds. the maps of the effect and standard error of the effect for each individual run were utilized to generate a 2log p. For this. 2004. the mean of the absolute value of translations and rotations across each run was calculated for each direction for each subject. For the stimuli used see the Appendix. 2002) was used with fixed effects. a map of the effect. effect.
T. vation and baseline for the two runs in each individual. The stimuli were presented in red lettering on a black background using E-Prime software (http:00www. a repeated measures MANOVA was computed with the percent signal change of each ROI as dependent variables. including PT) were bounded anteriorly by the most anterior Heschl’s sulcus. posteriorly by the most posterior point of the Sylvian fissure. A MANOVA was performed to examine differences in degree of AQ with group as the independent variable and AQ of frontal and temporal areas as dependent variables.3 seconds.5 seconds. with positive AQ indicating higher percent signal change in the left and negative AQ indicating higher percent signal change in the right.11) 5 2. a practice session in the mock scanner was done. which was overlaid on each subject’s respective high-resolution T1-image. The relationship between percent signal change and behavioral measures was examined within each group using Pearson correlations. during which each subject performed one run consisting of different stimuli from those used in the actual scanning.pstnet. The first 2 volumes of each run were discarded to allow for magnet stabilization. 2007). the output files from motion correction were examined to ensure that there was not significant motion. p . Subjects with movement in any direction greater than 2 mm or 2 degrees were excluded. 1992) was utilized (Fig..05). Correlations between percent signal change of each ROI with age.. The AQ was calculated as (left percent signal change 2 right percent signal change)0 (left percent signal change 1 right percent signal change). using a 3-D Gaussian kernel with 6-mm full width at half-max. The words task and a baseline plus drift were modeled. During scanning. and group as the independent variable. 1a). 1999). and a map of the standard error of the effect. hemisphere as the repeated measure.36. whether the letters were in upper or lower case. When the Sylvian fissure gently sloped upward. The relationship between percent signal change of each ROI was also examined. This task was chosen so that areas related to primary visual processing and motor areas related to the button press could be subtracted out of the language activation. In the ASD group. Bonferroni correction was used with the combined activation map for each participant thresholded to p 10 27 .
f MRI analyses were carried out using Neurolens (www. Prior to scanning. 2006. The motion correction parameters for translations and rotations in each direction were included as regressors to improve the model. Knaus et al. to combine acti-
. This revealed no significant group differences in motion for any direction for either run (F(12. Spatial smoothing was also performed on each functional run. 2 functional runs were completed. the knife-cut method (Witelson & Kigar.A. com0products0e-prime0). Each run was motion corrected using a volume registration algorithm in which each volume was co-registered to a target volume (volume 85 of each run) (Cox & Jesmanowicz. the two word choices or the words upper and lower displayed for 2 seconds. and standard error effect map for the two runs together. neurolens. activation within these regions was examined in individuals. An asymmetry quotient (AQ) of the percent signal change was calculated for each ROI. and superiorly by the horizontal ramus of the Sylvian fissure. Activation in both banks of all of these gyri was included (Fig. Frontal language areas (pars triangularis and pars opercularis) were bounded anteriorly by the anterior horizontal ramus of the Sylvian fissure.970 they had thought of.
the end of the Sylvian fissure and the superior boundary. (a) Boundaries used for frontal language ROI.048.. The activation map. ADOS social 1 communication scores were added as regressors to examine regions correlated with the ADOS. All activation between pre-central sulcus and anterior horizontal ramus. age was added as a regressor and regions correlated with age were examined. 2000).014). Accuracy measures for this subject were not included. we computed a transformation matrix by fitting the first functional run of each subject to a group averaged EPI brain in MNI space. perceptual) as factors and accuracy or reaction time as dependent variables.001. (b) Boundaries used for temporal language ROI. This transformation matrix. p 5 . in a separate analysis. see Table 2. demonstrating faster reac-
. 1995). for each group was overlaid onto an individual’s T1 scan. To examine differences in accuracy and reaction time. but reaction time data were included for this subject for all trials in which a response was recorded. Using an individual voxel probability threshold of p 5 . For accuracy and reaction time data. No subject made errors on more than 2 trials across both runs. For reaction time. 1. to help with localization of activation.21) 5 7. bounded superiorly by the inferior frontal sulcus. thresholded using a cluster size of 51 voxels. was included. during the practice run this subject made no errors. DS 5 Diagonal Sulcus. which resampled the f MRI acquisition voxels to 2-mm isotropic voxels. however. Program AlphaSim (Ward. There was a significant effect of group (F(1.22) 5 130. Dashed lines indicated posterior boundary. For the ASD group.
