Like,it's important: The frequency and use of the discourse marker like in older autistic children

Abstract

Background & Aims

Discourse markers, such as well or like, serve a variety of functions to support conversational reciprocity: filling pauses, aiding word-finding, and modulating turn-taking by holding the conversational floor. Previous research shows that autistic individuals use discourse markers less frequently than non-autistic (NonAu) peers; however, the discourse marker like has not been included in that research, despite its ubiquitous use by NonAu individuals, and despite the fact that like serves important pragmatic functions that are not encoded by any other discourse marker. Specifically, like signals to the listener that the content of upcoming speech is 1) Important/new; 2) Loose/approximate; 3) Reformulative; or 4) Quotative. The current study addresses this gap in the literature by comparing the frequency of discourse marker like use between older autistic and non-autistic children as well as exploring patterns of usage between the four like functions.

Methods

Twenty-one 10-to-17-year-old children on the autism spectrum and 20 NonAu peers—statistically matched on age, sex, IQ and language scores—engaged in a semi-structured interview with a researcher. Uses of discourse-marker like were identified from written transcripts of interviews and each use was categorized into one of the four functions.

Results

There were no significant differences in like frequencies between groups, nor were there differences in relative proportions of functions used by each group.

Conclusions

Research consistently indicates that autistic individuals use discourse markers significantly less often than their NonAu counterparts, but the findings from our study suggest that this pattern does not persist to all such markers. This group of older autistic children use like as often as their peers and use it to signify similar information about upcoming speech to their listener.

Keywords

autism spectrum disorders expressive language pragmatics adolescents school-age children

Use of the word like as a filler (e.g., “I’m, like, starving!”) emerged in the mid-to-late 1900s as a dialectal feature of adolescents and young adults in Southern California and then quickly spread to the dialects of most American English speakers, regardless of geographic region and age (Siegel, 2002). However, despite the fact that like was quickly adopted into the speech of Americans of all ages and from all regions of the country, its association with adolescence, generally, and “Valley Girl” speech, specifically, persisted for years, and still affects perception of its use (Johnson, 1998; Mehren, 1999; McWhorter, 2016). In support of this, Hesson and Shellgren (2015) found that a speaker's like frequency negatively correlates with how others perceive their intelligence.

Like as a discourse marker

Despite the stigmatization like has received, there is evidence that it is quite useful. Like functions as a discourse marker, which means that it signals nuanced pragmatic information to one's listener (D’Arcy, 2005; Fox Tree, 2007; Fuller, 2003; Siegel, 2002; Underhill, 1988). American English speakers utilize many discourse markers, including but, so, and, then, because, you know, well, and I mean, and each of these markers serve pragmatic functions that are fundamental to discourse maintenance and conversational reciprocity (Gorman et al., 2016). Specifically, discourse markers fill pauses, aid in word-finding, relay uncertainty, and hold one's conversational turn by indicating that the speaker is making an intra-turn pause (Brennan & Schober, 2001; Goodwin & Goodwin, 1986; Irvine et al., 2016; Maclay & Osgood, 1959; Swerts, 1998). Although like is often left out of discourse marker research (Crible, 2017; Geelhand et al., 2020; Kyrstzis & Ervin-Tripp, 1999), analyses of like use show that the pragmatic information it encodes is not represented by other discourse markers, and, correspondingly, speakers use discourse marker like contrastively to others (Fox Tree, 2007; Odato, 2013). Specifically, like is used to convey four different messages about upcoming speech: looseness, focus, quotation, and revision. We detail each of these functions in the following.

First, discourse marker like can be used to indicate looseness, signaling that upcoming speech is not exact or approximate. For example, a speaker may say “I have like 100 pairs of shoes” to communicate that they have a relatively large collection of shoes. There are two interpretations of the phrase “like 100” in this example. The first is where like is synonymous with “approximately” (Fuller, 2003), and the speaker is attempting to provide an actual estimate of how many shoes they have (e.g., perhaps they own 96 pairs of shoes). The second interpretation of “like 100” is of exaggeration (i.e., the speaker owns many fewer than 100 pairs). This interpretation shows that the approximation use of looseness like can be used much more broadly than “approximately” can. For example, if a speaker owns 40 pairs of shoes (which is more pairs of shoes than many people own), it is appropriate for them to say they have “like 100 pairs” despite their not having anywhere near 100 pairs. However, it would be inappropriate for this same speaker to say that they owned “approximately 100 pairs of shoes.” In this case, the difference between the actual number of shoes and the approximated number is beyond the appropriate reach of “approximately” so that the phrase “approximately 100 pairs” in this case could be interpreted as a lie (unless the speaker otherwise noted that they were being facetious).

In other instances, looseness uses of the word like can signal hedging of an upcoming statement. A hedge “leave[s] [a] statement slightly open” by taking the exactness out of a speaker's word choice and, thus, “shield[ing them] in the case of refusal” (Underhill, 1988, p. 241). An example of hedging is the following: “Jim is, like, my favorite person.” In this case, the speaker is not suggesting that Jim is approximately their favorite person nor are they exaggerating their fondness for Jim, but they are also not committing to saying that Jim is absolutely their favorite person. Similarly, the hedging use of looseness like can simply express uncertainty (Fuller, 2003), e.g., “I think Sarah works as a, like, dentist?” Just as the previous speaker did not want to commit to Jim being their favorite person, the speaker here is not committing to the fact that Sarah is a dentist.

A second function of discourse marker like is to signal focus. Focus markers are used to draw attention to upcoming speech, signaling that the speech represents information that is new to the listener, unusual, and/or important (Meehan, 1991; Underhill, 1988). For example, in the question “Could you, like, pay the water bill tomorrow, because they called and we’re, like, in arrears?” both likes are serving to focus the listener on the information that immediately follows each like, as that information is important and—in the case of the latter like—the information is possibly new to the listener. Focus like can be used to signal information that elaborates, narrows in, and/or provides specification on a previous point. For example, consider like's role in “He was really upset. Like, he couldn't catch his breath he was crying so much.” Here, like is being used to elaborate on and exemplify the previous description (“really upset”). In other such examples, focus like is somewhat synonymous with “such as” (Meehan, 1991), and in these cases like is most obviously linked to its historical sense as a preposition (e.g., “I’m craving something salty. Like, popcorn.”).

An interesting aspect of focus like is the fact that it fills a gap that otherwise exists in the pragmatic-syntactic interface of English; this gap is occupied in other languages, including Japanese, Korean, Tagalog, West African languages, and many others, by grammatical morphology. Japanese, for example, has two morphemes, wa and ga, that are used grammatically to mark topic and subject, respectively, but they are also used pragmatically to highlight new information (Kuroda, 2005; Ono et al., 2000). In such cases, both morphemes function as focus markers, and this function represents an established interface between morphosyntax and pragmatics in Japanese (Hara, 2006). In English, before the 1950s, when like began to be used as a focus marker, there was no lexical or grammatical equivalent to the focus particles used in languages like Japanese. This void may have (at least partially) motivated its provenance.

