Impression formation and autism spectrum disorder: the role of interpersonal synchrony

doi:10.21203/rs.3.rs-2885684/v1

Download PDF

Article

Impression formation and autism spectrum disorder: the role of interpersonal synchrony

https://doi.org/10.21203/rs.3.rs-2885684/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 18 Sep, 2023

Read the published version in Scientific Reports →

You are reading this latest preprint version

When people meet, they almost instantaneously form an impression of each other. First impressions of character traits and rapport are less favourable when people with autism spectrum disorder (ASD) are judged compared to non-autistic people. Little is known about the behavioural differences that drive these altered impressions. In the present study, we investigated the influence of interpersonal synchrony on impression formation of autistic and non-autistic people. Specifically, we used lagged cross-correlations to assess how much each interactant’s motion energy, a measure which can be determined from video recordings, influenced the other interactant’s motion energy. In short, silent clips of dyadic conversations, we asked non-autistic participants to rate their impression of one of the two interactants, which was solely based on the outlines of both interactants. We expected that the amount of leading of the target interactant, their diagnostic status as well as the interaction of these factors would influence impression formation. We found that while the amount of leading had a positive effect on the impressions of non-autistic interactants, this was not true for interactants with ASD. This suggests that interpersonal synchrony of motion energy is one driver of less favourable impressions of autistic compared to non-autistic people.

Biological sciences/Psychology/Human behaviour

Health sciences/Medical research

Studies have shown that autistic individuals differ from non-autistic individuals both in terms of verbal and nonverbal communication [8–11]. These differences influence how non-autistic individuals perceive autistic people that they encounter. Indeed, multiple studies show that first impressions of autistic individuals are less favourable [12–16]. First impressions as automatic judgments are quick and based on limited information [17]. Nonetheless, they substantially influence subsequent interactions and judgements [18] and predict whether friendships are formed [19]. This has a profound impact on the lives of autistic individuals [15,20,21]. Therefore, it is imperative to understand why autistic people, routinely receive a less favourable first impression.

In several studies, Sasson and colleagues demonstrated that these findings regarding first impressions of ASD persist across different modalities [12,15]. First, they investigated how non-autistic participants rated audio sequences, silent videos, videos with tone, static images and transcripts of speech of individuals with and without ASD in a mock audition for a TV show [12]. They found that in all modalities, except transcribed speech content, individuals with ASD were rated less positively than individuals without ASD. This finding suggests that the judgements are not unfavourable due to content but rather the quality of communication; possibly due to differences in pronunciation, intonation, gestures, facial expressions, and other verbal and nonverbal features.

While these effects were robust across stimulus types and experimental settings, the effect of diagnostic status on impression formation was decreased when non-autistic participants were aware of the existing ASD diagnosis of the rated person [15]. This attenuation suggests that in real life, non-autistic people might be more willing to adjust their first impressions if they are aware of a diagnosis. Notably, the adjustment of first impressions seems to depend more on the rater than the rated person [16]. Specifically, a higher stigma against autistic individuals was associated with less favourable ratings. This bias also interacted with the positive effect of awareness of the diagnosis and reversed it such that non-autistic participants who showed greater stigma attribution gave less favourable ratings if they knew about a person’s diagnosis.

Interestingly, the unfavorable first impressions of individuals with ASD is not limited to the observations of non-autistic individuals but also extends to judgements made by other autistic individuals. DeBrabander et al. [14] showed brief videos of adults with and without ASD to participants with and without ASD. They found a main effect of stimulus diagnosis: autistic individuals were rated less favourably on average compared to non-autistic individuals, which was independent of the diagnostic status of the rater.

Despite the importance of impression formation for everyday life and typical social interactions, it is still unclear which mechanisms and behavours specifically lead to individuals with ASD receiving less favourable judgements. Impression formation often focuses on rapport, for example, asking raters to judge someone’s likeability [12,14,16]. Interpersonal synchrony has been used to indirectly but objectively measure rapport [22–26] and is the phenomenon where people temporally coordinate their behaviour and brain waves when interacting [27,28]. Such coordination helps establish a common ground between interaction partners. Interpersonal synchrony is assumed to help foster relationships and create social bonds, for example, by increasing rapport and empathy [22,23,25,26,29–31]. Measures of interpersonal synchrony often make use of lagged cross-correlations between interactants’ behaviours [32,33]. Lagged cross-correlations not only provide measurements of perfect coordination but also when one person leads the other person [34] (Figure 1). Here, interpersonal synchrony is composed of perfect coordination and the leading of each interactant. Dyads whose interaction elicit a high level of interpersonal synchrony may lead to both interaction partners making better impressions to observers than interactants whose behaviours are not in sync. Additionally, one person leading their interaction partner’s behaviour could influence first impressions of themselves and their partner. Therefore, reduced interpersonal synchrony could be one cause driving less favourable impressions, given that previous studies have shown that it is altered in interactions with autistic individuals (for a review, see [35]).

Interpersonal synchrony of motion, and its relevance for ASD, has been studied with various tasks ranging from synchronisation with objects to interpersonal synchrony of motion in natural interaction between individuals with and without ASD [35]. The extent of a person’s motion can be estimated by recording a video and analyzing the change of intensity values in any pixel. The more a person moves, the more does the intensity and color of the pixels change. Pixel value changes can be added up frame by frame resulting in a quantitative measure, the so called motion energy [33]. Georgescu and colleagues [36] used motion energy and compared homogeneous autistic, homogeneous non-autistic and heterogeneous dyads consisting of one autistic and one non-autistic interactant. All dyads involving autistic participants were less synchronised compared to homogeneous non-autistic dyads, despite comparable overall motion in all three dyad compositions. Importantly, reduced interpersonal synchrony of motion energy seems to be specific for ASD in comparison to a clinical control group with differences in social interaction [37]. This robust effect led to the application of machine learning techniques to differentiate between autistic and non-autistic individuals based on interpersonal synchrony in motion energy [38]. Interpersonal synchrony of motion energy comprises situations of perfect coordination of movements but also of one person leading the other who adapts their motion energy to them. A recent study used this adaptation of motion energy and facial expressions to successfully train a support vector machine classifier to distinguish between autistic and non-autistic interactions with 79.5% balanced accuracy [38]; showing that synchrony of both modalities differs between homogeneous non-autistic dyads and heterogeneous dyads consisting of an autistic and non-autistic person.