For the group analysis. AHR 5 Anterior Horizontal Ramus. were at ceiling. . which was converted to a 2log p map. Heschl’s sulcus was the anterior boundary.84. 2-factor ANOVAs were done with group and task (semantic.Language activation in ASD
Fig. Peaks of activation were identified for each contrast using a cluster threshold of 51 MNI transformed voxels.17. An image series of the effect maps for all subjects was made and
sorted into blocks with members in each group ordered together. Pre-CS 5 Pre-Central Sulcus. was then applied to the effect and standard error of the effect maps from the combined runs for each subject. HG 5 Heschl’s gyrus. Complete behavioral data were not available for 1 subject due to button box errors. Both groups’ performances. a Monte Carlo simulation that is part of AFNI was used to determine cluster size and significance. These transformed effect and error maps were used in the group analysis with mixed effects. with the ASD group having higher accuracy. Direct group comparison was done by using the transformed effect and standard error of the effect maps for all subjects. Cluster size thresholding was used to adjust for multiple comparisons (Forman et al. A block design GLM with regressors for each group type was then performed. We examined regions more active in the ASD group and regions more active in the control group. For each group. IFS 5 Inferior Frontal Sulcus. however.
RESULTS Behavioral Performance
All individuals were able to do the task easily and with only 1 practice session.001). transformed into MNI space. For accuracy. the horizontal ramus of the Sylvian fissure. indicated that using a minimum cluster size of 36 original voxels (51 MNI transformed voxels) resulted in an overall significance of p 5 . The effect size was then divided by the residual standard error to produce a t-map. A block design GLM was fitted to all the effect maps for an individual group. p . there was a significant effect of task (F(1. there were no significant effects of task or task by group interaction. AAR 5 Anterior Ascending Ramus.
.27 (102.89.001) and a significant group effect (F(2.36. 3). or IQ were significant in either group. p . In the control group.9 (2.001) and temporal (F(1. p . There were no regions correlated with age. The group difference was significant only for frontal ROIs (F(1. There was a significant group difference in degree of AQ (F(2.54) 97. . demonstrating stronger AQ in controls than ASD.83) 666. however.10.
tion times for the visual than semantic task. pars orbitalis.934.
Fig. Another large cluster was in left temporal regions. clusters were present in left orbito-frontal cortex. Because of the design of the task.29.11.031). . 4a).27) 861.6 (1.509 (. See Table 4 for correlations in each group. p 5 .66.273) .440)
Functional Activation — Individuals
Every subject in both groups had activation in left frontal language areas and left temporal language regions with very little activation in homologous right hemisphere areas. banks of the inferior frontal gyrus. This correlation was not significant in either hemisphere in the ASD group (Fig. The hemisphere by group interaction was close to significant (F(2. Mean accuracy presented as percent correct and reaction time in ms for the response-naming task and visual processing task for each group
Controls Semantic Task Visual Task Accuracy Reaction Time Accuracy Reaction Time 99. p 5 . They then chose the word that matched what they had generated.001) ROIs. reaction time was not a measure of time to generate a response. age. There was also a significant group effect (F(1.46 (134. At the univariate level.446) ASD (n 5 12) . .63) 949. 4b).22) 5 7.972
Table 2. In the ASD group. In addition. . Knaus et al. and left superior pre-central gyrus.001). Correlations with ADOS scores were also not significant. Mean (standard deviation) asymmetry quotient for each ROI in each group Controls (n 5 12) Frontal % Signal AQ Temporal % Signal AQ . 2).83. Large clusters were also in left anterior and posterior superior temporal regions. Subjects were instructed and given time to think of their response prior to the word choices. Again.001). After correcting for multiple comparisons.008).00) 541.21) 5 33. p . They also were not given specific instructions to respond as quickly as possible. p 5 .21) 5 6. the largest cluster was in the
Table 3.38 (114. with a larger difference between left and right hemispheres in controls than ASD.60 (103. but not in the right hemisphere. Follow-up univariate tests revealed significant differences in frontal percent signal change AQ (F(1. and parahippocampal area.22) 5 11. Mean AQs are presented in Table 3. 21) 5 15.22) 5 47. the hemisphere by group interaction was significant for the frontal region (F(1.
