Speech Feedback Research Articles

The role of auditory feedback in learning speech of second languages

Learning to speak requires coordination between auditory and motor systems. Motor programs need to be fine-tuned based on goals and feedback. The motor-tuning process includes establishing the sensory-to-motor transformation (generate motor codes from articulation) and motor-to-sensory transformation (generate predictions for speech control). In this study, we investigated the relations among the three factors (feedback, goals, and predictions) in speech production in the context of second language learning. Adult Mandarin speakers were asked to learn non-native vowels: /o/ and /ɵ/—the former being less similar than the latter to Mandarin vowels, in feedback available or feedback masked conditions. We found no improvement in learning when feedback was masked, suggesting that motor-based prediction could not directly compare with goals for adult second language acquisition. Furthermore, feedback helped to learn only when target sounds are distinct from existing sounds in one’s native speech (competition between prediction and goals is minimal). The results suggest prediction and goals may share a similar representational format, which could yield a competing relation in speech learning. The feedback can conditionally overcome such interference between prediction and goals. The study further probed the functional relations among key components (prediction, goals, and feedback) of sensorimotor integration in speech production and learning.Learning to speak requires coordination between auditory and motor systems. Motor programs need to be fine-tuned based on goals and feedback. The motor-tuning process includes establishing the sensory-to-motor transformation (generate motor codes from articulation) and motor-to-sensory transformation (generate predictions for speech control). In this study, we investigated the relations among the three factors (feedback, goals, and predictions) in speech production in the context of second language learning. Adult Mandarin speakers were asked to learn non-native vowels: /o/ and /ɵ/—the former being less similar than the latter to Mandarin vowels, in feedback available or feedback masked conditions. We found no improvement in learning when feedback was masked, suggesting that motor-based prediction could not directly compare with goals for adult second language acquisition. Furthermore, feedback helped to learn only when target sounds are distinct from existing sounds in one’s native speech (competition be...

Roles of articulatory movements and speech feedback in Japanese text comprehension during oral reading

Roles of articulatory movements and speech feedback in Japanese text comprehension during oral reading

Effects of COMT polymorphism on the cortical processing of vocal pitch regulation.

Recent evidence has shown that auditory-motor integration for speech production is influenced by cognitive functions such as working memory and attention, suggesting that speech motor control is likely modulated by mechanisms mediated by prefrontal regions. Catechol-O-methyltransferase (COMT) gene plays an important role in dopamine breakdown in the prefrontal cortex and has been associated with a variety of prefrontal cognitive functions. The present event-related potential study investigated the association between COMT ValMet polymorphism and auditory-motor processing of vocal feedback errors. A sample of 131 Chinese young female adults was genotyped for rs4680 and produced sustained vowels while hearing their voice unexpectedly shifted down in pitch by 50 or 200 cents. The behavioral results showed no effects of COMT ValMet on vocal compensations for pitch perturbations. However, individuals with the Met allele produced significantly larger P2 responses to -200 cents perturbations than individuals with the Val/Val genotype. These results suggest the existence of a relationship between COMT ValMet polymorphism and self-monitoring of speech feedback errors, and they provide insights into our understanding of the top-down modulations of speech motor control mediated by prefrontal regions.

Developing English learning program using speech recognition-based feedback.

Developing English learning program using speech recognition-based feedback.

Sermon Feedback as Facework: Task and Identity Goals in Mentoring Homiletical Theologians

The problem of in-class sermon feedback has vexed the teaching and learning of preaching for some time. For too long, however, the problem has been understood in more or less personal terms insofar as it tries to facilitate an environment where task mastery becomes psychologically feasible for students in a stressful feedback situation. This article argues that the problem of sermon feedback is actually more complex in that it is also tied to privilege, power dynamics, and the multiple identities of the pluralistic classroom and cultural identities in churches. The use of facework theory in communication studies offers an alternative by placing task mastery in relation to the tending to identities in the room. Here the work of Jeff Kerssen-Griep et al. posits a compelling vision for dealing with an analogous face-threatening situation of speech feedback in the college classroom that includes both positive (belonging, competency) and negative (autonomy) facework and thus locates task mastery in relation to tending to identities in a more self-reflective manner. In the process, a more mentoring-type relation between teacher and student is envisioned, one that might even build on Dale Andrews’ notion of apprenticeship in preaching education. The article concludes with homiletical-theological reflections about the relationship of face/Face and promise in reflecting on gospel in the task of teaching and learning preaching.