Like (along with a preceding copular verb) can also be used as a quotative marker (Blyth et al., 1990; Fuller, 2003; Romaine & Lange, 1991; Tangliamonte & D’Arcy, 2004). However, its use is broader than a verb such as ‘say’ is; not only can like signal a direct quotation, but it can also signal “[a speaker's] inner monologue, speaker attitude, or non-verbatim renditions of dialogue” (Fuller, 2003, p. 366). For example, in “I woke up with a headache and was like, ‘I can't go to work today,’” like introduces the speaker's feelings about going to work. It is unclear whether the quoted language represents something the speaker said out loud or if it is inner monologue; the use of like is appropriate either way. In fact, quotative like's meaning can vacillate not only between “say” and “think” but also “feel” or “behave/act.” As such, it can introduce nonverbal information, like facial expressions and/or gestures, e.g., “I tripped in front of everyone and was like [speaker makes an embarrassed face and then covers her face with her hands]¹.”

The three uses of discourse marker like described above—looseness, focus, and quotative—are included in all the literature on this word. But Fuller (2003) argues that there is yet a fourth sense of like, where it is used to signal reformulation. Participants in her study used like to revise and reformulate previously stated interview questions (e.g., “What kind of place do you live in? Like, do you rent an apartment, or do you own a house?”) Fuller (2003, p. 368²). Such uses mean something like, “What I mean to say is…” and can also occur in declarative utterances. For example, if a speaker is trying to provide directions but is having trouble formulating the message, they may use like as a signal to the listener that upcoming speech represents a revision of what preceded it. For example, “You take a left at the next light, like, not right at the light… like, you pass the light before turning left.” Each use of like marks a new attempt to rephrase what is previously said to convey the intended message. What makes this use different from the other three is that it not only signals something about the following speech (i.e., that it represents a revision or restatement), but it also signals something about previous speech (i.e., that it was unsatisfactory in some way).

In summary, the discourse marker like, which has historically been devalued by both listeners and speakers, serves to signal four aspects of upcoming discourse, and these rich pragmatic functions are not otherwise represented by any single English word. A speaker would need to utilize a litany of other lexical items and phrases if they wanted to eradicate like from their speech but still communicate the same information about upcoming discourse that like does.

The use of discourse markers in autism

Because discourse markers serve important (but arguably subtle) pragmatic functions, the appropriate use of such words may prove challenging for individuals who struggle with neurotypical conventions about conversation reciprocity and other aspects of pragmatics, such as individuals on the autism spectrum (APA, 2013).

Evidence that the use of discourse markers is divergent in autism comes from a substantial body of work on words that are used to fill pauses, specifically uh or um (Clark & Fox Tree, 2002; Engelhardt et al., 2017; Fox Tree, 2001; Lake et al., 2011; McGregor & Hadden, 2020). During a conversational exchange, if a speaker pauses, the listener may assume that the speaker has completed a turn and is ready for the listener to take the conversational floor. However, sometimes the speaker is pausing because they are formulating their next thought, or because they are attempting to access a word. In these cases, the speaker will fill or precede the pause with uh or um, and, by doing so, they are signaling that they are not ready to relinquish their conversational turn (Clark & Fox Tree, 2002; Fox Tree, 2007; Gorman et al., 2016; Lake et al., 2011). Because filled pauses provide information about the nature of upcoming speech (or the lack thereof, in the case of upcoming pauses) and information about the structure of the upcoming discourse, they should be considered a type of discourse marker (Fox Tree, 2007). Some work on uh and um use in autism finds that autistic³ speakers simply use both less often than non-autistic (NonAu) speakers do. For example, Lake et al. (2011) compared rates of uh and um between autistic children and non-autistic peers. They found that autistic children were less likely to fill their pauses (i.e., they used both uh and um less) than NonAu children. In fact, autistic children used silent pauses as frequently as NonAu children produced uh and um, which the authors interpreted as suggesting that autistic children used silent pauses “in the place of filled pauses” (Lake et al., 2011, p. 137). Based on the increased rate of unfilled pauses in autistic children's conversations, Lake et al. (2011) suggest that autistic children are less sensitive than NonAu children to the listener's mind, and specifically to the fact a silent pause can be interpreted by the listener as a signal that the speaker is relinquishing their turn. Thus, autistic children may unintentionally encourage their listeners to start talking, even when they want to hold the floor⁴.

In fact, uh and um are used in complementary distribution, where um typically precedes longer pauses than uh does (Clark & Fox Tree, 2002). Because of this distributional difference, research subsequent to Lake et al. (2011) elected to examine autistic speakers’ use of uh and um separately. The findings from this research have been fairly consistent: Autistics and non-autistics show comparable rates of uh use, while autistic speakers use um significantly less frequently (Gorman et al., 2016; Irvine et al., 2016; McGregor & Hadden, 2020). Further, um rate has also been shown to negatively correlate with autism traits (and, therefore, positively associate with neurotypical expectations of social-communication norms), while no such relationship exists for uh rate (Irvine et al., 2016). Because of the association between um use and autism characteristics, Irvine et al. (2016) explain differing patterns of usage for um vs. uh in autism as depending on the former being a “listener-oriented” filled pause while uh is a “speaker-directed” filled pause. They argue that autistic individuals struggle to attend to their listener's needs and are therefore less likely to utilize um.

There is very little work on how autistic speakers use other discourse markers besides uh and um. In fact, as far as we are aware, there is only one study that has done this. Geelhand et al. (2020) examined discourse marker⁵ use while children on the autism spectrum told narratives and compared their use to non-autistic peers. They found that autistic children used these markers less often than non-autistic peers. However, it remains to be seen whether this pattern extends to other types of discourse (like conversations or interviews) and importantly, for our purposes, whether these patterns extend to the discourse marker like, as these authors did not include like in their analysis.

Current study

The current study attempts to address a gap in the literature by examining the use of discourse marker like by autistic and non-autistic children and adolescents as they answer questions about themselves in a conversational context. We specifically focus on like for several reasons. One is simply its frequency; the discourse-marker like is prolific within the speech of older children and adolescents, which means any conversational interaction is likely to yield many tokens of like (at least in our NonAu group). Further, if we find that older autistic children and adolescents use like less frequently (as we hypothesize), this would have important implications for social integration. The use of like is part of the social-communication code of older children and adolescents; therefore, children who use it less often may be perceived as not fully belonging to their peer group. Another reason we focus on like is because it has four distinct pragmatic functions, which allows us to analyze and compare subtype use between diagnostic groups. If there are differences in the proportional uses of certain subtypes between participant groups, this would signal specific areas of pragmatics that are divergent between groups.

Not only can such a detailed analysis of like provide insight into specific areas of pragmatics that may differ across autistic and NonAu children, but it is hoped that this analysis and the resulting findings will yield clinical implications as well. Training manuals for language sample analysis procedures traditionally guide users to code discourse markers, including like, as mazes (see for example the standard transcription conventions for SALT® Software, Miller, 2010). This thereby prevents their inclusion in the calculation of MLU and other typical measures of expressive language ability (TNW, etc.). While this may be appropriate for measures of morphosyntactic language development in young children (like MLU), ignoring like's use entirely prevents clinicians from analyzing a potentially helpful indicator of expressive pragmatic and sociolinguistic skills.

We ask the following research questions: 1) Do older children and adolescents on the autism spectrum use like significantly less frequently than non-autistic (NonAu) peers? 2) Do autistic children show differing proportional frequencies in like subtypes (reformulation, focus, looseness, and quotative) as compared to NonAu peers?

