Cortical Activation during Cooperative and Competitive Joint Actions in Children With and Without Autism Spectrum Disorder (ASD) – An fNIRS Study

Children with Autism Spectrum Disorder (ASD) have social communication and perceptuomotor diculties that affect their ability to engage in various types of joint actions. In this study, we compared spatio-temporal action errors and fNIRS-related cortical activation between children with and without ASD during a Lincoln log joint action game requiring them to play leader or follower roles, move in synchrony or while taking turns, and move cooperatively or competitively with an adult partner. Children with ASD had greater motor, planning, and spatial errors and took longer to complete the building tasks compared to typically developing (TD) children. Children with ASD had lower superior temporal sulcus (STS) activation during Turn-take and Compete, and greater Inferior Parietal Lobe (IPL) activation during Lead and Turn-take compared to TD children. As joint action demands increased, TD children showed greater STS activation during Turn-take (vs. Synchrony) and Compete (vs. Cooperate) whereas children with ASD showed greater IPL activation during Lead and Compete (vs. Cooperate). Our ndings suggest that children with ASD rely on self-generated action plans (i.e., increased IPL activation) more than relying on their partner’s action cues (i.e., reduced STS activation) when performing joint actions. support planning of and between cortical activation and the Compete. VABS score better social, communication, or functional performance, hence, a positive correlation implies greater activation being associated with better VABS performance. In the TD children, greater VABS communication scores associated with greater right IFG and STS activation (r = 0.364 and 0.381, p < 0.01), greater VABS social scores were associated with greater left STS activation (r = 0.341, p < 0.01), and greater VABS daily living scores were associated with lower right IPL activation during Lead (r = -0.513, p < 0.001). Similarly, in TD children, greater VABS communication scores were associated with greater left MFG, IFG, and right PCG activation (r = 0.318 to 0.403, p < 0.05), and greater VABS social scores were with greater left STS activation during Follow (r = 0.341, p < 0.01). For Turn-take in TD children, greater VABS communication scores were associated with greater right IFG activation = 0.434, p = 0.001), and greater VABS daily living scores were associated with greater left IPL and right IFG activation 0.404 0.456, ps < 0.001; However, greater VABS daily living scores were associated with lower left PCG activation during Turn-take (r = -0.356, p < 0.01). studies focused on imitation and synchrony-based cooperative actions but not turn-taking or competition. Using fNIRS and motion tracking systems, we have reported differences in behavioral performance and cortical activation in healthy adults and school-age children with and without ASD during multiple interpersonal synchrony tasks involving reaching/body sway versus solo actions. 25–28 In this study, we extend our past work to a novel naturalistic, joint action building game using Lincoln Logs in children with and without ASD. We found that children with ASD had greater motor, planning, and spatial errors, and they took longer to complete the tasks compared to the TD children. For group-based activation differences, children with ASD had lower bilateral STS activation during Turn-take, and lower left STS activation during Compete and a similar statistical trend for Follow compared to TD children. In contrast, children with ASD had greater left IPL activation during Lead and Turn-take compared to TD children. For hemispheric differences, TD children had right lateralized IPL activation during Lead and Turn-take, whereas children with ASD had right lateralized IPL activation during Compete. For condition-related differences, TD children had a consistent pattern of greater right STS activation during joint action tasks involving greater social monitoring/intention inferring demands (Turn-take > Lead, Compete > Lead, and Compete > Follow). They also had greater left IPL activation during Compete vs. Turn-take. In contrast, children with ASD had a completely different strategy of greater left and/or right IPL or right MFG activation during Lead vs. Follow, Lead vs. Turn-take, Lead vs. Compete, and Compete vs. Follow. For correlations between the cortical activation and the adaptive functioning measure in both groups, better VABS social, communication, and daily living performance was associated with greater left MFG, right IFG, and bilateral STS activation. The associations between VABS scores and bilateral PCG/IPL activation were not consistent. For correlations between cortical activation and SRS scores, children with ASD’s poor social performance was also associated with lower right IFG, PCG, and STS activation and greater left/right IPL activation across multiple joint action


