Attention Bias in School-Age Children Who Stutter: Evidence From a Dot-Probe Task

Abstract

Purpose:

Cognitive models of anxiety attribute anxiety and ruminative thought patterns to selective processing of threat-related stimuli that automatically capture attention. We explored whether stuttering was associated with similar attentional biases by examining: (a) whether school-age children who stutter (CWS) differed from controls in selective processing of threat-related and neutral stimuli and (b) whether attentional biases in CWS were specific to threat stimuli that reflected stuttering-related experience.

Method:

Participants included 39 children (19 CWS), ages 8 to 15 years. Children completed a dot-probe task in which they responded as quickly as possible to on-screen probes that replaced threat-related or neutral words. Three types of threat words were presented: (a) general threat words; (b) words related to stuttering; and (c) personalized words on which participants anticipated stuttering. Attention bias (AB) was computed based on reaction times for congruent conditions (probe replaced threat stimuli) relative to incongruent conditions (probe replaced neutral stimuli) and compared across groups and stimulus types.

Results:

Strong evidence for an AB effect was observed for CWS but not for controls, as demonstrated by faster responses to congruent relative to incongruent trials. Within the stuttering group, AB effects were driven primarily by stuttering-related and personal words but not general threat words.

Conclusions:

Findings indicate that CWS preferentially allocate attention toward stimuli relevant to stuttering experiences. Further research is needed to clarify how such selective processing may contribute to the development of stuttering-related concerns, psycho-emotional reactions to stuttering, and associated behaviors, such as avoidance of sounds, words, or speaking situations.

Supplemental Material:

https://doi.org/10.23641/asha.29231180

Many people can relate to the experience of having their attention repeatedly drawn to an intrusive worry that disrupts their concentration. Such involuntary pulls on attention are particularly characteristic of anxious individuals, who tend to automatically direct attention toward sources of potential danger in their environment and struggle to stop attending to such stimuli, especially when the stimuli relate to personal concerns or fears (Mathews & MacLeod, 1994). This type of selective processing or cognitive bias toward perceived threat can promote a state of ongoing hypervigilance characterized by continuous monitoring of environmental stimuli for potential hazards. Such processes strongly resemble anticipatory responses associated with stuttering, in which speakers who stutter might scan planned speech output for words or sounds on which they expect to stutter (Jackson et al., 2015, 2018). Repetitive negative ideation, or the tendency to engage in recurrent thought focused on negative emotions (worry or rumination), is similarly rooted in cognitive processes that bias attention toward environmental input aligned with personal concerns (Mathews & MacLeod, 2005). Repetitive negative thinking is thought to be a transdiagnostic process, meaning that it manifests similarly across multiple emotional disorders (e.g., posttraumatic stress disorder, social phobia, anxiety, eating disorders, depression) with disorder-specific differences in the content of the intrusive thought (Ehring et al., 2011). Recent data indicate that adults who stutter are prone to repetitive negative ideation and that these tendencies are associated with higher adverse impact of stuttering (Tichenor & Yaruss, 2020). Below we elaborate on these and other conceptual ties linking the vast literature on attention bias (AB) to stuttering. Just as selective attention to environmental threat feeds emotional disorders in a cyclical manner, we propose that stuttering experiences, anticipated stuttering, and psycho-emotional responses to stuttering may be interconnected through similar cognitive underpinnings. Threatening stimuli (and stuttering-related stimuli, in particular) may automatically capture attention as a result of accumulated stuttering experiences, especially when these involve negative social reactions (teasing, bullying, rejection, stereotyping) or internal stigmatization of stuttering by the speaker. In this way, AB may reinforce associations between negative emotions and stuttering experiences over time, particularly in those more vulnerable to such attentional tendencies, as described further below. Although he was not referring to AB, Starkweather aptly captured this role of attentional processes in his description of the internal experience of stuttering: “The nonstutterer typically does not even notice his disfluencies … unless some oddity calls them to his attention. But for the stutterer, they are salient above all else at that moment. A source of fear, frustration, shame, loathing, and anger, they loom larger than anything else, huge, swollen, a figure of such magnitude that it can block out all other experience” (Starkweather, 1999).

In the present study, we took a first step toward exploring the association between AB and stuttering by examining whether children who stutter (CWS) and nonstuttering controls differ in the extent to which they preferentially attend to threat-related stimuli. We further sought to clarify the nature of this attentional bias by determining whether the bias differs across threat stimuli with varied levels of relevance to stuttering.

What Is AB?

AB refers to the tendency to prioritize processing of some stimuli over others. Selectively attending to certain environmental input is an important cognitive adaptation, allowing us to filter irrelevant input and thoroughly process a more limited amount of information needed to guide our behavior. In addition to this ongoing form of selective processing, paying more attention to obvious threat is a key function of our attention system and critical for survival. Whereas all individuals prioritize attention allocation for unequivocal hazards (an approaching bear, skidding car, armed assailant, intense physical pain), AB for low-risk stimuli, such as an image of a bomb or printed word “pain,” varies across individuals, with some being more vulnerable to having their attention involuntarily engaged by such stimuli.

Heightened AB to minor forms of threat is a well-established finding in clinical and subclinical forms of anxiety or phobia, and has been demonstrated in both adults and children (Bar-Haim et al., 2007; Fu & Pérez-Edgar, 2019; MacLeod & Mathews, 1988; Mogg & Bradley, 2018). Based on cognitive models of anxiety (Cisler & Koster, 2010; Mathews & Mackintosh, 1998; Mathews & MacLeod, 2005; Mogg & Bradley, 2018), the potential for AB occurs when simultaneously activated representations for two or more stimuli compete for attentional resources. When one of these representations is associated with danger, its activation level is increased by output from a threat evaluation system which reallocates resources so that more attention is directed toward the potential threat. These adjustments can be triggered by conditions of arousal that temporarily reset the threat evaluation system to a lower threshold. Threshold settings are also consistently lower in high trait-anxious individuals, resulting in frequent vigilance responses on a regular basis. In this way, AB feeds anxiety by promoting sensitivity toward threat and reinforcing hypervigilant scanning of environmental stimuli for potential danger. In other words, the more that anxious individuals attend to perceived threats in their environment, the more they encode information about potential danger, reinforcing a circular relationship between selective processing and anxious mood. This feedback loop creates an underlying vulnerability to emotional disorder (Mathews & MacLeod, 1994) and AB is therefore widely understood to play an important role in the onset and maintenance of anxiety and related disorders.

