R1.1. Embracing heterogeneity

Some commentators felt that we ignored the heterogeneity that is part and parcel of autism (Levy). Several seem to have interpreted our argument as suggesting that all autistic people are motivated to engage in social interactions with others at all times (Dawson & Cowen; Fletcher-Watson & Crompton; Gillespie-Lynch; Mundy; Yankowitz & Clements). Fletcher-Watson & Crompton (para. 4) worried that such an argument “carries risks for autistic people” if “well-meaning [non-autistic] individuals” insist on social contact when it is not wanted. Dawson & Cowen (last para.) went further, writing that our failure to acknowledge that “social interest and motivation in autism are both atypical and characterized by high variance across individuals” marginalized many autistic people and “consigned [them] to difficult fates.”

We have much more to say about variability in social motivation – and particularly about how experience and expectations contribute to that variability – in section R2. But here we want to be clear that we did not intend to suggest that all autistic people will always be motivated to interact with all other people. We recognize that there will be both inter- and intra-individual variability in social motivation in autistic people. We attempted to acknowledge this in section 3 of the target article and explained that our focus was on “autistic people who profess an interest in others because we have not seen their perspective or experiences well represented in the scientific literature on autism and because they present a challenge to social motivation accounts of autism” (see also Livingston, Shah, & Happé [Livingston et al.]). We did not intend to suggest that those perspectives or experiences were shared by all autistic people. We regret that we were not sufficiently clear on this point in our target article, especially if the effect could be interpretations that carry risk for and/or marginalize autistic people.

R1.2. Of straw men and mice

Uljarević, Vivanti, Leekam, & Hardan (Uljarević et al.) felt that our characterization of the social motivation theory created “straw man” arguments. For example, they suggested that we misleadingly asserted that proponents of the theory do not recognize that there are individual differences in social motivation in autism. We disagree. In section 1 of our target article, we highlighted the position of one prominent group of proponents, who noted that if the social motivation framework of autism were correct, “social motivation deficits, which are primary, ought to appear in all or nearly all individuals with ASD [autism spectrum disorder]” (Chevallier et al. Reference Chevallier, Grèzes, Molesworth, Berthoz and Happé2012b, p. 236, emphasis added). In section 6, we noted that “Kohls et al. (Reference Kohls, Chevallier, Troiani and Schultz2012) [four of the five same authors] acknowledge that some autistic individuals may show greater social interest than others,” though we raised concerns that Kohls et al. expected autistic people to show social interest in conventional ways. We could have been more explicit that proponents sometimes predict that social motivation “deficits” are universal and sometimes acknowledge individual differences, but we do not believe that we created a straw man argument on this point.

Uljarević et al. also felt that it was not appropriate for us to include motor stereotypies in our interrogation of social motivation accounts of autism. They noted that Chevallier et al. (Reference Chevallier, Grèzes, Molesworth, Berthoz and Happé2012b) explicitly excluded repetitive behaviors and restrictive interests in their seminal paper outlining their framework of social motivation in autism. But for better or worse, social motivation accounts of autism extend beyond the one offered by Chevallier et al. (Reference Chevallier, Grèzes, Molesworth, Berthoz and Happé2012b), and others have suggested that motor stereotypies may be related to a lack of social interest in infancy.

For example, in section 2.3, we pointed to research showing that non-human animals raised in restricted or confined environments often develop motor stereotypies, a finding some have considered relevant to autism (e.g., Lewis et al. Reference Lewis, Tanimura, Lee and Bodfish2007). In a review of this research on non-human animals (and other work), Leekam et al. (Reference Leekam, Prior and Uljarević2011) wrote that, “It can be argued that the early onset of deficits in social, communicative, and adaptive behavior (arising from extreme social withdrawal) in infants and young children could interfere with experience-dependent behavioral and brain development in early life, as children with ASD begin to create their own restricted environment” and that “the message from these neurobiological findings [in non-human animal research] supports the desirability of active and intensive intervention that acts upon that self-imposed constrained environment to enhance brain development and reduce stereotypies” (p. 577). In section 5.1 of the target article, we argued that Leekam et al.’s framing of the relation between motor stereotypies and social withdrawal was problematic, in part because it is inaccurate to describe any child's environment as “self-imposed.” In their commentary, Uljarević et al. explained that Leekam et al. did not intend to endorse the social withdrawal-motor stereotypy proposal themselves, and we apologize for misreading them in this way.

