Recent efforts to quantify clinical heterogeneity in autism spectrum disorder (ASD) indicate that verbal autistic girls and women behave differently than male counterparts, even when matched on social symptom severity [
38]. Certain behaviors, like mimicking other people’s facial expressions or gestures, making eye contact, and memorizing social scripts may serve as “camouflage” for social impairments [
63], and are thought to be utilized more often by autistic girls and women than autistic boys and men [
49]. Sex-specific differences in autistic behaviors, including camouflaging, are not explicitly measured by current gold standard diagnostic instruments [
87], leading to concerns that girls are systematically under-diagnosed compared to boys [
68]. Sex is a core biological difference that impacts children’s experiences before, during, and after ASD symptoms emerge [
25], so understanding the effects of sex on ASD expression has important implications for diagnostic and clinical practice. For instance, quantifying the precise nature of sex differences in ASD could help clinicians develop personalized interventions that are more effective than a one-size-fits-all approach to autism treatment.
Direct behavioral measurement is the primary method for diagnosing ASD [
69], but recent evidence points to a variety of “autistic behaviors” that present differently in girls. For example, atypical or reduced gesturing is common in ASD [
32,
52], but empirical studies suggest that verbal autistic girls produce gestures that are more vibrant and noticeable than autistic boys [
95]. Researchers have argued that autism is associated with unusual verbal disfluency patterns [
47,
51,
65], but autistic girls produce disfluency patterns that are sex-typical, and measurably distinct from the disfluency patterns of autistic boys [
82]. Autism is associated with diminished social attention [
60], but recent evidence from infrared eye tracking suggests that autistic girls may look more at faces than autistic boys (Harrop et al., under review). On the playground, autistic girls are more likely to hover near groups of other girls, whereas autistic boys are more likely to be isolated [
35]. As adults, autistic women show greater discrepancies between outward symptoms of ASD and their own internal experiences [
64]. Taken together, these differences suggest that the behavioral symptoms of ASD manifest differently in girls and women than they do in boys and men.
For verbal individuals, language is an important pathway to friendships, romantic relationships, jobs, and overall quality of life. Given population sex differences in a variety of linguistic domains [
75,
76,
108], and the core dimensions of social communication that are used to diagnose ASD [
1], understanding similarities and differences in language produced by autistic boys and girls could shed light on sex-specific differences in the clinical presentation of autism. In this study, we focus on sex and diagnostic group differences in the language children use during a brief storytelling task.
Narratives
Storytelling is an ancient social art that hinges on interpersonal skills. Reliance on oral histories has diminished over time, but brief daily storytelling is preserved as a central component of communal living. Even the simplest question, “How was your day?” provides an opportunity for short narratives to strengthen interpersonal connections. Storytelling is ubiquitous, and the basic elements of storytelling are acquired by most children in early childhood [
39,
85]. However, storytelling relies on much more than vocabulary and grammar. In fact, successful storytellers leverage a rich array of skills, including working memory [
18], executive function [
20], and a sense of social appropriateness, or knowing how much information to provide to different kinds of listeners [
105]. Practical language skills that use social context to facilitate effective communication (i.e., pragmatic language abilities) are centrally important for storytelling. For example, speakers must monitor whether listeners are engaged, and whether they understand the story. They must watch facial expressions and interpret nonverbal cues to guide them to explain further, pause, or otherwise act to get the listener back on track. Given this important pragmatic dimension to storytelling, it is unsurprising that narrative competence is closely related to social ability [
16,
104].
Narrative skills in autism
Pragmatic language skills are universally impaired in ASD [
1], with a large body of research showing that the narratives of autistic adults [
5,
8,
10,
66,
74] and children differ from typically developing (TD) peers in a variety of measurable ways [
3]. These differences include impoverished event explanations [
21,
58,
101], reduced story structure complexity [
83], reduced coherence [
71], reduced syntactic complexity, more ambiguous pronouns, fewer story grammar elements [
6], poorer inferencing, and a tendency to include extraneous information [
72]. In adolescence, even “optimal outcome” individuals who no longer meet ASD criteria show subtle language differences during narrative tasks, with higher rates of self-correction and idiosyncratic speech compared to controls [
19,
33,
56,
100]. Importantly, careful matching on language ability does not ameliorate diagnostic group differences; a number of studies found that the narratives of autistic children still differ on structural, evaluative, and global narrative features, including shorter stories and reduced causal statements, suggesting that other factors besides language ability must explain performance discrepancies [
36,
57,
58,
98,
101].
Narratives produced by children with ASD not only differ from narratives produced by typically developing peers, but from other clinical groups as well. Compared to narratives produced by children with specific language impairment (SLI), autistic children’s narratives show weaknesses in areas that rely on perspective-taking, such as mental state language (e.g.,
think,
know,
believe), referencing, and relevancy [
28]. However, children with ASD and comparison children with SLI produce similarly simplistic and semantically lean narratives that omit important story elements, relative to TD controls [
79] and both groups make more ambiguous references during storytelling [
78]. This suggests that social impairment and language deficits result in distinct narrative profiles [
41]. Compared to children with attention deficit/hyperactivity disorder (ADHD), autistic children refer less to cognitive states and provide less coherent narratives. However, both groups leave out key story components and produce shorter narratives than TD peers [
91]. Because individuals with ASD narrate in a way that is unique to their diagnostic group, narrative generation and retelling tasks are viewed as clinical tools that shed light on various aspects of atypical development [
13,
28,
62,
73].