Functional Activation – Group
For the control group. The largest cluster was in the left hemisphere in Broca’s area. p . Graph showing percent signal change of activation in left frontal language regions for each subject in each group. it included the pars triangularis.21) 5 3. p 5 . which included the inferior temporal gyrus.48) 100. there were eight clusters of activation (Table 6 and Fig.45. there was a significant correlation between frontal and temporal percent signal change in the left hemisphere (r 5 . seven clusters of activation were identified (Table 5 and Fig. For percent signal change there was a significant hemisphere effect (F(2.A.011) with the ASD group responding faster than controls for both tasks. p .716 (.11.08)
T.001). Follow-up univariate tests indicated that for the hemisphere effect there was higher percent signal change in left regions than right for frontal (F(1. indicating significantly higher percent signal change in ASD individuals compared to controls (Fig.066). and medial orbito-frontal regions.216 (.0 (0.421) .507 (.3 (1.22) 5 47.18. p 5 . . no correlations between percent signal change with CELF-III scores. left inferior temporal gyrus. 2.22) 5 32.22) 5 5.58) ASD 99. p .002). fusiform gyrus.
786 .232 .718 .639 .120 .581 — — .841 2.226 .925 2.231 Left Frontal % Signal .245 2.152 .831 — — Left Temporal % Signal 2.340 .336 2.392 .364 .540 — — — .175 .Language activation in ASD
Table 4. Correlations (r.079 .393 2.272 .172 .258 2.459 .030 .631 .296 .177 .065 .167 .934* .137 .388 — 2.208 2.337 .516 .350 2.592 .030 .244 .304 .856 .305 .069 .719 2.927 2.469 2.and p-values) between percent signal change of each ROI with behavioral measures for each group Controls (n 5 12) Left frontal % signal Age CELF Rec CELF Exp K-BIT VIQ K-BIT NVIQ ADOS Com ADOS Social Left Frontal % Signal Right Frontal % Signal r p r p r p r p r p r p r p r p r p 2.220 .718 .393 — Right Temporal % Signal 2.162 .550 .445 .808 — — — — Right frontal % signal 2.173 .088 .117 .656 .632 .672 2.086 — — — — Left temporal % signal 2.841
.615 .059 .057 .378 .272 .197 .162 .453 .730 .111 .000 — Right temporal % signal 2.629 .337 2.274 .419 .257 .374 .064 .605 2.473 .471 .861 .302 .116 .144 .028 2.615 2.355 .028 2.283 2.230 .493 — — ASD (n 5 12) Right Frontal % Signal .636 .590 .117 .027 .065 .139 2.587 .133 .151 .
0 33.0 211.0 237. pars orbitalis. and orbito-frontal. and superior parietal regions. Cluster size is in voxels. banks of the inferior frontal gyrus.77 5.37 Anatomical region Left inferior frontal gyrus. and right cerebellum. had more activation than controls in 11 clus-
ters (Table 7).42 4.
left hemisphere in Broca’s area. When groups were directly contrasted. pre-central gyrus. There were also clusters in left medial superior and middle frontal gyrus and medial Broca’s area. there was one small region near the posterior corpus callosum0cingulate that was more active in controls than the ASD group. Activation was also in the left in the brainstem. 3.5 z 14 212 6 216 212 240 50 Cluster size 778 220 168 165 88 64 54 t-value 6. inferior temporal and fusiform gyri. parahippocampus.
Fig. pars orbitalis. insula.10 4. ASD subjects. Smaller clusters were found in the right in pre-central gyrus. we found differences in activation between ASD and typically developing adolescents. medial orbito-frontal Left anterior superior temporal gyrus Left posterior superior temporal gyrus Left inferior temporal gyrus.47 4. Regions of activation in the typically developing group for response-naming compared to visual processing task. Other clusters were in the left in inferior frontal gyrus. The biggest clusters were located in right inferior and middle frontal gyri and right middle temporal gyrus.51 5. frontal and temporal activation was
Table 5. Clusters were also in left thalamus and caudate. and posterior superior temporal gyrus. hippocampus and fusiform gyrus. Knaus et al. pars triangularis. orbito-frontal. There was 1 region correlated with age located in right lateral parietal0postcentral gyrus and no regions were correlated with ADOS scores. extending into superior pre-central gyrus. however.43 5. parahippocampus Left medial orbito-frontal Left inferior temporal gyrus Left superior pre-central gyrus
. MNI coordinates are for peak activated voxel in each cluster Control group activations MNI coordinates x 250 254 258 232 232 236 244 y 25. left medial superior frontal gyrus.0 2. which included posterior middle and inferior temporal gyri.974
T. Correlations between left frontal percent signal change with left temporal percent signal change in the control group and in the ASD group. In the left hemisphere. fusiform gyrus.