Introduction
Several functional magnetic resonance imaging (fMRI) studies suggest potential neural mechanisms that support the aforementioned behavioral strategies when playing leader and follower roles. Using a mutually adaptive tapping synchrony paradigm, Fairhurst et al. found greater cortical activation in regions that are important for self-initiated movements, including supplementary motor area, premotor cortex, precuneus, and inferior parietal sulcus, in leaders compared to the followers 11 . When engaging in bimanual movement synchrony using haptic inputs, leaders showed more activation over the primary somatosensory, motor, supplementary motor, as well as dorsolateral prefrontal cortices / middle frontal gyrus (MFG), which are important for motor control and motor planning, while followers showed more activation over the temporoparietal junction or superior temporal sulcus (STS), a part of the mentalizing and social networks 12 . Similarly, an fNIRS study found greater activation in temporoparietal and sensorimotor regions when musicians played the second violin part as followers compared to when they played the rst violin part as leaders 13 .
Taken together, greater cortical activation over the sensorimotor or prefrontal cortices in the leaders may re ect efforts in controlling and planning their own actions, while the greater temporoparietal activation in the followers may re ect their efforts to adapt to partners, to monitor and infer their partner's actions, and to match their own actions to that of their partner's. In the present study, we compared behaviors and cortical activation patterns between Lead and Follow conditions during a Lincoln Log joint action game.
Temporal Components during Joint Actions -Synchrony or Turntaking Besides different roles, movement timing is also critical to achieve movement goals and ensure appropriate social interactions. While interpersonal synchrony is important for many cooperative tasks, such as moving a heavy object together, turn-taking is embedded in many everyday activities, such as playing games on playgrounds and engaging in back-and-forth conversations. Both synchrony and turn-taking require one to monitor the cues from their social partner, anticipate/predict partner's movements, and adjust one's own movements accordingly, therefore, the systems that support perceptuo-motor integration are of particular importance 14 . In contrast to turn-taking, interpersonal synchrony involves moment-to-moment synchronization and effort of online monitoring and adjustments 14 . Turn-taking, on the other hand, requires one to remember their partner's actions, wait for one's turn, and plan one's own actions; therefore, processes involving working memory, inhibition control, and motor planning will be important 15 .
Many neuroimaging studies suggest an important role for the observation-execution matching systems (OEMS), including inferior frontal gyrus (IFG), superior temporal sulcus (STS), and inferior parietal lobe (IPL), in matching movements with observed actions; a critical component in synchronous actions or turn-taking 16,17 . The STS region is reported to be more active during movement imitation compared to passive observation or execution, therefore, is said to represent visuomotor correspondences between one's own and another's actions 18 . The frontoparietal connections are important for multisensory integration and perceptuomotor control during joint actions 19 . Speci cally, the IFG region is important for goal understanding and inferring intentions of observed actions while the IPL region is important for predicting and planning the kinematics of goal-directed actions [20][21][22] . Other important brain regions include the pre-and post-central gyrus (PCG) and the prefrontal cortices/MFG. PCG includes the primary motor and somatosensory cortices that receive/process sensory information and execute actions 23 . The prefrontal regions, mainly, the MFG, are important for executive functions such as motor planning, working memory, cognitive shifting, and inhibition -a set of mental skills that are important during interpersonal synchrony and turn-taking 24 .
Using fNIRS, we have reported greater activation over the IFG, STS, and IPL regions in healthy adults and children during interpersonal synchrony compared to solo conditions during reaching and postural sway tasks [25][26][27][28] . Similarly, during turntaking while having conversations or when playing piano duets, healthy adults showed differential frontotemporal activation suggestive of social monitoring [29][30] . Similar ndings have been found when examining cortical activation during turn-taking interactions. During a table setting task, adults showed greater IPL activation during turn taking with another partner vs. moving solo or when observing their partner's actions 31 . In the current study, we compared behaviors and cortical activation during naturalistic, Lincoln Log-based joint actions involving synchrony (Lead and Follow conditions) and turn-taking (Turn-take condition) in children with and without ASD.