With regards to our first research question, we predict lower rates of like, overall, in our autistic group because previous literature reports less frequent use of listener-directed discourse markers in this population. The second research question builds on the first, by exploring whether autistic children and adolescents use like differently than NonAu peers, in terms of how frequently they use like to signal its four distinct discourse functions. This question is purely exploratory, as there is not enough relevant background literature for us to form hypotheses about which subtypes might be more or less frequent in either group, let alone how proportions might differ between groups. Finally, as a post-hoc analysis, we compare uh and um rate between groups to determine whether this specific set of participants in this particular discourse context show the same uh/um rate differences that have been captured in other examinations of autistic and NonAu groups in other types of discourse contexts (Irvine et al., 2016; Gorman et al., 2016; McGregor & Hadden, 2020). We predict proportionally higher um rates by NonAu speakers, as this has been reported previously. If this result is borne out, we can be more confident that like findings are not attributable only to the specific nature of the current study, including discourse type, discourse partner, and participant selection criteria.

Methods

Participants

Twenty-one (21) children on the autism spectrum (17 males; 4 females) and twenty (20) NonAu children (11 males; 9 females) participated, aged 10–17 years. Groups were statistically equivalent in age, sex, IQ scores, and standardized language scores (all ps > 0.1). See Table 1.

Table 1.
Data are shown as mean ± standard deviation or as ns and are compared by t-tests or fisher exact tests

Characteristic Autistic (n = 21) Non-autistic (n = 20) Test Statistic p

Age (years) 13.53 ± 2.30 13.24 ± 2.50 t = 0.45 .66

Sex (n = female:male) 4:17 9:11 n/a (Fisher exact test) .10

K-BIT 2 standard scores 115.90 ± 19.90 112.00 ± 14.33 t = 0.72 .47

CELF-5 standard scores 110.76 ± 17.80 112.95 ± 14.99 t = −0.43 .67

SCQ 20.10 ± 6.73 2.50 ± 2.70 t = 11.08 <.001

ADOS-2, module 3 (n = 14) 9.64 ± 2.68 n/a n/a n/a

ADOS-2, module 4 (n = 7) 11.29 ± 1.11 n/a n/a n/a

Participants were recruited via newspapers, local magazines, the Internet, local schools, local advocacy groups for families of children on the spectrum, and word of mouth. In order to participate, all children had to have acquired English as a native language, along with meeting the following criteria: earning a score above 85 on both the Kaufman Brief Intelligence Test-2nd Edition (K-BIT-2; Kaufman & Kaufman, 2004) and the Clinical Evaluation of Language Fundamentals, Fifth Edition (CELF-5; Wiig et al., 2013), no significant hearing loss, no psychiatric or developmental condition (aside from autism, for autistic participants). Participants recruited for the NonAu group were excluded if they had an autistic sibling. To confirm autism diagnosis for autistic participants, the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) was administered by a research-reliable ADOS administrator (Lord et al., 2012). Additionally, caregivers from both groups completed the Social Communication Questionnaire-Lifetime (SCQ; Rutter et al., 2003). Potential participants in the NonAu group were excluded if they earned a score above 15 on the SCQ.

This study was approved by the Emerson College Institutional Review Board in Boston, Massachusetts. Informed consent was obtained in writing from participants who were over the age of 12 and all participants’ parents or guardians. All participants were asked for verbal consent before each research task was conducted and were compensated with Amazon gift cards for participation.

Procedures

Participants engaged in a double interview (Winner, 2002), where they first were interviewed by a research assistant (RA) and then reversed roles and interviewed the RA. During the portion when they served as the interviewee, participants were asked the same series of interview prompts in identical order. The interview prompts were as follows:
Can you tell me what you like to do for fun?

Tell me about your family.

Can you tell me about a cool trip or vacation you have been on?

Tell me about the hardest or your least favorite part of school.
RAs were encouraged to respond naturally to participants’ answers to these four prompts (asking follow-up questions and making comments based on participant responses), which meant that the four above prompts represented the only part of the interview/conversation that was scripted. Once a natural end to this portion of conversation was reached, the participant was asked to look at a series of photographs of the RA as prompts for potential topics. The participants then asked questions of the RA until they did not have any further questions and the interview was concluded. A video camera was positioned on each side of the interview table so that both the RA and the participant were video recorded.

Analysis

Coding

After all participant interviews were completed, RAs transcribed the interviews from video recordings. Only transcriptions from the first half of the interview, in which the participant was the interviewee, were included for like coding. We excluded the second half of the interviews (when participants interviewed RAs) from like coding for several reasons. First, in the participant-led portion of the interviews, there was less uniformity in both the length (ranging from 4 to 14 min) and semantic content (as participants could ask about anything they wanted), as compared to the RA-led portion. Second, participants talked much less in the second half of the interviews, as their job was to ask (rather than answer) questions. Thus, there were less opportunities for like use in the second portion of interviews.

Initially, the first author identified all discourse marker likes in all transcripts of the first halves of the double interview. Afterwards, she and another researcher coded each like for one of the four subtypes. To train the second researcher to distinguish between these subtypes, the first author created a mock transcript that contained several examples of discourse marker like being used to convey each of the four functions. Together, the first author and other coder went through the mock transcript and discussed why each of the likes represented each subfunction. Once the second researcher was confident in their ability to recognize all four subtypes, both coders began coding likes in earnest.

To prevent bias, neither coder was aware of participants’ diagnoses during coding. The first author completed all coding before reliability coding began. Another researcher coded like in a random assignment of 20% of the total number of transcripts (n = 8). We estimated reliability between like does by calculating Cohen's Kappa (McHugh, 2012) for each subtype. We selected this relatively conservative measure of reliability due to the novelty of this coding scheme. Cohen's Kappa for the four categories ranged from fair to almost perfect agreement: Codes for quotative like showed strong agreement (0.816); loose codes showed substantial agreement (0.689); focus showed moderate agreement (0.538); reformulation codes achieved fair agreement (0.346). Table 2 provides examples of each of the four subtypes from participant transcripts.

Table 2.
Example participant utterances coded as different subtypes

Main Subtypes Subcategories Examples

1. Looseness 1a. Approximation (1a) “I’ve been playing hockey for about like six, five years”
(1a) “In her like…twenties…”

1b. Hedging (1b) “She grew up like outside Boston…as like…a Catholic I think I don't know”
(1b) “It [is] like my favorite city”
(1b) “…I think she like…found hippie culture and like…fell in love with it so she became like a Hindu”

2. Focus 2a. New information (2a) “Like I wouldn't really say anything in school is hard for me to do”
(2a) “Like the uh when we were crossing the border um… the guard was there”
(2a) “Um… Talking Tom… it's like you have to take care of this cat”

2b. Example/elaboration (2b) “I skateboard because it's fun like obviously if I go to a camp and there's a competition … I’ll do it”
(2b) “Uh Lego video games like Lego batman or Lego batman 3”
(2b) “Well I actually like I like to play video games uh…like Pokémon and Super Smash Brothers”

3. Quotative (3) “I was like (makes funny face)⁹“
(3) “But when I told my mother about it she's like…he didn't tell me about this”

4. Reformulation (4) “I was trying to…jump…into a ramp like instead of just dropping in I jumped before I was going down”
(4) “… I wouldn't really say anything in school is hard for me …like if I\\¹⁰ if a teacher gives me something … that doesn't… like if I’m not learning anything from it?”
(4) “Can this be a trip coming up or something… like in the future?”

Statistical analysis

The total frequency of like, per child, was calculated in proportion to the total number of words (TNW) that each participant produced, resulting in a proportional rate of like (i.e., like/all words). Converting like use to proportions (vs. raw frequencies) allowed us to control for differences in interview length. To compare like use between groups, we conducted a repeated-measures 2 × 4 (Participant Group × like Subtype) ANOVA.

