Nadine Lavan

While a reader might imagine the narrator's and characters' voices in a written book, a spoken narrative solidifies the relationship between the narrator and the text. Here, we report two experiments investigating how perceived narrator-to-story fit affects narrative listening enjoyment and other listening outcomes. In a first experiment (N=400 participants), we manipulated the gender-related congruence between the expected and heard narrator of four spoken narratives in a willingness-to-wait paradigm. We found that listeners were more likely to continue listening to congruent than incongruent stories, but did not enjoy them more or comprehend them more accurately. However, we observed a dissociation in the observed effects of perceived narrator-to-story fit: specifically, higher individual evaluations significantly increased enjoyment, while both higher individual and (aggregated) mean fit ratings predicted greater motivation to continue listening. This suggested that crowd-sourced data may be less effective than personal taste when predicting narrative listening enjoyment. Indeed, a second experiment using a larger number of stories showed that participants’ enjoyment of spoken narratives is most strongly predicted by within-sample individual fit ratings, but not by within- or between-group aggregate ratings. Our findings provide evidence to support personalisation of narration as a way to promote enjoyment of spoken narratives and, potentially, audiobooks.

Auditory verbal hallucinations are often experienced as the voices of entities who have a character and personality of their own. We asked whether hallucination-proneness in listeners affects how they evaluate the personality of external voices. 48 participants who self-reported a diagnosis of a psychotic disorder, and 50 participants without a similar diagnosis rated brief vowel recordings on attractiveness, competence, dominance, friendliness and trustworthiness. Hallucination-proneness was measured with the Cardiff Anomalous Perceptions Scale (CAPS) and the auditory hallucinations subscale of the Psychotic Symptom Rating Scales (PSYRATS-AH). Despite clearly higher symptom levels in listeners diagnosed with psychotic disorders, there was no group-level difference in average social trait evaluations, nor were average ratings associated with hallucination-proneness. There was, however, a numerical trend for listeners diagnosed with psychotic disorders to report more extreme social trait evaluations, where additionally, greater hallucination-proneness was also significantly associated with more extreme ratings across all social traits, on both measures. This association was, however, primarily present in controls but not in listeners diagnosed with psychotic disorders. Hallucination-proneness therefore appears to amplify, rather than directionally bias, social trait evaluations of voices. Our findings link proneness to experiencing internal voices, i.e. auditory verbal hallucinations, to how external voices are perceived. This finding is broadly consistent with previous work reporting that hallucination-prone individuals may be more likely to detect a signal (here: a salient social evaluation) in relatively ambiguous stimuli, speaking to mechanistic accounts that propose that stronger priors imbue stimuli with meaning.

Humans often struggle to accurately determine whether two different unfamiliar voice samples or face images belong to the same person. One method shown to boost face-matching performance is to have participants perform the task in pairs, which results in subsequent improvements for the poorer performers. This is a phenomenon known as ‘the pairs training effect’. Previous studies have explored the pairs training effect using simultaneous face matching tasks whereby the two face images are presented side by side at the same time. Here we tested the efficacy of the pairs paradigm in unfamiliar voice matching (Experiment 1) which requires sequential presentation of the voices. We also tested sequential face matching (Experiment 2) to align the face task with the temporal demands of the voice task. In both experiments, participants completed one third of trials individually, one third in a pair, and a final third individually again. Across both tasks, we found that participants who initially performed poorly in the first individual task improved when completing the task in a pair and, importantly, retained this improvement in the final individual phase. Our findings show that the pairs training effect extends to identity matching of unfamiliar voices and remains even when stimuli are presented sequentially.

When we hear a voice, we routinely form an impression of the speaker’s personality and characteristics. Previous work has primarily modelled first impression formation as a bottom-up, feed-forward process, where acoustic cues drive how different person characteristics, such as trustworthiness or competence, are perceived from unfamiliar voices. Contributions from top-down processes to impression formation, such as the influences of social stereotypes and other conceptual beliefs, have to date received less attention in voice research. Here, we quantify how much such top-down processes could contribute to first impression formation by mapping representational spaces for (1) first impression judgments from voices and (2) conceptual beliefs about person characteristics in other people in a behavioural study. Using Representational Similarity Analysis (RSA), we observe substantial structural similarities between these spaces. This association shows that conceptual beliefs could play a role in shaping impressions, via top-down influences. This effect generalises across voice types (e.g., typical, AI-generated, pathological voices), listener samples, and even across modalities to facial impressions. Aligning with recent theoretical advances in social psychology, our findings thus call for vocal impression research to start to explicitly model and explore the interplay between top-down and bottom-up processes during impression formation.