R1.3. Do we really believe data about social motivation are useless?

Dawson & Cowen (para. 3) asserted that we “dismiss as irrelevant (‘unlikely to yield data that are useful in theory or practice,’ sect. 3, para. 3) any attempt to test claims about social motivation in autism.” This is incorrect. We love data and are most definitely in favor of attempts to test claims about social motivation in autism.

What we wrote was that “attempts to measure whether autistic people are, on average, less socially motivated than non-autistic people are unlikely to yield data that are useful in theory or practice” (sect. 3). This was in a section about autistic testimony, and we had just described a study showing that autistic adolescents report, on average, less pleasure in social situations than non-autistic adolescents (Chevallier et al. Reference Chevallier, Kohls, Troiani, Brodkin and Schultz2012a). We do not find average differences in self-reports of social motivation to be “useful in theory or practice” because, as we tried to explain in the target article and as we describe in more detail in section R2, whether someone reports themselves to be socially motivated at any given time depends on much more than their innate predisposition toward social stimuli and interaction. We agree that studies like those Dawson & Cowen summarized in their commentary, showing average differences between autistic and non-autistic participants on tasks related to the prioritization of social information and reputation, can be useful and important in efforts to understand the nature of social motivation in autism.

In our target article, we questioned whether neuroimaging data that show average differences in reward network activity between autistic and non-autistic people were useful in understanding autistic people's social motivation. Yankowitz & Clements mounted a vigorous defense of these neuroimaging efforts, suggesting that their utility lies in being able to “detect processes underlying cognition and behavior that are difficult to access by direct report or behavioral experiments” (para. 2). We agree that neuroimaging can uncover interesting and possibly important underlying processes. But we worry when average differences in neural responses to reward in autistic people compared with non-autistic people are used to make claims about differences in psychological constructs like social motivation (e.g., Poldrack Reference Poldrack2011; see also Mottron).

This is not an idle worry. Yankowitz & Clements explained in their commentary that “when neuroimaging studies report that a group of autistic individuals shows lower social motivation than a group of non-autistic individuals, these effects are averaged across each group” (para. 4, emphasis added), and that the effect size of 0.25 they found in their meta-analysis of reward processing differences means that “only 60% of autistic individuals show lower social motivation than the average non-autistic individual” (para. 4, emphasis added). But the neuroimaging studies included in their meta-analysis showed diminished reward processing, not lower social motivation. In their commentary, Yankowitz & Clements conflated an average difference in activity in particular brain regions on a particular set of tasks having to do with reward processing with differences in what they acknowledged is “the broader, multifaceted process that produces the mental state we call ‘motivation’” (para. 3). They are not the same thing.

Dawson & Cowen cautioned that our argument was based on “selective popular narratives about autism” that were provided at the expense of “a complex literature on social and nonsocial cognition, based on data from thousands of autistic study participants” (last para.). We disagree. We certainly made use of autistic testimony, but throughout we also described and relied on a complex literature on social and nonsocial cognition, with both autistic and non-autistic participants. Uljarević et al. similarly suggested that many of the arguments we offered were “not scientifically testable” (para. 3), including our proposal that social motivation may be manifested in unconventional ways. We disagree. We laid out the start of such a research agenda for investigating this issue in section 5.2 of the target article.

R2.1. Wanting versus wanting plus

Keifer, Dichter, McPartland, & Lerner (Keifer et al.) rightly pointed out that the construct of social motivation in the scientific literature is poorly defined. In their view, there are four components: reward motivation (“wanting”), reward processing (“liking”), reward learning, and habit formation. Each may be subserved by distinct neurobiological circuitry. Keifer et al. suggested that our argument that many autistic people say they want social interaction (despite behaviors that get interpreted otherwise) may present a challenge to the reward motivation (“wanting”) component of social motivation, but it may not challenge the other three components.

We agree that we used “social motivation” in our target article (and use it in this response) primarily in the sense of wanting social interaction or social connection (see also Kissine). This is arguably what most people who would like to make use of the science but are not embedded in it themselves understand social motivation to be. This is reflected in, for example, a Centers for Disease Control (2018) fact sheet on autism, which asserts that “some people with ASD might not be interested in other people at all,” and an Autism Speaks (n.d.) “learn the signs” web page, which suggests a “red flag” for autism “at any age” is a “persistent preference for solitude.” Some readers may object that a lay definition of a construct should not drive scientific discussions of that construct. But “wanting” is not just the sense of social motivation in lay terms; it also has been used as the default definition in professional and scientific discourse on autism.

