Target Speech Research Articles

Differences in speech intelligibility in noise between native and non-native listeners under ambisonics-based sound reproduction system

The current paper examines how native and non-native listeners of New Zealand English differ in terms of speech intelligibility in noise in a number of room acoustics reproduced by a first-order Ambisonics-based sound reproduction system. Speech intelligibility test was conducted under three room acoustics environments (living room, lecture theatre and church) using the sound reproduction system, where a pink noise masker was played from one of five azimuthal angles (0, 45, 90, 135, 180°) while the target speech was always played from 0°. We found significant two-way interactions between language nativeness and speech-noise separation, language nativeness and room acoustics, as well as between room acoustics and speech-noise separation. This suggests that native and non-native listeners respond differently to the virtually reproduced acoustic environments and they benefit from spatial release from masking in a different manner. Post-hoc results showed the native listeners performing significantly better than their non-native counterparts for all the angles of speech-noise separation and the room acoustics.

Analyzing the Targets of Hate in Online Social Media

Social media systems allow Internet users a congenial platform to freely express their thoughts and opinions. Although this property represents incredible and unique communication opportunities, it also brings along important challenges. Online hate speech is an archetypal example of such challenges. Despite its magnitude and scale, there is a significant gap in understanding the nature of hate speech on social media. In this paper, we provide the first of a kind systematic large scale measurement study of the main targets of hate speech in online social media. To do that, we gather traces from two social media systems: Whisper and Twitter. We then develop and validate a methodology to identify hate speech on both these systems. Our results identify online hate speech forms and offer a broader understanding of the phenomenon, providing directions for prevention and detection approaches.

Spatial release from masking in varying spatial acoustic under higher order ambisonic-based sound reproduction system

A previous study found that spatial release from masking (SRM) could be observed under virtual reverberant environments using a first order Ambisonic-based sound reproduction system, however, poor localisation accuracy made it difficult to examine effect of varying reverberation time on SRM. The present study follows on using higher order Ambisonics (HOA) to examine how benefits from SRM vary in different spatial acoustics. Subjective speech intelligibility was measured where four room acoustics:reverberation time (RT)= 0.7 s (clarity (C50)= 16 dB, 7 dB); RT= 1.8 s (C50= 8 dB, 2 dB) were simulated via a third order Ambisonic system with a 16 channel spherical loudspeaker array. The masker was played from 8 azimuthal angles (0, +-45, +-90, +-135, 180 degrees) while the target speech was played from 0 degree. The listeners are deemed to benefit from SRM if their intelligibility scores were higher when the masker comes from a different angle than that of the target. We found while listeners could benefit from SRM at C50 = 16 dB and 8 dB, the benefit starts to diminish at C50 = 7 dB, and listeners could no longer benefit from SRM at C50 = 2 dB.

Speech Understanding in Modulated Noise and Speech Maskers as a Function of Cognitive Status in Older Adults.

Objectives The purpose of this study was to investigate the impact of different types of maskers on speech understanding as a function of cognitive status in older adults. The hypothesis tested was that individuals with a diagnosis of mild cognitive impairment (MCI) or mild dementia would perform like their age- and hearing status-matched control counterparts in modulated noise but would perform more poorly in the presence of competing speech. Design Participants (n = 39; age range: 55-77 years old) performed a speech-in-noise task and completed two cognitive screening tests and a measure of working memory. Sentences were presented in the presence of two types of maskers (i.e., speech envelope-modulated noise and two-talker, same-sex competing speech). Two analyses were undertaken: (a) a between-groups comparison of individuals diagnosed with MCI/dementia, individuals who failed both cognitive screeners (possible MCI), and age- and hearing status-matched neurologically healthy control individuals and (b) a mixed-model analysis of variance of speech perception performance as a function of working memory capacity. Results The between-groups comparison yielded significant group differences for speech understanding in both masking conditions, with the MCI/dementia group performing more poorly than the neurologically healthy controls and possible MCI groups. A single measure of working memory (Size Comparison Span [SICSPAN]) was correlated with performance on the speech perception task in the competing speech conditions. Conclusions Adults with a diagnosis of MCI or mild dementia performed more poorly on a speech perception task than their age- and hearing status-matched control counterparts in the presence of both maskers, with larger group mean differences when the target speech was presented in a two-talker masker. This suggests increased difficulty understanding speech in the presence of distracting backgrounds for people with MCI/dementia. Future studies should consider how to target this potentially vulnerable population as they may be experiencing increased difficulty communicating in challenging environments.