During semantic processing. and right lateral inferior frontal gyrus.0 237. including the pars triangularis. A large cluster was also present in the anterior and posterior superior temporal gyrus.A. The ASD group had significantly more activation in Broca’s area.0 213. which was less lateralized compared to controls.
correlated in controls but not in ASD individuals. McAlonan et al.. 2005. There are important differences between our study and previous language studies. 2004. (a) Sagittal slices showing control group activation of Broca’s and Wernicke’s areas during response-naming minus perceptual task. In the current study. Because of its critical role in language. which may account for this discrepant finding. de Fossé et al. parahippocampus.. superior pre-central gyrus.63 5. 2005). 2006). white) indicate positive responses. orbito-frontal Left anterior and posterior superior temporal gyrus. MNI coordinates are for peak activated voxel in each cluster ASD group activations MNI coordinates x 248 262 28 24 36 28 232 62 y 23 247 217 7 281 221 239 31 z 210 8 12 60 226 210 217 18 Cluster size 4117 1976 741 605 313 256 249 63 t-value 8. One difference is task performance. Just et al. red. purple. 1999. 4. insula. caudate Left medial superior frontal gyrus Right cerebellum Left brainstem. Broca’s area is involved in semantic encoding and retrieval and on-line manipulation of semantic representations (Blumenfeld et al. 2007. however.94 5.68 4. The ASD group relative to the typically developing group also had additional regions activated during semantic processing. however... Harris et al. both groups performed at ceiling levels..43 4..09 7. posterior middle temporal gyrus. pars triangularis. 2004.27 7. Colder colors (blue.40 6. Kana et al.. we did not expect an increase in activation in ASD. Broca’s has been examined. Regions of activation in the ASD group for response-naming compared to visual processing task. hippocampus Left fusiform gyrus Right lateral inferior frontal gyrus
. Cluster size is in voxels. (b) The same sagittal slices showing Broca’s and Wernicke’s area activation in the ASD group for response-naming minus perceptual processing.32 Anatomical region Left inferior frontal gyrus. pars orbitalis. Gaffrey
et al. in some prior studies. 2006. We predicted differences in this critical language area.. Herbert et al. posterior inferior temporal gyrus Left thalamus. the ASD group had significantly lower
Table 6. 2006. black) indicate negative responses and hotter colors (yellow. 2002.Language activation in ASD
Fig.. with structural and functional language studies demonstrating differences in ASD (Abell et al.
Harris et al. 2006) or sentences (Just et al. 2000). 2001) in ASD. By using a task with ceiling performance.. (2004) found decreased integrity of white matter in ASD adolescents. 2006). involve understanding words in context. Control group for response-naming compared to visual task. Just et al.. our sample involved only adolescents and therefore included a small age range.. with our task involving language generation.. Our finding supports Just and colleagues (2004) who were the first to show decreased functional connectivity during a language task in adults with autism.A. Imaging studies have also suggested different strategies in semantic processing. approaches.73 3. but may reflect regions that are compensating for not being able to do the task or associated with errors (Bookheimer. We only found one small region in the ASD group associated with age. 2001). Other studies have involved making decisions about individual words (Gaffrey et al. whereas our study was comprised of adolescents. 2007) in ASD. The sample has also differed between studies.73 3.43 3. 2003. 2007. Harris et al. Gaffrey and colleagues (2007) demonstrated some regions of activation that correlated with increasing semantic errors. however.71 3. 2007. activation differences are not related to differences in task performance.. fusiform gyrus Left pre-central gyrus Left posterior superior temporal gyrus Left medial superior0middle frontal gyrus Left medial pars triangularis0pars opercularis Right pre-central gyrus Right orbito-frontal Right superior parietal
accuracy than controls (Gaffrey et al. activation may not be a measure of regions involved in the task. behavioral studies have not found semantic deficits but rather point to differences in semantic organization. Because accuracy on the task was equivalent. Our task involved semantic integration and word generation and Broca’s area is particularly important in language production (Foundas. these activation differences suggest differences in semantic organization or approach to semantic processing. Making decisions about single words may require limited semantic processing and does not require integration of information and sentence tasks are more complicated. Schlaggar et al.. Similarly. 2004).