Consider a series of events that helped motivate us to write the target article and to begin studying unconventional ways that people can show their social motivation. A few years ago, one of us (Jaswal) met with the parents of a 9-year-old nonspeaking autistic boy who had received an evaluation to help determine what kind of augmentative and alternative communication device the school district might provide him. As part of that assessment, his teacher had been interviewed about his ability to communicate in various ways. The first line of a section called “communication interaction skills” included the following summary: “Per teacher report, [this child] does not have a desire to communicate.” That is, according to his teacher, this child was not motivated to engage in the social act of communication. Such a conclusion came as a surprise to the child's parents, who could list a dozen or more ways in which he showed his desire to communicate, and who were understandably concerned about the negative effects this teacher's essentialized belief about their son's motivation (see Bartsch & Estes) would have on how she treated him.

Coincidentally, at around the same time, we came across a paper on nonspeaking autistic school-aged children, in which the authors described some factors that may predict why about 30% of autistic people do not acquire fluent spoken language (Tager-Flusberg & Kasari Reference Tager-Flusberg and Kasari2013). Among the factors they described was a lack of social motivation: “In some cases, the almost complete absence of any social motivation may be associated with no spoken language” (p. 469). The authors did not specify what component of the scientific construct of social motivation they thought was absent. But we think a fair reading is that they meant it in the “wanting” sense: Some individuals may not speak because they are not motivated to engage in the social act of communication.

We have gone on at some length here because Keifer et al. noted that “most ASD research does not specify any component as decisively representing social motivation” (para, 4). We agree, but the “wanting” component seems to be the default interpretation – the component that, in the absence of any specification, is most often assumed to be lacking or somehow deficient when this term is used in the context of autism (see Brown & Foxley-Webb and Gillespie-Lynch for historical accounts of why this may be). Like Keifer et al., we are not sure what implications our challenge to the wanting component of social motivation has for the other components they describe; we look forward to additional research that investigates how they are related to one another. Regardless of how the components are related, we believe (and expect that Keifer et al. would agree) that it is a mistake to assume that any of the components are static traits of individuals.

R2.2. Social motivation is dynamic, emergent, embedded, and perceived

Social motivation (in the sense of a desire for social interaction) is frequently talked about as if it were a fixed trait. Even when heterogeneity is acknowledged, it tends to be discussed as variability across individuals, as in “some people are more socially motivated than others.” But this characterization misses two critical features of social motivation.

First, an individual's motivation to interact with other people is not fixed; it varies over the course of their lives (e.g., Carstensen Reference Carstensen and Jacobs1993) and even over the course of a day, depending on a host of variables, including their other goals, who is available for interaction, their ability to tune out distracting sensations and thoughts, the effort required, and importantly, how they have been treated in the past. Indeed, Brown & Foxley-Webb argued that at least part of the reason there is variability in social interest in autistic (and non-autistic) people has to do with variability in how they have been treated. They proposed that frequent adverse social experiences – being socially excluded, for example, or routinely being prevented from engaging in self-regulating motor stereotypies – could result in epigenetic changes to genes thought to be relevant to social interaction (e.g., Gordon et al. Reference Gordon, Martin, Feldman and Leckman2011). Mitchell, Cassidy, & Sheppard (Mitchell et al.) pointed out that “how one grows socially and emotionally depends not just on an immutable aspect of the individual's constitution but also on how the behaviour of others shapes how you behave, which in turn shapes how others behave towards you, which shapes how you behave towards them, and so on” (para. 2).

A second critical feature of social motivation we tried to highlight in our target article is that an individual's social motivation depends not just on how much that person wants social interaction at a particular point in time, but also on how others interpret that person's behavior. If others see a person's behavior as a bid for connection – say, that person's echolalia or a sidelong glance – they are more likely to act on it than if they do not interpret it in that way, which, as described above, will affect how much that person wants to interact, and so on.

