Eyewitness testimony by individuals who stammer: Evidence, experience and perceived credibility · Maras K, Park S, Grafton P et al.

STUDY 2

When considering witness credibility, jurors often (erroneously) rely on perceived behavioural indicators of veracity and truthfulness to inform their judgements (Vrij & Turgeon, 2018). Witnesses are typically also judged on their ability to testify (Cuddy et al., 2009; Neal et al., 2012), how likeable they appear (e.g. how friendly, respectful, well mannered and pleasant) and how confident or certain a witness is regarding the accuracy of their testimony (McClure et al., 2013; Wixted et al., 2015).

Various elements of speech, including fluency, tone, pacing and emotional expression, can shape a jury's perception of the witness (Cramer et al., 2009; Iudici et al., 2023). For example, witnesses who speak ‘powerlessly’, with pauses, hesitations and filler words, are often rated as less credible than witnesses who speak ‘powerfully’ and fluently (Gojkovich et al., 2019; Ruva & Bryant, 2004). Although no research to date has examined juror perceptions of PWS specifically, research indicates that other groups with speech and communication differences, such as individuals with developmental language disorder (Hobson et al., 2022), intellectual disability (Henry et al., 2011) and autism (Lim et al., 2022; but see Maras et al., 2019), are often rated as less credible, even when they are accurately recalling events.

The current study examined the perceived credibility of PWS. Perceived speech quality was also assessed to determine whether credibility perceptions are associated with perceived differences in the fluency and quality of their speech. Based on previous research with other populations, it was predicted that PWS would be perceived as less credible by mock jurors than witnesses who do not stammer. We also predicted that PWS would be scored lower on speech quality than witnesses who do not stammer. After the scale ratings, mock jurors were asked open‐ended explanatory follow‐up questions to contextualize their responses, including whether/how their perceptions changed after being provided with information about stammering.

Study 2: Method

Design

One‐hundred‐and‐fifty mock juror participants were randomly allocated to view and rate one of 30 videos from Study 1 of either a PWS (n = 15) or a non‐stammering witness (n = 15). Half (n = 75) of jurors viewed a PWS, and the other half (n = 75) viewed a non‐stammering witness (between participants). This resulted in each of the 30 videos being viewed by five different jurors. The dependent variables were the four subscales of perceived credibility and seven subscales of speech quality. Additionally, mock jurors provided free text responses explaining their reasoning for the credibility and speech scale ratings, which were subject to an inductive content analysis.

The study design, hypotheses and planned analyses were pre‐registered on the Open Science Framework (https://osf.io/wfb2z?view_only=2640b4aaf6da480c9de1a6803523f0a5).

Participants

An a priori power analysis using G*Power version 3.1.9.7 (Faul et al., 2007) indicated that a total sample size of 126 participants would give 80% power to detect a medium‐to‐large effect size (Cohen, 1988). One hundred and fifty mock juror participants took part via QuestionPro or subsequently (due to a technical issue) Qualtrics. They were recruited through Prolific.com using convenience sampling and were paid according to Prolific's payment principles (approximately £4.50). To be eligible for the study, participants were required to be age 18 or over, able to read, write and understand English as a first language or equivalent, and reside in the United Kingdom. Due to an oversight in survey development, demographic information was not collected.

Ethical approval was obtained from the University of Bath Social Science Research Ethics Committee (reference number: 5309‐5733).

Materials and procedure

Eyewitness testimony videos

A total of 30 videos (15 PWS and 15 comparison witnesses) were selected from the wider Study 1 sample of 26 PWS and 41 comparison witnesses. Multiple interview videos were used to capture a diverse range of stammers, while trimming witness groups from the wider Study 1 sample to ensure they were matched more closely on age, gender and ethnicity.1 All 30 witnesses gave consent for their videos to be used in the current study. The videos were edited to contain only the free recall phase of the interview (which lasted between 2 and 5 min).

Mock juror participants were randomly allocated to watch a video of either a PWS or non‐stammering witness. They were instructed to watch the entire video and have the page set to full‐screen mode.