Measuring AB to Threat

A wide range of paradigms and methodologies are available for measuring AB, but most traditional measures use computerized cognitive tasks that infer AB based on differences in reaction time (RT) to threat versus neutral stimuli (see Azriel & Bar-Haim, 2020; Mathews & MacLeod, 1994; Mogg & Bradley, 2016; Van Bockstaele et al., 2014, for detailed reviews). Unlike some tasks that feature threat stimuli as distractors (as in the emotional Stroop) or as targets (e.g., visual search tasks), which can make it difficult to interpret findings with certainty (Bar-Haim et al., 2007), the dot-probe task (DPT) incorporates threat stimuli in a different capacity and is the most extensively studied AB paradigm. Originally described by MacLeod and colleagues (MacLeod et al., 1986), the task involves pairs of neutral and threat-related stimuli that are presented simultaneously for a brief time, followed by a probe (e.g., arrow) at the location of either the neutral or threat stimulus. Participants respond as quickly as possible to the location or identity of the probe. AB toward threatening stimuli is inferred when RT is faster on trials in which the probe appeared at the location of the threatening cue (congruent trials) compared to trials in which the probe replaced the neutral stimulus (incongruent trials), reflecting the involuntary pull of attention toward the location of perceived threat.

Several considerations related to AB measurement guided the design of the present study. We selected the DPT paradigm because it is the most well-researched behavioral task in the AB literature and is better at disentangling AB effects from other response effects related to arousal, such as interference. For example, when participants name the ink color and ignore content of threat and nonthreat words in the emotional Stroop task, some question whether slower color naming for threat versus neutral stimuli reflects attentional capture or a negative emotional state that is induced by the threat cue and impairs RT (Bar-Haim et al., 2007; Mathews & Mackintosh, 1998). An advantage of the DPT is that participants are always responding to neutral stimuli (probes), eliminating the possibility that slow RT reflects general arousal.

The DPT also allows for manipulation of presentation durations for the critical stimuli (threat and neutral pairs) preceding the probe, making it possible to examine different stages of attentional processes. In most studies, stimuli are presented for 500 ms or longer, allowing the stimuli to be consciously perceived; however, some studies present stimuli very briefly (subliminal exposure) to capture early, automatic processes outside participants' awareness. Longer stimulus duration times (e.g., 1,250 ms) can also be used to engage top-down cognitive mechanisms that induce AB away from threat (avoidance) as a fear reduction strategy following initial vigilance (e.g., see Haft et al., 2019, for this pattern of results in school-age children; see Cisler & Koster, 2010, for a review). Manipulations of this parameter therefore target relatively distinct AB components and patterns of effects. For example, vigilance effects (automatic, facilitated detection of threat) can be detected at short (e.g., 500 ms or less), but not long (e.g., 1,250 ms), durations. When exposure to threat stimuli is long, high trait-anxious participants will tend to show patterns of attentional avoidance to threat rather than facilitated detection. In the current study, we presented critical stimuli (threat and neutral word pairs) for 500 ms in order to focus on AB patterns that reflected largely automatic responses to stimuli that were within participants' awareness. These timing parameters have been used in prior dot probe studies with children (e.g., Burris et al., 2017; McAllister et al., 2015; Susa et al., 2012) with some presenting word stimuli to children at even shorter durations (e.g., Beck et al., 2011).

Finally, meta-analyses of behavioral (Pergamin-Hight et al., 2015) and neurophysiological data (Botelho et al., 2023) consistently indicate that AB effects are largest for threatening stimuli that are disorder-congruent, or relevant to personal concerns or fears of participants (e.g., spider images for individuals with spider phobia). Content specificity for AB effects is also observed in children. For example, Haft and colleagues found that children with learning disorders showed attentional avoidance of word stimuli related to reading (e.g., book, grammar) but not general threat (GT) stimuli (Haft et al., 2019). Meta-analytical results further indicate that naturalistic images and words are equally potent in inducing threat-related AB, particularly for stimuli with supraliminal exposure (Bar-Haim et al., 2007). We used words in the present study (see Haft et al., 2019, and Beck et al., 2011, for examples of prior AB studies that used word stimuli with school-age children) so that we could incorporate stimuli that were as content-specific as possible without sacrificing on the magnitude of effects of interest.

Threat-Related AB and Stuttering

Theoretical links and some empirical data (Bauerly, 2022; Hennessey et al., 2014; Lowe et al., 2016; McAllister et al., 2015; Rodgers et al., 2020) suggest a potential role of AB in stuttering, specifically in the development of psycho-emotional responses to the experience of stuttering. Cognitive models of anxiety attribute threat-related AB to individual differences in temperament and attention control (Cisler & Koster, 2010; Eysenck et al., 2007; Lonigan et al., 2004; Lonigan & Vasey, 2009; Mogg & Bradley, 2018). In particular, the behavioral inhibition system has been identified as a core mechanism involved in detecting danger and directing attention toward threat (Henderson et al., 2015; Mathews & Mackintosh, 1998). Aspects of temperament, such as negative affect (NA) and effortful control (EC) also impact attention from an early age. For example, infants with high NA shifted attention away from threatening stimuli more slowly than those with low NA on a Baby Dot Probe task with eye tracking (Pérez-Edgar et al., 2017). EC has been shown to moderate this relation between NA and threat-related bias, such that at school age, only children with high NA and low EC showed AB to threat (e.g., Lonigan & Vasey, 2009; see Valadez et al., 2022, for a review). Thus, high levels of NA serve as a risk factor for threat-related AB (and ultimately, psychopathology); however, high levels of EC enable children to override the attentional bias, offering protection against the development of emotional disorders.

Many studies have demonstrated associations between stuttering and aspects of temperament, with increased behavioral inhibition, increased NA, and decreased EC among individuals who stutter relative to controls (Ambrose et al., 2015; Eggers et al., 2010; Felsenfeld et al., 2010; Karrass et al., 2006; Ntourou et al., 2020). Available data span different age groups, with similar results across preschoolers, school-age children, and adults. Although some of these studies focus on relationships between temperament and overt stuttering characteristics (Choi et al., 2013; Kraft et al., 2014, 2019), several also link temperament traits and adverse impact of stuttering. For example, stuttering impact was positively correlated with NA in adolescents who stutter (Eggers et al., 2021). Additionally, among adults who stutter, those with high NA and low EC engaged in more frequent repetitive negative thoughts and reported more significant adverse stuttering impact (Tichenor & Yaruss, 2020). Recent research also demonstrates an association between stuttering and weaknesses in attention control across different age groups (Anderson et al., 2020; Eichorn et al., 2018; Eichorn & Pirutinsky, 2021; Ofoe et al., 2018; Rocha et al., 2019; Tichenor et al., 2022).