Trait Self-Compassion as a Buffer Against Post-event Processing Following Performance Feedback

Post-event processing (PEP) refers to a prolonged and negative rumination following anxiety-provoking social events. Because PEP may maintain social anxiety over time, it is important to reduce this repetitive, negative thinking. Past research has shown that PEP can be reduced through self-compassion. As such, the primary purpose of the present study was to examine the circumstances under which self-compassion buffers against PEP. Given that PEP may be exacerbated by negative performance feedback, we examined whether self-compassion would buffer against PEP under these circumstances (i.e., receiving negative performance feedback). Participants (N = 66) provided an impromptu speech and were randomly assigned to receive either positive or negative speech feedback. As expected, negative performance feedback led to significantly more PEP than positive feedback. However, whereas this effect was particularly pronounced amongst those low on self-compassion, there were no significant differences between conditions on PEP amongst those high on self-compassion. The findings from the present study suggest that trait self-compassion serves to limit PEP in situations where negative performance feedback is provided. This work builds on the benefits of self-compassion in the context of social stress.

Robust Sensorimotor Learning during Variable Sentence-Level Speech

Sensorimotor learning has been studied by altering the sound of the voice in real time as speech is produced. In response to voice alterations, learned changes in production reduce the perceived auditory error and persist for some time after the alteration is removed [1-5]. The results of such experiments have led to the development of prominent models of speech production. This work proposes that the control of speech relies on forward models to predict sensory outcomes of movements, and errors in these predictions drive sensorimotor learning [5-7]. However, sensorimotor learning in speech has only been observed following intensive training on ahandful of discrete words or perceptually similar sentences. Stereotyped production does not capture the complex sensorimotor demands of fluid, real-world speech [8-11]. It remains unknown whether talkers predict the sensory consequences of variable sentence production to allow rapid and precise updating of speech motor plans when sensory prediction errors are encountered. Here, we used real-time alterations of speech feedback to test for sensorimotor learning during the production of 50 sentences that varied markedly in length, vocabulary, and grammar. Following baseline production, all vowels were simultaneously altered and played back through headphones in near real time. Robust feedforward changes in sentence production were observed that, on average, precisely countered the direction of the alteration. These changes occurred in every participant and transferred to the production of single words with varying vowel sounds. The results show that to maintain accurate sentence production, the brain actively predicts the auditory consequences of variable sentence-level speech.

Considerations on Effective Feedback in Computerized Speech Training for Dysarthric Speakers.

Introduction: In this article, we consider the role of feedback in computerized speech training for patients with dysarthric speech due to acquired neurological disorders such as Parkinson's disease and stroke. Methods: Based on literature from different fields, we review several issues that play a role when designing feedback for computerized speech training, including serious games. Results: Which speech dimensions (e.g., articulation, loudness) to target in computerized speech feedback, and at what level of detail to provide the feedback, depend on the focus of the specific speech therapy and the technical feasibility of providing reliable speech feedback. Furthermore, feedback that provides knowledge of results generally appears equally effective as knowledge of performance feedback, and is more easily provided in computerized speech training systems. Implicit feedback can be more motivating than explicit feedback (e.g., in serious games), but may be harder to understand. Multimodal feedback is preferred over unimodal feedback to ensure that the feedback message is optimally perceived. Mild negative task-oriented feedback can enhance effort but should be balanced with positive feedback, such as feedback referring to patients' progress. Finally, a balance should be found between the motivational advantages of high-frequency immediate feedback (e.g., in serious games) and the advantages of low-frequency and delayed feedback in preventing patients from becoming dependent on augmented feedback. Conclusions: Several aspects play a role in designing feedback for computerized speech training for dysarthric speakers. Decisions on these aspects should be made carefully, accounting for the specific characteristics of the patient group that is targeted, the focus of the specific speech therapy, and the technological possibilities that are available.