Our emphasis on the observer's interpretation of behavior resonated with a number of commentators. For Rutherford, for example, social motivation “exists in the space between persons” (para. 5). Friedner noted that autistic individuals’ actions must be seen as “interactional and embedded within enabling or disabling social worlds” (para. 1). Heasman & Gillespie noted that “actors and observers may have differing perspectives on the extent to which behavior is social” (para. 1). Delafield-Butt, Trevarthen, Rowe, & Gillberg (Delafield-Butt et al.) wrote that “meaningful social relations require sensitive appreciation and forms of response that respect all forms of expression and seek to share experiences” (last para.). Moseley & Sui appreciated our focus on the role of “the non-autistic participant in a dyad, whose beliefs may markedly affect, or worse reduce, interaction” (para. 1).

In section 5 of the target article, we advocated for research into unconventional, idiosyncratic ways social interest could be manifested and perceived in autism as a step in changing unhelpful non-autistic beliefs. We suggested that lessons might be learned from how social interest is manifested and perceived in disabilities other than autism (e.g., blindness) and across cultures. Gliga & Elsabbagh agreed that additional research to characterize environments in which social motivation is likely to be high in autism (what they call “optimal environments”) is important. But they disagreed “with the suggestion that this requires revisiting the classical way of measuring motivation” (last para.). They argued that attempts to “bring a social stimulus closer and maintain engagement with it” represent the critical features of social motivation and that “the nature of the behaviour that brings about the reward was never key to theories of motivated behaviour” (last para.).

Even though the nature of the behavior may not be key to theories of social motivation, whether an observer perceives that behavior as socially motivated is key. For reasons having nothing to do with their social interest, some autistic people may not behave in ways that non-autistic people interpret as indicating social interest (e.g., eye contact), and/or they may behave in ways that non-autistic people sometimes misinterpret as indicating social disinterest (e.g., echolalia). These “errors” of omission and commission can lead some observers to mistakenly assume that they are not socially motivated, which may have cascading effects.

Gliga & Elsabbagh described a number of important studies demonstrating, for example, early similarities and differences in overt orientation toward faces and eyes between infants who later receive a diagnosis of autism and those who do not. But we think studying how infants respond to typical social stimuli like faces is only part of the equation; it is also important to consider how infants (and others) are responded to. This requires a consideration of what behavior “counts” as indicating social motivation. It is in this sense that we believe that “revisiting the classical way of measuring motivation,” per Gliga & Elsabbagh, may be useful.

One lovely example of such a revisitation comes from work on attachment described by Oppenheim, Koren-Karie, & Joels (Oppenheim et al., paras. 4-6). They described how one 3.5-year-old nonspeaking autistic boy did not show the conventional interactive behaviors indicative of secure attachment on reunion after separation from his mother in the Strange Situation Procedure (SSP): He did not engage in eye contact, “vocalize, or ask to be held, and in fact was turning his back to her throughout.” Nevertheless, they explained, he kept his mother close, pulling her by the hand behind him as he returned to the playroom and later glancing at her once briefly “as if confirming her whereabouts.” This constellation of behaviors may very well have been unique to this particular child, but Oppenheim et al. interpreted it (in combination with his distress during the separation period) as indicative of a secure attachment to his mother – a pattern they say is associated with “a history of having a sensitive caregiver who sees and feels things from the child's point of view” even when the child may not act in expected ways (see also Delafield-Butt et al.). Note also that Oppenheim et al. interpreted the boy's brief glance toward his mother in a rich, intentional way, not as an accidental or meaningless aside. In the context of what had just transpired during the SSP, it appeared to them a meaningful bid for connection.

Because our target article focused on experiential, contextual, and interpretive factors that contribute to differences in social motivation, a reader might wonder whether we believe that there are any inherent aspects of an individual that contribute to these differences. We do, but here too we want to emphasize the role the environment plays in how such predispositions are expressed. Consider temperament, for example. Some infants are highly reactive: They tend to respond strongly to novel people, places, and things, and they become distressed more easily or more quickly than less reactive infants (Rothbart Reference Rothbart2011). One could interpret their distress on meeting new people as indicating diminished social motivation; they do not appear to find this kind of social situation very rewarding. But in some contexts, some of these infants can appear highly sociable; for example, if they are introduced to new people one at a time and in familiar contexts (Thomas & Chess Reference Thomas and Chess1977). Autistic individuals may have different thresholds for social (and sensory) stimulation than non-autistic people (see Kapp, Goldknopf, Brooks, Kofner, & Hossain [Kapp et al.]; Olderbak, Geiger, & Wilhelm; Pastore, Dellantonio, Mulatti, & Esposito [Pastore et al.]; Perrykkad; Riva, Di Lernia, & Dakanalis), and this may lead them to withdraw from certain social situations. But that does not mean that they are not socially interested.