Given the close association of interpersonal synchrony and regular use of the rapport to assess impression formation, this study aimed to investigate the contribution of interpersonal synchrony of motion energy to the altered impression formation of people with ASD. Interpersonal synchrony, by definition, is a measurement that takes more than one person’s actions into account. However, it is possible to deconstruct interpersonal synchrony to focus on the individuals’ leading each other in an interaction. Given that previous research has established a link between synchrony and power perception [39], we were interested in the influence of one interactant leading the other, as well as being led by their interaction partner, on impression formation. We created silent videos, which show solely the outlines of two people interacting, and asked non-autistic participants to rate their impression of one of the interactants. This approach allowed us to directly measure the mere effect of body motions in an interaction on impression ratings. Specifically, we were interested in the influence of each interactant’s motion energy on the other interactant, thereby assessing leading of each other’s motion energy. Based on prior impression formation and interpersonal synchrony research, we proposed the following hypotheses. First, we expected to replicate the effect of ASD on the ratings, such that interactants with ASD are rated less favourably than interactants without ASD (H1). Second, we hypothesised that interpersonal synchrony of motion energy between the two interactants, such that one person leads the other, would significantly influence the impression judgement of the leading person (H2). Third, we anticipated that diagnostic status might interact with this influence (H3). Furthermore, we observed participants’ eye movements to assess whether they focused on the target interactant whose impression they were asked to judge [40]. We considered that diagnostic status might influence gaze, such that autistic interactants lead to increased fixations due to idiosyncratic gestures associated with ASD (H4). In addition to our explicit hypotheses H1 to H4, we explored how autism-like traits in raters may influence the impression formation of interactants with and without ASD, as well as differences in nuanced aspects of impression formation, e.g., judging whether a person is likeable or trustworthy.

This study was preregistered on OSF: https://osf.io/wtm3q. We added the interaction hypothesis (H3) to the preregistered hypotheses. Concerning the analysis, we applied a Bayesian framework to allow for interpretation of the evidence for and against a hypothesis. Stimulus evaluation, preprocessing and analysis were performed with R 4.2.2 [41] in RStudio 2022.12.0 [42] and Python 3.9.13 in Spyder 5.3.3 [43]. All code used in the analysis of the stimuli, the pilot and the experimental data is available on OSF (https://osf.io/whgx6/).

Participants

We aimed for a sample size of 195 non-autistic participants. The sample size was determined with a simulation-based power analysis as implemented in the R package mixedpower [44] using pilot data from 37 participants. We determined a threshold of 2 and 1,000 simulations to achieve 90% power for detecting an effect of interpersonal synchrony of motion energy on impression formation. Inclusion criteria were age between 18 and 60 years, no psychiatric or neurological diagnoses, normal or corrected-to-normal vision and informed consent. Exclusion criteria were high scores on the short form of German translation of the Autism Quotient questionnaire (above 5 of 10 points, AQ-10 [45]) and low scores on a verbal intelligence test, the Wortschatz-Test (below 6 of 42 points, WST [46]). Additionally, we asked participants in a post-experimental debriefing questionnaire whether they have answered the questions conscientiously. They could either choose “Yes - my answers can be used for research without any problems” or “No - my answers should rather not be used”. This gave participants who were distracted or chose random ratings the opportunity to indicate that their data should not be used in the study. If they indicated that their data should not be used, their data was excluded from the analysis. We continuously preprocessed collected data along all inclusion and exclusion criteria. In total, we collected data from 247 participants of which one dataset was excluded because the participant advised against it and one because of a low WST score. Additionally, 49 participants scored above five in the AQ-10 (age: 18 to 59 years old, mean = 25.00 ± 4.59; 35 female). Therefore, 196 participants (age: 18 to 59 years old, mean = 26.56 ± 6.97; 142 female) were included to test H1, H2 and H3. For H4, we had to exclude additional participants due to insufficient gaze data quality, resulting in a sample of 91 participants (age: 19 to 59 years old, mean = 25.62 ± 5.93; 61 female). For more details on the sample, refer to the sample description in the supplementary materials. Participants were compensated for their participation with 10€ or course credit. This study was conducted following the Declaration of Helsinki and approved by the Ethics committee of the Medical Faculty, LMU Munich.

Experimental procedure

Data collection was conducted in German using the Gorilla Experiment Builder on GORILLA^TM (www.gorilla.sc), an online platform to create experiment designs [47]. The experiment consisted of study information, a consent page, demographic questions, AQ-10, WST, videos with their associated ratings (impression formation task) and a post-experiment debriefing questionnaire. In the impression formation task, participants saw 44 ten-second-long videos of two people interacting with each other. The videos did not contain any audio, and only outlines of the interactants were shown (Figure 2). The outlines on half of the screen were coloured in green to indicate the target interactant, on whom participants should base their ratings. Each video was followed by six ratings on a scale from 0 (not at all) to 100 (very much): intelligent, awkward, likeable, trustworthy, Would you start a conversation with the green person? (abbr.: conversation) and Do you think the green person has many friends? (abbr.: friends; for the original German versions see Figure 2). All but the last question were taken from Sasson et al. [12,15]. Participants had to respond to each rating to continue to the next trial. During the presentation of the videos, webcam-based eye tracking was used to measure gaze patterns. First, a calibration was performed. GORILLA^TM uses support vector machines to track the participant’s face and, specifically, their eyes with a frequency of 60Hz. For each ten-second-long video, this should yield 600 samples under optimal conditions. Then, two metrics were calculated: the proportion of fixations on each of the screen halves and switches between the two halves.