Results

Table 3 presents raw frequencies and proportions of like use within each group. See also Figure 1.

Figure 1.
Group comparison of overall and subtype use

Table 3.
Means and standard deviations for raw frequencies of like use (and proportions of TNW in parentheses) within groups

like subtype

looseness focus quotative reformulation total

Group M SD M SD M SD M SD M SD

Autistic 7.33 (0.01) 5.62 (0.01) 6.86 (0.01) 7.29 (0.01) 1.05 (<0.00) 2.01 (<0.00) 1.62 (<0.00) 3.44 (<0.00) 16.86 (0.03) 14.05 (0.02)

Non-autistic 9.05 (0.02) 10.63 (0.02) 8.80 (0.01) 11.97 (0.02) 0.80 (<0.00) 1.70 (<0.00) 1.50 (<0.00) 1.54 (<0.00) 20.15 (0.03) 21.90 (0.03)

A 2 × 4 (Group × like Subtype) repeated-measures ANOVA showed neither a significant main effect of group (F_{(1, 39)} = 0.811, p = .373, ges = 0.010) nor a significant interaction between group and like subtype (F_(4,156) = 0.915, p = .461 after using a Greenhouse-Geisser correction, ges = 0.010). There was a significant effect of like subtype (F_(4,156) = 49.630, p < .001 after using a Greenhouse-Geisser correction, ges = 0.398), reflecting lower rates of reformulation and quotative likes, as compared to focus and looseness like in both groups. ANOVA results are presented in Table 4. A post-hoc Tukey HSD test found that rates of quotative and reformulation like were each significantly lower than looseness and focus (p < .0001 for all four comparisons).

Table 4.
ANOVA results

Predictor df_Num df_Den Epsilon F p η²_g

Group 1.00 39.00 0.81 .373 .01

Like Type 1.75 68.09 0.44 49.63 .000 .40

Group × Like Type 1.75 68.09 0.44 0.75 .461 .01

There was no significant difference between rates of focus and looseness like (p = .713) or between rates of reformulation and quotative like (p = .924).

Finally, post-hoc t-tests were conducted to compare the use of uh and um between participant groups. We calculated um:uh + um ratios, based on previous analyses (Gorman et al., 2016; Parish-Morris et al., 2017), and then used a two-tailed t-test to compare these ratios between groups. Results showed that um:uh + um ratios were significantly smaller for autistic participants than NonAu participants (autistic m = 0.60; NonAu m = 0.72, t = −2.265, p = .029), even though the number of uh and um tokens were not statistically different between groups (uh: autistic group m = 7.904; NonAu group m = 4.45, t = 1.589, p = .1204; um: autistic group m = 13.714; NonAu group m = 13.85, t = −0.025, p = .98).

Discussion

As described in the introduction, there is research examining discourse marker like usage by non-autistic speakers (D’Arcy, 2005; Fox Tree, 2007; Fuller, 2003; Meehan, 1991; Miller & Weinert, 1995; Odato, 2013; Siegel, 2002; Underhill, 1988), but to our knowledge there are no studies examining its use by autistic speakers. The current research has addressed this gap by examining like use, overall, and like subtypes by older autistic children. By focusing our efforts on like, we hoped to highlight the pragmatic significance and versatility of a word that has traditionally been coded as a filler and/or maze by both clinicians and researchers who use language sample analysis techniques to assess linguistic skills in clinical populations. Our results find—in contrast to other research on discourse markers in autism—that there is no significant difference in the overall use of like between diagnostic groups or in the relative proportions of like subtypes between groups. We address each of these findings in the following.

Overall frequency

Previous research finds differences between use of discourse markers in children on the autism spectrum compared to NonAu peers (Gorman et al., 2016; Irvine et al., 2016; Lake et al., 2011; McGregor & Hadden, 2020). Specifically, research on um and uh has found that um is used significantly less frequently by autistic children compared to NonAu peers (Gorman et al., 2016; Irvine et al., 2016; McGregor & Hadden, 2020). In contrast, our findings suggest that like is as prolific among the speech of older autistic children and adolescents as it is in the speech of their NonAu peers. This suggests that there are differences between um and like, which encourages the use of the latter (but not the former) by autistic individuals. In support of this, the children in our study on the spectrum did use um proportionally less often than their NonAu peers (as has been found in previous literature), despite using like as often.

As discussed in the introduction section of this manuscript, um is a listener-oriented discourse marker (as opposed to speaker-oriented), which means that it is used to benefit the listener (Engelhardt et al., 2017; Lake et al., 2011). This aspect of um (the fact that it is listener-oriented) has been used as an explanation for why it is less commonly used by autistic speakers. However, like is also a listener-oriented discourse marker (Engelhardt et al., 2017; Irvine et al., 2016; Lake et al., 2011), so our findings suggest that listener-oriented discourse markers are not generally elusive to autistic individuals. Fox Tree (2007) demonstrates that like and um are in complementary distribution, which shows that these terms are not interchangeable. Thus, one potential explanation for different patterns of use between um and like is that there are differences in the specific listener-oriented functions these terms convey.

Specifically, um's listener-oriented role is to hold the conversational floor when the speaker needs to pause without relinquishing their turn (Lake et al., 2011). In contrast, like does not help organize turn-taking, but instead provides the listener with information about the content of the upcoming message (e.g., looseness like informs the listener that the following language represents an approximation or an exaggeration). Therefore, it is possible that um's specific use to signal turn maintenance—and, as such, its use to manage turn-taking—is what makes it less accessible for this population. This seems aligned with a previous findings that autistic individuals show differences with turn-taking (Bone et al., 2013; Kaczmarek, 2002; Kim & Clarke, 2015; Ying Sng et al., 2018). Similarly, a recent paper from our lab finds that this same group of autistic children also produce back-channeling less often than their NonAu counterparts (Matthewman et al., 2021). One explanation for this is that—when they are acting as the listener—they do not pick up on moments when the speaker pauses for a moment while still holding the floor. These are the moments when a backchannel is warranted; the speaker pauses to formulate thoughts and/or to check in with their listener's comprehension, and the listener fills that gap with a brief verbal/nonverbal signal (e.g., saying “mmhmm” or nodding) to convey to the speaker that: a) they are following along with the speaker's message; b) they recognize that the speaker is still holding the conversational floor.

Combining the results of Matthewman et al. (2021) and the current study, we find that whether these children are acting as listeners or as speakers, they respond differently to turn-taking cues, specifically those where a speaker is pausing between ideas/utterances/clauses but still desires to hold the floor. In contrast, our like findings show that older autistic children and adolescents are adept at using the discourse marker like, despite its representing complex and nuanced pragmatic information (Fox Tree, 2007; Fuller, 2003; Underhill, 1988). We argue that our results represent a pragmatic strength in autism that has heretofore gone unnoticed.

This brings us to another possible explanation for the fact that we find as frequent like use (but less frequent um use) by our participants on the spectrum, as compared to NonAu peers: interactional context. Not only do um and like serve different functions, but they also seem differently affected by interactional context. For example, speakers use like more frequently when they are talking with friends (i.e., people who are perceived as peers) than they do with speakers who hold a position of authority, while uh and um is used more generally, with both friends and authority figures (Fox Tree, 2007).