Intentions during Joint Actions -Cooperative or Competitive
Cooperation and competition are important social behaviors for humans. When engaging in cooperative tasks, social partners work towards a shared goal to improve their group performance 32 . In contrast, during competitive tasks, the competitors focus on individual goals and would either optimize one's own performance or undermine the performance of their competitor 33 . For both cooperative and competitive behaviors, it is important for one to consider/refer to their competitor's intentions 34 . Social regions such as the bilateral temporoparietal junction and the inferior frontal/prefrontal cortices will again become important to monitor partner's behaviors and to understand the goals and intentions of their Using a computerized pattern-building game, an fMRI study found common activation over the frontoparietal network during cooperative and competitive behaviors, however, greater orbitofrontal activation was found during cooperative, while greater IPL and medial frontal activation was found during competitive behaviors 35 . Similarly, Liu et al. found fNIRSrelated differential activation in the right IFG during competitive and cooperative disc-building games 36 . Using hyperscanning techniques (i.e., simultaneous scanning of partners), the same research group found signi cant interbrain neural synchronization over right STS during cooperative and competitive conditions, as well as greater right IPL activation during the competitive condition of the disc-building game 37 . These results support differential activation of IFG for intention understanding during both competitive and cooperative behaviors, and competition-speci c increases in IPL activation to support planning of self-initiated actions and self-other distinctions. In the present study, we compared behaviors and cortical activation during Lincoln Log-based cooperative (Lead and Follow and Turn-take) and competitive (Compete) conditions in children with and without ASD.

ASD-related Di culties in Joint Actions
Children with ASD have poor perceptuomotor control, executive functioning, and intention understanding, that might lead to di culties in various types of joint actions [2][3][4][5][6][7]9 . During a joint improvisational mirroring game that required participants to take lead or follow the leader, children with ASD spent less time in synchrony with their partner, especially when they are in the follower role 38 . They also spent less time synchronizing with the tester during rhythmic actions such as joint marching, clapping, postural sway, and pendulum swaying tasks [3][4][5][6]9 . These di culties have been attributed to their poor visuomotor and inter-limb coordination within solo and social contexts [5][6]9 . Children with ASD also showed di culties during turn-taking tasks [39][40] . During back-and-forth conversations, children with ASD showed longer turntaking gaps and reduced temporal variability, suggesting poor response inhibition/executive functioning 39 . During cooperative/competitive joint actions, children with ASD may have di culties inferring intentions of others which might affect their joint action performance 40 .
In terms of cortical activation, children with ASD have atypical activation over the regions important for OEMS (including IFG, STS, and IPL), executive functioning (including the prefrontal cortices/MFG), and intention understanding (including temporoparietal junction/STS and the prefrontal cortices) that might re ect their di culties in performing different types of joint actions 17,41,42 . Most fMRI studies have investigated ASD-related cortical activation that asked participants to imitate/follow others and reported atypical activation over OEMS regions [16][17] . Using fNIRS, our research group has also reported hypoactivation in the IFG and STS regions along with hyperactivation in the IPL region when children with ASD engaged in synchronous reaching or whole-body sway motions while following the lead of an adult partner [27][28] . Although activation differences between leading and turn-taking joint actions are not well-studied; studies have found reduced prefrontal activation in individuals with ASD during executive functioning tasks requiring inhibition control and motor planning 41 . Such atypical prefrontal activation might also present in children with ASD during leading and turn-taking because these tasks require signi cant motor planning and response inhibition. When parent-child dyads of children with and without ASD engaged in cooperative and competitive actions there were no cortical activation differences even if there was reduced behavioral synchrony in children with ASD 43 . Given the known ASD-related differences in social monitoring/intention inferring, perceptuo-motor integration, and response inhibition/executive functioning, we hypothesize that children with ASD will show atypical activation over inferior frontal/prefrontal, STS, and IPL regions during cooperative and competitive joint actions.
Using fNIRS, the current study investigated differences in cortical activation in children with and without ASD during a Lincoln log-based joint action game. This game incorporated conditions with different roles (Lead vs. Follow), movement timing (Synchrony-Lead and Follow vs. Turn-take), and with/without shared goals (Cooperate included Lead, Follow, and Turn-take vs. Compete). We hypothesize that children with ASD will show greater spatio-temporal errors during joint action and differences in cortical activation over the OEMS (i.e., IFG, STS, IPL) and prefrontal cortices (MFG) for all joint action conditions. Speci cally, they may have motor inhibition, control, and planning-related differences affecting MFG, PCG, and IPL activation or action matching/social monitoring differences affecting STS activation.  Table S1). Hand preferences did not differ between the two groups as indicated by similar proportions of log pickups using right, left, or both hands (p > 0.05, Fig. 1C).