Stimulus creation

Stimuli were created from videos collected in previous studies investigating the influence of diagnostic status on interpersonal synchrony of motion energy in dyadic interactions. Specifically, 24 videos were excerpts from Georgescu et al. [36], and 20 videos were excerpts from Koehler et al. [37]. Half of the videos showed two non-autistic interactants, in which one of the two interactants was highlighted green as the target to be judged. The other half of the videos presented mixed interactions between one autistic and one non-autistic interactant. In these videos, it was always the autistic interactant who was highlighted in green to be rated. Therefore, the design was balanced with regard to the independent variable diagnostic status (autistic, non-autistic).

Videos were muted and visually reduced to the outlines of the interactants in MATLAB R2020b (Natick, Massachusetts: The MathWorks Inc; Figure 2). First, a Gaussian filter with σ = 2.0, as implemented in the imgaussfilt function, was applied to the greyscale versions of the videos in a frame-wise fashion to remove personally-identifying details that could be used to identify individuals. In two videos showing the same dyad, a Gaussian filter of σ = 2.5 was used which made the outlines slightly rougher but was hardly distinguishable from σ = 2.0 for the observer. Then, the edge function was used to determine edges in the filtered frames, so that only a rough outline of the individuals featured in the original video remained in the stimulus material. Last, one-half of the frames were coloured in green before exporting the edited video versions that finally were shown in the experiment. The green colouring indicated the target interactor who was to be rated.

Second, all videos were analysed using Motion Energy Analyses (MEA) [33]. We focused on head motion because the videos from Koehler et al. [37] included one person holding a clipboard, thereby artificially restricting motion in the upper body and arms. Resulting MEA values were further preprocessed in rMEA [48]. From each of the five-minute-long videos, two ten-second-long excerpts were chosen: one from 90 to 100 seconds, representing the introductory phase, and the other from 240 to 250 seconds, representing the body of the conversation. Outliers from the head region were removed using the rMEA::MEAoutlier function, which defines outliers as values more than ten times the standard deviation. Then, they were scaled using the rMEA::MEAscale function. Lastly, we computed lagged cross-correlations of the interactants’ MEA values within the same dyad using the rMEA::MEAccf function. We used pseudosynchrony to ensure that this measure captures interpersonal synchrony [49]. We evaluated 1) whether there is time dependency and 2) whether there exists synchrony in 10-second sections of the full videos as proposed by Moulder et al. [32]. Bayesian one-sample t-tests, as implemented by the BayesFactor::ttestBF function, revealed extreme evidence in favour of 1) and strong evidence in favor of 2), each based on 1,000 iterations of pseudosynchrony per dyad (for details see the supplementary materials).

This investigation showed that lagged cross-correlation was an effective measure of interpersonal synchrony of motion energy in this sample. Additionally, lagged cross-correlations not only provide an estimate of the total interpersonal synchrony but also deconstructs the contributions of the two interactants by considering either lags where the target interactant is leading their interaction partner (Green leading) or vice versa (White leading). This approach has been successfully used to investigate relationship quality in the therapeutic context [34]. Since our participants were asked to rate only one interactant in each video, we decided to use these estimations of each interactant leading their counterpart as predictors in our analysis, rather than an interpersonal synchrony score describing the coordination of both interactants with each other. We log-transformed both leading scores to achieve normal distribution, which allows us to scale all our predictors.

Concerning the gaze patterns, we first excluded all trials where the face of the participant was tracked with 50% accuracy or less (value recommended by Gorilla Support, 2022). Additionally, we set the minimum threshold for fixation duration to 50 ms. Trials with less than 400 samples were excluded from the analysis. One hundred and five participants with less than 50% of the trials left were excluded from the analysis of the gaze patterns resulting in a sample of 91 participants.

Analysis

We used a combination of Bayesian linear mixed models, as implemented in the brms package [51], and Bayesian t-tests, as implemented in the BayesFactor package [52]. For all random effects, we followed the guidelines by Barr et al. [53,54]. The Bayesian linear mixed models were run with four Markov chains with a total of 10,000 iterations each (50% warm-up). To draw conclusions on the significance of estimated parameters and differences, we used the brms::hypothesis function and Jeffrey’s scheme to interpret Bayes Factors [55]. The brms::hypothesis function computes an evidence ratio and a posterior probability under the hypothesis against its alternative. We chose one-sided testing in the direction of the respective estimate and adjusted α to 2.5%, since we preregistered non-directional testing with α = 0.05. Therefore, all parameter estimates with a posterior probability of above 97.5% for non-directional hypotheses (H1, H2, H3) and 95% for the directional hypothesis (H4) were considered significant.

For H1, H2 and H3, we averaged the six ratings to create a composite impression score. Before computing the average, we reversed the the awkwardness rating, since here higher values correspond with a more negative impression. The unidim function of the psych package revealed a high factor fit of fa.fit = 0.98 suggesting that all ratings measure the same underlying concept. We entered the impression score into a Bayesian linear mixed model with three fixed effects of interest: diagnostic status (autistic, non-autistic), leading of the green target interactant (Green leading) and leading of the white non-target interactant (White leading). We also added three regressors of no interest: overall motion estimates of both interactants and the source of the videos [36,37]. All parametric predictors were scaled to allow for comparison of the estimates. Lastly, we added random intercepts for stimulus and participants, as well as random slopes for diagnosis and video source for the participants. We used treatment contrast coding [56], with non-autistic interactants and videos from Georgescu et al. [36] as the reference in relation to which all effects are evaluated.