Thus. Cluster size is in voxels. When 1 group is not as able to do the task. Our finding also supports studies of white matter. Regions of activation for ASD group . The more diffuse activation pattern in the ASD group is consistent with our hypothesis and previous behavioral findings. Different approaches to the semantic task could also result in different regions being activated.08 3. Barnea-Goraly et al. 2004). 11–19 years old..05 4.75 3.25 24 214 56 Cluster size 388 238 196 163 161 149 140 105 96 56 55 t-value 3... Kana
.. which is consistent with priming studies showing increased priming to pictures (Kamio & Toichi. 2006).. with increased white matter in toddlers but decreased white matter in adolescents with ASD compared to controls. task differences could account for our contrasting finding of increased Broca’s activation and decreased lateralization of activation. suggests differences in communication between critical language areas.. Another important difference is the semantic task. and may require syntactic processing.37 Anatomical region
T. 2007. 2006. Courchesne et al. Knaus et al.. 2003) and priming studies (Kamio et al.. In addition. which is in line with long-term memory (Toichi & Kamio. Holland et al.976
Table 7. resulting in different brain regions being used. and0or use of semantic information.. The lack of correlation between activation in left frontal and temporal language areas in the ASD group. 2006.97 3. implying less efficient connectivity in ASD than in controls. Just et al. Control group activations MNI coordinates x 62 72 254 262 234 258 22 236 38 38 42 y 31 257 19 265 3 269 5 17 13 39 271 z 20 26 18 28 34 12 60 2. Szaflarski et al. Studies of age effects on language activation have suggested differences. (2001) found abnormalities in white matter development. but results have been inconsistent (Gaillard et al. Kana et al..
Right inferior0middle frontal gyrus Right middle temporal gyrus Left inferior frontal gyrus Left inferior temporal gyrus. Wood et al. 2000) and correlations with non-verbal tasks (Toichi & Kamio. It is also consistent with a finding of decreased functional connectivity between language and visuospatial regions in ASD (Kana et al.. 2004. 2001. Previous studies have involved adults (Gaffrey et al.61 4.29 4. MNI coordinates are for peak activated voxel in each cluster ASD . Kana et al. 2002. 2004. 2006). Semantic information may be organized differently in individuals with ASD.
Forman. McKinney. M. whereas controls did not. Foundas. P. Kennedy.. Chabane. R. T. C. A. N. N. Boddaert. Malow. Lotspeich. S... E.. 1–19.. Neurology. (2007).. A.. Herbert. J. The authors thank all of their research assistants and Lin Themelis for help with screening and scheduling the participants.. & Burman. N. (2003).R..M. W. Gaillard. R. 161. T.D.. Brunelle. H. Campbell... Mental Retardation and Developmental Disabilities Research Reviews. Magnetic Resonance in Medicine. larger numbers of participants may be more revealing of within-group variability in activation patterns. M. J.D. Real-time 3D image registration for functional MRI. Carper. We did not. D.. Frazier. & Grandin. Boddaert.. M. Hunter.B.. Fedio. B. M.. S..J.L. Chisum. M. Gaffrey. Caviness. have a behavioral measure of processing efficiency in this study to examine this possibility. Tager-Flusberg.. (2006). Philippe. Happé. & Jesmanowicz.. (1995). (1997). 33. but not controls. M. Perception of complex sounds in autism: Abnormal auditory cortical processing in children. C. Lincoln. (1999). 45. S. this sample size is similar to other f MRI studies of ASD. Mintun..Y. U. Bookheimer. They also extend their sincere gratitude to the children and families who participated in this study. N. S.H. (1999). A typical participation of visual cortex during word processing in autism: An f MRI study of semantic decision.. 42. Fitzgerald. Schreibman. Magnetic Resonance in Medicine. & Noll.H. Sachs. R.M. White matter structure in autism: Pre-
. Karns. Barthélemy. 6. L.. J. 56. Akshoomoff.. and language abilities within the ASD group could also be explored.. M. Tigue. & Reiss. Ziccardi... Coez. which is part of the NICHD0NIDCD funded Collaborative Programs on Excellence in Autism. R. M. E. W. Hodge. Differential prefrontal-temporal neural correlates of semantic processing in children. Brain and Language. Future studies should include larger more heterogeneous samples so that differences in activation related to age. 10. C. Ziegler..R. (2007) found visual imagery regions activated during their semantic task in ASD. Diagnostic and Statistical Manual of Mental Disorders. American Journal of Psychiatry. N... Biological Psychiatry. 1014–1018..M.. 160. Eliez.. R. F. Neuroreport. 