Group Differences in Cortical Activation
Children with ASD had lower activation in left and right STS regions during Turn-take (ps < 0.01, Fig. 3C) and lower left STS activation during Compete compared to the TD children (p = 0.001, Fig. 3D). They also had lower left STS activation (p = 0.01 but did not survive FDR correction, Fig. 3B) during Follow compared to the TD children. In contrast, children with ASD had greater activation in the left IPL region during Lead and Turn-take compared to the TD children (ps < 0.001, Figs. 3A and 3C).

Hemispheric Differences in Activation
For both children with and without ASD, the hemispheric differences were found only in the IPL region, however, the condition for hemispheric differences differs between groups. Speci cally, greater right than left hemispheric activation (i.e., right lateralization) was found in the TD children during Lead and Turn-take (ps < 0.001, Fig. 4A), whereas in the children with ASD, a similar right lateralization pattern was found during Compete (p < 0.001, Fig. 4B).

Condition-related Differences in Cortical Activation
For condition-related differences, we conducted post-hoc analyses based on our original questions. For Lead vs. Follow differences, the differences were only present in children with ASD (Figs. 5A and 5B). Children with ASD had greater bilateral IPL and right MFG activation during the Lead compared to the Follow (ps < 0.001, Fig. 5B). For Synchrony-Lead & Follow vs. Turn-take, TD children had greater right STS activation during Turn-take compared to Lead (p = 0.01, Fig. 5C), while children with ASD had greater left IPL activation during Lead compared to Turn-take (p < 0.01, Fig. 5D). There was no signi cant difference between Follow and Turn-take in TD children and children with ASD (no p-value survived the FDR corrections, Figs 5J). There were no conditional differences between Turn-take and Compete in children with ASD (p-values did not survive FDR corrections, Fig. 5L).

Correlations between Cortical Activation and Behavioral Performance and Vineland Adaptive Behavioral Scales (VABS) scores
Correlations between cortical activation and behavioral errors did not survive FDR corrections (Supplementary Table S4). In terms of relations with VABS, the adaptive functioning measure TD children showed several positive/negative correlations between cortical activation and the VABS scores during Lead, Follow, and Turn-Take, but not Compete. A higher VABS score indicates better social, communication, or functional performance, hence, a positive correlation implies greater activation being associated with better VABS performance. In the TD children, greater VABS communication scores were associated with greater right IFG and STS activation (r = 0.364 and 0.381, p < 0.01), greater VABS social scores were associated with greater left STS activation (r = 0.341, p < 0.01), and greater VABS daily living scores were associated with lower right IPL activation during Lead (r = -0.513, p < 0.001). Similarly, in TD children, greater VABS communication scores were associated with greater left MFG, right IFG, and right PCG activation (r = 0.318 to 0.403, p < 0.05), and greater VABS social scores were associated with greater left STS activation during Follow (r = 0.341, p < 0.01). For Turn-take in TD children, greater VABS communication scores were associated with greater right IFG activation (r = 0.434, p = 0.001), and greater VABS daily living scores were associated with greater left IPL and right IFG activation (r = 0.404 to 0.456, ps < 0.001; Table 1). However, greater VABS daily living scores were associated with lower left PCG activation during Turn-take (r = -0.356, p < 0.01).

Discussion
Previous fMRI studies of joint actions have been limited to simple hand movements and unnatural environments. Most studies focused on imitation and synchrony-based cooperative actions but not turn-taking or competition. Using fNIRS and motion tracking systems, we have reported differences in behavioral performance and cortical activation in healthy adults and school-age children with and without ASD during multiple interpersonal synchrony tasks involving reaching/body sway versus solo actions. [25][26][27][28] In this study, we extend our past work to a novel naturalistic, joint action building game using Lincoln Logs in children with and without ASD. We found that children with ASD had greater motor, planning, and spatial errors, and they took longer to complete the tasks compared to the TD children. and SRS scores, children with ASD's poor social performance was also associated with lower right IFG, PCG, and STS activation and greater left/right IPL activation across multiple joint action conditions.
We found that children with ASD had greater motor, spatial, and planning errors and took longer to complete the tasks compared to their TD peers. Children with ASD have poor social awareness, visuo-motor coordination, and executive functioning skills, which might affect their joint building abilities [44][45][46][47] . Poor social monitoring is a fundamental diagnostic impairment and is widely reported in children with ASD. Children with ASD are less likely to follow an adult partner's gaze or gestural bids to observe objects in the environment 44,47 . Toddlers with ASD who were shown 2D clips of a complex scene involving objects and people pay less attention to interacting adults and paid more attention to the surrounding background 48 . Moreover, dyspraxia (i.e., di culties performing skilled motor sequences) is often reported in children with ASD with greater spatio-temporal errors and greater time to task completion compared to those without ASD 3-6 . Together, these motor coordination/planning and social impairments could impair joint action performance in children with ASD.

