About

Arthur Samuel is a researcher whose work focuses on speech perception, auditory processing, and psycholinguistics. His research explores how listeners perceive, process, and recognize spoken language, including the effects of dialect, accent, lexical tones, and phonemic variation. He investigates the cognitive and perceptual mechanisms underlying speech recognition, including the influence of social and contextual factors, as well as the neural processes involved in auditory and linguistic processing. Throughout his career, Samuel has contributed to understanding the role of knowledge-based expectations in music perception, the use of rhythm in speech attention, and the perceptual adjustments listeners make to speaker variability. His work also examines how visual cues, such as lip reading, interact with auditory information, and how speech perception is affected by factors like ethnicity, musicianship, and language experience. Samuel's research has implications for models of spoken word recognition, language learning, and social communication, and he has published extensively on these topics, advancing the scientific understanding of auditory and linguistic cognition.

Research topics

• Computer Science
• Psychology
• Artificial Intelligence
• Natural Language Processing
• Cognitive psychology
• Linguistics
• Developmental psychology
• Speech recognition
• Communication

Selected publications

• How sleep redraws phonemic categories after auditory selective adaptation

  Psychonomic Bulletin & Review · 2026-01-01

  articleOpen accessSenior author

  After information has either been perceived or brought into working memory from long-term memory, it may remain active for hours or days. There is extensive evidence that sleep can consolidate newly learned material into long-term memory, and some recent work shows that sleep may also help clear out either unneeded or already established information. We examine the effect of sleep on a third type of information: adjustments to established speech categories caused by repeated exposure to a speech sound-selective adaptation. We find that sleep does not consolidate selective adaptation per se. Instead, sleep implements a change in phoneme category frequency to reflect the properties of the input-the many instances of the adapting sound that had been presented repeatedly. While adaptation temporarily reduces the perception of tokens similar to the repeating sound, sleep increases their perception, producing a "reverse adaptation" pattern. The results constrain models of phoneme category adjustment, favoring those that have separate mechanisms for assimilative versus contrastive effects over those with a single mechanism for both types of effects.

  PublisherOA PDFDOI

• Speech perception consistency facilitates initial lexical activation, but not speech perception flexibility

  Scientific Reports · 2026-04-09

  articleOpen access

  Speech perception consistency refers to the similarity between a listener’s responses to repeated presentations of the same speech sound (e.g., along a /ba/–/pa/ continuum). Although research has demonstrated multiple advantages of consistency, its role in initial lexical activation (i.e., early activation of word candidates) and speech perception flexibility (e.g., the ability to recover from misleading speech input) has not been tested. We investigated the role of consistency in spoken-word recognition among Spanish (L1)–English (L2) bilingual listeners. Focusing on the bilabial stop contrast (/b/–/p/), consistency was measured in a task where participants repeatedly rated speech sounds on a continuous scale (Visual Analog Scale), whereas initial lexical activation and speech perception flexibility were assessed using a word-to-picture matching task with eye-tracking (Visual World Paradigm). Seventy Spanish–English bilinguals completed these tasks in both languages. Listeners with higher consistency exhibited stronger initial lexical activation for acoustically compatible word candidates across /b/ and /p/ in L1 Spanish and for /p/ in L2 English. However, consistency was not associated with speech perception flexibility. These findings suggest that speech perception consistency primarily supports early lexical activation during spoken-word recognition, while playing a more limited role in later processes involved in revising initial misinterpretations.

  PublisherDOI

• The Role of Speech Perception Gradiency in L1 versus L2 Spoken Word Recognition

  2025-05-07

  preprintOpen access

  Speech perception gradiency reflects sensitivity to subphonemic differences. Prior research has shown that gradiency facilitates recovery from misperceptions (i.e., speech perception flexibility) in L1 (Kapnoula et al., 2021), but whether and how gradiency contributes to speech perception flexibility in L2 remains unknown. This study investigated the role of gradiency in spoken word recognition among Spanish (L1)-English (L2) bilinguals. Gradiency was assessed using a Visual Analogue Scale (VAS) with stop consonants (/b/-/p/), and initial lexical activation and speech perception flexibility were assessed using an eye-tracking Visual World Paradigm (VWP) task. Seventy Spanish-English bilinguals completed these tasks in both languages. Following previous results in L1 English, gradiency facilitated speech perception flexibility in L1 Spanish. In contrast, gradiency did not facilitate L2 speech perception, and a different pattern emerged: participants relied more heavily on lexical (top-down) than subphonemic (bottom-up) information, especially lower-L2-proficiency listeners. Additionally, a positive correlation between L1 and L2 gradiency was observed only among higher-proficiency listeners. Overall, these findings suggest that the functional role of gradiency in L1 versus L2 speech perception is modulated by the differential reliance on bottom-up versus top-down information.