161–165. K. 2117–2120. W. B. This study was also supported by NINDS F30 NS055511. Passingham. Barthélemy.A. M. & Zilbovicius. Eddy. Annals of Neurology.. A.. Frackowiak. F. J. Frith. Perception of complex sounds: Abnormal pattern of cortical activation in autism.A. Abnormal cortical voice processing in autism.D. Neuropsychologia. H. Gaffrey et al. 99. C. functioning level.M. C.. Belin. S..K.A. Mouren-Simeoni.. de Fossé.. F. Makris. B. Zeffiro.) Washington.A... W. 2057–2060. Gaillard. (2001). because of the task design. Moses.W. C.. (2001). Barthélemy.C. 7. A. (2006) also found decreased differences in activation between semantic and visual tasks in the ASD group than controls. Brunelle. Braniecki..C...D.. Neurology. The neuroanatomy of autism: A voxel-based whole brain analysis of structural scans... 57. Belin.. Barnea-Goraly.. which may reflect differences in semantic organization or approaches to the semantic task. M.R.. (2003). Bookheimer. The anatomical basis of language. J. 18..B.S. D. McGrath. Menon. Balsamo. (2004). Gervais. American Journal of Psychiatry. 151–188.. American Psychiatric Association (1994).. K.. Annual Review of Neuroscience. In summary. 25. R. Y. Brunelle. Cox. F. V. Kwon. (2004). & Frith. V. J. L.. as well as funding for the GCRC at Boston University School of Medicine (M01RR0533).. H. C. Friston. D. DC: American Psychiatric Association Press.A. Samson.. Human Brain Mapping. Pizzo. There are several limitations to this study.Language activation in ASD et al. Nature Neuroscience. Samson. P.L. M.D. 801–802.C. Another potential limitation is that although some lower-functioning individuals with ASD were included.. Pierce... Harris et al. L.. (2002).H. Y. Topics in Language Disorders.. Unusual brain growth patterns in early life in patients with autistic disorder: An MRI study. H. L. Poline. K. (2004).. H.Y.
This study was supported by a program project grant from the National Institute on Deafness and Other Communication Disorders (U19 DC 03610). Boddaert. N.-A. Language-association cortex asymmetry in autism and specific language impairment. however. Methodological issues in pediatric neuroimaging. & Zilbovicius. & Courchesne. L. Steele.
liminary evidence from diffusion tensor imaging.. Developmental aspects of language processing: f MRI of verbal fluency in children and adults.. Z.. N.A. 21. Cohen.. Booth. V. given the large variability in ASD.A.
Abell.S.. Petrella. Haist. Courchesne. Haas.and low-imagery sentences. Royer. Samson. N. Kleinhaus. F.R.. & Müller. Another limitation is the relatively small sample size. (4th ed.. 1672–1684. 48. most of our subjects were high-functioning. A. & Theodore.C. A. One limitation is that we do not have a measure of processing efficiency. & Zilbovicius. A.. 1647–1651.. P. Davis. F. R. Bourgeois.. Although reaction time was collected.R. 55. Chabane.. I. Blumenfeld. Krams. 226–235. The more diffuse activation in ASD could also be related to efficiency in semantic processing. N.F.D.. Sfaello. P. Ashburner.. 176–185. J. 636– 647... Z.... 757–766. (2004). Akshoomoff... 1056–1065. S.. Similarly. (2006) found that ASD individuals used the same regions to process high... S. D. S.J. Whitnah. especially given that the ASD group had significantly lower receptive and expressive language scores. C.. R. Improved assessment of significant activation in functional magnetic resonance imaging (f MRI): Use of a cluster-size threshold. Sato. H. N. Kufta. Y.Y. S.. S.. (2000). Bookheimer. A direct comparison of PET activation and electrocortical stimulation mapping for language localization. Blaxton. we found differences in activation between ASD and typically developing adolescents during a semantic task involving language production.N.. Mouren-Simeoni. & Harris. P. Lord. Xu.. Courchesne.. C.. Functional MRI of language: New approaches to understanding the cortical organization of semantic processing.M. Jr.. 245–254. M. Ahmad... Belin.. Leboyer.. this was not a measure of time to generate a response and is therefore not a meaningful measurement. Although.J. G. 323–326.
Clinical Evaluation of Language Fundamentals. Brain asymmetries in autism and developmental language disorder: A nested whole-brain analysis. Chakraborty. 964–969...K. Lambrecht. Herbert. 127. D.. R. Y. Harvey. Vigneau. Hodgson. Long-term memory in highfunctioning autism: Controversy on episodic memory in autism reconsidered. V. M.. (2001). B.K. Dual access to semantics in autism: Is pictorial access superior to verbal access? Journal of Child Psychology and Psychiatry. 619– 630.M.D. N. Visscher. 1– 4. A. R.. & Byars.. Murphy. Volkmar...J. (2001). (2004). Meta-analyzing left hemisphere language areas: Phonology. Rothermel.F. I. N.C.N. Witelson. Plante. G. (3rd ed. J. 296.S. (2006).M.L. D. Cheung..A. Kennedy. Lord.A. V. O. Crivello. Handedness and intelligence: Patterns of hand preference in gifted and nongifted children. T. 245. & Chugani. J..