Auditory cortical activity drives feedback-dependent vocal control in marmosets

Vocal communication is a sensory-motor process requiring auditory self-monitoring to correct errors and to ensure accurate vocal production. When presented with altered speech feedback, humans rapidly change their speech to compensate. Although previous evidence has demonstrated suppression of auditory cortex during both speech and animal vocalization, the specific role of auditory cortex in such feedback-dependent control is unknown. Here we show the relationship between neural activity in the auditory cortex and feedback-dependent vocal control in marmoset monkeys. We demonstrate that marmosets, like humans, exhibit feedback control of vocal acoustics. We further show that feedback-sensitive activity of auditory cortex neurons predict such compensatory vocal changes. Finally, we demonstrate that electrical microstimulation of auditory cortex rapidly evokes similar changes in vocal production. These results are evidence for a causal role of auditory cortex in vocal self-monitoring and feedback-dependent control, and have implications for understanding human speech motor control.

Group and individual variability in speech production networks during delayed auditory feedback.

Altering reafferent sensory information can have a profound effect on motor output. Introducing a short delay [delayed auditory feedback (DAF)] during speech production results in modulations of voice and loudness, and produces a range of speech dysfluencies. The ability of speakers to resist the effects of delayed feedback is variable yet it is unclear what neural processes underlie differences in susceptibility to DAF. Here, susceptibility to DAF is investigated by looking at the neural basis of within and between subject changes in speech fluency under 50 and 200 ms delay conditions. Using functional magnetic resonance imaging, networks involved in producing speech under two levels of DAF were identified, lying largely within networks active during normal speech production. Independent of condition, fluency ratings were associated with midbrain activity corresponding to periaqueductal grey matter. Across subject variability in ability to produce normal sounding speech under a 200 ms delay was associated with activity in ventral sensorimotor cortices, whereas ability to produce normal sounding speech under a 50 ms delay was associated with left inferior frontal gyrus activity. These data indicate whilst overlapping cortical mechanisms are engaged for speaking under different delay conditions, susceptibility to different temporal delays in speech feedback may involve different processes.

The impact of perilaryngeal vibration on the self-perception of loudness and the Lombard effect

The role of somatosensory feedback in speech and the perception of loudness was assessed in adults without speech or hearing disorders. Participants completed two tasks: loudness magnitude estimation of a short vowel and oral reading of a standard passage. Both tasks were carried out in each of three conditions: no-masking, auditory masking alone, and mixed auditory masking plus vibration of the perilaryngeal area. A Lombard effect was elicited in both masking conditions: speakers unconsciously increased vocal intensity. Perilaryngeal vibration further increased vocal intensity above what was observed for auditory masking alone. Both masking conditions affected fundamental frequency and the first formant frequency as well, but only vibration was associated with a significant change in the second formant frequency. An additional analysis of pure-tone thresholds found no difference in auditory thresholds between masking conditions. Taken together, these findings indicate that perilaryngeal vibration effectively masked somatosensory feedback, resulting in an enhanced Lombard effect (increased vocal intensity) that did not alter speakers' self-perception of loudness. This implies that the Lombard effect results from a general sensorimotor process, rather than from a specific audio-vocal mechanism, and that the conscious self-monitoring of speech intensity is not directly based on either auditory or somatosensory feedback.

F149. NEUROBIOLOGY OF SELF-AGENCY DURING REALITY MONITORING AND SPEECH FEEDBACK MONITORING: IMPLICATIONS FOR TREATMENT DEVELOPMENT IN SCHIZOPHRENIA