Children with ASD had lower bilateral STS activation during Turn-take and lower left STS activation during Compete and
Follow compared to TD children. In contrast, they had greater left IPL activation during Lead and Turn-take than the TD children. Recent studies have reported that followers showed greater STS activation whereas leaders showed greater supplementary motor area and sensorimotor activation during synchronous movement or playing of the violin [12][13] . When planning joint actions, children with ASD may rely more on internal or self-generated plans than being externally driven (i.e., using social information from their partners to plan their actions). For hemispheric differences, TD children had right lateralized IPL activation during Lead and Turn-take conditions whereas children with ASD had a similar right lateralized IPL activation during the Compete condition. Right IPL is important in making self-other distinctions when engaging in synchrony or competition vs. cooperation tasks 35,49 . Less right IPL activation is expected when cooperating as it may involve greater merging of self and other whereas greater right IPL activation is expected when competing with a partner 35 . In the present study, children with ASD also showed a similar pattern of greater right lateralized IPL activation during Compete vs. Cooperate. Interestingly, TD children showed a similar pattern of greater right-lateralized IPL activation in Lead and Turn-take; this may also be attributed to greater deactivation in left IPL across multiple joint action conditions. The left IPL region is considered part of the Default Mode Network (DMN) and is said to deactivate when performing externally directed processing (i.e., tasks that are cognitively demanding, goal-directed, or requiring greater social awareness) 28,50 . The DMN is said to be important during social as well as imitation tasks 51 . In fact, being imitated led to greater DMN deactivation, compared to when imitating others suggesting that an individual is perhaps more socially aware of their partner's actions when they are able to regulate the social interaction 52  For correlations between activation and the adaptive functioning in both groups, better VABS social, communication, and daily living performance was generally associated with greater left MFG, right IFG, and bilateral STS activation. For correlations between cortical activation and SRS scores, children with ASD that had better social performance had higher right IFG, PCG, and STS activation and lower left/right IPL activation across multiple joint action conditions. During joint actions, one must anticipate their partner's actions by observing them, infer their intentions, plan one's own actions, and execute the action plan 14 . The networks formed by MFG, IFG, and STS regions are implicated in each of these processes (i.e., social awareness (STS), intention inferring (IFG), planning actions (MFG and IPL), and action execution (PCG) along with many other important brain regions.
Finally, this is a pilot study involving a relatively small sample size (N = 30); yet several signi cant and meaningful ndings were revealed. Despite matching multiple factors across groups, we included children with ASD with a broad range of functioning, which may have increased the variability of our study sample. During naturalistic play, it is di cult to control the time to task completion and standardize the duration of the stimulation period. While we followed consistent probe placement, variation in participant head sizes and probe placement could have led to inconsistency in our spatial registration output.
Our study identi ed multiple behavioral and fNIRS-based neurobiomarkers during a Lincoln Log-based joint action game across prefrontal, frontal, temporal and parietal cortices. Children with ASD had greater behavioral errors (motor, spatial, and planning) and took greater time to complete tasks. In addition, children with ASD showed reduced STS activation and increased IPL activation as well as a lack of differential activation in the STS region compared to TD children. We also found different patterns of activation in children with ASD compared to TD children suggesting that both groups used different mechanisms to process social-perceptual information for motor planning/execution of joint actions. In the future, we will use the aforementioned fNIRS neurobiomarkers to assess changes in cortical activation following a bout of socially embedded motor intervention focused on imitation, synchronization, and cooperation. Overall, fNIRS appears to be a valid and powerful child-friendly tool to examine cortical activation during joint play in both children with and without ASD. From a learning standpoint, clinicians must consider utilizing opportunities for leading and competition to improve social awareness in children with ASD apart from following and cooperation.