Concerning the gaze patterns, we first used Bayesian paired t-tests to check whether participants focused on the green partner of the video led to increased fixation times. Then, we tested H4 by comparing fixation durations, as well as the number of switches per 100 samples, between the halves of the videos showing autistic and non-autistic targets for impression judgments. For each outcome variable, the interquartile method was used to detect and remove outliers.

For our explorative analyses, we performed similar Bayesian linear mixed models as described for testing H1, H2 and H3. We added binarised AQ-10 scores as a categorical predictor (0 = low, 1 = high) to the original model predicting the impression score to explore differences between participants with high and low autism-like traits. Additionally, we ran six Bayesian linear mixed models of the same structure for each of the six ratings to determine which nuanced aspects of impression formation are influenced by diagnostic status and the interactants leading one another.

Influence of diagnostic status and leading on impression formation

The Bayesian linear mixed model estimated the influence of diagnostic status of the target interactant, leading of the target interactant (Green leading), leading of their interaction partner (White leading) and their interactions on the composite impression score of the rater (Figure 3). Additionally, we added the influence of motion and video source to the model. Visual inspection suggests that autistic interactants were rated less favourably; however, H1 was not significant (estimate = -4.93, posterior probability = 0.97). More leading by the target interactant led to better impression ratings; thus, confirming H2 (Green leading; estimate = 4.46, posterior probability = 0.98). This effect interacted with diagnostic status suggesting that ASD decreases the positive effect of leading on the impression score; thus, confirming H3 (estimate = -6.70, posterior probability = 0.99; Figure 4). For the full summary of the model, please see the supplementary materials). To further dissect the nature of the interaction, we divided the data into items featuring an autistic target interactant and items featuring a non-autistic target interactant. Then, we computed correlations between the composite impression scores and the amount of the target interactant separately for these two datasets. While there is a positive correlation between the composite impression scores and the amount of non-autistic target interactants leading (r = 0.55), there is no correlation between the composite impression scores and the amount of autistic target interactants leading (r = -0.04). This reveals that leading only seems to improve impressions for non-autistic interactants.

Gaze patterns during impression formation

First, we used a Bayesian paired samples t-test to assess whether participants focused on the green target interactant. There was extreme evidence that this was the case (BF = 1.19e+21), with participants fixating, on average, 64.95% of the time on the green half of the screen. However, neither the fixation proportion nor the switches between the two halves of the screen differed due to the diagnostic status. In fact, there was moderate evidence that diagnostic status does not affect the fixation proportion (BF = 0.29; mean_autistic = 65.00 ± 10.43%, mean_autistic = 64.00 ± 10.49%) or the number of switches per 100 samples (BF = 0.16; mean_autistic = 5.55 ± 1.73, mean_autistic = 5.49 ± 1.71%).

Exploring the influence of autism-like traits on impression formation

We added another categorical predictor to the Bayesian linear mixed model distinguishing between participants with low and high autism-like traits. Participants with high autism-like traits were excluded from the linear mixed model testing H1, H2 and H3. Therefore, this exploratory analysis includes 245 participants, of which 49 were in the high AQ-10 group. While there were no differences between the two groups (estimate = 0.83, posterior probability = 0.75), there was a trend for the interaction effect of the AQ-10 group and diagnostic status (estimate = 1.81, posterior probability = 0.97), such that the people with high autism-like traits rated autistic and non-autistic interactants more similarly than people with low autism-like traits (Table 1).

Table 1: Mean and standard deviation for the impression score separated by the rater’s autism-like traits and diagnostic status of the observed target interactants.

	autistic interactant	non-autistic interactant	difference
Low autism-like traits	48.86 ± 17.18	58.10 ± 16.47	9.24
High autism-like traits	51.18 ± 15.94	58.96 ± 15.05	7.78

Exploring nuanced aspects of impression

We combined six ratings of different nuanced aspects of impression into one impression score to test our hypotheses. Additionally, we explored the effects of our predictors on the separate scores with six Bayesian linear mixed models (Table 2). While the ratings awkward and friends were significantly predicted by diagnostic status with autistic interactants being rated less favourably, the other ratings depended on the green person leading and the diagnostic status in interaction with the green person leading. This means that the ratings intelligent, likeable, conversation, and trustworthy were higher when the target interactant was leading, but this effect was decreased or even fully diminished for autistic interactants. Additionally, the rating trustworthy depended on the white person leading, with higher values corresponding to the target interactant being judged as less trustworthy.

Table 2: Estimates for all explorative Bayesian linear mixed models with the nuanced aspects of the impression rating as outcomes. Asterisks mark significant estimates.

	awkward	friends	intelligent	likeable	conversation	trustworthy
Diagnostic status	-12.51*	-7.51*	-1.96	-2.33	-3.4	-2.26
Green leading	2.07	2.46	6.03*	5.08*	5.79*	5.68*
White leading	0.82	-0.56	-2.71	-4.34	-4.07	-4.92*
Diagnostic x Green leading	-8.14	-5.92	-7.94*	-5.53*	-7.2*	-5.92*
Diagnostic x White leading	1.55	0.21	2.68	2.85	2.94	4.21
Diagnostic x Green leading x White leading	3.13	-1.44	0.96	-1.54	-0.91	-0.58

When non-autistic individuals form first impressions of autistic individuals, they often judge them less favourably compared to non-autistic people [12, 15, 16]. Since autistic and non-autistic behaviour varies on multiple levels, it is paramount to carve out which behaviours might lead to less favourable impressions. In the current study, we collected first impressions of non-autistic raters based on ten-second-long videos. These videos showed a dyadic interaction in which only the outlines of the interactants were visible. Therefore, the only information participants could use to form their impressions of the target interactant was based on their motion. We analysed motion energy in these videos and computed lagged cross-correlations to measure how much each interactant led the other. This allowed us to assess the interactant’s individual contributions to interpersonal synchrony of motion energy. We found that overall impression was significantly predicted by the extent to which the target interactant led their counterpart in the social interaction, i.e. how strongly their motion energy influenced the motion energy of their interaction partner. Yet, impression, as a result of leading, was only improved for non-autistic interactants, whereas leading did not influence the impression formation for autistic interactants. The previously reported robust effect of diagnostic status [12, 15, 16] was only observable as a trend in the current study. This further corroborates that the specific aspects of behaviour isolated in the current study are at least partly responsible for the less favourable impression non-autistic individuals derive from observing autistic interactants.