BackgroundSelf-agency is the experience of being the agent of one’s own thoughts and motor actions. The intact experience of self-agency is necessary for successful interactions with the outside world (i.e., reality monitoring). Reality monitoring is the ability to distinguish internally self-generated information from outside reality (externally-derived information). We found that healthy control (HC) participants recruit medial prefrontal cortex (mPFC) during encoding of self-generated information, which is also activated during accurate retrieval of self-generated information. By contrast, patients with schizophrenia (SZ) have specific self-agency impairments and do not show mPFC activation during encoding or retrieval of self-generated information. These findings indicate that SZ may rely more on environmental externally-derived information, rather than on internal self-generated information, to guide reality monitoring. Here, we relate the experience of self-agency during a lower-level speech feedback monitoring (i.e., monitoring what we hear ourselves say) to our higher-level cognitive reality monitoring task. We examine whether the sense of self-agency during speech feedback monitoring and reality monitoring are driven by the same fundamental mechanism that we hypothesize underlies the capacity to experience self-agency—the ability to make reliable predictions about the outcomes of self-generated actions.MethodsDuring speech feedback monitoring we assess self-agency by altering environmental auditory feedback so that participants listen to a perturbed version of their own speech. When subjects hear minimal perturbations in their auditory feedback while speaking, they make corrective responses, indicating that they judge the perturbations as errors in their speech output. These corrective responses are modulated by subjects’ reliance on internal predictions about the outcome of their speech output; the more subjects rely on their internal predictions, i.e. their sense of self-agency, the less they rely on external auditory feedback, resulting in smaller corrective responses. Thus, subjects who produced smaller corrective responses manifested an enhanced sense of self-agency in that they relied more on their internal predictions to generate their own actions (i.e., their speech output).ResultsWe found that self-agency judgments in the reality-monitoring task were higher in people who had smaller corrective responses (p=.05) during minimal speech perturbations of their auditory feedback. These results provide support for a unitary process for the experience of self-agency resulting from the ability to reliably predict the outcomes of self-generated actions, that governs low-level speech control and higher level reality monitoring.DiscussionThese findings have important therapeutic implications in SZ, suggesting that the more participants rely on internal predictions to guide their actions, the smaller their corrective responses in their speech output, and the more likely they are to make correct judgments of self-agency during reality monitoring. In conclusion, these findings, therefore, indicate that reality monitoring and speech monitoring paradigms provide quick and robust markers of the experience of self-agency, indicating which subjects followed their internal predictions to guide their own actions. Together, these findings have important therapeutic implications for potentiating improvements in self-agency judgments not only in HC, but in patients with schizophrenia who suffer from critical self-agency impairments.

Reality Monitoring and Feedback Control of Speech Production Are Related Through Self-Agency.

Self-agency is the experience of being the agent of one’s own thoughts and motor actions. The intact experience of self-agency is necessary for successful interactions with the outside world (i.e., reality monitoring) and for responding to sensory feedback of our motor actions (e.g., speech feedback control). Reality monitoring is the ability to distinguish internally self-generated information from outside reality (externally-derived information). In the present study, we examined the relationship of self-agency between lower-level speech feedback monitoring (i.e., monitoring what we hear ourselves say) and a higher-level cognitive reality monitoring task. In particular, we examined whether speech feedback monitoring and reality monitoring were driven by the capacity to experience self-agency—the ability to make reliable predictions about the outcomes of self-generated actions. During the reality monitoring task, subjects made judgments as to whether information was previously self-generated (self-agency judgments) or externally derived (external-agency judgments). During speech feedback monitoring, we assessed self-agency by altering environmental auditory feedback so that subjects listened to a perturbed version of their own speech. When subjects heard minimal perturbations in their auditory feedback while speaking, they made corrective responses, indicating that they judged the perturbations as errors in their speech output. We found that self-agency judgments in the reality-monitoring task were higher in people who had smaller corrective responses (p = 0.05) and smaller inter-trial variability (p = 0.03) during minimal pitch perturbations of their auditory feedback. These results provide support for a unitary process for the experience of self-agency governing low-level speech control and higher level reality monitoring.

Fluency-Enhancing Strategies for Hypokinetic Dysarthria Exacerbated by Subthalamic Nucleus Brain Stimulation: A Case Study

Introduction Speech disorders associated with Parkinson's disease (PD) and the pharmaceutical treatments of PD are well documented. A relatively recent treatment alternative for PD is deep brain stimulation (DBS) of the subthalamic nucleus (STN), which is used to manage the symptoms of PD as the disease progresses. This case study documented the speech characteristics of a unique client with PD STN-DBS and reported initial findings on a variety of fluency- and intelligibility-enhancing strategies. Method A speech-language pathologist referred a 63-year-old man, previously diagnosed by a speech-language pathologist with neurogenic stuttering as a result of an STN-DBS battery change, for a speech evaluation, reporting lack of success with traditional stuttering treatment strategies. The client's speech was assessed, and a variety of fluency- and intelligibility-enhancing techniques were tested during trial therapy. Results The client's speech exhibited the hallmark characteristics of hypokinetic dysarthria, including speech disfluencies. A variety of pacing and prosthetic strategies were tested, revealing that auditory and tactile prosthetic speech feedback provided optimal improvements in fluency and intelligibility. Discussion These results suggest that the prosthetic speech feedback provided optimal intelligibility and fluency enhancement and could potentially improve articulation and speech volume, which are also common in cases of hypokinetic dysarthria.