To our knowledge, six published studies have examined information-processing biases in individuals who stutter to date (Bauerly, 2022; Hennessey et al., 2014; Lowe et al., 2016; McAllister et al., 2015; Rodgers et al., 2020). Lowe et al. (2012) tracked eye movements of adult stuttering and nonstuttering speakers while they delivered an impromptu speech to prerecorded audience members displaying varied emotional expressions (positive, negative, and neutral). Relative to controls, speakers who stutter showed shorter looking times for audience members with positive expressions compare to neutral or negative expressions. Authors interpreted these findings as evidence that adults who stutter may be biased toward negative environmental stimuli and neglect positive social cues that may help disconfirm fears and relieve anxiety over time. Another study (Hennessey et al., 2014) measured AB using an emotional Stroop task and found that adults who stutter were slower (by an average of 21 ms) to name colors of threat words relative to neutral words, whereas control participants did not show this effect. This pattern was limited to spoken (vs. manual) responses and was elicited using GT words (e.g., failure) drawn from prior studies of social anxiety. Performance on the emotional Stroop task can reflect various cognitive and emotional responses other than AB (e.g., freezing in response to threat, mood effects), as mentioned above, and may therefore not provide an optimal measure of vigilant attention patterns (Bar-Haim et al., 2007). This paradigm also primarily reflects later, more conscious attentional processes and does not reliably capture facilitated detection of threat which is the most established and well understood component of AB (Cisler & Koster, 2010).

The study by Lowe et al. (2016) compared vigilance effects in adults who stutter and controls using a dot-probe paradigm with images of faces displaying neutral, positive, and negative expressions (sadness, disgust, anger, fear) paired with neutral images of household items (e.g., vacuum cleaner). Results revealed no group differences in AB, suggesting that threat-related AB is not characteristic of adults who stutter. The authors acknowledge, however, that their sample was largely drawn from posttreatment maintenance courses and support groups, and did not differ from controls on social anxiety. These sample characteristics could explain why AB effects were not observed with stimuli known to elicit the bias in socially anxious individuals.

The remaining three studies of AB in individuals who stutter used the DPT, but like Hennessey et al. (2014) and Lowe et al. (2016), included GT stimuli consisting of faces with negative expressions. Rodgers et al. (2020) used a slightly modified version of the standard DPT in which critical stimuli (threatening and neutral faces) were presented between an initial “anchor” probe and subsequent target probe. The authors distinguished between engagement trials, in which the threatening face appeared in a different location than the initial anchor, and disengagement trials, in which the threat cue appeared in the same location as the anchor. The two trial types were designed to assess vigilance and avoidance patterns, respectively, and both types of threat-related AB were evident in adolescents who stutter but not in controls. Bauerly (2022) likewise found evidence of AB toward threatening faces in adults who stutter on a traditional DPT following social threat induction (an anticipated public speaking task). Across available studies, none incorporated content-specific stimuli directly relevant to stuttering, making it difficult to determine whether attentional effects are primarily driven by stuttering-related concerns and experiences. Only one published study focused on CWS (McAllister et al., 2015) but did not include a control group. Thus, available data regarding AB effects in individuals who stutter are limited and data for CWS are largely inconclusive.

Present Study

The present study assessed whether CWS show facilitated detection of threat stimuli to a greater extent than children who do not stutter (CWNS) on a standard DPT. We incorporated stimuli that varied in relevance to personal concerns and experiences of CWS in order to clarify the nature of potential AB patterns in children and to distinguish between stuttering-specific concerns and more general forms of anxiety. We examined two primary hypotheses:

1. Based on preliminary evidence of AB patterns in adults who stutter from prior studies, as well as temperament traits associated with stuttering that predispose individuals to threat-related AB (increased NA, decreased EC), we hypothesized that AB to threat would be evident in CWS but not in controls.

2. Given robust findings of content-specific effects in AB to threat, we further predicted that AB to threat would be evident for threat stimuli related to stuttering but not for GT associated with social or general forms of anxiety.

Method

Participants

Participants included 19 CWS (five female) and 22 CWNS (10 female) between the ages of 8–15 years. We focused on school-age children in order to better understand the potential role of AB in the development of stuttering and children's responses to stuttering. School-age children are old enough to read words reliably (e.g., Haft et al., 2019); many school-age CWS have also developed anticipation (Jackson et al., 2018) and other adverse responses to stuttering. Our sample size was based on similar published studies with children (Haft et al., 2019), as well as a power analysis estimate using an alpha level of .05 (two-sided), expected effect size of .45 (Haft et al., 2019; Roy et al., 2008), and power of .80 to reject a null hypothesis (AB of 0). Children were recruited via an e-mail blast to National Stuttering Association membership, and from local schools, speech-language clinics, and the community via flyers, newspaper advertisements, social media posts, and word of mouth. Participants completed all study procedures online and received compensation in the form of a $40 Amazon gift card. All children were proficient in English and scored within at least the average range on standardized tests of nonverbal intelligence and language ability, as measured by the Test of Nonverbal Intelligence–Fourth Edition (Brown et al., 2010) and Clinical Evaluation of Language Fundamentals–Fifth Edition (CELF-5; Wiig et al., 2013), respectively. Absence of other speech/language disorders, developmental disorders, and significant medical conditions was confirmed for all participants based on parental report. Parents additionally completed the Parent Rating Scale of the Behavior Assessment System for Children–Third Edition (BASC-3; Reynolds & Kamphaus, 2015), a widely used measure that provides information regarding children's emotional, cognitive, and behavioral characteristics. Items on the scale are designed with a four-choice response format (Never, Sometimes, Often, or Almost Always) and form a total of 23 scales, organized into three scale types: (a) clinical scales, which measure maladaptive behaviors such as attention problems; (b) adaptive scales, which measure behavioral strengths such as social skills; and (c) content scales, which measure theoretically based areas of interest such as executive functioning. We included the BASC to control for variables such as anxiety and attention problems that might differ between groups and could influence AB measures but were not our primary outcome of interest in the present study. The racial/ethnic makeup of the groups was similar, with 20 and 17 White children in the CWNS and CWS groups, respectively, one Hispanic or Latino child per group, one Asian CWNS, and one African American CWS. More than half of the participants (25/41) were from Tennessee, with four non-TN residents in the CWNS group, and 12 in the CWS group (one Canadian). Nearly all participants (n = 38) were monolingual. One participant in each group was bilingual (one CWNS spoke and read Spanish well; one CWS spoke and read French very well). Two additional CWNS had a small amount of exposure and experience in other languages.

