Talker and background noise specificity in spoken word recognition memory

Abstract

Prior research has demonstrated that listeners are sensitive to changes in the indexical (talker-specific) characteristics of speech input, suggesting that these signal-intrinsic features are integrally encoded in memory for spoken words. Given that listeners frequently must contend with concurrent environmental noise, to what extent do they also encode signal-extrinsic details? Native English listeners' explicit memory for spoken English monosyllabic and disyllabic words was assessed as a function of consistency versus variation in the talker's voice (talker condition) and background noise (noise condition) using a delayed recognition memory paradigm. The speech and noise signals were spectrally-separated, such that changes in a simultaneously presented non-speech signal (background noise) from exposure to test would not be accompanied by concomitant changes in the target speech signal. The results revealed that listeners can encode both signal-intrinsic talker and signal-extrinsic noise information into integrated cognitive representations, critically even when the two auditory streams are spectrally non-overlapping. However, the extent to which extra-linguistic episodic information is encoded alongside linguistic information appears to be modulated by syllabic characteristics, with specificity effects found only for monosyllabic items. These findings suggest that encoding and retrieval of episodic information during spoken word processing may be modulated by lexical characteristics.

Talker-familiarity benefit in non-native recognition memory and word identification: The role of listening conditions and proficiency

“…The present study is the first study that investigated talker familiarity benefit in recognition memory and word identification for words spoken in a non-native language, and the role of background noise and non-native proficiency on this talker familiarity effect. After successfully learning to recognize four previously unfamiliar talkers over the course of four days, the familiar talker benefit in nonnative listening was observed for recognition memory in line with other studies for native (Bradlow & Pisoni, 1999;Cooper & Bradlow, 2017;Goh, 2005;Goldinger, 1996;Luce & Lyons, 1998;Palmeri et al, 1993;Sheffert, 1998) and non-native listening (Trofimovich, 2005;Winters et al, 2013). At the same time, no talker familiarity benefit was observed during non-native word identification in 2 Following Levi et al (2011) and Nygaard and Pisoni (1998), a second analysis was conducted including voice recognition performance on the last training day as a potential predictor of relative progress.…”

“…The discrepancies in the emergence of the effect of indexical information in these two different types of tasks have been previously observed by Luce and Lyons (1998), Kittredge et al (2006), and Lee and Zhang (2015), who found the effect of indexical information in a recognition memory task and repetition priming, but not in a lexical decision task and during semantic priming. While the recognition memory task requires minimal contact with the mental lexicon and does not abstract away indexical information in spoken words (Cooper & Bradlow, 2017), the word identification task focuses on sound and lexical processing which does abstract from voice-specific characteristics (Cutler et al, 2010a, b). Indeed, even listeners not familiar with a language demonstrate effects of indexical information in recognition memory (Winters et al, 2013) but not in word identification (Levi et al, 2011).…”

“…A speech signal is extremely variable, and not only contains linguistic but also so-called indexical information (Abercombie, 1967 sensitive to changes in indexical information: words are better recognized when they are spoken by one talker than when they are spoken by multiple talkers (Mullennix et al, 1989;Ryalls & Pisoni, 1997;Sommers et al, 1994). Moreover, words repeated by the same talker and even words combined from phonemes repeated by the same talker are recognized more quickly and more accurately than words repeated by a different talker (Bradlow & Pisoni, 1999;Cooper & Bradlow, 2017;Goh, 2005;Goldinger, 1996;Jesse et al, 2007;Luce & Lyons, 1998;Palmeri et al, 1993;Sheffert, 1998), indicating that surface details of words, such as indexical information, are retained in some form in the memory of the listeners and subsequently facilitate speech processing.…”

Voices in the mental lexicon: Words carry indexical information that can affect access to their meaning

“…Many of the aforementioned studies start with an initial encoding phase, during which participants are presented with a series of word stimuli. During the second phase of the experiment, participants are presented with a mix of new and old (i.e., previously presented) words and are asked to indicate whether each stimulus is new or old, and/or whether the voice in which it is spoken is the same or not (Bradlow, Nygaard, & Pisoni, 1999;Cooper & Bradlow, 2017;Palmeri, Goldinger, & Pisoni, 1993). Other studies have used more implicit measures of processing, such as word identification in noise (Goldinger, 1996;González & McLennan, 2007) and lexical decision (Luce & Lyons, 1998).…”

“…These studies indicate that extra-phonological information is encoded in some form along with phonological information. Interestingly, such information may not be limited to talker-specific information (Goldinger, 1998;Luce, McLennan, & Charles-Luce, 2012; McLennan & Luce, 2005;Vitevitch & Donoso, 2011), but may also include other aspects of the auditory context in which a word was encountered, such as background noise (Cooper & Bradlow, 2017;Pufahl & Samuel, 2014).…”