ASD is consistently associated with decreased levels of interpersonal synchrony [31, 35–37], and this decrease in interpersonal synchrony seems to be an important factor in impression formation. This study has identified a key feature that explains altered first impressions between individuals with and without ASD by showing an association between aspects of interpersonal synchrony and impression formation. Specifially, it shows the importance of how much one person’s motion energy is leading their interaction partner’s motion energy in a dyadic interaction. However, it is important to note that upon closer inspection, the attribution of leading seems to have an effect only benefiting non-autistic individuals. In fact, the presently reported statistical interaction between diagnostic status and leading of motion energy on impression scores suggests that behaviours improving first impressions for non-autistic individuals do not have the same positive effect for individuals with ASD. Interestingly, this reduced beneficial effect of interpersonal synchrony on impressions shows parallels to the finding that link interpersonal synchrony and cognitive empathy. Koehne and colleagues showed that perceived interpersonal synchrony increased cognitive empathy in non-autistic but not in autistic participants [29]. Given that interpersonal synchrony has also been linked with increased levels of rapport [22, 23, 25, 26], it should be investigated whether beneficial effects of interpersonal synchrony on rapport also exist when people without ASD interact with people with ASD.

It is unlikely that the results obtained in the current study are merely due to a bias against individuals with ASD, given that our participants were not aware of the diagnostic status. In fact, less than 5% of participants guessed that the study’s topic was related to ASD at post-experimental debriefing. By reducing available information about motion, we showed the main effect of diagnostic status on impression formation shrunk to a trend, which means that the effect of diagnostic status could arguably be explained by distinctive features of motion in ASD.

Based on our results, it is unclear which further aspects of the video properties explain the remaining variance in impressions. In future studies, experimenters may want to include ratings of the videos showing only the target interactant, and, therefore, only half of the interaction. This would allow them to assess impressions based on motion alone instead of motion in interaction as we did in this study. Additionally, other interactive aspects of motion may explain variance in the impression scores between people with and without ASD. For example, Fujiwara and Daibo [57] recently proposed distinguishing behaviour matching, referring to the similarity in body postures, from interactional synchrony, referring to simultaneous movements and interaction rhythms. They showed that only behaviour matching enhanced empathic accuracy but not interactional synchrony. Therefore, future studies should aim to carve out how other aspects of motion may influence impression formation, along with interpersonal synchrony.

In addition to testing our preregistered hypothesis regarding the overall impression score, we also explored how diagnostic status, the leading role and their interaction influence nuanced aspects of impression formation by evaluating the separate ratings. The overall impression was based on judging the target interactant’s awkwardness, intelligence, likeability and trustworthiness, as well as how many friends the participants believed them to have (friends) and whether or not they would like to strike up a conversation with them (conversation). Both the rating of awkwardness and friends were only significantly predicted by diagnostic status, with autistic target interactants being rated less favourably compared to non-autistic target interactants. The effect of diagnostic status on awkwardness was the strongest effect present in this study: target interactants with ASD were on average judged 12.5 of 100 points more awkward than target interactants without ASD. The ratings of intelligence, likeability, conversation and trustworthiness were not significantly predicted by diagnostic status. These four ratings were predicted by a positive effect of the target interactant leading and the interaction between leading by the target interactant and the diagnostic status where interactants with ASD did not benefit from leading. This paints a similar picture to the overall impression score and is in contrast to previous studies that found influences of diagnostic status, especially on likeability and conversation ratings [12]. However, these studies did not show outlines of people in dyadic interactions, nor did they investigate the influence of interpersonal synchrony of motion energy on impression ratings. Trustworthiness was the only rating that was negatively predicted by how much the target interactant was led by their counterpart. Similarly, Sasson and colleagues found in their research that trustworthiness was rated differently which they argued could be due to it being a character trait rather than social appeal or competence [12]. This explorative analysis shows that not all aspects of impression formation are influenced in the same way by diagnostic status and leading. Additionally, it shows a strong effect of aspects of interpersonal synchrony impacting impression formation, particularly leading by the target interactant.

Participants in our study were always asked to rate their impression of one of two people in a dyadic interaction. We captured their gaze with webcam-based eye tracking to ensure that they focus on the target interactant. Our data shows that participants spent nearly two-thirds of the time fixating on the respective target field. Therefore, we are confident that they focused more on the target interactant, while still taking into account the full interaction, when giving their impression judgements. In addition to the relationship between diagnostic status and impression formation, we expected that diagnostic status would influence gaze patterns. However, neither fixation duration, nor switches between the two interactants were significantly predicted by diagnostic status. Based on this study, it is unclear whether this is due to the low spatial resolution of webcam-based eye-tracking or whether diagnostic status does not yield different gaze patterns. It is possible that more fine-grained gaze patterns would differ between forming an impression of an autistic or non-autistic individual and, thus, should be explored further.