Allocation of attention to real-time visual speech feedback in a digital mirror

The role of sensory feedback in speech motor control is typically investigated by observing the behavioral and/or neurological response to experimental feedback perturbation. The function of feedback and its use in typical control is therefore inferred from responses under atypical conditions. The present study avoided perturbation, using eye tracking and a digital mirror (continuous video relay from a webcam) to examine how speakers attend to real-time visual feedback across normal speech and non-speech tasks. Fixation locations and durations were recorded during a period of rest prior to task onset (no-task condition), a recitation of the ABC’s, a re-telling of a popular children’s story (Goldilocks and the Three Bears), and parallel clear speech versions of the latter two tasks. Analysis of gaze showed that participants overall seemed to avoid fixating on their self-image, and that avoidance increased during speech tasks compared to non-speech “rest.” Across the two content areas (ABC’s vs. storytelling), participants’ self-gaze was concentrated more heavily on speech-relevant areas during storytelling, and no differences in gaze were observed between casual and clear speech. These data suggest an increased role for sensory feedback during complex linguistic tasks, as well as indicating an overall aversion towards realistic visual self-feedback.

Supra-Segmental Changes in Speech Production as a Result of Spectral Feedback Degradation: Comparison with Lombard Speech.

Perturbations to acoustic speech feedback have been typically localized to specific phonetic characteristics, for example, fundamental frequency (F0) or the first two formants (F1/F2), or affect all aspects of the speech signal equally, for example, via the addition of background noise. This paper examines the consequences of a more selective global perturbation: real-time cochlear implant (CI) simulation of acoustic speech feedback. Specifically, we examine the potential similarity between speakers' response to noise vocoding and the characteristics of Lombard speech. An acoustic analysis of supra-segmental characteristics in speaking rate, F0 production, and voice amplitude revealed changes that paralleled the Lombard effect in some domains but not others. Two studies of speech intelligibility complemented the acoustic analysis, finding that intelligibility significantly decreased as a result of CI simulation of speaker feedback. Together, the results point to differences in speakers' responses to these two superficially similar feedback manipulations. In both cases we see a complex, multi-faceted behavior on the part of talkers. We argue that more instances of global perturbation and broader response assessment are needed to determine whether such complexity is present in other feedback manipulations or if it represents a relatively rare exception to the typical compensatory feedback response.

Sensitivity and specificity of auditory feedback driven articulatory learning in virtual speech

The current work presents articulatory kinematic and acoustic data characterizing how the form of synthesized auditory feedback in virtual speech affects the sensitivity and specificity of articulatory learning. The term “virtual speech” refers to talker-manipulated synthesized speech controlled in real time using electromagnetic articulography (EMA). In the current work, 36 participants (4 with dysarthria) participated in a learning experiment requiring them to control an articulatory speech synthesizer using movements of the tongue, lips, and jaw. Participants were divided among two experimental conditions: (1) an “unmatched” condition, during which all participants received auditory feedback based on common articulatory synthesis settings (neither formant working space nor fundamental frequency were distinguishable between talkers); and (2) a “matched” condition, during which the articulatory synthesis parameters were adjusted to mimic the formant working space and average fundamental frequency of the learner. Analyses focused on the kinematic and acoustic differences in learning between the two conditions. Results suggest that the sensitivity and specificity of articulatory learning is affected by the extent to which the auditory feedback matches the learner's familiar acoustic working space. Findings have implications regarding how the acoustic characteristics of perceived speech affect sensorimotor integration and learning in typical talkers and individuals with dysarthria.

Introduction of real-time visual feedback impairs intelligibility when other senses are degraded

Sensory feedback allows talkers to accurately control speech production. Under ordinary circumstances, real-time speech feedback is available in the auditory and somatosensory modalities. When sensation in these modes is degraded, the accuracy of speech motor control is impaired and intelligibility decreases. In the present study, we documented this drop in intelligibility for speech produced with degraded sensory feedback in the auditory (8-channel cochlear implant simulation) and somatosensory (1 mL benzocaine topical application) domains. Speakers (N = 15) were also asked, however, to produce speech in front of a mirror while their senses were degraded, giving them access to an alternate fully-specified but atypical source of sensory feedback. Given the large body of work showing the benefits of visual speech information in perception, we predicted that speakers would use the visual feedback to improve real-time control. Contrary to that prediction, naïve participants (N = 38) found speech produced with visual feedback (and degraded acoustic/somatosensory feedback) to be less intelligible overall than speech produced without it (and other senses still degraded). This deleterious effect of visual feedback was found for both aural intelligibility and visual (lipread) intelligibility.