. Neurology.T. S. 213–226. Duffau. R.M. Tai. Schlaggar. Bollich. Knaus et al. Annals of Neurology. Suckling... 29.. P. 837–843. B.S. L.C. Knaus. 129.. R.. & Kaufman..A. 335–364). & Chua. 61.... H. Autism diagnostic interview—revised: A revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. Cohen (Eds. B. D. V. Worsley.M. Condouris.. Jr..S. Semel. Cortical activation and synchronization during sentence comprehension in high-functioning autism: Evidence of underconnectivity. C. Müller. M. Ward. D. Just.. (2002). Chugani. Brain.E. Müller. M. A. & Just. A. F. G. Journal of Comparative Neurology. Developmental Neuropsychology. 7. Minshew. H. Journal of Autism and Developmental Disorders.G. & Belleville. T. Chabris. Journal of Autism and Developmental Disorders. Liao. Current directions in research on autism.. Neuroimage.. Y.. Journal of Autism & Developmental Disorders. C. Kelley. 19–31. V. (2005). Neuroscience Letters.H.. Schmithorst. (2006). V. Joseph. S.A.. A.. A.R.L. A. M. C. H.J. Urban. Chugani. Neuroimage. D. & Foundas. J. 859–867. S. Paul. S. f MRI study of language lateralization in children and adults...S. 738–747.. Variability in perisylvian brain anatomy in healthy adults.T. Morasse.. M. N.H. Kean.T. Y..F.A. (1999). Tager-Flusberg.. R. E. C... 27.E. Brain.. & Caviness. J. A. & Kamio..L. K. Beaucousin. D. E. (2002).W. O’Brien.S. A. Keller. Bollich... (2004). Knaus.. H. Kamio.. & Tager-Flusberg. Mangner. Holland.) San Antonio. 24.... Petre. 1–15. & Fein.. M. C. & Kamio.. Tager-Flusberg. 15. H. N. S.. Morales. 1414–1432. Deutsch.Y. 323. R. Impairment of dentato-thalamo-cortical pathway in autistic men: Language activation data from positron emission tomography. (1998).L. Lenventhal. Saling. 236–340.R. Miezin.D.S. MN: AGS Publishing. D.. 1811–1821.
Mottron. 253–260.L. Handbook of autism and pervasive developmental disorder (3rd ed. Schmithorst. L. Duncan.D. K.J. Kennedy.. R.. Long-term memory and levels-ofprocessing in autism. Filipek....L. D. (2007). Takeoka. R. T. Mental Retardation and Developmental Disabilities. Adrien. Sentence comprehension in autism: Thinking in pictures with decreased functional connectivity. & Ball. Byars.L..N. Pickles. Neuropsychology. 43.). N. J.T. Jr. N.L. (2000).H. & Evans. TX: The Psychological Corporation.. Cherkassky. Toichi.. McAlonan. L. Lemen.J.F. P. Piro.. A. Brain activation during semantic processing in autism spectrum disorders via functional magnetic resonance imaging. 1476–1479.. 97.. Journal of Autism & Developmental Disorders. Brain.. P. & Kamio. 33.. 40.. & Rutter.A. K.. Abbott. 205–223. M.. L.. 659– 685. Takeoka.Y. 128..G.-A. Rutter. 202–212. Kaufman Brief Intelligence Test. 63.. Language and communication in autism.H. Y. 31. 268–276. H.. 52. Neuropsychologia. R.) Circle Pines. (1994)... Cherkassky.J. M... Mapping the brain in autism. T.. C. Behen.. Ziegler. Behen. 128. (2007).M. M.J... Lemen. Brain mapping of language and auditory perception in high-functioning autistic adults: A PET study. Paul. Journal of Child Psychology and Psychiatry and Allied Disciplines. (1998)..W.E. Lange. Science. (2006). G. Kaufman.P. K.K. & Foundas.C. Brain and Cognition. T.D. (2003). Journal of Autism & Developmental Disorders. New York: Wiley. D. Hodgson.. Sylvian fissure morphology and asymmetry in men and women: Bilateral differences in relation to handedness in men.M. Harcourt Brace and Co. Clark. (2004). W.. & Secord.