Importantly, there are limitations to consider when interpreting the present results. First, webcam-based eye-tracking only offers rough estimates of gaze patterns because of coarse resolution and high data loss. In this study, we excluded more than half of the sample because gaze could not be reliably tracked for at least two thirds of the videos. Second, our sample was predominantly young (average 26.56 years old) and female (72.45%). Therefore, it is unclear if these effects extend to the general population. Future studies should aim for a more representative sample to ensure generalisability. Third, we used videos from two studies as source material for our stimuli [36, 37]. Both studies assessed conversations of heterogeneous and homogeneous dyads, and their setups closely matched. However, they also differed: Georgescu et al. [36] asked participants to engage in a five-minute, unstructured conversation planning a meal with foods and drinks that both participants dislike[58]. The participants were randomly paired adults with and without ASD who did not know each other before participating in the study. In contrast, Koehler et al.[37] conducted diagnostic interviews with participants from a clinical population referred for autism diagnostics. In this case, participants were always paired with the first author of the paper who was holding a clipboard. Autistic interaction partners in the videos were those who received an ASD diagnosis, wherein an ASD diagnosis was ruled out for clinical control interaction partners [59]. To account for these differences, we added video source as a regressor to our models and found that interactants in videos from Koehler et al.[37] were rated more favourably than interactants in videos from Georgescu et al.[36]. Lastly, the videos we used were reduced to outlines of two people engaging in conversations. Therefore, participants had limited information about the social interactions and the people engaged in them. This has the advantage that all effects found in this study can be traced back to the motion of two people interacting. However, this limits generalisability of these effects to more rich and natural social interactions.

In this study, we investigated the link between interpersonal synchrony and impression formation of autistic and non-autistic people. We measured interpersonal synchrony by using lagged cross-correlations to estimate the amount of leading of each interactant in a dyadic conversation. While there was no significant overall difference between impressions formed of autistic and non-autistic interactants, impressions were more favourable when the target interactant led their counterpart. This beneficial effect of leading only applied to non-autistic target interactants, suggesting that less favourable impressions of autistic people could be due to differences in interpersonal synchrony.

Authors’ note: We acknowledge the different preferences concerning the language used to describe the autism community (i.e., person-first vs. identity-first), which widely varies across languages and cultures. Therefore, we alternate between “autistic/non-autistic adults” and “adults with/without ASD” throughout this article. [1–7]

Acknowledgements

We want to thank Carola Bloch for acting as a liaison between the team in Cologne and the team in Munich. This work was supported by the FöFoLe-Programme 2020 (LMU Munich, Medical Faculty, Promot 12/2020) awarded to CFW and LT and the German Research Council awarded to CFW and KV (FA 876/3-1) and CFW (FA 876/5-1).

Author contributions

C.F.W., J.K., A.G., K.V. and L.T. conceptualised the study. L.T., A.N., J.K., R.T., K.V. and A.G. created the stimuli. C.F.W. secured research funding. I.P., L.T. and C.F.W. developed the analysis plan and preregistered the study on OSF. L.T. and S.L. created the experiment on GORILLA^TM, recruited participants and collected the data. I.P., L.T. and A.G. preprocessed the data. I.P. analysed the data and wrote the first draft of the manuscript. All authors contributed to the revision of the manuscript.

Data availability statement

Data and scripts to reproduce the results are available in the Open Science Framework repository: https://osf.io/whgx6/.

Additional information

The authors declare no competing interests.