Another inherent aspect of an individual that could contribute to differences in social motivation involves motor differences. Delafield-Butt et al. made a convincing case that the production and perception of motor movements are disrupted in autism, which could disrupt the ability to appear (and, perhaps in the long term, be) communicatively and socially engaged (see also Torres et al. Reference Torres, Brincker, Isenhower, Yanovich, Stigler, Nurnberger, Metaxas and Jose2013). If the timing of the production of movement in an interaction is perturbed, for example, this “can be misread as absence of sociability by persons with whom an autistic child is seeking meaningful engagement and shared learning” (para. 5). Additionally, when motor development is delayed or different, there can be cascading effects on the number and quality of opportunities for social interaction and communication a child is provided (e.g., LeBarton & Iverson Reference LeBarton and Iverson2016).

R2.3. Valuing different forms of sociality

According to Friedner, one reading of our target article is as an attempt to “render autistic people commensurable to non-autistic people” (para. 3) (see also Rutherford). Despite surface-level differences in social interest, we seem to be saying that autistic and non-autistic people are actually quite similar on this dimension. But Friedner (para. 3) poses a provocative question: “What if (some) autistic people were actually quite different from non-autistic people in their world views and in their ontologies?”

Indeed, a number of commentators suggested that autistic people do not just express social interest differently than non-autistic people, as we argued in our target article; autistic people also experience and value it differently. For example, Kapp et al. (para. 3) wrote that “autistic people report relating to others differently,” and Silverman (para. 4) pointed out that “autistic sociability and friendships may need to be defined differently from conventional (neurotypical) expectations” (see also Dawson & Cowen; Forgeot d’Arc & Soulières; Mottron). In our target article, we focused on challenging the view that autistic people are not socially interested; we did not consider whether their sociality could be different from that of non-autistic people. These commentaries have convinced us that it can be.

For example, Heasman & Gillespie provided an insightful description of sociality that may seem foreign to many non-autistic people. They found that when playing video games together, autistic people engaged in behavior that, to a non-autistic person, could be interpreted as indicative of social indifference – echolalia, shouting, lengthy monologues, and so on. But autistic participants did not interpret each other's behavior in this way. They were open to “a different type of sociality, one that permits periods of incoherent and fragmented dialogue in favor of pockets of intense rapport, reciprocation, and humor” (para. 5).

In fact, as Livingston et al. noted, some autistic people find having to adopt non-autistic social rules and interactive styles to be highly effortful, “comparing it to physical exercise or mental arithmetic, thus draining resources required for daily functioning” (para. 3). Livingston et al. described a group of autistic people who are so highly socially motivated that they routinely put themselves in social situations with non-autistic people even though they find these situations to be stressful. Unfortunately, the payoff may not be worth the cost: Livingston et al. described links between this approach to navigating the non-autistic social world and some of the same negative health outcomes associated with loneliness in autistic people, including depression and suicidal ideation (Kapp et al.; Mitchell et al.; Moseley & Sui). Livingston et al. argued that some autistic people may respond adaptively to the effort required to conform to non-autistic social norms by withdrawing from or avoiding social situations, instead seeking out “environments where non-social skills are valued over social skills” (para. 4). We add that in line with Heasman & Gillespie, another adaptive response might involve seeking out environments where a different type of sociality is valued.

R2.4. Valuing and problematizing asociality

Is it possible to carve out a space for valuing not just a different type of sociality in autism, but also asociality (Friedner; Gillespie-Lynch)? We embrace Fletcher-Watson & Crompton’s call for acceptance of “all people regardless of how enthusiastic they are about spending time with other people” (para. 4). As Friedner wrote, insisting that everyone be socially interested could be seen as “coercive and does not allow for diverse ways of being in the world” (para. 4). We certainly do not want to defend a position that is seen as coercive or intolerant of diverse ways of being.