CWS all self-identified as people who stutter, began stuttering at or before the age of 5 years, had been formally diagnosed by a speech-language pathologist, and reported a perceived life impact of stuttering within at least the mild range on the Student (ages 7–12 years) or Teen (ages 13–18 years) versions of the Overall Assessment of the Speaker's Experience of Stuttering (OASES), a subjective measure of perceived life impact of stuttering (Yaruss et al., 2016). Based on the OASES, nine of the CWS rated their life impact of stuttering as mild–moderate, eight as moderate, and two as moderate–severe. The majority of CWS (17/19) had received speech therapy in the past (one for less than a year, eight between 2 and 5 years, and eight for > 5 years); seven of these children were still receiving speech therapy at the time of the study. Group characteristics for CWS and CWNS were compared via a series of t tests; results confirmed that CWS and CWNS were closely matched on age and did not differ significantly in gender distribution, nonverbal intelligence, or language ability. Demographic details, results of standardized testing, and group comparisons are summarized in Table 1.

Table 1.

Participant characteristics and group comparisons.

Variable	CWNS		CWS		Difference
	Min–Max	M (SD)	Min–Max	M (SD)	t testa	db
Age in years	8–14	10.96 (2.06)	9–15	11.89 (1.59)	t(38.6) = 1.65, p = .108	0.51
TONIc	89–143	107.4 (11.97)	91–128	108.37 (9.55)	t(38.8) = 0.3, p = .767	0.09
CELFc	92–134	112.64 (11.92)	86–150	113.74 (17.99)	t(30.5) = 0.23, p = .822	0.07
BASC: Anxietyc	36–77	51.14 (9.35)	37–70	55 (8.64)	t(37.9) = 1.36, p = .183	0.43
BASC: ADHD	37–67	48.43 (7.51)	35–80	51.53 (11.26)	t(30.9) = 1.01, p = .319	0.32
OASESd	—	—	1.68–3.69	2.32 (0.52)	—	—

Note. Em dashes indicate that assessment is designed for speakers who stutter and was not administered to CWNS. CWNS = children who do not stutter; CWS = children who stutter; TONI = Test of Nonverbal Intelligence; CELF = Clinical Evaluation of Language Fundamentals; BASC = Behavior Assessment System for Children; ADHD = attention-deficit/hyperactivity disorder; OASES = Overall Assessment of the Speaker's Experience of Stuttering.

^aWelch two-sample t test (2-tailed).

^bCohen's d effect size.

^cStandard scores, see text for a description of the variables.

^dOverall impact score.

General Procedure

All procedures were completed remotely in two sessions conducted via Zoom and were approved by The University of Memphis Institutional Review Board (ID # PRO-FY2021–157). Sessions were led by graduate students trained as research assistants. In the initial “setup” session (approximately 15 min), research assistants confirmed that the children were using appropriate devices (laptops or desktop computers and not iPads or tablets), guided the children or parents through installation of the Inquisit Player (Millisecond, 2021) needed to run the experiment, helped the child locate a quiet area for their testing session, and obtained informed parental consent and child assent. CWS also engaged with the research assistant in a brief interview designed to elicit a list of words on which the child anticipated stuttering. Elicitation procedures followed protocols described in other studies (Bowers et al., 2012; Jackson et al., 2020; see Goldfarb et al., 2023, for application of the elicitation procedure to children and adolescents). The subsequent session, generally scheduled within 1 week of the setup session, lasted approximately 1.5 hr and included administration of standardized tests (TONI and CELF) as well as completion of the experimental task. Tasks were sequenced such that children first completed the TONI, followed by the experimental task and the CELF. We separated the standardized measures to avoid fatigue and boredom; we also did not want to begin the session with the experimental task in order to avoid linking the task with the elicitation procedure in which the children provided anticipated words at the end of their previous session. Parents provided demographic and developmental information via a written questionnaire; they also completed the Parent Rating Scale of the BASC-3, as described above.

Experimental Task

The experimental task was modeled after the classic visual–spatial paradigm described by Macleod and colleagues (MacLeod et al., 1986) and was administered via Inquisit by Millisecond (Millisecond, 2021) in a Zoom session. Individual trials consisted of a central fixation cross presented for 500 ms, followed by paired words (threat and neutral) presented one above the other (based on configurations described in Haft et al., 2019; MacLeod et al., 1986) for 500 ms. A single arrow pointing right or left replaced one of the words and remained on the screen until the participant responded by pressing a letter key (F or J) corresponding to the arrow's direction. Children were instructed to place their left pointer finger on the “F” key on their keyboard and their right pointer finger on the “J” key. All words and arrows appeared in black print on a white screen.

Participants completed four practice trials, followed by four blocks of 48 trials each, with a resulting total of 192 trials (64 trials per condition). Children had the option of repeating practice trials, if needed; all practice trials consisted of neutral word pairs that did not appear elsewhere in the experiment and were omitted from later analyses. Breaks were provided after each set of 48 trials. Threat words were general (e.g., war, bully), specific to stuttering (e.g., speech, block), or specific to each child's personal stuttering experiences (i.e., anticipated words). All threat categories included 16 exemplars, and each threat word was paired with a neutral word matched on various parameters, as described below. AB to threat was quantified as the difference in RT between congruent trials (in which probe replaced threat stimulus) and incongruent trials (in which probe replaced neutral stimulus). Individual word pairs were presented four times each to counterbalance the two word positions (upper and lower), and probe locations (congruent and incongruent). Threat types were not batched by condition but were instead interspersed across the 192 trials in a random sequence; locations of words and probes were likewise randomly sequenced. See Figure 1 for a representative sample of a GT and neutral word pair in a congruent trial.

Figure 1.

Representative sample of a general threat and neutral word pair in a congruent trial.