Two aspects of the current paradigm may have made it less amenable for like use (at least for some participants): 1) each participant's interlocutor was a research assistant, rather than a friend; 2) participants engaged in a double interview (Winner, 2002), rather than a conversation. Combined, these factors may have discouraged participants from using like as often as they would in an informal conversation with peers. Beginning with discourse partner: Research assistants were all at least five years older than participants. Further, their role as researchers required them to engage with participants in an authoritative manner; they directed participants to perform a variety of tasks and even paid them for their time (with Amazon gift cards) at the end of the research session. Moving on to discourse type: the use of an interview (it was even labeled as such to participants) may have created a more formal context than a conversation would. Although previous research does find that individuals frequently use like during interviews (Fuller, 2003), this research does not compare like frequency between interviews and other less formal contexts, so it is unclear whether the same person would use like as often in a formal interview as they would talking with friends.

Putting these factors together, the power differential between participants and their interlocutor, along with the discourse type (an interview rather than a conversation) may have discouraged the use of like by some participants, and perhaps more strongly for NonAu participants. While we have not discovered research that examines how autistic individuals adjust linguistic form/content (i.e., shift registers) to meet the needs of different sociocultural contexts, research on other types of communicative behaviors, like emotional facial expressions, finds that autistic individuals are less likely to follow “display rules” that neurotypicals use to guide the degree to which one should express emotions in a particular context (Barbaro and Dissanayake, 2007; Scheeren et al., 2010; Zane et al., 2018). Thus, another explanation for similar like usage between participant groups in the current study may be that NonAu participants may have used like less than they would in other contexts, while participants on the spectrum used like as often as they always do. That is, if we had examined like use while participants engaged with not only a peer, but specifically a friend, we may have found the group differences we predicted, where autistic participants use like less than NonAu counterparts, simply because NonAu individuals would have increased their usage of like in such a context.

Subtype proportions

Our overall-use findings show that like is used as frequently by older autistic children as it is by their NonAu peers. The subtype findings strengthen this picture, since the relative proportions of each like subtype are statistically equivalent between groups. This suggests that children on the autism spectrum not only use like as often as their NonAu peers do, but also that each group uses like in similar ways. In the current study, children in both groups used focus and loose like significantly more often than quotative and/or reformulation like. Although we did not have hypotheses regarding which subtypes would be more or less common in autistic discourse, there are two patterns of use that surprised us, nonetheless: 1) Infrequent use of reformulation like; 2) Frequent use of focus like.

Despite not having specific hypotheses related to subtype proportions in either group, there were several reasons to suspect that—if autistic participants were going to favor any ‘like’ subtype—they would show a preference for reformulation. Research finds increased rates of repetitions, reformulation, and/or revision in the speech of children on the spectrum when they are performing various language tasks, including storytelling and conversation (Irvine et al., 2016; Lake et al., 2011; Shriberg et al., 2001; Suh et al., 2014⁶). Because discourse markers, specifically like, are one way a speaker can signal that they are struggling with word finding or formulating an upcoming thought (Swerts, 1998), this would suggest that autistic individuals, who have been shown in studies to frequently revise and repeat, would rely on the reformulation function of like. Further, studies find that autistic individuals exhibit lower executive functioning (EF) skills than NonAu peers (Blijd-Hoogewys et al., 2014; South et al., 2007; Van Eylen et al., 2015), and research shows that filled pauses, reformulations, and revisions negatively correlate with EF ability (Engelhardt et al., 2013). In short, autistic individuals tend to reformulate/revise their speech more frequently than neurotypical individuals (perhaps because of challenges with EF), and this would seem to motivate a relatively high reliance on reformulation like for this group. However, our results did not support this preference.

We offer two possible (and related) explanations for this. First, rates of reformulation like were strikingly low in both groups, suggesting that marking reformulation is not a common way for anyone to use the discourse marker like⁷; thus, low rates of reformulation like by autistic participants may simply reflect a larger pattern where no one uses this like subtype very often. Secondly, we did not measure overall rates of revision/reformulation in this study. Therefore, it is possible that autistic children were indeed revising/reformulating their language more often than the NonAu group was (as has been found in previous work), but they just did not use like to signal these revisions. They may have used other discourse markers to do so⁸), or they could have revised speech after an unfilled pause. In short, because neither group seemed to use like to signal a revision, we interpret the lack of difference in rates of reformulation like to reflect the fact that this is simply a rare use for this word and not to reflect low rates of reformulation, generally (for either group).

In addition to low frequencies of reformulation like amongst autistic participants, another surprising finding was their high rates of use of focus like. Focus markers, including like, introduce information that is new and important to the discourse. As such, using them requires that the speaker be aware that upcoming information is new to the conversation and new to the listener. This awareness not only entails an accurate and up-to-date representation of the conversation thus far, but it also arguably entails theory of mind (ToM) skills, since the speaker must attend to what the listener already knows and what is new information to them. All of this should prove challenging for autistic individuals. As mentioned earlier, research finds that autistic individuals show weaker EF skills than NonAu peers, including working memory skills (Engelhardt et al., 2013; Kercood et al., 2014). Thus, maintaining a precise representation of previous discourse should be challenging for autistic individuals, simply because of a lower working memory capacity. This has been used as an explanation for other discourse differences in autism, including inappropriate and ambiguous referencing (Arnold et al., 2009). Beyond challenges with working memory, a large body of research finds that autistic individuals show differences on ToM tasks (Baron-Cohen, 1995; Kimhi, 2014). Like EF skills, ToM differences have also been implicated in research on referential skills in discourse (Kuijper et al., 2015; Novogrodsky & Edelson, 2015). Such differences might also affect autistic individuals’ recognition of when a focus marker is needed, as they may not identify upcoming information as new or important to their listener.

Despite the aforementioned reasons for suspecting challenges with focus in autism, our results show that older children on the spectrum use the focus subtype at similar rates to that of their NonAu peers. We provide two possible explanations for this finding. First, we did not test either EF or ToM skills for any of our participants. Thus, even though previous research demonstrates lower scores on EF and ToM metrics by individuals on the spectrum as compared to non-autistic individuals, it is possible that this was not true for our specific group of autistic participants. Further, because we did not measure either EF or ToM, we cannot actually determine whether there were correlations between either factor, and the use of focus like in either group. That is, while there are reasons to suspect that better EF/ToM skills would support the use of focus markers, including focus like, we do not have any evidence to show that these factors in fact relate to one another. Further, it is possible that higher EF/ToM skills would not result in higher frequencies of focus like but would instead result in more appropriate use of focus like. This brings us to our second explanation for why we do not see differences in frequencies of focus like between groups. It is possible that autistic individuals—perhaps due to lower EF and/or differing ToM capacity—do not use focus like in the same way that NonAu participants do, even though they use focus like frequently. As we did not code for the appropriateness of like use or the use of any like subtype, we cannot speak to whether autistic participants produced focus like (or any like, for that matter) in expected contexts (i.e., contexts when NonAu participants would use it). Thus, some participants could have been using focus like to signal information that was not actually new to the conversation, and—vice versa—there may have been opportunities for participants to use focus like when they neglected to do so. Future studies should examine the appropriateness of like use in autism: not only whether its use is functionally appropriate in the discourse, but also whether its position within an utterance (according to both syntactic and prosodic structure) is predictable. Frequent use of like in functionally and/or structurally inappropriate ways could overwhelm the message and cause breakdowns in communication between autistic and NonAu individuals.