T.H. Corey. (2000).C. Toichi. 42.. G. (2006). & D.... Simultaneous Inference for f MRI Data [Computer software].A. R. L. Wellard..J.M. 30. D. F. Verbal association for simple common words in high-functioning autism. Lam. Brown. A.. 18. 7.M.. G.C... L. V. 151–161.. (2005). Aharon. Muzik.F.. In F. & Lord.M. S. Cheung. Brain and Language. A voxel-based MRI study of volumetric differences and intercorrelations in autism. A. The autism diagnostic observation schedule—generic: A standard measure of social and communication deficits associated with the spectrum of autism. Cook.. L. Kana. O. semantics.-A. Human Brain Mapping. M. Normal f MRI brain activation patterns in children performing a verb generation task. McGrath. V. & Kigar. Risi. Bakardjiev. D. Yip. Lemen. Aston. S. W. D. V. M.. & Tzourio-Mazoyer. (1992).A. Herbert. (2004). 1035–1044. J. (1995). A general statistical analysis for f MRI data. & Foundas.. & Le Couteur..E.. 588–596. A. K.M. M. V. K..K.. H. Anderson. S.A. T. Functional neuroanatomical differences between adults and school-age children in the processing of single words. Journal of Autism and Developmental Disorders. Makris.. B.. & Chugani.. 21–29. P. 219–232. Harris. L. D. Makris. M.J.M. C. Lugar. A...M. Steele. Tager-Flusberg.. Jr.. Language cortex activation in normal children. Houdé.. Abnormal asymmetry in language association cortex in autism. A. 37.I. (2005).A. Swainson. (2002). & Caviness. (2nd ed.L. Wiig. T. B. Brain.. Bollich. A. Chabris. Robins..A. T. G..M. Wood. Holland.... Strawsburg.S.J. & Toichi.. and sentence processing. M. Hervé. 14. & Minshew. (2002).C.. & Jackson.M. Toichi.. E. 499–510. Sex-linked differences in the anatomy of perisylvian language cortex: A volumetric MRI study of graymatter volumes.. Bakardjiev. A.. Klin..E. Muzik. Ziegler. Cortex. A study of memory functioning in individuals with autism. 14. Anatomical asymmetries of anterior perisylvian speech-language regions.K. 2484–2493. Corey. N. D. & Folstein.T..A. 54– 68. O. H. J.... (2000). DiLavore. A. & Petersen.. 1116–1122. Lord.. Mazoyer. Rothermel. S. V. Neuroimage. E.. Szaflarski. Kamio. Knaus. 483– 490. M. E. 30. (2001). pp. (2001).. F.M. M. Keller. (2006). M. Atypical lexical0semantic processing in high-functioning autism spectrum disorders without early language delay. Y.. Corey.978
.Language activation in ASD
APPENDIX Stimuli Used in the Scanner
DBCD FMNTL RSVK gnc pqkst vdrh bvmnld crtpjs lkjsdf CBRD SMSSM WTY meal in morning people live there keeps hands warm drink out of tyrmnb cpldry wqkjdh YPFV CQT DGSG LVKBW NPPRT BDCS srtmnc jbwp kqcnv sit on it wear on feet people worship there eat soup with RFD YGKH SSP hgtst plmvsr sdfjkl VBMNCV JLHSDR RWPYL kbtx wmrv pljf criminals go there write with it keeps head warm people drive them UPPER UPPER UPPER UPPER breakfast store gloves plate UPPER UPPER UPPER UPPER desk pants church spoon UPPER UPPER UPPER UPPER library pencil shirt cars lower lower lower lower dinner house socks cup lower lower lower lower chair shoes store knife lower lower lower lower jail scissors hat planes bpl mxwy stb QXJFGH KRNBW PCTBD ybmcr cdlk rtpss NKF DNNP GRTS keeps neck warm helps you see tells the time kids learn there btkx mrwv ljpf MNVBCV HSJLDR PYRWL TSHGT MVPLSR FJSDKL FRD GKYH SPS buy food here keeps rain off sleep on it people read them MCYBR DLCK PSRTS KNF NNDP RTGS pbl xwmy tsb jfqxgh nbkrw tbpcd jewelry for finger borrow books from lock door with people fly them UPPER UPPER UPPER UPPER scarf glasses lamp hospital UPPER UPPER UPPER UPPER office sweater table books UPPER UPPER UPPER UPPER pin library key planes lower lower lower lower shoes hat clock school lower lower lower lower store umbrella bed radio lower lower lower lower ring park nail buses
Bold indicates correct response.