Stimuli

Threat Words

Stimulus selection was guided by prior studies, published databases, and clinical resources related to stuttering. GT words were selected from stimulus lists used in similar studies with children (Haft et al., 2019; Vasa et al., 2006). We also adapted criteria from these studies for identifying words as “threat” or “neutral” and referred to emotional valence ratings from a published database with nearly 14 thousand English lemmas (Warriner et al., 2013). Valence refers to the pleasantness of emotions invoked by a word and is commonly measured on a scale from 1 (unhappy) to 9 (happy). We defined GT stimuli as words with a valence rating lower than 4; the mean valence for our final GT stimuli was 2.25 (min = 1.6; max = 3.66). Stutter-specific threat (ST) words were drawn from an emotional Stroop task used with individuals who stutter (van Lieshout et al., 2014) and from published assessments that measure negative attitudes and reactions toward stuttering in CWS (Ayre & Wright, 2009; Yaruss et al., 2016). Valence values for these stimuli tended to fall in the neutral range but ranged from 3.81 (interruption) to 6.73 (read), with a mean valence rating of 5.71 for the final 16 words comprising the ST stimuli. Personalized threat (PT) words on which CWS anticipated stuttering were elicited individually for each CWS using procedures described in earlier studies (Bowers et al., 2012; Jackson et al., 2020). Although these methods were previously used for adults, available evidence indicates that school-age children also anticipate stuttering (Jackson et al., 2018) and develop feared sounds and words as early as age 8 years (Bloodstein, 1960). CWNS did not provide PT words but were presented randomly selected PT words from the running total list of stimuli elicited from CWS. Valence values for PT stimuli varied widely, with ratings spanning almost the entire scale (min = 2.0, max = 8.11) and a mean valence value of 5.96. A large proportion of elicited PT stimuli (182 out of 367) could not be associated with a value because they did not represent lemmas and were not included in the Warriner et al. database (e.g., Allie, with, here).

Neutral Matches

Each threat word (across categories) was matched to a neutral word based on word length (in characters and syllables), part of speech, and frequency of occurrence in English (Brysbaert & New, 2009). Neutral matches for the GT stimuli had valence ratings between 4 and 6, with a mean valence of 5.05 for the final set of 16 words (min = 4.43, max = 5.73). For ST words, neutral words were matched on valence (< 1 point valence difference based on Warriner et al., 2013) to isolate perceived threat associated with general stuttering experiences. Neutral matches for PT words were drawn from a subset of words with valence values between 4 and 6 from the Warriner et al. (2013) database and matched to elicited words based on remaining parameters (characters, syllables, part of speech, frequency). In selecting neutral matches for PT words, we avoided matches that began with any reported anticipated sounds or that resembled any anticipated words phonetically or semantically.

Validity of the ST stimuli was verified using a brief piloting task in which a separate group of 10 CWS and 20 CWNS provided valence ratings for 17 neutral words, 34 GT words, and 17 ST words (selected and matched as described above) using a 1–9 scale on which 1 = completely unhappy and 9 = completely happy. Analysis of variance results indicated a significant Group × Word Type interaction, F(2, 67) = 3.16, p = .043, such that group ratings differed only for the ST words, with lower (more negative ratings) provided by CWS relative to controls. This finding strongly suggests that ST stimuli captured stuttering-related associations that were specific to the CWS. All threat and neutral words had a maximum age of acquisition rating below age 8 years (Kuperman et al., 2012). A full list of threat and neutral stimuli used in the GT and ST conditions, as well as sample PT stimuli for a single participant is provided in Supplemental Material S1 (see Table S1).

Statistical Analysis

Data were analyzed using Bayesian mixed-effects regression in R (Version 4.4.1; R Core Team, 2024), using the stan_lmer function from the rstanarm package (Version 2.32.1; Goodrich et al., 2024). The Bayesian approach offers several advantages over traditional frequentist statistics. These include the incorporation of prior knowledge, reduced bias in estimates, particularly with sparse data and small sample sizes, and a direct, transparent method for assessing how strongly empirical results support the research hypothesis. For a more detailed discussion, we refer readers to the dedicated literature on this topic (Etz & Vandekerckhove, 2018; Gelman et al., 2014; Haendler et al., 2020).

The final model described in the results section includes effects and interactions of group (CWS, CWNS), condition (GT, ST, PT), and congruence (Congruent, Incongruent), along with effects of age, sex, nonverbal intelligence (TONI standard score), language ability (CELF Core Language Score), anxiety (gender-normed standardized score on BASC-3 Anxiety subscale), and attention (gender-normed standardized score on BASC-3 Attention-Deficit/Hyperactivity Disorder Index). The dependent variable consisted of RT to probes replacing critical stimuli (threat and neutral words). The random effect structure was the maximal justified by the study design (Barr et al., 2013) and included random intercepts for subjects, random slopes for within-subject predictors (condition and congruence), and their interactions. For details on alternative models and the model selection approach adopted, please refer to Tables S2 and S3 in Supplemental Material S1.

We used weakly informative priors for both fixed effects (normal distribution for the coefficients and exponential for sigma; Goodrich et al., 2024) and the correlation matrix of the random effects' variance–covariance matrix (Lewandowski–Kurowicka–Joe prior, with regularization = 2; Lewandowski et al., 2009). Prior distributions were centered around the coefficients' means and had standard deviations one order of magnitude greater than the parameters' standard deviations. The use of weakly informative priors helps regularize the model by providing some structure to the parameter estimates without significantly influencing the model estimates, and therefore the final results (Gelman et al., 2014).

Models were run with four chains, 3,000 iterations each, including 1,000 warm-up iterations, resulting in a total of 4 × 2,000 = 8,000 samples, which proved to be adequate. Convergence was checked both visually (e.g., via traceplots and posterior predictive checking) and by examining various numerical convergence indices, specifically, for each parameter: effective sample size, Monte Carlo standard error, and R-hat values (Gelman et al., 2014; Stan Development Team, 2024; Vehtari et al., 2021). No convergence concerns emerged.

Prior to statistical analysis, we excluded observations with incorrect responses (4.6% of responses for CWS and 4.4% for CWNS). All errors involved incorrect selection of F or J, with no selection of keys other than those designated as acceptable responses. Following procedures in prior dot-probe literature, we used a two-step approach to additionally remove extreme values. First, we excluded observations with implausible values due to impulsive/anticipatory responses or lapses in performance: these included observations with a response time < 200 ms for all children (1% of the data), longer than 2,000 ms for CWNS (0.6% of the data), and responses longer than 3,000 ms for CWS (0.5% of the data). We used different cutoffs for CWS and CWNS due to the significant difference in average response times between the two groups. Next, we removed observations beyond ±2 SDs of each individual participant's mean (2.8% of the data). To ensure that our results were not dependent on any specific outlier detection approach (Price et al., 2015), we examined effects of alternative values, for example, 150 and 300 ms for the RT lower bound, and 2.5 and 3 SDs instead of 2. In no case did the pattern of results change significantly. In the results below, a probability (p value) of less than .05 will be considered strong evidence for an effect, while a probability between .05 and .10 will be interpreted as weak evidence for the effect.