At the same time, as we explained in section R2.2, we do not see social motivation as a fixed trait residing within an individual. Sometimes people want to interact, sometimes they don't. Sometimes the environment can be more conducive to interaction, sometimes less so. Sometimes potential interlocutors are appealing social partners; sometimes they are not. How you have been treated in the past will affect how much you want to (or try to) interact. We would be surprised if any of this is controversial. But we think the existence of inter- and intra-individual variability and especially its causes have not been given sufficient attention in traditional approaches to thinking about social motivation in autism. A preference for solitude (and/or aversion to social contact) could be the result of one's social experiences.

And so, how do we reconcile our desire to accept all ways of being regardless of how much they involve social interaction with what we see as the malleability of social motivation? Here is a preliminary answer that is surely not going to be satisfactory to many people. It seems reasonable to us to presume that from birth, humans are motivated to form social bonds, just as they are motivated to seek nourishment; being part of a community confers considerable survival and reproductive benefits (e.g., Baumeister & Leary Reference Baumeister and Leary1995; Dweck Reference Dweck2017; Stevens & Fiske Reference Stevens and Fiske1995). To be clear, this is not to suggest that all newborns are driven to form social bonds to the same extent or in the same ways. As we noted earlier, temperamental and other differences can influence both the amount and kind of social motivation they experience, as well as how they express it. Also, as we keep emphasizing, how they are responded to will interact with whatever drive toward the social world they may have, changing that drive in the process.

Over the course of life (or at various points in life), some people may decide that they prefer solitude even when they have been treated well by others. Some people may decide they prefer solitude because they have been treated badly (“to hell with you people,” Robledo et al. Reference Robledo, Donnellan and Strand-Conroy2012, p. 6). Both of these are acceptable ways of being if they were voluntarily chosen and/or are ways of being that a person feels comfortable with. But we were motivated to write the target article by autistic people who end up in solitude even though they say they want social interaction. They have been forced into an asocial way of being that they say they are not comfortable with.

R2.5. Why does it matter?

The reason we are so insistent that a desire for social relationships should be considered the default (recognizing that desire can be expressed, experienced, and valued by autistic people differently from non-autistic people) is that there are clear benefits to having them. As House et al. (Reference House, Landis and Umberson1988, p. 541) put it, “Social relationships, or the relative lack thereof, constitute a major risk factor for health – rivaling the effect of well-established health risk factors such as cigarette smoking, blood pressure, blood lipids, obesity and physical activity.” The mechanisms for this relation are beyond our scope, but proposals have been made that social support may, for example, buffer neuroendocrine and behavioral response to stressors (e.g., Coan & Sbarra Reference Coan and Sbarra2015; Holt-Lunstad et al. Reference Holt-Lunstad, Smith and Layton2010).

The health effects of social support among autistic people could be different, but we are not aware of data on this point. Interaction is not a panacea for loneliness (Moseley & Sui), and many autistic people prefer different kinds of interaction than non-autistic people (sect. R2.3). But given the elevated risk of premature mortality in autism compared to the non-autistic population (Hirvikoski et al. Reference Hirvikoski, Mittendorfer-Rutz, Boman, Larsson, Lichtenstein and Bolte2016); the elevated levels of social isolation among autistic adults compared to other groups (Orsmond et al. Reference Orsmond, Shattuck, Cooper, Sterzing and Anderson2013); feelings of loneliness among autistic people and its link to depression, self-injury, suicidal ideation, and suicidality (Moseley & Sui); and the fact that some autistic people say they want connection even though their behavior is sometimes misinterpreted to mean that they do not (Pellicano, den Houting, Du Plooy, & Lilley [Pellicano et al.]), we believe a low-cost and potentially high-impact strategy is to presume that people want some kind of social connection until they inform you otherwise.

Do we think everyone should be motivated to be social? No. But we do think everyone should be presumed to want some kind of social interaction until they make it clear that they do not. This, of course, raises a number of questions. For example, what would serve as conclusive evidence that someone has chosen and/or feels comfortable with solitude? We have argued that autistic people say their behavior is sometimes mistakenly interpreted to mean that they prefer to be alone, so should we rely on their words instead? What about children who express in actions and/or words a desire for solitude (e.g., Calder et al. Reference Calder, Hill and Pellicano2013), but whose caregivers want to provide them with social opportunities? We have argued that social motivation is not fixed, so should repeated attempts be made to engage someone who has said they are not interested? Answering these questions in a way that applies to everyone under all circumstances is, of course, impossible. But understanding that social motivation can be expressed and experienced in a variety of ways and that it is labile may help people come to more informed answers when these kinds of questions arise.