Taboas A, Doepke K, Zimmerman C. Short report: Preferences for identity-first versus person-first language in a US sample of autism stakeholders. Autism. 2022. doi:10.1177/13623613221130845
Buijsman R, Begeer S, Scheeren AM. ‘Autistic person’ or ‘person with autism’? Person-first language preference in Dutch adults with autism and parents. Autism. 2022. doi:10.1177/13623613221117914
Keating CT, Hickman L, Leung J, Monk R, Montgomery A, Heath H, u. a. Autism-related language preferences of English-speaking individuals across the globe: A mixed methods investigation. Autism Res. 2023;16: 406–428. doi:10.1002/aur.2864
Kenny L, Hattersley C, Molins B, Buckley C, Povey C, Pellicano E. Which terms should be used to describe autism? Perspectives from the UK autism community. Autism. 2016;20: 442–462. doi:10.1177/1362361315588200
Lei J, Jones L, Brosnan M. Exploring an e-learning community’s response to the language and terminology use in autism from two massive open online courses on autism education and technology use. Autism. 2021;25: 1349–1367. doi:10.1177/1362361320987963
Tepest R. The Meaning of Diagnosis for Different Designations in Talking About Autism. J Autism Dev Disord. 2021;51: 760–761. doi:10.1007/s10803-020-04584-3
Bury SM, Jellett R, Haschek A, Wenzel M, Hedley D, Spoor JR. Understanding language preference: Autism knowledge, experience of stigma and autism identity. Autism. 2022. Abgerufen: https://doi.org/10.1177/13623613221142383
Ochi K, Ono N, Owada K, Kojima M, Kuroda M, Sagayama S, u. a. Quantification of speech and synchrony in the conversation of adults with autism spectrum disorder. PLoS One. 2019;14: 1–22. doi:10.1371/journal.pone.0225377
Noel JP, De Niear MA, Lazzara NS, Wallace MT. Uncoupling between Multisensory Temporal Function and Nonverbal Turn-Taking in Autism Spectrum Disorder. IEEE Trans Cogn Dev Syst. 2018;10: 973–982. doi:10.1109/TCDS.2017.2778141
de Marchena A, Kim ES, Bagdasarov A, Parish-Morris J, Maddox BB, Brodkin ES, u. a. Atypicalities of Gesture Form and Function in Autistic Adults. J Autism Dev Disord. 2019;49: 1438–1454. doi:10.1007/s10803-018-3829-x
Kissine M, Geelhand P. Brief Report: Acoustic Evidence for Increased Articulatory Stability in the Speech of Adults with Autism Spectrum Disorder. J Autism Dev Disord. 2019;49: 2572–2580. doi:10.1007/s10803-019-03905-5
Sasson NJ, Faso DJ, Nugent J, Lovell S, Kennedy DP, Grossman RB. Neurotypical Peers are Less Willing to Interact with Those with Autism based on Thin Slice Judgments. Sci Rep. 2017;7: 1–10. doi:10.1038/srep40700
Morrison KE, DeBrabander KM, Jones DR, Faso DJ, Ackerman RA, Sasson NJ. Outcomes of real-world social interaction for autistic adults paired with autistic compared to typically developing partners. Autism. 2020;24: 1067–1080. doi:10.1177/1362361319892701
DeBrabander KM, Morrison KE, Jones DR, Faso DJ, Chmielewski M, Sasson NJ. Do First Impressions of Autistic Adults Differ Between Autistic and Nonautistic Observers? Autism in Adulthood. 2019;1: 250–257. doi:10.1089/aut.2019.0018
Sasson NJ, Morrison KE. First impressions of adults with autism improve with diagnostic disclosure and increased autism knowledge of peers. Autism. 2019;23: 50–59. doi:10.1177/1362361317729526
Morrison KE, DeBrabander KM, Faso DJ, Sasson NJ. Variability in first impressions of autistic adults made by neurotypical raters is driven more by characteristics of the rater than by characteristics of autistic adults. Autism. 2019;23: 1817–1829. doi:10.1177/1362361318824104
Willis J, Todorov A. Making Up Your Mind After a 100-Ms Exposure to a Face. 2015;17: 592–598.
Uleman JS, Newman LS, Moskowitz GB. People as Flexible Interpreters: Evidence and Issues from Spontaneous Trait Inference. Adv Exp Soc Psychol. 1996;28: 211–279. doi:10.1016/S0065-2601(08)60239-7
Human LJ, Sandstrom GM, Biesanz JC, Dunn EW. Accurate First Impressions Leave a Lasting Impression: The Long-Term Effects of Distinctive Self-Other Agreement on Relationship Development. Soc Psychol Personal Sci. 2013;4: 395–402. doi:10.1177/1948550612463735
Brosnan M, Mills E. The effect of diagnostic labels on the affective responses of college students towards peers with „Asperger’s Syndrome“ and „Autism Spectrum Disorder“. Autism. 2016;20: 388–394. doi:10.1177/1362361315586721
Van Roekel E, Scholte RHJ, Didden R. Bullying among adolescents with autism spectrum disorders: Prevalence and perception. J Autism Dev Disord. 2010;40: 63–73. doi:10.1007/s10803-009-0832-2
Vacharkulksemsuk T, Fredrickson BL. Strangers in sync: Achieving embodied rapport through shared movements. J Exp Soc Psychol. 2012. doi:10.1016/j.jesp.2011.07.015
Tickle-Degnen L, Rosenthal R. The Nature of Rapport and Its Nonverbal Correlates. Psychol Inq. 1990;1: 285–293. doi:10.1207/s15327965pli0104_1
Miles LK, Nind LK, Macrae CN. The rhythm of rapport: Interpersonal synchrony and social perception. J Exp Soc Psychol. 2009;45: 585–589. doi:10.1016/j.jesp.2009.02.002
Lakin JL, Chartrand TL. Using Nonconscious Behavioral Mimicry to Create Affiliation and Rapporte. Psychol Sci. 2003;14: 334–339.
LaFrance M. Nonverbal Synchrony and Rapport: Analysis by the Cross-Lag Panel Technique. Soc Psychol Q. 1979;42: 66. doi:10.2307/3033875
Holroyd CB. Interbrain synchrony: on wavy ground. Trends Neurosci. 2022;45: 346–357. Abgerufen: https://pubmed.ncbi.nlm.nih.gov/35236639/
Mayo O, Gordon I. In and out of synchrony—Behavioral and physiological dynamics of dyadic interpersonal coordination. Psychophysiology. 2020;57: 1–15. doi:10.1111/psyp.13574
Koehne S, Hatri A, Cacioppo JT, Dziobek I. Perceived interpersonal synchrony increases empathy: Insights from autism spectrum disorder. Cognition. 2016;146: 8–15.
Behrends A, Müller S, Dziobek I. Moving in and out of synchrony: A concept for a new intervention fostering empathy through interactional movement and dance. Arts Psychother. 2012;39: 107–116. doi:10.1016/j.aip.2012.02.003