As in TD children, research shows that narrative ability is far from a standalone skill in autistic children; rather, it has been linked to a broader set of social, cognitive, and communicative abilities, including Theory of Mind [
4,
98], working memory [
62], emotional understanding [
70], and conversational competence [
4,
98]. The relative centrality of narrative ability for social competence in ASD [
104], as well as for academic success [
99], has made it a popular intervention target [
44,
84,
110].
Word choice during narration
Words are necessary for conveying the contents of a story. In addition, word choice and frequency shed light on what a speaker finds important enough to describe [
103], and thus may be interpreted as a measure of preference or motivation [
59]. Word choice is particularly interesting in ASD, as autistic individuals regularly produce idiosyncratic words or phrases during narratives [
27,
70], when describing videos [
59], and during clinical interviews [
80,
81]. Among the most widely studied word-based differences in ASD are (1) concrete/literal language, generally reported to be more common in ASD than matched controls, and (2) cognitive/mental state words, often described as diminished and reflecting poor Theory of Mind in ASD.
Concrete/literal language
The first published accounts of verbal autistic children included descriptions of overly formal and pedantic language [
2,
53,
54], which made children sound like “little professors.” Thinking and speaking patterns described as “concrete” and “literal” soon followed [
48,
88], as well as reports that autistic individuals have difficulty understanding irony, sarcasm, metaphor, and deceit [
7,
54,
92]. Research shows that concrete words are more likely to be nouns than any other word class [
11], although some nouns are abstract (e.g., “justice”). In the present study, we use the number of nouns produced by children during narratives to indicate object orientation or concreteness, and aim to replicate prior research showing that children with ASD produce language that is more concrete and object-focused than matched typical peers.
Cognitive process words
Autistic children’s narratives have been found to contain fewer cognitive process words like
think and
know (also referred to as mentalizing words or internal state language), than narratives produced by typical peers [
9,
14,
21,
55,
83]. Reduced reliance on cognitive process words is argued to index diminished social cognition in autism [
9,
59], and indeed, the proportion of cognitive process words produced during autistic children’s narratives predicts their ability to understand the thoughts and feelings of others (Theory of Mind) [
98]. However, unexplained heterogeneity in autistic children’s cognitive process language still exists across samples and tasks, as some studies do not report this effect [
58,
96,
102]. Possible explanations for these mixed findings include heterogeneous sampling and failure to consider the influence of relevant factors like biological sex on word choice during storytelling. In this study, we test the hypothesis that children with ASD produce fewer words about cognitive processes than matched peers, with an eye toward potential moderating effects of biological sex on word choice.
Sex differences in narration
From an early age, typically developing girls and boys have different narrative experiences. Parental narratives directed at girls include more references to emotions and internal states than narratives directed toward boys [
43]. Subsequently, girls tell narratives that are distinct from boys’ narratives [
12,
17,
67], including longer narratives that are more emotionally laden and more likely to reference others’ internal states [
77,
97]. However, it is largely unknown whether the narratives of boys and girls with ASD also differ from one another.
To our knowledge, one study has examined sex differences in the narratives of children with ASD. In this small study, German-speaking autistic girls (
N = 11) used more internal state language than autistic boys when telling a story from a wordless picture book [
55]. However, the clinical autism symptom severity of the boys and girls was not reported, leaving open the possibility that autistic girls were less severely socially impaired, or more socially motivated, than autistic boys. Many otherwise large narrative studies included insufficient numbers of girls with ASD to assess sex differences in this domain [
42,
66]. Due to this paucity of research, current narrative interventions are not sex-sensitive. A lack of sex-sensitive narrative interventions is especially problematic in light of recent research suggesting that storytelling is a critical social medium for school-aged autistic girls who experience peer rejection when they violate storytelling norms [
34].
The current study
This study explores sex differences in the narrations of autistic girls and boys matched on age, intelligence quotient (IQ), and social challenges. In particular, we focus on the relative frequency of two kinds of words: nouns (words that indicate object-oriented storytelling and tend to be concrete) and cognitive process words (words like
think and
know that indicate cognitive orientation). First, based on prior research suggesting that the speech of autistic children is often literal and dominated by concrete words [
2,
48,
53,
88] and nouns are likely to be concrete [
11], we expect a main effect of diagnosis on noun use. Specifically, we expect that autistic children will produce more nouns (labels for objects or characters) in their narrations compared to TD children. Second, given population sex differences in the production of cognitive process words during narrations [
77,
97] and emerging research suggesting that autistic girls produce more of these words than autistic boys across a variety of tasks, including narratives [
45,
55], we expect a main effect of sex on cognitive process words, such that girls use more cognitive process words than boys. Finally, consistent with prior research showing an inverse relationship between autism symptoms and mentalizing words [
98], we hypothesized that cognitive word use would negatively correlate with social impairment across the sample as a whole.