Results

To address our research questions, we examined the effects of group (CWS, CWNS), condition (GT, ST, PT), and congruence (Congruent, Incongruent). The model also included age, sex, nonverbal intelligence, language ability, anxiety, and attention as covariates, as described above.

For the model covariates, we found a strong effect of age (p < .001) in the expected direction, with older children being faster than younger children overall. Language abilities, as assessed by the CELF-5 core language score, also showed a strong, positive effect on response time, with higher core language scores related to overall faster responses (p < .001). There was a weak effect of nonverbal intelligence (p = .083), as assessed by TONI standard scores, again in the expected direction; higher scores were associated with faster responses. There was no evidence of any effect of sex, anxiety, or attention. Parameter estimates and uncertainty intervals are reported in Table 2.

Table 2.

Summary of Bayesian mixed-effects regression: Parameter estimates and uncertainty intervals for the model covariates.

Parameter	5%a	25%a	Mdn	75%a	95%a	p(β < 0)b	p(β > 0)b
Age	−129.08	−110.93	−98.55	−86.38	−69.05	1	0
Sex	−21.47	−4.57	6.87	18.66	35.59	.341	.659
TONIc	−54.78	−36.75	−24.61	−12.63	4.69	.917	.083
CELFc	−90.15	−72.76	−60.37	−48.37	−30.83	1	0
BASC: Anxietyc	−50.33	−29.78	−16.58	−3.72	16.27	.81	.19
BASC: ADHDc	−32.28	−12.13	1.32	14.72	33.97	.476	.524

Note. TONI = Test of Nonverbal Intelligence; CELF = Clinical Evaluation of Language Fundamentals; BASC = Behavior Assessment System for Children; ADHD = attention-deficit/hyperactivity disorder.

^aLower and upper bounds of 50% (25th and 75th percentiles) and 90% (5th and 95th percentiles) uncertainty intervals.

^bp(β < 0) and p(β > 0): Posterior probabilities of the parameter being less than or greater than zero, respectively.

^cSee text for description. Sex was coded 0 (male) and 1 (female). All predictors were standardized. Note that the estimates in Tables 2 and 3 are from the same mixed-effects regression model; they are presented in separate tables to facilitate readability.

The responses of CWS as a group were, on average, almost 400 ms longer than those of the CWNS. As illustrated in Table 3, Parameter 3, the probability for this difference being greater than zero was 100%, indicating a very strong and reliable difference between the two groups.

Table 3.

Summary of Bayesian mixed-effects regression: Parameter estimates and uncertainty intervals for the effects and interactions of Group, Condition, and Congruence.

Parameter	5%a	25%a	Mdn	75%a	95%a	p(β < 0)b	p(β > 0)b
A. Group effects
1. CWS	846.68	872.25	888.79	904.82	929.21	0	1
2. CWNS	453.4	476.35	492.67	509.34	533.23	0	1
3. Difference: 1–2	336.3	370.51	395.52	420.54	456.2	0	1
B. Group × Congruence Effects
4. CWS Incongruent	852.92	878.39	895.36	911.71	936.04	0	1
5. CWS Congruent	840.55	865.71	882.22	898.23	922.47	0	1
6. CWNS Incongruent	451.03	474.27	490.91	507.86	532.05	0	1
7. CWNS Congruent	454.72	477.9	494.37	511.04	534.6	0	1
8. Difference: 4–5 (AB in CWS)	4.78	9.62	12.99	16.46	21.45	.006	.994
9. Difference: 6–7 (AB in CWNS)	−11.32	−6.7	−3.47	−0.26	4.62	.767	.233
10. Difference: 8–9	4.87	11.69	16.47	21.19	28.09	.01	.99
C. Group × Congruence × Condition Effects
11. CWS Incongruent PT	854.09	879.57	896.75	913.37	938.2	0	1
12. CWS Congruent PT	841.55	866.88	883.93	900.02	924.24	0	1
13. CWS Incongruent ST	860.39	886	903.34	920.08	945.31	0	1
14. CWS Congruent ST	841.48	867.66	884.45	901.3	926.01	0	1
15. CWS Incongruent GT	842.99	868.27	886.01	902.61	927.93	0	1
16. CWS Congruent GT	836.16	861.49	878.56	895.12	919.63	0	1
17. CWNS Incongruent PT	446.95	471.06	487.45	504.96	529.95	0	1
18. CWNS Congruent PT	448.37	472.43	488.78	505.9	529.49	0	1
19. CWNS Incongruent ST	449.86	473.49	490.6	507.89	532.8	0	1
20. CWNS Congruent ST	454.42	477.71	495.01	511.92	536.25	0	1
21. CWNS Incongruent GT	453.29	477.32	494.5	511.52	535.29	0	1
22. CWNS Congruent GT	459.54	482.72	499.45	516.31	540.93	0	1
23. Difference: 11–12 (AB in CWS PT)	−0.53	7.31	12.97	18.53	26.22	.058	.942
24. Difference: 13–14 (AB in CWS ST)	5.08	13.03	18.81	24.4	32.78	.014	.987
25. Difference: 15–16 (AB in CWS GT)	−7.24	1.38	7.41	13.43	22.2	.202	.798
26. Difference: 17–18 (AB in CWNS PT)	−13.56	−6.11	−1.07	4.16	11.72	.551	.448
27. Difference: 19–20 (AB in CWNS ST)	−17.07	−9.34	−4.04	1.24	8.96	.696	.304
28. Difference: 21–22 (AB in CWNS GT)	−19.12	−11.02	−5.22	0.5	8.61	.732	.268

Note. Note that estimates in Tables 2 and 3 are from the same mixed-effects regression model; they are presented in separate tables to facilitate readability. CWS = children who stutter; CWNS = children who do not stutter; Congruent = congruent trials; Incongruent = incongruent trials; AB = attention bias; PT = personal threat words; ST = stuttering-related threat words; GT = general threat words.

^aLower and upper bounds of 50% (25th and 75th percentiles) and 90% (5th and 95th percentiles) uncertainty intervals.

^bp(β < 0) and p(β > 0): Posterior probabilities of the parameter being less than or greater than zero, respectively.