Witness credibility scale

Perceived credibility was assessed using a 20‐item adapted version of the ‘Witness Credibility Scale’ (Brodsky et al., 2010). Each item contains bipolar adjectives that are rated on a 10‐point Likert scale (e.g. unreliable [1] to reliable [10]). A score of 1 indicates a strong agreement with the negative attribute while a score of 10 indicates a strong agreement with the positive attribute. The scale measures credibility across four subscales (Likeability, Trustworthiness, Confidence and Knowledge), each made up of five items. As the scale was originally developed for expert witnesses, four items were modified and the Knowledge subscale was renamed Capability (see Supplementary Material D). Subscale scores (out of 10) are calculated by averaging the ratings of the five corresponding items. The Witness Credibility Scale has good reliability (α = .95), and the four subscales yield consistently high overall internal consistency in credibility scores (.91–.98; Brodsky et al., 2010).

Two attention check questions were included in the Witness Credibility Scale, asking participants to select a specific Likert score and submit a specified word for the free‐text follow‐up. If participants failed to submit the correct score or word, they were removed from the dataset.

Speech quality

The ‘Public Speaking Scale’ (Iberri‐Shea, 2017) is an 11‐item measure that asks respondents to rate various indicators of an individual's public speaking ability (e.g. projection). Items are rated on a 5‐point Likert scale (1 = least favourable evaluation of the speaker; 5 = most favourable). For each item, descriptions of the five points are provided to help guide the rater's decision. Four items were removed from the scale (Introduction, Body, Conclusion and Topic Choice) as they were not relevant to the aims of the current study (see Supplementary Material E). An overall averaged score between 1 and 5 could be obtained, with 5 indicating the highest speech quality. The Public Speaking Scale demonstrates high inter‐rater reliability for measuring overall public speaking ability (α = .93), with high inter‐rater reliability for the individual indicators (.87 to 1.0; Iberri‐Shea, 2017).

After completing the scales, jurors in the PWS condition answered open‐ended free text questions probing whether they noticed anything about the witnesses' speech or communication style. They were then informed that the witness had a stammer and were provided information about stammering, taken from STAMMA's website. This included a definition, common characteristics and strategies that PWS often use to mask their stammering (see Supplementary Material F). After reading the information about stammering, participants were asked a final free text question about whether/how this information changed their perception of the witness.

Study 2: Results

Perceived credibility

To examine differences between groups in perceived credibility scores, a one‐way MANOVA was conducted, with type of video shown (PWS vs no stammer) as the independent variable and the composite juror scores for confidence, likeability, trustworthiness and capability as dependent variables. Table 3 displays the mean scores for each of the four perceived credibility outcome variables.

Table. Mean credibility and speech quality scores within each witness video condition (standard deviations in parentheses). (See the original article for the full table.)

There was a significant multivariate effect of type of witness video viewed on the combined confidence, likeability, capability and trustworthiness scores, F(4, 145) = 4.39, p = .002, ηp 2 = 0.11. However, separate univariate tests on the outcome variables revealed no significant effects of witness group on confidence, F(1, 148) = 3.81, p = .053, ηp 2 = 0.03, likeability, F(1, 148) = 3.02, p = .084, ηp 2 = 0.02, capability, F(1, 148) = 0.14, p = .711, ηp 2 = 0.01 or trustworthiness, F(1, 148) = 1.94, p = .165, ηp 2 = 0.01. Although not pre‐registered, we also modelled each measure of perceived credibility using a linear mixed‐effects analysis to account for variation across videos. These models included a fixed effect for type of video (coded PWS = 0.5, no stammer = −0.5) and a random intercept for video. Models were fit in R (R Core Team, 2024, Version 4.3.3) using the lme4 package (Bates et al., 2015, Version 1.1–35.5), and degrees of freedom were approximated using Satterthwaite's method via the lmerTest package (Kuznetsova et al., 2017, Version 3.1.3). These analyses led to the same conclusions for confidence, B = −0.53, t(28) = −1.57, p = .129, likeability, B = 0.30, t(28) = 1.72, p = .096, capability, B = 0.09, t(28) = 0.32, p = .755 and trustworthiness, B = 0.33, t(148) = 1.39, p = .165.