  PublisherDOI

• The Role of Speech Perception Gradiency in L1 versus L2 Spoken Word Recognition

  2025-11-20

  articleOpen access

  Speech perception gradiency reflects sensitivity to subphonemic differences. Prior research has shown that gradiency facilitates recovery from misperceptions (i.e., speech perception flexibility) in L1 (Kapnoula et al., 2021), but whether and how gradiency contributes to speech perception flexibility in L2 remains unknown. This study investigated the role of gradiency in spoken word recognition among Spanish (L1)-English (L2) bilinguals. Gradiency was assessed using a Visual Analogue Scale (VAS) with stop consonants (/b/-/p/), and initial activation of a lexical competitor and speech perception flexibility were assessed using an eye-tracking Visual World Paradigm (VWP) task. Seventy Spanish-English bilinguals completed these tasks in both languages. Following previous results in L1 English, gradiency facilitated speech perception flexibility in L1 Spanish. In contrast, gradiency did not facilitate L2 speech perception; instead, a different pattern emerged: participants relied more heavily on lexical (top-down) than subphonemic (bottom-up) information, as would be expected given the less robust category representations in L2. Additionally, a positive correlation between L1 and L2 gradiency was observed only among higher-proficiency listeners. Overall, these findings suggest that the functional role of gradiency in L1 versus L2 speech perception is modulated by the differential reliance on bottom-up versus top-down information.

  PublisherDOI

• The Role of Speech Perception Gradiency in L1 versus L2 Spoken Word Recognition

  2025-11-20

  articleOpen access

  Speech perception gradiency reflects sensitivity to subphonemic differences. Prior research has shown that gradiency facilitates recovery from misperceptions (i.e., speech perception flexibility) in L1 (Kapnoula et al., 2021), but whether and how gradiency contributes to speech perception flexibility in L2 remains unknown. This study investigated the role of gradiency in spoken word recognition among Spanish (L1)-English (L2) bilinguals. Gradiency was assessed using a Visual Analogue Scale (VAS) with stop consonants (/b/-/p/), and initial activation of a lexical competitor and speech perception flexibility were assessed using an eye-tracking Visual World Paradigm (VWP) task. Seventy Spanish-English bilinguals completed these tasks in both languages. Following previous results in L1 English, gradiency facilitated speech perception flexibility in L1 Spanish. In contrast, gradiency did not facilitate L2 speech perception; instead, a different pattern emerged: participants relied more heavily on lexical (top-down) than subphonemic (bottom-up) information, as would be expected given the less robust category representations in L2. Additionally, a positive correlation between L1 and L2 gradiency was observed only among higher-proficiency listeners. Overall, these findings suggest that the functional role of gradiency in L1 versus L2 speech perception is modulated by the differential reliance on bottom-up versus top-down information.

  PublisherDOI

• Perceptual Rhythmic Ability Predicts Cognitive Load in Perception of Accented Second Language Speech

  2025-11-20

  articleOpen access

  Languages differ in their rhythmic properties; for instance, English has higher temporal variability (is more stress-timed) than Spanish. Although previous research has shown that perceptual rhythmic ability can influence the intelligibility of accented speech in L1 English listeners, it remains unclear whether this benefit extends to L2 listening and to cognitive load. Therefore, we examined how perceptual rhythmic ability predicts cognitive load during listening in L1 Spanish–L2 English bilinguals, while controlling for other factors that may influence it. Cognitive load was indexed via pupil size as participants listened to English sentences spoken with Spanish, American, and Mandarin accents. Overall, individuals with stronger perceptual rhythmic ability exhibited greater cognitive load when processing L2 English, for which rhythm diverges from Spanish, compared to those with weaker perceptual rhythmic ability. Notably, this difference was reduced in the American-accented English condition, suggesting that the rhythmic alignment between accent and target language facilitates processing.

