Online Spoken Word Recognition Research Articles

Perception of Different Tone Contrasts at Sub-Lexical and Lexical Levels by Dutch Learners of Mandarin Chinese.

This study explores the difficulties in distinguishing different lexical tone contrasts at both sub-lexical and lexical levels for beginning and advanced Dutch learners of Mandarin, using a sequence-recall task and an auditory lexical decision task. In both tasks, the Tone 2-Tone 3 contrast is most prone to errors for both groups of learners. A significant improvement in the advanced group was found for this tone contrast in the sub-lexical sequence recall task, but not in the lexical decision task. This is taken as evidence that utilizing tones in on-line spoken word recognition is more complex and demanding for L2 learners than in a memory-based task. The results of the lexical decision task also revealed that advanced learners have developed a stronger sensitivity to Tone 1 compared to the other three tones, with Tone 4 showing the least sensitivity. These findings suggest different levels of robustness and distinctiveness for the representation of different lexical tones in L2 learners’ lexicon and consequently different levels of proficiency in integrating tones for lexical processing. The observed patterns of difficulty are potentially related to the acoustic characteristics of different lexical tone contrasts as well as to the interference of the suprasegmental features of learner’s native language (i.e., the tonal contrasts of Dutch intonation) on the acquisition of the Mandarin lexical tone contrasts.

Re-reconsidering the role of temporal order in spoken word recognition.

Spoken word recognition models incorporate the temporal unfolding of word information by assuming that positional match constrains lexical activation. Recent findings challenge the linearity constraint. In the visual world paradigm, Toscano, Anderson, and McMurray observed that listeners preferentially viewed a picture of a target word's anadrome competitor (e.g., competitor bus for target sub) compared with phonologically unrelated distractors (e.g., well) or competitors sharing an overlapping vowel (e.g., sun). Toscano et al. concluded that spoken word recognition relies on coarse grain spectral similarity for mapping spoken input to a lexical representation. Our experiments aimed to replicate the anadrome effect and to test the coarse grain similarity account using competitors without vowel position overlap (e.g., competitor leaf for target flea). The results confirmed the original effect: anadrome competitor fixation curves diverged from unrelated distractors approximately 275 ms after the onset of the target word. In contrast, the no vowel position overlap competitor did not show an increase in fixations compared with the unrelated distractors. The contrasting results for the anadrome and no vowel position overlap items are discussed in terms of theoretical implications of sequential match versus coarse grain similarity accounts of spoken word recognition. We also discuss design issues (repetition of stimulus materials and display parameters) concerning the use of the visual world paradigm in making inferences about online spoken word recognition.

Influence of speaking style adaptations and semantic context on the time course of word recognition in quiet and in noise

This study examines the effects of different listener-oriented speaking styles and semantic contexts on online spoken word recognition using eyetracking. In Experiment 1, different groups of listeners participated in a word-identification-in-noise and in a pleasantness-rating task. Listeners heard sentences with high- versus low-predictability semantic contexts produced in infant-directed speech, Clear Speech, and Conversational Speech. Experiment 2 (in silence) and 3 (in noise) investigated the time course of visual fixations to target objects when participants were listening to different speaking styles and contexts. Results from all experiments show that relative to conversational speech, both infant-directed speech and Clear Speech improved word recognition for high-predictability sentences, in quiet as well as in noise. This indicates that established advantages of infant-directed speech for young listeners cannot be attributed only to affect; the acoustic enhancements in infant-directed speech benefit adult speech processing as well. Furthermore, in silence (Experiment 2) lexical access was facilitated by contextual cues even in conversational speech; but in noise (Experiment 3) listeners reliably focused the target only when a combination of contextual cues and listener-adapted acoustic–phonetic cues were available. These findings suggest that both semantic cues and listener-oriented acoustic enhancements are needed to facilitate word recognition, especially in adverse listening conditions.

Low-frequency fine-structure cues allow for the online use of lexical stress during spoken-word recognition in spectrally degraded speech.

English listeners use suprasegmental cues to lexical stress during spoken-word recognition. Prosodic cues are, however, less salient in spectrally degraded speech, as provided by cochlear implants. The present study examined how spectral degradation with and without low-frequency fine-structure information affects normal-hearing listeners' ability to benefit from suprasegmental cues to lexical stress in online spoken-word recognition. To simulate electric hearing, an eight-channel vocoder spectrally degraded the stimuli while preserving temporal envelope information. Additional lowpass-filtered speech was presented to the opposite ear to simulate bimodal hearing. Using a visual world paradigm, listeners' eye fixations to four printed words (target, competitor, two distractors) were tracked, while hearing a word. The target and competitor overlapped segmentally in their first two syllables but mismatched suprasegmentally in their first syllables, as the initial syllable received primary stress in one word and secondary stress in the other (e.g., "'admiral," "'admi'ration"). In the vocoder-only condition, listeners were unable to use lexical stress to recognize targets before segmental information disambiguated them from competitors. With additional lowpass-filtered speech, however, listeners efficiently processed prosodic information to speed up online word recognition. Low-frequency fine-structure cues in simulated bimodal hearing allowed listeners to benefit from suprasegmental cues to lexical stress during word recognition.

Two-year-olds' sensitivity to subphonemic mismatch during online spoken word recognition.