The MANOVA was followed up with discriminant analysis, which revealed one discriminant function. This explained 100% of the variance, canonical R 2 = .11. This discriminant function significantly differentiated the type of witness video viewed between groups, Λ = .89, c 2(4) = 16.70, p = .002. The correlations between outcomes and the discriminant function revealed that likeability (r = .41) and trustworthiness (r = .33) both loaded moderately onto the function. Capability loaded weakly onto the function (r = .09), while confidence loaded moderately but negatively onto the function (r = −.46). This indicates that the witness groups were distinguished by a variate that has opposite effects on different dimensions of credibility. Considering the means from Table 3 and the correlations between outcomes and the discriminant function, it appears that PWS were viewed as less confident but more likeable and trustworthy. Perceptions of capability were similar between witness groups.

Speech quality

To examine differences between groups in perceived speech quality, a second one‐way MANOVA was conducted. Type of video shown (PWS vs no stammer) was the independent variable, and the juror scores for projection, pace, intonation, diction, language use, vocabulary and purpose were dependent variables. There was no significant multivariate effect of witness video viewed on speech quality, F(7, 142) = 1.54, p = .158, ηp 2 = 0.07 (Table 3).

Relationship between overall speech quality and perceived credibility

A simple linear regression analysis found that overall speech quality was positively associated with higher ratings of overall witness credibility, B = 0.74, β = .54, t(148) = 7.80, p < .001, r 2 = .29. A linear mixed‐effects model predicting perceived credibility from overall speech quality, with a random intercept for video, revealed the same conclusion: B = 0.77, t(142.89) = 7.90, p < .001. We also tested a model that included a random slope by video for the relationship between credibility and speech quality, but it was flagged as a singular fit, indicating that the random slope variance was near zero and did not meaningfully vary across videos.

Content analysis of responses to free‐text questions

The free‐text responses to the final open‐ended questions from jurors who viewed a PWS video were analysed by the first and fifth authors using an inductive qualitative content analysis, following Mayring's (2015) guidelines.

Speech distinctions

When reflecting on whether they noticed anything about the witness's speech or communication style (prior to being informed the witness had a stammer), jurors frequently mentioned the stammer itself (with some referring to it as ‘delayed’ or ‘awkward’). Few mentions were made of other speech aspects, such as nervousness and pauses (see Table 4).

Table. Code frequencies and example quotes for the open‐ended question on noticeable speech distinctions in PWS group. (See the original article for the full table.)

Effect of receiving information about stammering

Finally, analysis of jurors' free text responses after they had received knowledge about the witness's stammer indicated they felt that this information generally improved their credibility perceptions of the witness. Specifically, jurors indicated that they realized that the stammering was not due to nervousness or a lack of credibility but a condition that the witness was struggling with. Jurors also reported becoming more empathetic and understanding of the witness's mental and emotional state due to their stammer (see Table 5).

Table. Code frequencies and example quotes for the open‐ended question on effect of receiving stammer information on perceptions of PWS. (See the original article for the full table.)

Study 2: Discussion

Study 2 examined the perceived credibility and speech quality of PWS and witnesses with no stammer. Mock juror ratings of witness credibility (measured on confidence, likeability, trustworthiness and capability) were significantly affected by the type of witness video they viewed; however, not as initially expected. PWS were perceived as less confident, but more likeable and trustworthy than witnesses who do not stammer, while perceived capability did not differ between groups. The difference in perceived confidence does, however, mirror Study 1 findings, whereby PWS reflected how concerns about others' perceptions of them diminished their confidence in their own ability to provide testimony. This is consistent with the notion that anticipated negative evaluation can lead to internalized self‐stigma in PWS (Fitzgerald et al., 2021; Nang et al., 2018).

Speech quality was positively associated with higher ratings of overall perceived credibility but, perhaps critically, mock juror ratings of speech quality did not significantly differ between PWS and comparison witnesses, which may explain the inconsistent findings regarding credibility. Nevertheless, juror participants' responses to the open‐ended follow‐up questions indicated that they did notice witnesses' stammers, but that formal disclosure of the stammering would have improved their credibility ratings of them.