Sensitivity to meter in auditory feedback during music performance.

Performance of a complex sequential task, such as speech or music, occurs simultaneously with its perceived outcome, known as auditory feedback. The importance of auditory feedback to the fluency of performance is demonstrated by the disruptive effects of altered auditory feedback (AAF). For instance, it has been know for decades that sequence production (in music and speech) can be disrupted by delayed auditory feedback (DAF), a slight asynchrony between actions and resulting sounds (for reviews see Finney, 1999; Howell, 2004; Pfordresher, 2006; Yates, 1963). Fluency in performance clearly relies on temporal coordination between planned actions and concurrent sounds. It is less clear whether performance relies on the content of auditory feedback, which in a musical context refers to the pitches associated with individual notes.For some time, it appeared as though performance may not rely on feedback content. For instance, Howell and Archer (1984) found that disruption from DAF was unchanged when speech feedback was converted to an amplitude-modulated square-wave tone. Finney (1997) further demonstrated that pianists' performances of Bach inventions did not suffer when pitch content was altered in a random-like way. These observations are consistent with Howell's claim that DAF effects result from displacement of auditory rhythms, relative to the rhythmic coordination of actions (Howell, 2001, 2004; Howell, Powell, & Khan, 1983; Howell & Sackin, 2002).However, Pfordresher (2003) introduced an AAF manipulation of pitch content that can disrupt production, referred to as a serial shift. In contrast to other manipulations of content, serial shifts involve presenting pitches from a planned melody (i.e., a sequence of actions that are associated with a sequence of anticipated pitch events) at an alternate serial position. For instance, a lag-1 serial shift causes the feedback pitch heard at each keypress to match what the performer produced at the previous keypress. As such the feedback melody is serially shifted relative to the planned melody. Serial shifts lead to increased pitch errors in performance, while sparing produced timing (Pfordresher, 2003), can disrupt performance of both pianists and nonpianists (Pfordresher, 2005), can lead to disruption of music performance via keyboard or singing (Pfordresher & Mantell, 2012), and can also disrupt the accuracy of speech production (Pruitt & Pfordresher, 2015). It is important to note that serial shifts are distinct from traditional DAF, even though serial shifts may refer to previous events. For instance, hearing serially shifted pitches from future positions also disrupts performance (Pfordresher & Palmer, 2006).Although the effect of serial shifts does demonstrate sensitivity to pitch content, this effect alone does not rule out the possibility that their disruptive effect is to some extent based on rhythmic relationships between perception and action. It has often been observed that the patterning of pitch content can convey rhythmic information through melodic accents (Cooper & Meyer, 1960; Ellis & Jones, 2009; Jones, 1987), including breaks or pivots in the melodic contour (for examples, see Handel, 1989, p. 398), and the use of parallelism in broader pitch patterns (Acevedo, Temperley, & Pfordresher, 2014; Deutsch & Feroe, 1981; Steedman, 1977; Temperley & Bartlette, 2002).We adopted these features to create stimuli for the present study. Notation for the four melodies we used are shown in Figure 1 along with their meter, represented as a grid. Metrical accents are associated with positions that have higher columns of X's above them. Melodies were created to match either a binary (4/4) or a ternary (3/4) metrical structure. The first binary melody (Figure IA) used breaks in the melodic contour to signal metrical boundaries, whereas the second binary melody includes a recurring pattern structure based on an alternating melodic contour that is inverted in the second measure. …

Linguistically guided adaptation to foreign-accented speech.

Adaptation to foreign-accented sentences can be guided by knowledge of the lexical content of those sentences, which, being an exact match for the target, provides feedback on all linguistic levels. The extent to which this feedback needs to match the accented sentence was examined by manipulating the degree of match on different linguistic dimensions, including sub-lexical, lexical, and syntactic levels. Conditions where target-feedback sentence pairs matched and mismatched generated greater transcription improvement over non-English speech feedback, indicating listeners can draw upon sources of linguistic information beyond matching lexical items, such as sub- and supra-lexical information, during adaptation.