Sensitivity to noncontrastive subphonemic detail plays an important role in adult speech processing, but little is known about children's use of this information during online word recognition. In two eye-tracking experiments, we investigate 2-year-olds' sensitivity to a specific type of subphonemic detail: coarticulatory mismatch. In Experiment 1, toddlers viewed images of familiar objects (e.g., a boat and a book) while hearing labels containing appropriate or inappropriate coarticulation. Inappropriate coarticulation was created by cross-splicing the coda of the target word onto the onset of another word that shared the same onset and nucleus (e.g., to create boat, the final consonant of boat was cross-spliced onto the initial CV of bone). We tested 24-month-olds and 29-month-olds in this paradigm. Both age groups behaved similarly, readily detecting the inappropriate coarticulation (i.e., showing better recognition of identity-spliced than cross-spliced items). In Experiment 2, we asked how children's sensitivity to subphonemic mismatch compared to their sensitivity to phonemic mismatch. Twenty-nine-month-olds were presented with targets that contained either a phonemic (e.g., the final consonant of boat was spliced onto the initial CV of bait) or a subphonemic mismatch (e.g., the final consonant of boat was spliced onto the initial CV of bone). Here, the subphonemic (coarticulatory) mismatch was not nearly as disruptive to children's word recognition as a phonemic mismatch. Taken together, our findings support the view that 2-year-olds, like adults, use subphonemic information to optimize online word recognition.

Effects of orthographic consistency and homophone density on Chinese spoken word recognition

Studies of alphabetic language have shown that orthographic knowledge influences phonological processing during spoken word recognition. This study utilized the Event-Related Potentials (ERPs) to differentiate two types of phonology-to-orthography (P-to-O) mapping consistencies in Chinese, namely homophone density and orthographic consistency. The ERP data revealed an orthographic consistency effect in the frontal-centrally distributed N400, and a homophone density effect in central-posteriorly distributed late positive component (LPC). Further source analyses using the standardized low-resolution electromagnetic tomography (sLORETA) demonstrated that the orthographic effect was not only localized in the frontal and temporal-parietal regions for phonological processing, but also in the posterior visual cortex for orthographic processing, while the homophone density effect was found in middle temporal gyrus for lexical-semantic selection, and in the temporal-occipital junction for orthographic processing. These results suggest that orthographic information not only shapes the nature of phonological representations, but may also be activated during on-line spoken word recognition.

The interaction of lexical tone, intonation and semantic context in on-line spoken word recognition: An ERP study on Cantonese Chinese

In two ERP experiments, we investigate the on-line interplay of lexical tone, intonation and semantic context during spoken word recognition in Cantonese Chinese. Experiment 1 shows that lexical tone and intonation interact immediately. Words with a low lexical tone at the end of questions (with a rising question intonation) lead to a processing conflict. This is reflected in a low accuracy in lexical identification and in a P600 effect compared to the same words at the end of a statement. Experiment 2 shows that a strongly biasing semantic context leads to much better lexical-identification performance for words with a low tone at the end of questions and to a disappearance of the P600 effect. These results support the claim that semantic context plays a major role in disentangling the tonal information from the intonational information, and thus, in resolving the on-line conflict between intonation and tone. However, the ERP data indicate that the introduction of a semantic context does not entirely eliminate on-line processing problems for words at the end of questions. This is revealed by the presence of an N400 effect for words with a low lexical tone and for words with a high-mid lexical tone at the end of questions. The ERP data thus show that, while semantic context helps in the eventual lexical identification, it makes the deviation of the contextually expected lexical tone from the actual acoustic signal more salient.

Expectations from preceding prosody influence segmentation in online sentence processing

Previous work examining prosodic cues in online spoken-word recognition has focused primarily on local cues to word identity. However, recent studies have suggested that utterance-level prosodic patterns can also influence the interpretation of subsequent sequences of lexically ambiguous syllables (Dilley, Mattys, & Vinke, Journal of Memory and Language, 63:274-294, 2010; Dilley & McAuley, Journal of Memory and Language, 59:294-311, 2008). To test the hypothesis that these distal prosody effects are based on expectations about the organization of upcoming material, we conducted a visual-world experiment. We examined fixations to competing alternatives such as pan and panda upon hearing the target word panda in utterances in which the acoustic properties of the preceding sentence material had been manipulated. The proportions of fixations to the monosyllabic competitor were higher beginning 200 ms after target word onset when the preceding prosody supported a prosodic constituent boundary following pan-, rather than following panda. These findings support the hypothesis that expectations based on perceived prosodic patterns in the distal context influence lexical segmentation and word recognition.

Examining the role of variability in emotional tone of voice during online spoken word recognition

Previous studies demonstrate that listeners are faster to recognize words recently spoken by the same talker, relative to a different talker, when processing is relatively slow. The purpose of the present study was to examine intra-talker variability in emotional tone of voice on listeners' online perception of spoken words. Based on previous work, we predicted reduced priming in the mismatch condition, when words are repeated in a different emotional tone of voice (e.g., sad and frightened), relative to the match condition, when words are repeated in the same emotional tone of voice (e.g., frightened)—when processing is relatively slow (experiment 2, hard lexical decision) and equivalent priming in the match and mismatch conditions when processing is relatively fast (experiment 1, easy lexical decision). The present study should provide a greater understanding of the role that variability in emotional tone of voice plays in listeners' online perception of spoken words.

Gradient sensitivity to within-category variation in words and syllables.

Five experiments monitored eye movements in phoneme and lexical identification tasks to examine the effect of within-category subphonetic variation on the perception of stop consonants. Experiment 1 demonstrated gradient effects along voice-onset time (VOT) continua made from natural speech, replicating results with synthetic speech (B. McMurray, M. K. Tanenhaus, & R. N. Aslin, 2002). Experiments 2-5 used synthetic VOT continua to examine effects of response alternatives (2 vs. 4), task (lexical vs. phoneme decision), and type of token (word vs. consonant-vowel). A gradient effect of VOT in at least one half of the continuum was observed in all conditions. These results suggest that during online spoken word recognition, lexical competitors are activated in proportion to their continuous distance from a category boundary. This gradient processing may allow listeners to anticipate upcoming acoustic-phonetic information in the speech signal and dynamically compensate for acoustic variability.