Participants
Thirty children with and without ASD (Average ± SE: ASD: 11.5 ± 0.8, 12 males, 3 females; TD: 12.2 ± 0.9, 8 males, 7 females) participated. There were no signi cant group differences in age or ethnicity. Children were recruited through online postings, phone calls, and iers sent to ASD advocacy groups, and Simons Powering Autism Research (SPARK) participant research match service. SPARK informs their family database about research studies (https://www.sfari.org/resource/spark/). Before participation, we completed screening interviews with potential participants to obtain their demographic information and to con rm their eligibility. The inclusion criteria for children with ASD were (i) should hold a professionally con rmed ASD diagnosis, supported by school records, an Individualized Education Plan for ASD-related services, or medical/neuropsychological records from a psychiatrist or clinical psychologist using the Autism Diagnostic Observation Schedule; there is a growing trend of using professionally con rmed diagnostic records for ASD cohort studies 55 and (ii) met criteria for a social communication delay (> 12 points) on the Social Communication Questionnaire (SCQ) 56 . Children with ASD were excluded if they had any behavioral/sensory issues that prevented them from completing the test activities. The age-matched TD children were excluded if they had any neurological or developmental disorder/delay or a family history of ASD.
Parents of all children completed the Coren handedness survey to assess hand preferences 57 , the VABS measure to assess adaptive functioning, 58 and SRS to assess social responsiveness impairment 59 . Additionally, we administered the BOT-2 MD to assess ne motor skills 60 . Compared to TD children, children with ASD had signi cantly lower VABS, BOT-2 MD scores, and greater SRS total scores indicating impaired adaptive functioning, manual dexterity performance, and social responsiveness ( Table 2). All study procedures were carried out in accordance with the Declaration of Helsinki. All informed consent and assent forms as well as all study procedures were approved by the University of Delaware Institutional Review Board (UD IRB, Study Approval #: 930721). Prior to study participation, written informed consent was obtained from parents who gave approval for their child's study participation as their legal guardians and written and verbal assent was obtained from the children. Written parental permission and experimenter informed consent has been taken to use pictures for this publication.

Experimental Procedures
Each child sat at a table across from an adult tester and was tted with a 3×11 fNIRS probe set (Fig. 6A). A container of Lincoln logs consisting of four plain brown logs and four multi-colored (green, yellow, purple, blue) supporting logs was placed on the table and a cue card was placed facing one or both participants. We used a randomized block design comprised of 16 trials in 4 blocks and 4 conditions (Lead, Follow, Turn-take, Compete; Fig. 6B). In the Lead condition, the child built the con guration according to the cue card (shown to the child only) while making sure that the follower/adult tester followed their actions simultaneously. In the Follow condition, the child mimicked the building actions and moved synchronously with the tester who was the only one shown the cue card. In the Turn-take condition, the cue card was visible to both partners. They took turns to make the next move and built the log con gurations together. In the Compete condition, non-identical cue cards were given to both, tester and child, and they were asked to quickly and independently build the structure shown on their cue card. Each trial included a 10-second pre-stimulation, 15-second stimulation, and a 15-second post-stimulation period (Fig. 6B). During the pre-and post-baseline periods, participants were asked to observe a crosshair on the wall.

Data Collection
The Hitachi ETG-4000 system was used to capture the hemodynamic changes during the joint action tasks (Hitachi Medical Systems, Tokyo, Japan, Sampling Rate: 10 Hz). A cap embedded with a 3×11 probe set (including 17 infrared emitters and 16 receivers) was positioned over frontal, temporal and parietal regions of the brain (See Supplementary  Figures S1A and S1B). The midline of the probe set was aligned with the nasion and the lower border of the probe set was aligned just above the eyebrow and the ears. The adjacent pairs of probes, located 3 cm apart, acted as emitters and receivers for two wavelengths of light (695 and 830 nm). Light waves travel from the emitter through the skull, creating a banana-shaped arc reaching the capillary bed of the cortical tissue of the brain. Some of the light waves are absorbed/scattered while the remaining waves are transmitted back to the receivers. Using the Modi ed Beer-Lambert law, change in light attenuation is used to determine changes in the concentration of HbO 2 and deoxygenated hemoglobin (HHb) at the midpoint of two probes, also termed a channel. When a certain cortical region is more active, there will be an increase of metabolic demand/oxygen consumption and blood ow to the capillary bed supplying that cortical region, which in turn leads to an increase in HbO 2 , and a slight decrease in HHb 61 . E-prime 2.0 software was used to trigger the ETG system and mark the baseline and stimulation periods. The session was videotaped using a camcorder that was synchronized with the ETG-4000 system.