Overall AB effects, averaged across the three conditions (GT, ST, and PT), are reported in Parameters 8 and 9 in Table 3. There was strong evidence for an AB effect among CWS, as demonstrated by faster responses on congruent trials compared to incongruent trials within this group (p = .006). In contrast, CWNS showed no evidence of AB (p = .767). The evidence supporting a difference in AB effects between the two groups was strong (p = .01), indicating a greater AB effect in CWS relative to the CWNS group (Parameter 10 in Table 3).

Lastly, we examined AB effects by group and condition. There was no evidence of any AB effects in the CWNS group, consistent with the complete lack of an overall AB effect in this group, as described above (Parameters 26–28 in Table 3 and Figure 2). In contrast, results for the CWS indicated varying patterns of AB depending on condition. Analysis by condition showed that overall, AB effects for CWS were driven primarily by the ST and PT conditions (p = .014 and p = .058, respectively), with no evidence of AB in the GT condition (p = .202; Parameters 23–25 in Table 3 and Figure 2). Descriptive statistics for RT by speaker group and condition are provided in Table S4 (Supplemental Material S1).

Figure 2.

Density plots for attention bias (AB) effects by group and condition, corresponding to Parameters 23 to 28 in Table 3. Shaded areas and superimposed numbers indicate the area and the probability of the AB effect being greater than zero. CWS = children who stutter (orange); CWNS = children who do not stutter (blue); PT = personal threat words; ST = stuttering-related threat words; GT = general threat words.

Discussion

A key finding emerging from this study was that CWS prioritized attention toward threat-related stimuli, whereas CWNS did not show this pattern of attention allocation. This effect was observed for stimuli with direct relevance to stuttering and stimuli that reflected personal stuttering-related concerns of CWS, but not on general threat stimuli. These results provide the first evidence of disorder-congruent AB effects in individuals who stutter and have important implications for the way we understand anxiety-like characteristics associated with stuttering.

AB, Anxiety, and Stuttering

Our results are similar to those reported by other researchers examining AB related to stuttering (most recently, Bauerly, 2022; Rodgers et al., 2020) but is the first to extend this finding to children. The varied effects based on condition also highlight a facet of stuttering-related AB that has not been examined in prior studies. Based on the extensive literature examining AB in relation to anxiety, it is well established that words associated with general physical or social threat elicit AB effects in individuals with anxiety disorders (Bar-Haim et al., 2007; Mogg & Bradley, 2016; Morales et al., 2017; Roy et al., 2008; Valadez et al., 2022; Van Bockstaele et al., 2014), including children. In contrast to this demonstrated result, our findings showed no threat-related AB for GT stimuli in CWS, indicating that the attention allocation patterns demonstrated by CWS do not mirror patterns characteristic of generalized anxiety. Instead, CWS showed a tendency to have their attention captured specifically on words related to stuttering (e.g., speech, block) or words on which they anticipated stuttering. This finding suggests that AB effects observed for CWS reflect personal speech-related fears that develop over time in response to stuttering experiences.

Distinguishing between generalized and speech-related anxiety is important for clarifying the nature of anxiety in stuttering. Evidence indicates an association between stuttering and anxiety, particularly social and trait anxiety, and particularly in adult samples (Craig et al., 2003; Ezrati-Vinacour & Levin, 2004; Iverach, O'Brian, et al., 2009; Iverach & Rapee, 2014). Relative to population norms, prevalence rates of social anxiety are higher than expected among adolescents and adults who stutter, ranging from 21% to 60% compared to 8% to 13% in the general population (Iverach, O'Brian, et al., 2009). Similar findings are reported for children, with school-age CWS showing increased odds of social and generalized anxiety (Iverach et al., 2016), as well as a higher prevalence of clinically significant anxiety relative to population data (Eggers et al., 2022). In their systematic review, Smith et al. (2014) found higher levels of anxiety in CWS relative to controls in seven out of 13 studies and suggested that anxiety risk increases as CWS approach adolescence and adulthood. These findings were corroborated in a recent meta-analysis (Bernard et al., 2022), which indicated variability among studies but overall increased risk of anxiety symptoms among stuttering children and adolescents relative to nonstuttering peers (g = 0.42, 95% confidence interval [0.10, 0.74]).

Social anxiety is debilitating and interferes with healthy social development, relationships, educational achievement, and occupational performance (Lipsitz & Schneier, 2000). Among adults who stutter, anxiety also predicts perceived life impact of stuttering (Manning & Beck, 2013) and is associated with greater risk of posttreatment relapse (Craig & Hancock, 1995; Iverach, Jones, et al., 2009). Our finding that CWS showed AB effects for threat stimuli with general or personal relevance to stuttering suggests that anxiety-like tendencies (e.g., worry, anticipatory concerns) in this age group may, at least initially, be limited to their stuttering experiences and concerns. Like Haft et al. (2019), who reported similar content-specific effects for words related to reading in children with learning disabilities, CWS perceived ST and PT words as threatening because there was more sensitivity surrounding these words but not words related to social or general threat (e.g., lonely, bomb). It is unclear whether adolescents and adults would show similar content-specific AB patterns or whether their attentional performance more closely resembles anxious individuals. Based on available studies of AB (particularly Bauerly, 2022; Rodgers et al., 2020), threat-related AB can be elicited for a broader range of threat stimuli in adolescents and adults who stutter. The different patterns of AB in children and adults who stutter, along with more variable associations between anxiety and stuttering in younger relative to older speakers who stutter (Alm, 2014; Smith et al., 2014), suggests that stuttering-related anxiety may initially emerge in a more focused form but became more general and consistent over time. Our results underscore the need for meaningful interventions that can address early forms of anxiety before they impact children more broadly.

It is also worth noting that AB effects in the ST condition were particularly compelling, given that these stimuli were generally neutral in valence (e.g., speech = 5.62; overall mean valence for list = 5.36) and were paired with stimuli that had equivalent valence values. If perceived as threatening to CWS, this response must reflect participants' personal associations with these concepts, as the word pairs were matched on all other relevant parameters and were therefore essentially neutral−neutral pairs. Unsurprisingly, group differences in AB were more robust for ST relative to PT words. Like ST stimuli, the intent of the PT stimuli was to isolate threats related to stuttering concerns; however, because they were personalized, PT words spanned a broader range of valence values and some did not have associated valences (e.g., Mississippi). Overall mean valence for PT words (across participants) was 6.96, whereas matched stimuli, which were all drawn from the neutral range (valence values 4–6; Warriner et al., 2013), had a mean valence of 4.7. These differences between stuttering-specific threat stimuli may explain why AB effects were stronger for ST relative to PT words (see Limitations section for additional elaboration on this point).