Auditory attention decoding from electroencephalography based on long short-term memory networks

Available researches have confirmed that the target speech stream of human listener's selective attention in a multi-speaker scenario can be decoded from ongoing electroencephalography (EEG) signals. Due to its great potential application for the neuro-steered hearing devices, auditory attention decoding has recently received increasing attention. In practice, development of high-accuracy decoding performance and use of EEG signals with as short duration as possible are urgently needed for auditory attention decoding in real-time scenarios. However, these are still great challenges for the existing decoding methods to achieve such performances. In this paper, we propose a nonlinear modeling approach between EEG signals and speech envelopes by exploiting long short-term memory networks (LSTM), a special structure of recurrent neural networks (RNN), to decipher selective auditory attention from single-trial EEG signals, which is capable of learning long-term dependencies between EEG responses and continuous auditory stimuli. Unlike the existing linear modeling approaches, the LSTM RNN-based auditory attention decoding method is an end-to-end nonlinear decoding framework that subsumes the explicit similarity determination as part of feature learning in deep neural networks. The proposed decoding method is evaluated using the real EEG data collected from 21 subjects. Experimental results show that our proposed decoding method can achieve the state-of-the-art performance to identify the target speech in a competing two-talker scenario. When used the EEG signals with 1 s duration, the average accuracy for the 21 subjects' attention decoding is reached to 96.12%. The results show great potentials for the real-time auditory attention detection in neuro-steered hearing aids.

Inhibition Effect of Audio-Visual Semantic Interference in Chinese Interface: An ERP Study of Concrete Icons and Chinese Characters

In this study, event-related potential (ERP) was used to examine whether the brain has an inhibition effect on the interference of audio-visual information in the Chinese interface. Concrete icons (flame and snowflake) or Chinese characters (暖 and 凉) with opposite semantics were used as target carriers, and colors (red and blue) and speeches (热 and 冷) were used as audio-visual intervention stimuli. In the experiment, target carrier and audio-visual intervention were presented in a random combination, and the subjects needed to determine whether the semantics of the two matched quickly. By comparing the overall cognitive performance of two carriers, it was found that the brain had a more significant inhibition effect on audio-visual intervention stimuli with different semantics (SBH/LBH and SRC/LRC) relative to the same semantics (SRH/LRH). The semantic mismatch caused significant N400, indicating that semantic interference in the interface information would trigger the brain’s inhibition effect. Therefore, the more complex the semantic matching of interface information was, the higher the amplitude of N400 became. The results confirmed that the semantic relationship between target carrier and audio-visual intervention was the key factor affecting the cognitive inhibition effect. Moreover, under different intervention stimuli, the ERP’s negative activity caused by Chinese characters in frontal and parietal-occipital regions was more evident than that by concrete icons, indicating that concrete icons had a lower inhibition effect than Chinese characters. Therefore, we considered that this inhibition effect was based on the semantic constraints of the target carrier itself, which might come from the knowledge learning and intuitive experience stored in the human brain.

Negative impacts from latency masked by noise in simulated beamforming.

Those experiencing hearing loss face severe challenges in perceiving speech in noisy situations such as a busy restaurant or cafe. There are many factors contributing to this deficit including decreased audibility, reduced frequency resolution, and decline in temporal synchrony across the auditory system. Some hearing assistive devices implement beamforming in which multiple microphones are used in combination to attenuate surrounding noise while the target speaker is left unattenuated. In increasingly challenging auditory environments, more complex beamforming algorithms are required, which increases the processing time needed to provide a useful signal-to-noise ratio of the target speech. This study investigated whether the benefits from signal enhancement from beamforming are outweighed by the negative impacts on perception from an increase in latency between the direct acoustic signal and the digitally enhanced signal. The hypothesis for this study is that an increase in latency between the two identical speech signals would decrease intelligibility of the speech signal. Using 3 gain / latency pairs from a beamforming simulation previously completed in lab, perceptual thresholds of SNR from a simulated use case were obtained from normal hearing participants. No significant differences were detected between the 3 conditions. When presented with 2 copies of the same speech signal presented at varying gain / latency pairs in a noisy environment, any negative intelligibility effects from latency are masked by the noise. These results allow for more lenient restrictions for limiting processing delays in hearing assistive devices.