R3.1. Is autistic testimony important to understanding social motivation in autism?

One type of evidence we used in the target article to argue against the assumption that autistic people lack social interest was testimony from several autistic people who expressed a desire to interact with others and/or who offered explanations unrelated to social motivation for their unusual behaviors. To be honest, we were somewhat apprehensive about how this approach would be received. After all, as Pellicano et al. noted, “these qualitative, subjective reports have for the most part been eschewed by scientists, who often perceive them as contributing no more than anecdotal evidence” (para. 2). Silverman (para. 1) pointed out that researchers sometimes consider people diagnosed with autism (and other conditions) to be “unreliable narrators of their own experiences.”Footnote ¹ But we needn't have been concerned: At least among the authors of the commentaries on our target article (even some of the more critical ones), there seems to be agreement that autistic testimony has an important role to play in understanding social motivation in autism.

That said, some of the commentators raised concerns that the testimony we highlighted was, in Mitchell et al.’s words, “unsystematic and prone to various biases associated with the context in which the testimony was given and the motivations of those who gave it” (para. 5). For example, Silverman (para. 2) noted that people who write autobiographies “may be more invested in reframing conventional accounts of autism than others” (see also Fletcher-Watson & Crompton). We did not intend to suggest that the personal accounts we detailed were representative of all or even most autistic people. Instead, our goal in using such accounts was to highlight a perspective that has not been given sufficient attention in the autism literature, namely, that of autistic people who say they are socially interested (see also Livingston et al.).

We join with the many commentators who called for additional work that can provide a more systematic study of autistic perspectives on social motivation than our target article was able to provide (Fletcher-Watson & Crompton; Kapp et al.; Livingston et al.; Moseley & Sui; Silverman). As noted earlier, we do not think that comparisons of average levels of self-reported social motivation at a given point in time are all that useful. But systematic efforts to learn, for example, how a wide range of autistic people characterize their own social motivation, whether and how they have experienced its lability, and the factors they think contributed to their experience of it would be fascinating and important.

R3.2. How should autistic perspectives be obtained and used?

As Pellicano et al. noted, there is “growing acknowledgment within the scientific community that autistic people possess insight into autism that has been too frequently overlooked” (para. 3). Autistic people are, after all, the ones living and experiencing autism. This positive development raises fundamental questions about how those insights should be solicited and used in the research process.

A number of commentators wrote about the need to overhaul the research process in autism entirely – from the current system where the extent of autistic people's involvement tends to be as participants in studies that non-autistic researchers have designed on questions that non-autistic researchers have decided are important to one where autistic people are involved in all stages of the research process, from identifying the questions to disseminating the results (Heasman & Gillespie; Kapp et al.; Pellicano et al.; Silverman). We enthusiastically embrace this call to arms. Authentically including autistic people will lead to better, more respectful autism science and to the development of better, more respectful support for autistic people.

Turning to how autistic perspectives can be conveyed via testimony, Dinishak correctly inferred that our position is that “autistic testimony contains valuable phenomenological data that should be taken seriously by autism researchers” (para. 4). She asked whether taking autistic testimony seriously is a recommendation or obligation, pointing out that the answer to this question has implications for the theory and practice of autism science. If it is merely a recommendation, scientists would be free to ignore testimony that was not consistent with the prevailing theory or other kinds of data. If it is an obligation, Dinishak pointed out, there are a number of ways scientists could fulfill it. They could, for example, use testimony as (a) evidence supporting (or refuting) a theory of autism; (b) an explanandum, the thing that a theory ought to explain; (c) the inspiration for a particular research question; or (d) at the extreme, the only way to gain insight about autism.

We think that autism science should be motivated by efforts to understand the lived experience of autistic people, and this requires, as Rutherford put it, “foreground[ing] what autistic people think is going on” (para. 4). Thus, although we do not believe that testimony is the only way to gain insight about autism (sect R1.3), we do believe that scientists have an obligation to seek out and use it in all of the other ways detailed above. This may be controversial: In anthropology, Rutherford explained, self-report is an opportunity; in psychology, Pellicano et al. explained, it is often considered a problem. But like Pellicano et al., we think that in addition to participatory research, another important aspect in the overhaul of the research process in autism should be a greater emphasis on “methodological triangulation,” whereby introspective reports like testimony are valued and integrated with behavioral and brain measures (see also Kapp et al.; Moseley & Sui; Power et al. Reference Power, Velez, Qadafi and Tennant2018).