Limitations

There are several limitations to the current study. The first is the fact that data used for coding were written transcripts, rather than audio or video recordings. This prevented intonational and prosodic cues from being considered. While pauses of one second or longer were included in the transcript, as well as facial expression and several suprasegmentals (e.g., exaggeration, dramatic slowed speech, purposeful change in pitch), shorter pauses and more subtle pitch contours were not transcribed. This meant that it was impossible to code for more subtle aspects of like use, including whether it followed/preceded pauses of similar lengths in each group and whether the speech that preceded/followed a particular prosodic contour. Future work should examine like similarly to the ways uh and um have been analyzed, including pause length differences and syntactic environments. Further, we urge future researchers to consider ways of capturing how the speaker's intonation impacts interpretation of like's subfunction.

Secondly, the double interview paradigm used in the current study represents a very particular discourse context that may not offer a reliable sample of language use in other contexts (e.g., conversations with peers). Previous research on discourse markers in autism (usually um and uh) has often suffered from a similar limitation. While researchers have utilized a wide variety of discourse tasks to explore usage, almost all research examines the way that participants engage with either a research assistant or test administrator (in the case of ADOS conversations, for example) (de Marchena & Eigsti, 2016; Engelhardt et al., 2017; Geelhand et al., 2020; Gorman et al., 2016; Irvine et al., 2016; Lunsford et al., 2010; Parish-Morris et al., 2017; Wechsler, 1997), just as we did in the current study. These authors, like us, find that um use is less frequent in autism. Even though um is argued to be used generally, no matter the conversational partner (Fox Tree, 2007), it is possible that um usage patterns in each group would look different if participants were recorded speaking to someone who is not in a position of authority. Future research on the use of other discourse markers by autistic speakers should more carefully consider features of the context and the interlocutor rather than making conclusions about autistic usage overall.

Conclusions

These limitations notwithstanding, the current study provides evidence that older autistic and non-autistic children demonstrate similar use of a discourse marker that serves multiple, nuanced pragmatic functions, despite well-established differences in other pragmatic behaviors between these two groups. This result represents a heretofore unrecognized communicative and pragmatic strength in this population, since like is highly frequent in the speech of adolescents and young adults and since like indexes multiple important discourse functions. These findings suggest that instead of simply viewing discourse markers as mazes or speech disruptions in language samples, clinicians and researchers should consider that the use of these words may represent a potential strength in pragmatic skills. The four distinct functions of like (to revise, to quote, to focus, and to approximate) points to the value of analyzing how and why and with whom an individual is using this discourse marker. A clinician can determine specific areas of pragmatic strengths and weaknesses by determining which one(s) a child uses, which they do not, and which they possibly overuse or use inappropriately. Further research is needed to explore the potential of discourse marker analysis as a means to enrich our understanding of the pragmatic profile of autistic individuals and to inform assessment and intervention planning. Researchers and clinicians are encouraged to examine the use of like and other discourse markers in varying discourse contexts and with different discourse partners.

Characteristic	Autistic (n = 21)	Non-autistic (n = 20)	Test Statistic	p
Age (years)	13.53 ± 2.30	13.24 ± 2.50	t = 0.45	.66
Sex (n = female:male)	4:17	9:11	n/a (Fisher exact test)	.10
K-BIT 2 standard scores	115.90 ± 19.90	112.00 ± 14.33	t = 0.72	.47
CELF-5 standard scores	110.76 ± 17.80	112.95 ± 14.99	t = −0.43	.67
SCQ	20.10 ± 6.73	2.50 ± 2.70	t = 11.08	<.001
ADOS-2, module 3 (n = 14)	9.64 ± 2.68	n/a	n/a	n/a
ADOS-2, module 4 (n = 7)	11.29 ± 1.11	n/a	n/a	n/a

Main Subtypes	Subcategories	Examples
1. Looseness	1a. Approximation	(1a) “I’ve been playing hockey for about like six, five years” (1a) “In her like…twenties…”
1b. Hedging	(1b) “She grew up like outside Boston…as like…a Catholic I think I don't know” (1b) “It [is] like my favorite city” (1b) “…I think she like…found hippie culture and like…fell in love with it so she became like a Hindu”
2. Focus	2a. New information	(2a) “Like I wouldn't really say anything in school is hard for me to do” (2a) “Like the uh when we were crossing the border um… the guard was there” (2a) “Um… Talking Tom… it's like you have to take care of this cat”
2b. Example/elaboration	(2b) “I skateboard because it's fun like obviously if I go to a camp and there's a competition … I’ll do it” (2b) “Uh Lego video games like Lego batman or Lego batman 3” (2b) “Well I actually like I like to play video games uh…like Pokémon and Super Smash Brothers”
3. Quotative	(3) “I was like (makes funny face)⁹“ (3) “But when I told my mother about it she's like…he didn't tell me about this”
4. Reformulation	(4) “I was trying to…jump…into a ramp like instead of just dropping in I jumped before I was going down” (4) “… I wouldn't really say anything in school is hard for me …like if I\\¹⁰ if a teacher gives me something … that doesn't… like if I’m not learning anything from it?” (4) “Can this be a trip coming up or something… like in the future?”

	like subtype
Autistic	7.33 (0.01)	5.62 (0.01)	6.86 (0.01)	7.29 (0.01)	1.05 (<0.00)	2.01 (<0.00)	1.62 (<0.00)	3.44 (<0.00)	16.86 (0.03)	14.05 (0.02)
Non-autistic	9.05 (0.02)	10.63 (0.02)	8.80 (0.01)	11.97 (0.02)	0.80 (<0.00)	1.70 (<0.00)	1.50 (<0.00)	1.54 (<0.00)	20.15 (0.03)	21.90 (0.03)

Predictor	df_Num	df_Den	Epsilon	F	p	η²_g
Group	1.00	39.00		0.81	.373	.01
Like Type	1.75	68.09	0.44	49.63	.000	.40
Group × Like Type	1.75	68.09	0.44	0.75	.461	.01

Footnotes

Acknowledgments

Thank you to the children and families who participated in this research. We would also like to offer our sincere gratitude to Dr. Inge-Marie Eigsti, who provided extensive, helpful feedback and encouragement after reading an early version of this manuscript. Thank you also to Drs. Geralyn Timler and Jaime Lee, who served on Rebekah's thesis committee, and to Vivian Eccleston, who assisted with coding.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Research was funded by a grant from the NIH (NIDCD 1R01DC012774-01, Grossman PI). Open Access publication fees were paid for by James Madison University libraries.

ORCID iD

Rebekah Jones

Notes

References

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.).

Arnold

J. E.

Bennetto

Diehl

J. J

. (2009). Reference production in young speakers with and without autism: Effects of discourse status and processing constraints. Cognition, 110(2), 131–146. https://doi.org/10.1016/j.cognition.2008.10.016

Barbaro

Dissanayake

. (2007). A comparative study of the use and understanding of self-presentational display rules in children with high functioning autism and asperger’s disorder. Journal of Autism and Developmental Disorders, 37(7), 1235–1246. https://doi.org/10.1007/s10803-006-0267-y

Baron-Cohen

(1995). Mindblindness: An essay on autism and theory of mind. MIT Press.