In perception experiments, researchers often collect multiple perceptual judgements from the same stimuli and participants to answer their research questions. While the way in which perceptual judgements are collected may affect the data, research rarely considers task design when investigating this topic. We therefore investigated how two frequently used methods of collecting multiple ratings can affect perceptual judgements, focusing on first impressions of faces. One participant group provided ratings of seven person characteristics (e.g. femininity, youthfulness, trustworthiness) blocked by characteristic, rating faces for one person characteristic at a time, across seven blocks. Another participant group completed one block in which they rated all seven characteristics simultaneously via a list of rating scales. The listed presentation format reduced task duration by 22%, but affected the perceptual ratings in several ways, pointing to reduced data quality, potentially as a result of satisficing behaviours. Specically, inter-rater agreement was lower for some person characteristics (youthfulness, femininity, and trustworthiness) in the listed format. Variance in ratings was also reduced for youthfulness, femininity, and dominance, with ratings clustering closer to the middle of the scale. Importantly, correlations between different person characteristics became universally positive in the listed format, indicating reduced independence of judgements. These findings highlight that while presenting multiple judgements at the same time may offer efficiency, this approach can introduce systematic biases and potentially reduce the reliability of perceptual data. We therefore suggest using a blocked presentation format and consider how these trade-offs would impact experiments looking at multiple perceptual judgements collected from the same participants.

In perception experiments, researchers often collect multiple perceptual judgements from the same stimuli and participants to answer their research questions. While the way in which perceptual judgements are collected may affect the data, research rarely considers task design when investigating this topic. In our experiment, we investigated how two frequently used ways of collecting multiple ratings for affect perceptual judgements, focusing on first impressions of faces. One participant group provided ratings of seven person characteristics (e.g. femininity, youthfulness, trustworthiness) blocked by characteristic, rating faces for one person characteristic at a time, across seven blocks. Another participant group completed one block in which they rated all seven characteristics simultaneously via a list of rating scales. The listed presentation format reduced task duration by 22%, but affected the perceptual ratings in several ways, pointing to reduced data quality. Inter-rater agreement was lower for some person characteristics (youthfulness, femininity, and trustworthiness) in the listed format. Variance in ratings was also reduced for youthfulness, femininity, and dominance, with ratings clustering closer to the middle of the scale. Importantly, correlations between different person characteristics became universally positive in the listed format, indicating reduced independence of judgments. These findings highlight that while presenting multiple judgements at the same time may offer efficiency, this approach can introduce systematic biases and potentially reduce the reliability of perceptual data. We therefore suggest using a blocked presentation format and consider how these trade-offs would impact experiments looking at multiple perceptual judgements collected from the same participants.

Listeners rapidly form first impressions from voices, inferring multiple person characteristics within milliseconds. We employed a novel method, Self-Steered Sound Synthesis (S4), to identify and compare the acoustic signatures underlying these impressions. Participants interactively used S4 to synthesise voices expressing six person characteristics - age, masculinity, health, attractiveness, dominance, and trustworthiness - by manipulating four perceptually salient acoustic dimensions: mean pitch, pitch excursion, breathiness, and formant spacing. Masculinity, older age, and dominance were conveyed by lowering mean pitch and formant spacing, consistent with projecting the impression of a large person, and flattening the intonation. Physical health, attractiveness, and trustworthiness were conveyed by choosing less extreme and more "typical" acoustic properties. A second perceptual experiment confirmed that the synthesised voices from Experiment 1 indeed conveyed the intended person characteristics to an independent sample of listeners, and that listeners relied on similar acoustic cues for their evaluations. From a methodological perspective, we draw similar conclusions from two drastically different approaches, thus providing a comprehensive account of first impression formation that bridges across voice production (or synthesis) and perception. Our converging findings furthermore demonstrate the methodological robustness of S4. From a theoretical perspective, our findings furthermore extend theoretical frameworks that conceptualise first impressions within a continuous “trait space”, highlighting the graded and intercorrelated nature of different person characteristics on a perceptual and conceptual level. We extend this framework by showing that not only perceptual judgments, but also the acoustic signatures of person characteristics show intercorrelations, thus integrating acoustic cues into perceptual models of voice perception.

Hearing our own voice is commonly assumed to be an unpleasant experience. In contrast, some studies of voice perception have reported that we actually find our own voice to be more attractive than other voices. It remains however unclear whether previously observed enhancements of self-voice attractiveness were mediated by the recognition of the self-voice during the task, because recognition was never explicitly and rigorously examined in previous studies. We therefore conducted a voice attractiveness rating task under conditions that limited the likelihood of self-voice recognition, in order to directly compare the perceived attractiveness of one’s own voice when recognised versus when not. We found that both participants who did (N=59) and did not (N=80) recognise their own voice rated their own voice differently from other voices. However, crucially, the direction of this difference depended on whether the self-voice was recognised during the task or not: listeners who spontaneously recognised their voice rated it as more attractive than other voices, whereas listeners who did not recognise their voice rated it as less attractive. The findings help to explain one source of contradictory perspectives in the voice perception literature and beyond. highlighting the intricate interplay between self-perception, first impressions, and identity perception.