Theoretical Implications

Our findings suggest that attention may play an important role in linking factors known to contribute to stuttering. As proposed by contemporary multifactorial theories of stuttering (Smith & Weber, 2017), stuttering arises from dynamic interactions among multiple factors (e.g., cognitive, emotional, genetic, motor, environmental). Evidence from the broader AB literature indicates that threat-related attention moderates the relationship between vulnerable temperament traits and anxiety outcomes, such that children with high levels of behavioral inhibition are more likely to develop anxiety when they also show a strong threat bias (Valadez et al., 2022). EC (or attentional control more generally) has been identified as a protective factor that can help children override threat-related AB and shield them from developing emotional disorders. In this way, attention plays a critical role in framing a child's social world for processing and shaping how they respond to their environments. For CWS, this gate-keeping function of AB may determine how emotional factors, such as innate vulnerabilities based on temperament, and environmental factors, in the form of stuttering experiences, interact over time and whether temperamental characteristics ultimately manifest in emotional psychopathology. Pérez-Edgar and colleagues (Pérez-Edgar et al., 2017) aptly described attentional mechanisms as a “developmental tether” that keeps vulnerable children on a trajectory toward clinical anxiety, in place of ameliorative processes that normally smooth away these early risks. It is also worth noting that according to some viewpoints (e.g., Eysenck et al., 2007), the presence of anxiety disrupts attention control and increases the influence of stimulus-driven (i.e., bottom-up) attentional systems; thus, the relationship between aspects of attention and emotion processing may be bidirectional. Further research is needed to clarify the nature and role of AB in stuttering and to determine whether AB in individuals who stutter is primarily associated with overt stuttering characteristics (observable behaviors), covert features (perceived impact, psychological reactions to stuttering, communication attitude), or both.

Clinical Implications

Negative perceptions of stuttering among CWS can be heavily influenced by responses and reactions to their stuttering from caregivers, siblings, peers, teachers, and others. Experiences of teasing, bullying, or other forms of public stigma may be internalized by CWS (Boyle et al., 2023) and reinforce patterns of selective attention to stuttering-related threats, thereby feeding psycho-emotional responses to stuttering in a cyclical manner. In this way, allies of people who stutter, therapists, friends, and all listeners can play an important role in shaping how stuttering is experienced, how the speaker's attentional patterns are influenced by stuttering experiences, and the types of emotional sequelae that develop over time.

Many researchers are also exploring the clinical utility of treatment approaches to modify threat-related AB, particularly when attentional patterns are associated with clinical symptoms (Bar-Haim, 2010; Gober et al., 2021; Mogg & Bradley, 2016). Such interventions may hold promise as a cost-effective and effective way of retraining attentional biases to encourage more resilient and adaptive responses to stuttering experiences. Further research is needed to pinpoint precise components of AB (e.g., vigilance, difficulty disengaging, avoidance) affected in speakers who stutter and develop targeted interventions to retrain underlying cognitive mechanisms.

Limitations

Several limitations of the study warrant consideration. First, we acknowledge that although online data collection provided us with access to a larger sample size and to participants spanning a broader geographic region, it may have introduced more variability in our data.

A number of additional limitations relate specifically to the PT condition. Although we observed the expected effect for PT stimuli, procedures for selecting these words and their neutral matches might not have been optimal. Because we anticipated that many elicited words would not have associated valence values (e.g., names of people or places), our planned procedures limited potential match words to a subset of stimuli from the Warriner database with neutral valence values (between 4 and 6). Our final PT list included many words without valence values, as expected, but a considerable portion of elicited words did have associated values and these ranged from 2 (pollution, negatively valenced) to 8.11 (excited, positively valenced). Our resulting pairs therefore included some PT words that differed in valence from the neutral matches in either direction. This inconsistency may have contributed to the weaker AB effect observed for PT relative to ST words. PT stimuli might also not have been associated with sufficient experience to elicit a strong AB effect, given that some children were still quite young. Although children do experience stuttering anticipation (Jackson et al., 2018) and were all able to report feared words in our study, it is possible that the words were not as strongly associated with stuttering experiences as they might be in adults. We also had no way of objectively verifying the validity of reported words. We followed an established protocol for eliciting words on which participants anticipated stuttering (Goldfarb et al., 2023) and the notion of feared words is well accepted by researchers and clinicians (Bloodstein, 1960; Bowers et al., 2012; Vanryckeghem et al., 2004); however, findings would be strengthened with evidence indicating that reported words were in fact feared by the CWS. Last, we provided control participants a randomized subset of words elicited from CWS based on a cumulative, running list. It is possible that this approach resulted in threat stimuli that always had personal relevance to individual CWS but not to CWNS. Eliciting random words that simply “came to mind” for CWNS may have resulted in more comparable PT stimuli for this group that better controlled for their potential personal salience.

Several recent papers have also called the reliability of the DPT into question, with concerns related to internal consistency and test–retest reliability (Schmukle, 2005; Staugaard, 2009; Van Bockstaele et al., 2014). The task also cannot fully distinguish between vigilance (automatic capturing of attention) and slow disengagement. Other AB tasks or additional manipulations within the DPT, may be better able to pinpoint roles of specific AB components and their associated cognitive mechanisms. Finally, although the DPT can yield interesting insights regarding the end points of attention, it does not provide dynamic, time-course details related to attention. Future studies incorporating eye tracking methodologies in the DPT or other AB paradigms may yield more fine-grained information about dynamic aspects of attention during individual trials.

Conclusions

Overall, results extend our understanding about the role of attention in stuttering, providing new data that indicate a pattern of selective attention toward threat-related stimuli among CWS but not CWNS. CWS demonstrated this pattern on stimuli related to stuttering and on elicited words reflecting personal stuttering-related concerns. These disorder-congruent effects suggest that stuttering-related anxiety in CWS may initially be quite focused and can be distinguished from AB to general social or physical threats that characterize social and general anxiety. Additional studies are needed to clarify the extent to which innate characteristics such as NA and EC contribute to the development of AB in CWS, and extent to which AB predicts overt and covert features of stuttering. Future research exploring cognitive interventions that strengthen top-down attentional processes or that directly reshape AB patterns may introduce novel approaches for targeting cognitive contributors to stuttering and optimizing socioemotional outcomes in CWS.

Data Availability Statement

The data sets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Funding Statement

This research was supported, in part, by a National Institute on Deafness and Other Communication Disorders R21 Early Career Research Grant (PAR-21-107) and School of Communication Sciences and Disorders Faculty Research Grant to the first author.