Dinishak argued that taking autistic testimony seriously has both epistemological value (in that it leads to more accurate knowledge) and moral value (in that it affirms autistic people's position as reliable knowers, helping to prevent epistemic injustice; Fricker Reference Fricker2007). We agree with this thoughtful characterization. As Li & Koenig explained, “if we discount the credibility of others’ testimony on the grounds of who they are or what they look like, we may give researchers and clinicians more credibility than warranted and less credibility to certain speakers than they deserve” (para. 4).

R3.3. Should testimony from some autistic people be excluded?

About 30% of autistic people cannot use speech reliably or at all (DiStefano et al. Reference DiStefano, Shih, Kaiser, Landa and Kasari2016), a disability likely caused, at least in part, by significant difficulties in motor planning, coordination, and/or execution (Gernsbacher et al. Reference Gernsbacher, Sauer, Geye, Schweigert and Goldsmith2008a). Some nonspeaking autistic people have learned – after years of instruction and practice – to communicate by typing on a keyboard or pointing to letters on an alphabet board. Typically, these individuals rely on assistance from another person as they spell, someone who provides physical, emotional, and/or attentional support (Cardinal & Falvey Reference Cardinal and Falvey2014).

Vyse, Hemsley, Lang, Lilienfeld, Mostert, Schlinger, Shane, Sherry, & Todd (Vyse et al.) objected to our inclusion of testimony from several nonspeaking autistic people who communicate with assistance. In experimental settings, messages produced by people who communicate in this way have been influenced by those assisting them (but see Cardinal et al. Reference Cardinal, Hanson and Wakeham1996). Vyse et al. argued that the testimony of the nonspeaking people we quoted therefore could not have been their own.

We recognize that nonspeaking individuals who communicate with assistance could unintentionally or deliberately be led to produce testimony that is not their own. Indeed, there is always a possibility that a person's testimony – spoken, signed, written, or written with assistance – could have been influenced in some way. But there are a number of reasons to believe that testimony produced with assistance, including the testimony used in our target article, can reflect a nonspeaking autistic person's own thoughts. First, some people who learned to communicate with assistance have developed the ability to communicate independently (United for Communication Choice 2018); the perspectives they communicate independently are consistent with those they communicated with assistance (e.g., Kedar Reference Kedar2012; Reference Kedar2018). Second, nonspeaking people who communicate with assistance routinely convey – in naturalistic, everyday settings – information that the assistant could not have known (e.g., Bigozzi et al. Reference Bigozzi, Zanobini, Tarchi, Cozzani and Camba2012). Third, analyses of text produced by nonspeaking autistic people who communicate with assistance have shown their writing styles to be different from those of the assistants (e.g., Tuzzi Reference Tuzzi2009).

Finally, the speed and accuracy with which some nonspeaking autistic people spell make it unlikely those individuals are merely responding to cues from an assistant. For example, one of us (Jaswal) has collected timing and accuracy data from an individual who points to letters on an 8.5- by 11-inch alphabet board held vertically by an assistant. In one 47-letter utterance (“I feel like world is waiting on me not the other way around”), he pointed to a letter, on average, every 1,085 ms (SD = 381 ms) and made no spelling errors. It seems unlikely that in the span of about a second, he could use a subtle cue produced by the assistant to accurately select 1 of 26 possible letter targets, and then repeat this process 46 additional times consecutively without error.

Given that about one-third of autistic people are nonspeaking, that some have learned to communicate with assistance, and that some of them have written on the topic of our target article, we believe our inclusion of their testimony was reasonable and appropriate. Interestingly, in most cases, and although this was not intentional, when we quoted a nonspeaking person to make a particular point, we also quoted a speaking person who made a similar point. For example, on eye contact, we quoted Higashida (Reference Higashida2013), a nonspeaking autistic person, who wrote that it “feels a bit creepy, so I tend to avoid it” (p. 25) and Tammet (Reference Tammet2006), a speaking autistic person, who explained that it feels “strange and uncomfortable” (p. 75). The content of the quotes from nonspeaking autistic people was consistent with the content of the quotes from speaking autistic people. Our arguments about social motivation in autism would have been the same had we limited ourselves to quoting only speaking people.