Blijd-Hoogewys

E. M. A.

Bezemer

M. L.

van Geert

P. L. C

. (2014). Executive functioning in children with ASD: An analysis of the BRIEF. Journal of Autism and Developmental Disorders, 44(12), 3089–3100. https://doi.org/10.1007/s10803-014-2176-9

Blyth

Recktenwald

Wang

. (1990). I’m like, “say what?!”: A new quotative in American oral narrative. American Speech, 65(3), 215–227. https://doi.org/10.2307/455910

Bone

Lee

C. C.

Chaspari

Black

M. P.

Williams

M. E.

Lee

Levitt

Narayanan

(25–29 August, 2013). Acoustic-prosodic, turn-taking, and language cues in child-psychologist interactions for varying social demand. [Paper presentation]. ISCA 2013, Lyon, France.

Bottema-Beutel

Kapp

S. K.

Lester

J. N.

Sasson

N. J.

Hand

B. N

. (2021). Avoiding ableist language: Suggestions for autism researchers. Autism in Adulthood, 3(1), 18–29. https://doi.org/10.1089/aut.2020.0014

Brennan

S. E.

Schober

M. F

. (2001). How listeners compensate for disfluencies in spontaneous speech. Journal of Memory & Language, 44(2), 274–296. https://doi.org/10.1006/jmla.2000.2753

10.

Clark

H. H.

Fox Tree

J. E

. (2002). Using uh and um in spontaneous speaking. Cognition, 84(1), 73–111. https://doi.org/10.1016/s0010-0277(02)00017-3

11.

Crible

(2017). Towards an operational category of discourse markers: A definition and its model. In Fedriani

Sansó

(eds), Pragmatic markers, discourse markers and modal particles: New perspectives (pp. 99–124). John Benjamins.

12.

D’Arcy

(2005). Like: Syntax and Development. Ph.D. diss., University of Toronto.

13.

de Marchena

Eigsti

I. M

. (2016). The art of common ground: Emergence of a complex pragmatic language skill in adolescents with autism spectrum disorders. Journal of Child Language, 43(1), 43–80. https://doi.org/10.1017/S0305000915000070

14.

Engelhardt

P. E.

Alfridijanta

McMullon

M. E. G.

Corley

. (2017). Speaker-versus listener-oriented disfluency: A re-examination of arguments and assumptions from autism spectrum disorder. Journal of Autism and Developmental Disorders, 47(9), 2885–2898. https://doi.org/10.1007/s10803-017-3215-0

15.

Engelhardt

P. E.

Nigg

J. T.

Ferreira

. (2013). Is the fluency of language outputs related to individual differences in intelligence and executive function? Acta Psychologica, 144(2), 424–432. https://doi.org/10.1016/j.actpsy.2013.08.002

16.

Fox Tree

J. E.

(2001). Listeners’ uses of um and uh in speech comprehension. Memory & Cognition, 29(2), 320–326. https://doi.org/10.3758/BF03194926

17.

Fox Tree

J. E.

(2007). Folk notions of um and uh, you know, and like. Text & Talk, 27(3), 297–314. https://doi.org/10.1515/TEXT.2007.012

18.

Fuller

J. M

. (2003). Use of the discourse marker like in interviews. Journal of Sociolinguistics, 7(3), 365–377. https://doi.org/10.1111/1467-9481.00229

19.

Furkó

Abuczki

. (2014). English Discourse markers in mediatized political interviews. Brno Studies in English, 40(1), 45–64. https://doi.org/10.5817/BSE2014-1-3

20.

Geelhand

Papastamou

Deliens

Kissine

. (2020). Narrative production in autistic adults: A systematic analysis of the microstructure, macrostructure and internal state language. Journal of Pragmatics, 164, 57–81. https://doi.org/10.1016/j.pragma.2020.04.014

21.

Goodwin

. (1986). Gesture and coparticipation in the activity of searching for a word. Semiotica, 62(1–2), 51–75. https://doi.org/10.1515/semi.1986.62.1-2.51

22.

Gorman

Olson

Hill

A. P.

Lunsford

Heeman

P. A.

van Santen

J. P. H.

(2016). Uh and um in children with autism spectrum disorders or language impairment. Autism Research, 9(8), 854–865. https://doi.org/10.1002/aur.1578

23.

Hara

. (2006). Grammar of knowledge representation: Japanese discourse items at interfaces. Ph.D. thesis, University of Delaware.

24.

Hesson

Shellgren

. (2015). Discourse marker like in real time: Characterizing the time-course of sociolinguistic impression formation. American Speech, 90(2), 154–186. https://doi.org/10.1215/00031283-3130313

25.

Irvine

C.A.

Eigsti

I. M.

Fein

D. A

. (2016). Uh, um, and autism: Filler disfluencies as pragmatic markers in adolescents with optical outcomes from autism spectrum disorder. Journal of Autism and Developmental Disorders., 46(3), 1061–1070. https://doi.org/10.1007/s10803-015-2651-y

26.

Johnson

(1998, August 9). Correspondence/duh; Today’s kids are, like, killing the English Language. Yeah, right. New York Times. https://www.nytimes.com/1998/08/09/weekinreview/correspondence-duh-today-s-kids-are-like-killing-the-english-language-yeah-right.html

27.

Kaczmarek

L. A.

(2002). Assessment of social communicative competence: An interdisciplinary model. In Goldstein

Kaczmarek

L. A.

English

K. M.

(Eds.), Promoting social communication: Children with developmental disabilities from birth to adolescence (pp. 55–116). Brookes.

28.

Kaufman

A. S.

Kaufman

N. L.

(2004). Kaufman Brief Intelligence Test, Second Edition. Pearson, Inc.

29.

Kercood

Grskovic

J. A.

Banda

Begeske

. (2014). Working memory and autism: A review of literature. Research in Autism Spectrum Disorders, 8(10), 1316–1332. https://doi.org/10.1016/j.rasd.2014.06.011

30.

Kim

Clarke

. (2015). Case study: An iPad-based intervention on turn-taking behaviors in preschoolers with autism. Behavioral Development Bulletin, 20(2), 253–264. https://doi.org/10.1037/h0101314

31.

Kimhi

. (2014). Theory of mind abilities and deficits in autism spectrum disorder. Topics in Language Disorders, 34(4), 329–343. https://doi.org/10.1097/TLD.0000000000000033

32.

Kuijper

S. J. M.

Hartman

C. A.

Hendriks

. (2015). Who is he? Children with ASD and ADHD take the listener into account in their production of ambiguous pronouns. PLoS One, 10(7), 1–18. https://doi.org/10.1371/journal.pone.0132408

33.

Kuroda

S. Y

. (2005). Focusing on the matter of topic: A study of wa and ga in Japanese. Journal of East Asian Linguistics, 14(1), 1–58. https://doi.org/10.1007/s10831-004-2701-5

34.

Kyratzis

Ervin-Tripp

S. M

. (1999). The development of discourse markers in peer interactions. Journal of Pragmatics, 31(10), 1321–1338. https://doi.org/10.1016/S0378-2166(98)00107-6

35.

Lake

J. K.

Humphreys

K. R.

Cardy

. (2011). Listener vs. Speaker-oriented aspects of speech: Studying the disfluencies of individuals with autism spectrum disorders. Psychonomic Bulletin & Review, 18(1), 135–140. https://doi.org/10.3758/s13423-010-0037-x

36.

Lord

Rutter

DiLavore

Risi

Gotham

Bishop

S. L.

(2012). Autism Diagnostic Observation Schedule–Second Edition (ADOS-2). Western Psychological Services.

37.

Lunsford

Heeman

P.A.

