The Duration of Auditory Sensory Memory for Vowel Processing: Neurophysiological and Behavioral Measures

Abstract

Speech perception behavioral research suggests that rates of sensory memory decay are dependent on stimulus properties at more than one level (e.g., acoustic level, phonemic level). The neurophysiology of sensory memory decay rate has rarely been examined in the context of speech processing. In a lexical tone study, we showed that long-term memory representation of lexical tone slows the decay rate of sensory memory for these tones. Here, we tested the hypothesis that long-term memory representation of vowels slows the rate of auditory sensory memory decay in a similar way to that of lexical tone. Event-related potential (ERP) responses were recorded to Mandarin non-words contrasting the vowels /i/ vs. /u/ and /y/ vs. /u/ from first-language (L1) Mandarin and L1 American English participants under short and long interstimulus interval (ISI) conditions (short ISI: an average of 575 ms, long ISI: an average of 2675 ms). Results revealed poorer discrimination of the vowel contrasts for English listeners than Mandarin listeners, but with different patterns for behavioral perception and neural discrimination. As predicted, English listeners showed the poorest discrimination and identification for the vowel contrast /y/ vs. /u/, and poorer performance in the long ISI condition. In contrast to Yu et al. (2017), however, we found no effect of ISI reflected in the neural responses, specifically the mismatch negativity (MMN), P3a and late negativity ERP amplitudes. We did see a language group effect, with Mandarin listeners generally showing larger MMN and English listeners showing larger P3a. The behavioral results revealed that native language experience plays a role in echoic sensory memory trace maintenance, but the failure to find an effect of ISI on the ERP results suggests that vowel and lexical tone memory traces decay at different rates.Highlights:We examined the interaction between auditory sensory memory decay and language experience.We compared MMN, P3a, LN and behavioral responses in short vs. long interstimulus intervals.We found that different from lexical tone contrast, MMN, P3a, and LN changes to vowel contrasts are not influenced by lengthening the ISI to 2.6 s.We also found that the English listeners discriminated the non-native vowel contrast with lower accuracy under the long ISI condition.

Formation and decay of auditory short-term memory in the macaque monkey

“…Whereas the substrate of EM remains unclear, the EEG literature has embraced the notion that certain EEG phenomena such as MMN or refractoriness are reliable markers of EM. This notion is so prevalent that properties of these EEG phenomena, such as their time constants, are used as estimates of the properties of EM itself (Bartha-Doering et al 2015;Grau et al 2001;Ioannides et al 2003;Yu et al 2018). Thus, it has been argued that the gradual strengthening of MMN with increased repetition of the standard reflects the gradual strengthening of EM.…”

“…Yu et al (2017) have studied the effect of long-term representation of lexical tone in native speakers of a tonal language They show that such long-term representations can slow the decay of MMN compared with speakers of a nontonal language. Because MMN is often equated with echoic memory, this might imply a slower decay of echoic memory in the presence of a more stable categorical memory trace (see, however, Yu et al 2018). The study also showcases the challenges in ruling out alternative interpretations of the behavioral findings: In the absence of a neural correlate that is modulated exclusively by the content of echoic memory (and it is not clear whether this is actually the case for MMN), the behavioral findings alone cannot distinguish whether subjects perform the task based on the stable categorical memory trace or a somewhat less volatile sensory trace.…”

“…It has been noted that this attenuation fades with similar time constants as information decays from EM (Lu et al 1992). Consequently, these shortlived contextual modulations have been viewed as markers of EM (Atienza et al 2002;Bartha-Doering et al 2015;Grau et al 2001;Ioannides et al 2003;Lindín et al 2013;Lu et al 1992;Näätänen et al 1989;Sams et al 1993;Haenschel et al 2005;Yu et al 2018). Interestingly, all of these short-lived contextual phenomena depend in one way or another on how often a specific sound has been presented.…”

Neural Representation of the English Vowel Feature [High]: Evidence From /ε/ vs. /ɪ/

“…We chose to use this simple paradigm of one token per category rather than using multiple tokens (e.g., Eulitz and Lahiri, 2004;Hisagi et al, 2010;Hestvik and Durvasula, 2016;Yu et al, 2017Yu et al, , 2018 because we wanted to be able to directly relate the findings to the previous studies where we used these stimuli. Importantly, speech sounds, whether presented as a single token, or as one of a set of tokens are processed at a phonological level.…”

Neurophysiological Correlates of Asymmetries in Vowel Perception: An English-French Cross-Linguistic Event-Related Potential Study

“…Because the MMN is thought to reflect “pre-attentive” processes, it may be too early of a cortical response to reflect asymmetries, which do not appear to derive from early stages of acoustic processing ( Polka and Bohn, 2011 ; Masapollo et al, 2018b ; Polka et al, 2019 ). Recently it has also been demonstrated that MMNs recorded in an oddball paradigm with a longer ISI (e.g., 600 vs. 2,600 ms) reflect sensitivity to language-specific phonological information rather than the acoustic information in speech sounds ( Yu et al, 2017 , 2018 ). An additional goal of the current study was to go beyond examination of the classic MMN response and track cortical oscillations.…”