Bloch C, Vogeley K, Georgescu AL, Falter-Wagner CM. INTRApersonal Synchrony as Constituent of INTERpersonal Synchrony and Its Relevance for Autism Spectrum Disorder. Front Robot AI. 2019;6: 1–8. doi:10.3389/frobt.2019.00073
Moulder RG, Boker SM, Ramseyer F, Tschacher W. Determining synchrony between behavioral time series: An application of surrogate data generation for establishing falsifiable null-hypotheses. Psychol Methods. 2018;23: 757–773. doi:10.1037/met0000172
Ramseyer FT. Motion energy analysis (MEA): A primer on the assessment of motion from video. J Couns Psychol. 2020;67: 536–549. doi:10.1037/cou0000407
Ramseyer F, Tschacher W. Nonverbal synchrony in psychotherapy: Coordinated body movement reflects relationship quality and outcome. J Consult Clin Psychol. 2011;79: 284–295. doi:10.1037/a0023419
McNaughton KA, Redcay E. Interpersonal Synchrony in Autism. Curr Psychiatry Rep. 2020;22. doi:10.1007/s11920-020-1135-8
Georgescu AL, Koeroglu S, De AF, Vogeley K, Falter-Wagner CM, Tschacher W. Reduced nonverbal interpersonal synchrony in autism spectrum disorder independent of partner diagnosis: a motion energy study. Mol Autism. 2020;11: 1–14. doi:10.1186/s13229-019-0305-1
Koehler JC, Georgescu AL, Weiske J, Spangemacher M, Burghof L, Falkai P, u. a. Brief Report: Specificity of Interpersonal Synchrony Deficits to Autism Spectrum Disorder and Its Potential for Digitally Assisted Diagnostics. J Autism Dev Disord. 2021. doi:10.1007/s10803-021-05194-3
Koehler JC, Dong MS, Nelson AM, Fischer S, Späth J, Plank IS, u. a. Machine Learning Classification Of Autism Spectrum Disorder Based On Reciprocity In Naturalistic Social Interactions. 2022; 1–30.
Kuschefski M, Falter-Wagner CM, Bente G, Vogeley K, Georgescu AL. Inferring power and dominance from dyadic nonverbal interactions in autism spectrum disorder. Autism Res. 2019;12: 505–516. doi:10.1002/aur.2069
Capozzi F, Human LJ, Ristic J. Attention promotes accurate impression formation. J Pers. 2020;88: 544–554. doi:10.1111/jopy.12509
R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2021. Abgerufen: https://www.r-project.org
RStudio Team. RStudio: Integrated Development Environment for R. Boston, MA; 2020. Abgerufen: http://www.rstudio.com/
Raybaut P. Spyder Documentation. Spyder Proj. 2009.
Kumle L, Võ ML, Draschkow D. Estimating power in (generalized) linear mixed models: an open introduction and tutorial in R. Behav Res. 2021. Abgerufen: doi:10.3758/s13428-021-01546-0
Allison C, Auyeung B, Baron-Cohen S. Toward brief „red flags“ for autism screening: The short Autism Spectrum Quotient and the short Quantitative Checklist in 1,000 cases and 3,000 controls. J Am Acad Child Adolesc Psychiatry. 2012;51: 202–212.e7. doi:10.1016/j.jaac.2011.11.003
Lehrl S. Mehrfachwahl-Wortschatz-Intelligenztest: MWT-B. Spitta; 1999.
Anwyl-Irvine AL, Massonnié J, Flitton A, Kirkham N, Evershed JK. Gorilla in our midst: An online behavioral experiment builder. Behav Res Methods. 2020;52: 388–407. doi:10.3758/s13428-019-01237-x
Kleinbub JR, Ramseyer FT. rMEA: An R package to assess nonverbal synchronization in motion energy analysis time-series. Psychother Res. 2021;31: 817–830. doi:10.1080/10503307.2020.1844334
Ramseyer F, Tschacher W. Nonverbal synchrony or random coincidence? How to tell the difference. COST 2102. 2010;5967: 182–196. doi:10.1007/978-3-642-12397-9_15
Gorilla Support. Eye Tracking in Gorilla. 2022. Abgerufen: https://support.gorilla.sc/support/reference/eye-tracking#analysis
Bürkner PC. brms: An R package for Bayesian multilevel models using Stan. J Stat Softw. 2017;80. doi:10.18637/jss.v080.i01
Morey RD, Rouder JN, Jamil T, Urbanek S, Forner K, Ly A. BayesFactor. 2022. Abgerufen: https://richarddmorey.github.io/BayesFactor/
Barr DJ, Levy R, Scheepers C, Tily HJ. Random effects structure for confirmatory hypothesis testing: Keep it maximal. J Mem Lang. 2013;68: 255–278. doi:10.1016/j.jml.2012.11.001
Barr DJ. Random effects structure for testing interactions in linear mixed-effects models. Front Psychol. 2013;4: 3–4. doi:10.3389/fpsyg.2013.00328
Goss-Sampson M. Bayesian Inference in JASP. 2020. S. 1–120. doi:10.17605/OSF.IO/CKNXM
Brehm L, Alday PM. Contrast coding choices in a decade of mixed models. J Mem Lang. 2022;125: 104334. doi:10.1016/j.jml.2022.104334
Fujiwara K, Daibo I. Empathic accuracy and interpersonal coordination: behavior matching can enhance accuracy but interactional synchrony may not. J Soc Psychol. 2022;162: 71–88. doi:10.1080/00224545.2021.1983509
Tschacher W, Rees GM, Ramseyer F. Nonverbal synchrony and affect in dyadic interactions. Front Psychol. 2014;5: 1–13.
Arbeitsgemeinschaft der Wissenschaftlichen Medizinischen Fachgesellschaften. Autismus-Spektrum-Störungen im Kindes-, Jugend- und Erwachsenenalter, Teil 1: Diagnostik: Interdisziplinäre S3-Leitlinie der DGKJP und der DGPPN sowie der beteiligten Fachgesellschaften, Berufsverbände und Patientenorganisationen. 2016.

No competing interests reported.

IMPASsuppanalysisv2.docx

Download PDF

Journal Publication

published 18 Sep, 2023

Read the published version in Scientific Reports →

Editorial decision: Major revision
04 Aug, 2023
Reviews received at journal
03 Jul, 2023
Reviewers agreed at journal
26 Jun, 2023
Reviewers agreed at journal
24 Jun, 2023
Reviewers invited by journal
24 Jun, 2023
Editor assigned by journal
24 Jun, 2023
Editor invited by journal
08 May, 2023
Submission checks completed at journal
08 May, 2023
First submitted to journal
02 May, 2023

You are reading this latest preprint version

Impression formation and autism spectrum disorder: the role of interpersonal synchrony

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Methods

Participants

Experimental procedure

Stimulus creation

Analysis

Results

Influence of diagnostic status and leading on impression formation

Gaze patterns during impression formation

Exploring the influence of autism-like traits on impression formation

Exploring nuanced aspects of impression

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1