Black

van Saten

J. P.H

. (2010, September 25–26). Autism and the use of fillers: Differences between ‘um’ and ‘uh’ [Paper Presentation]. DiSS-LPSS Joint Workshop, Tokyo, Japan.

38.

Maclay

Osgood

C. E

. (1959). Hesitation phenomena in spontaneous English speech. Word-Journal of the International Linguistic Association, 15(1), 19–44. https://doi.org/10.1080/00437956.1959.11659682

39.

Matthewman

Zane

Grossman

. (2021). Comparing frequency of listener responses between adolescents with and without ASD during conversation. Journal of Autism and Developmental Disorders, 52(3), 1007–1018. https://doi.org/10.1007/s10803-021-04996-9. Epub 2021 Apr 10.

40.

McGregor

K. K.

Hadden

R. R

. (2020). Brief report: “um” fillers distinguish children with and without ASD. Journal of Autism and Developmental Disorders, 50(5), 1816–1821. https://doi.org/10.1007/s10803-018-3736-1

41.

McHugh

M. L

. (2012). Interrater reliability: The kappa statistic. Biochemia Medica, 22(3), 276–282. https://doi.org/10.11613/BM.2012.031

42.

McWhorter

. (2016, November 25). The linguistic evolution of ‘like’. The Atlantic. https://www.theatlantic.com/entertainment/archive/2016/11/the-evolution-of-like/507614/

43.

Meehan

. (1991). It’s like, “what’s happening in the evolution of ‘like’?”: A theory of grammaticalization. Kansas Working Papers in Linguistics 16:37–51. https://doi.org/10.17161/KWPL.1808.423

44.

Mehren

. (1999, March 22). Colleges, like, focus on speech. Los Angeles Times. https://www.latimes.com/archives/la-xpm-1999-mar-22-mn-19735-story.html

45.

Miller

(2010) Systematic Analysis of Language Transcripts (SALT): English version 2010 [computer software]. SALT Software, LLC.

46.

Miller

Weinert

. (1995). The function of ‘like’ in dialogue. Journal of Pragmatics 23(4), 365–393. https://doi.org/10.1016/0378-2166(94)00044-F

47.

Novogrodsky

Edelson

L. R

. (2015). Ambiguous pronoun use in narratives of children with autism spectrum disorders. Child Language Teaching and Therapy. 32(2), 241–252. https://doi.org/10.1177/0265659015602935

48.

Odato

C. V

. (2013) The development of children’s use of discourse like in peer interaction. American Speech, 88(2), 117–143. https://doi.org/10.1215/00031283-2346825

49.

Ono

Thompson

Suzuki

. (2000). The pragmatic nature of the so-called subject marker ga in Japanese: Evidence from conversation. Discourse Studies, 2(1), 55–84. https://doi.org/10.1177/1461445600002001003

50.

Parish-Morris

Liberman

M. Y.

Cieri

Herrington

J. D.

Yerys

B. E.

Bateman

Donaher

Ferguson

Pandey

Schultz

R. T

. (2017). Linguistic camouflage in girls with autism spectrum disorder. Molecular Autism, 8(48), 1–12. https://doi.org/10.1186/s13229-017-0164-6

51.

Romaine

Lange

. (1991). The use of like as a marker of reported speech and thought: A case of grammaticalization in progress. American Speech, 66(3), 227–279. https://doi.org/10.2307/455799

52.

Rutter

Bailey

Lord

(2003). SCQ. The Social Communication Questionnaire. Western Psychological Services.

53.

Scheeren

A. M.

Begeer

Banerjee

Terwogt

M. M.

Koot

H. M

. (2010). Can you tell me something about yourself?: Self-presentation in children and adolescents with high functioning autism spectrum disorder in hypothetical and real life situations. Autism, 14(5), 457–473. https://doi.org/10.1177/1362361310366568

54.

Shriberg

L. D.

Paul

McSweeny

J. L.

Klin

Cohen

D. J.

Volkmar

F. R

. (2001). Speech and prosody characteristics of adolescents and adults with high-functioning autism and asperger syndrome. Journal of Speech, Language, and Hearing Research, 44(5), 1097–1115. https://doi.org/10.1044/1092-4388(2001/087)

55.

Siegel

M. E. A

. (2002). Like: The discourse particle and semantics. Journal of Semantics, 19(1), 35–71. https://doi.org/10.1093/jos/19.1.35

56.

South

Ozonoff

McMahon

W. M

. (2007). The relationship between executive functioning, central coherence, and repetitive behaviors in the high-functioning autism spectrum. Autism, 11(5), 437–451. https://doi.org/10.1177/1362361307079606

57.

Suh

Eigsti

I. M.

Naigles

Barton

Kelley

Fein

. (2014). Narrative performance of optimal outcome children and adolescents with a history of an autism spectrum disorder (ASD). Journal of Autism and Developmental Disorders, 44(7), 1681–1694. https://doi.org/10.1007/s10803-014-2042-9

58.

Swerts

. (1998). Filled pauses as markers of discourse structure. Journal of Pragmatics, 30(4), 485–496. https://doi.org/10.1016/S0378-2166(98)00014-9

59.

Tangliamonte

D’Arcy

. (2004). He’s like, she’s like: The quotative system in Canadian youth. Journal of Sociolinguistics, 8(4), 493–514. https://doi.org/10.1111/j.1467-9841.2004.00271.x

60.

Underhill

. (1988). Like is, like, focus. American Speech, 63(3), 234–246. https://doi.org/10.2307/454820

61.

Van Eylen

Boets

Steyaert

Wagemans

Noens

. (2015). Executive functioning in autism spectrum disorders: Influence of task and sample characteristics and relation to symptom severity. European Child & Adolescent Psychiatry, 24(11), 1399–1417. https://doi.org/10.1007/s00787-015-0689-1

62.

Wechsler

(1997). Wechsler Adult Intelligence Scale (3rd ed.). The Psychological Corporation.

63.

Wiig

E. H.

Semel

Secord

W. A.

(2013). Clinical Evaluation of Language Fundamentals–Fifth Edition (CELF-5). Pearson.

64.

Winner

(2002). Thinking About You Thinking About Me. Michelle Garcia Winner Publishing.

65.

Ying Sng

Carter

Stephenson

. (2018). A systematic review of the comparative pragmatic differences in conversational skills of individuals with autism. Autism & Developmental Language Impairments, 3, 1–24. https://doi.org/10.1177/2396941518803806

66.

Zane

Neumeyer

Mertens

Chugg

Grossman

. (2018). I think we’re alone now: Solitary social behaviors in adolescents with autism spectrum disorder. Journal of Abnormal Child Psychology., 46(5), 1111–1120. https://doi.org/10.1007/s1082-017-0351-0

	like subtype
	looseness		focus		quotative		reformulation		total
Group	M	SD	M	SD	M	SD	M	SD	M	SD
Autistic	7.33 (0.01)	5.62 (0.01)	6.86 (0.01)	7.29 (0.01)	1.05 (<0.00)	2.01 (<0.00)	1.62 (<0.00)	3.44 (<0.00)	16.86 (0.03)	14.05 (0.02)
Non-autistic	9.05 (0.02)	10.63 (0.02)	8.80 (0.01)	11.97 (0.02)	0.80 (<0.00)	1.70 (<0.00)	1.50 (<0.00)	1.54 (<0.00)	20.15 (0.03)	21.90 (0.03)