  PublisherDOI

• The Role of Speech Perception Gradiency in L1 versus L2 Spoken Word Recognition

  2025-09-11

  preprintOpen access

  Speech perception gradiency reflects sensitivity to subphonemic differences. Prior research has shown that gradiency facilitates recovery from misperceptions (i.e., speech perception flexibility) in L1 (Kapnoula et al., 2021), but whether and how gradiency contributes to speech perception flexibility in L2 remains unknown. This study investigated the role of gradiency in spoken word recognition among Spanish (L1)-English (L2) bilinguals. Gradiency was assessed using a Visual Analogue Scale (VAS) with stop consonants (/b/-/p/), and initial activation of a lexical competitor and speech perception flexibility were assessed using an eye-tracking Visual World Paradigm (VWP) task. Seventy Spanish-English bilinguals completed these tasks in both languages. Following previous results in L1 English, gradiency facilitated speech perception flexibility in L1 Spanish. In contrast, gradiency did not facilitate L2 speech perception; instead, a different pattern emerged: participants relied more heavily on lexical (top-down) than subphonemic (bottom-up) information, as would be expected given the less robust category representations in L2. Additionally, a positive correlation between L1 and L2 gradiency was observed only among higher-proficiency listeners. Overall, these findings suggest that the functional role of gradiency in L1 versus L2 speech perception is modulated by the differential reliance on bottom-up versus top-down information.

  PublisherDOI

• Exposure to second language accent prompts recalibration of phonemic categories

  2025-03-14

  preprintOpen accessSenior author

  We examine how first language (L1) Spanish listeners with varying levels of experience with English recalibrate their phonemic category boundaries following exposure to second language (L2), American English-accented Spanish. Specifically, we examine changes to voice onset time (VOT) boundaries, which are often positively-shifted when produced by American English-accented Spanish speakers (as compared to L1 Spanish speakers). Our results demonstrate that listeners make adjustments to their phonemic category boundaries following exposure to accented words with the critical sounds in onset position (e.g., “bailar” and “parir”, meaning “to dance” and “to give birth”, for the /b/ and /p/ phonemic categories). In many cases generalization of phonemic learning was also observed, such that boundaries for categories that were not presented in training were also adjusted. Surprisingly, however, there were cases in which boundaries for trained categories did not show adjustments; for example, listeners trained with items for all places of articulation (POAs) showed recalibration of their bilabial and alveolar boundaries but not their velar boundary. Also notable was the role of the Spanish listeners' experience with English: More experienced listeners showed more positively-shifted (English-like) boundaries in the pre-test session. This suggests that more experienced listeners may have rapidly identified the American English-accented Spanish and applied their English category boundaries accordingly. We conclude that listener accommodation of L2 accent is supported by a phonemic recalibration mechanism, and that experience with the L1 of an L2-accented speaker facilitates rapid recalibration of phonemic categories.

  PublisherDOI

• Rhythmic Abilities and Cognitive Load in Perception of Accented Second Language Speech

  2025-07-02

  preprintOpen access

  Languages differ in their rhythmic properties; for instance, English has higher temporal variability than Spanish. While listeners with stronger nonnative accents may better understand speech with corresponding accents, it remains unclear whether this general benefit relates to suprasegmental aspects and whether it extends to reduced cognitive load. As such, we examined how perceptual rhythmic ability, production rhythmic ability, and temporal variability in English speech production predict cognitive load during listening in L1 Spanish–L2 English bilinguals. Cognitive load was indexed via pupil size as participants listened to English sentences spoken with Spanish, American, and Mandarin accents. Overall, individuals with stronger rhythmic abilities exhibited greater cognitive load when processing L2 English, whose rhythm diverges from Spanish, compared to those with weaker rhythmic skills. Notably, for perceptual rhythmic ability, this difference was reduced in the American-accented English condition, suggesting that the rhythmic alignment between the accent and the target language facilitates processing.

  PublisherOA PDFDOI

• Statistical learning subserves a higher purpose: Novelty detection in an information foraging system.

  Psychological Review · 2025-02-24 · 6 citations

  articleOpen access

  (IF) system. IF aims to detect and assimilate novel recurrent patterns in the input that deviate from randomness, for which SL supplies a baseline. The broad implications of this viewpoint and their relevance to recent debates in cognitive neuroscience are discussed. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

  PublisherDOI

Recent grants

• NIH Grant R01MH051663

  NIH · $2.5M · 2010

• NIH Grant R01HD059787

  NIH · $757k · 2012

Frequent coauthors

• Saioa Larraza

  University of the Basque Country

  60 shared

• Martijn Baart

  Tilburg University

  31 shared

• Efthymia C. Kapnoula

  Basque Center on Cognition, Brain and Language

  28 shared

• Mikhail Ordin

  26 shared

• Tanya Kraljic

  Google (United States)

  21 shared

• Donna Kat

  Stony Brook University

  19 shared

• Tiphaine Caudrelier

  Université Lumière Lyon 2

  18 shared

• Clara D. Martin

  Basque Center on Cognition, Brain and Language

  18 shared