Spatial Registration Approach
We recorded the 3D location of standard cranial landmarks (nasion, inion, right/left ear) and each fNIRS probe with respect to a reference coordinate system using a Polhemus motion analysis system. Using the anchor-based spatial registration method developed by our co-author, Tsuzuki, the 3D spatial location of each channel was transferred to the Montreal Neurological Institute's coordinate system 62 . The structural information from a database of 17 adults was then used to provide estimates of channel positions in a standardized 3D brain atlas and the LONI Probabilistic Brain Atlas was used to label estimated channel locations based on MRI scans of 40 healthy adults 62-64 . A channel was included if 55% or more of the channel area (i.e., each channel was modeled as the centroid of sphere) was within a given ROI and was excluded if it was not. A channel was also excluded if its homologue belonged to another ROI. Based on these rules, we assigned 38 out of 52 channels to ve ROIs in each hemisphere (See Supplementary Figures S1C and S1D  Hitachi POTATo to process the fNIRS data output 65-67 . The processing steps include: (i) band-pass ltering of the signal between 0.01 and 0.5 Hz to remove high-/low-frequency noise, (ii) wavelet method to remove movement artifacts, (iii) General Linear Modeling to estimate the hemodynamic response, (iv) correction for baseline drifts by subtracting the linear trend between the pre-and post-baselines from values in the stimulation period, and (v) averaging the HbO 2 values during stimulation period for each trial, along with visualization of the processed data at each step 65-67 . We report HbO 2 data only, as it has a greater signal to noise ratio than the HHb data and is more often reported in the literature 67 . The reader is referred to our earlier publications for additional details on fNIRS methodology [25][26][27][28] .

Behavioral Coding
A trained student researcher scored the behavioral performance of the children during task completion. Each session was scored for three error types: (i) Motor error: the child dropped a log before placing in the container or knocked over the container; (ii) Planning error: the child hesitated, and then changed placement location; and (iii) Spatial error: the log was placed incorrectly based on color or location. Furthermore, we coded hand preferences by scoring how the child picked up each log using their left, right, or both hands. Lastly, we coded the time in seconds to complete each building con guration.

Statistical Analyses
To assess group differences in frequency of behavioral errors of each type, we conducted non-parametric, Mann-Whitney-U tests. For group differences in time to completion and hand preference (i.e., proportion of right, left or both hand actions) we conducted independent t-tests for each action type. For cortical activation, we conducted a repeated measure ANOVA using within-group factors of condition (Lead, Follow, Compete, Turn-Take), region (MFG, PCG, IFG, STS, IPL), hemisphere (Left, Right), a between-group factor of group (ASD, TD) with BOT-2 MD score and hand preference as covariates. When our data violated Mauchly's test of sphericity, we applied Greenhouse-Geisser corrections. Lastly, Pearson's correlations were used to correlate cortical activation and behavioral performance (both groups), VABS (both groups) and SRS performance (ASD only). To control for multiple comparisons for post-hoc analyses and correlation runs, the Benjamini-Hochberg False Discovery Rate (FDR) method was used to adjust the statistically signi cant cut-off 68 . Speci cally, the unadjusted p-values were rank ordered from low to high and the statistical signi cance was declared if the unadjusted p-value was less than the p-value threshold which was determined by multiplying 0.05 with the ratio of the unadjusted p-value rank to the total number of comparisons (p-threshold for i th comparison = 0.05 × i/n; where n = number of comparisons).

Data Availability Statement
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.   Hemispheric differences between TD children (A) and children with ASD (B). *indicates signi cant differences between the ASD and TD groups. *indicates signi cant differences between the ASD and TD groups.