Emotionally Informed Hate Speech Detection: A Multi-target Perspective

Hate Speech and harassment are widespread in online communication, due to users' freedom and anonymity and the lack of regulation provided by social media platforms. Hate speech is topically focused (misogyny, sexism, racism, xenophobia, homophobia, etc.), and each specific manifestation of hate speech targets different vulnerable groups based on characteristics such as gender (misogyny, sexism), ethnicity, race, religion (xenophobia, racism, Islamophobia), sexual orientation (homophobia), and so on. Most automatic hate speech detection approaches cast the problem into a binary classification task without addressing either the topical focus or the target-oriented nature of hate speech. In this paper, we propose to tackle, for the first time, hate speech detection from a multi-target perspective. We leverage manually annotated datasets, to investigate the problem of transferring knowledge from different datasets with different topical focuses and targets. Our contribution is threefold: (1) we explore the ability of hate speech detection models to capture common properties from topic-generic datasets and transfer this knowledge to recognize specific manifestations of hate speech; (2) we experiment with the development of models to detect both topics (racism, xenophobia, sexism, misogyny) and hate speech targets, going beyond standard binary classification, to investigate how to detect hate speech at a finer level of granularity and how to transfer knowledge across different topics and targets; and (3) we study the impact of affective knowledge encoded in sentic computing resources (SenticNet, EmoSenticNet) and in semantically structured hate lexicons (HurtLex) in determining specific manifestations of hate speech. We experimented with different neural models including multitask approaches. Our study shows that: (1) training a model on a combination of several (training sets from several) topic-specific datasets is more effective than training a model on a topic-generic dataset; (2) the multi-task approach outperforms a single-task model when detecting both the hatefulness of a tweet and its topical focus in the context of a multi-label classification approach; and (3) the models incorporating EmoSenticNet emotions, the first level emotions of SenticNet, a blend of SenticNet and EmoSenticNet emotions or affective features based on Hurtlex, obtained the best results. Our results demonstrate that multi-target hate speech detection from existing datasets is feasible, which is a first step towards hate speech detection for a specific topic/target when dedicated annotated data are missing. Moreover, we prove that domain-independent affective knowledge, injected into our models, helps finer-grained hate speech detection.

Spatially target adaptive speech masking – A pilot study on masking effect and annoyance

The present paper proposes a speech masking system framework that adapts to the characteristics of the target speech. The characteristics are observed in three signal domains: temporal, spectral (frequency), and spatial, and are integrated into the design of the masker via how the masker is projected from loudspeakers. The effectiveness of the prototype design as a proof of concept was examined using a subjective listening test. Results show the proposed framework has potential to reduce the level of annoyance cused by the addition of masker to the environment while maintaining the effect of masking the target speech.

Common Brain Substrates Underlying Auditory Speech Priming and Perceived Spatial Separation.

Under a “cocktail party” environment, listeners can utilize prior knowledge of the content and voice of the target speech [i.e., auditory speech priming (ASP)] and perceived spatial separation to improve recognition of the target speech among masking speech. Previous studies suggest that these two unmasking cues are not processed independently. However, it is unclear whether the unmasking effects of these two cues are supported by common neural bases. In the current study, we aimed to first confirm that ASP and perceived spatial separation contribute to the improvement of speech recognition interactively in a multitalker condition and further investigate whether there exist intersectant brain substrates underlying both unmasking effects, by introducing these two unmasking cues in a unified paradigm and using functional magnetic resonance imaging. The results showed that neural activations by the unmasking effects of ASP and perceived separation partly overlapped in brain areas: the left pars triangularis (TriIFG) and orbitalis of the inferior frontal gyrus, left inferior parietal lobule, left supramarginal gyrus, and bilateral putamen, all of which are involved in the sensorimotor integration and the speech production. The activations of the left TriIFG were correlated with behavioral improvements caused by ASP and perceived separation. Meanwhile, ASP and perceived separation also enhanced the functional connectivity between the left IFG and brain areas related to the suppression of distractive speech signals: the anterior cingulate cortex and the left middle frontal gyrus, respectively. Therefore, these findings suggest that the motor representation of speech is important for both the unmasking effects of ASP and perceived separation and highlight the critical role of the left IFG in these unmasking effects in “cocktail party” environments.

Novel Approaches to Measure Spatial Release From Masking in Children With Bilateral Cochlear Implants.

To investigate the role of auditory cues for spatial release from masking (SRM) in children with bilateral cochlear implants (BiCIs) and compare their performance with children with normal hearing (NH). To quantify the contribution to speech intelligibility benefits from individual auditory cues: head shadow, binaural redundancy, and interaural differences; as well as from multiple cues: SRM and binaural squelch. To assess SRM using a novel approach of adaptive target-masker angular separation, which provides a more functionally relevant assessment in realistic complex auditory environments. Children fitted with BiCIs (N = 11) and with NH (N = 18) were tested in virtual acoustic space that was simulated using head-related transfer functions measured from individual children with BiCIs behind the ear and from a standard head and torso simulator for all NH children. In experiment I, by comparing speech reception thresholds across 4 test conditions that varied in target-masker spatial separation (colocated versus separated at 180°) and listening conditions (monaural versus binaural/bilateral listening), intelligibility benefits were derived for individual auditory cues for SRM. In experiment II, SRM was quantified using a novel measure to find the minimum angular separation (MAS) between the target and masker to achieve a fixed 20% intelligibility improvement. Target speech was fixed at either +90 or -90° azimuth on the side closer to the better ear (+90° for all NH children) and masker locations were adaptively varied. In experiment I, children with BiCIs as a group had smaller intelligibility benefits from head shadow than NH children. No group difference was observed in benefits from binaural redundancy or interaural difference cues. In both groups of children, individuals who gained a larger benefit from interaural differences relied less on monaural head shadow, and vice versa. In experiment II, all children with BiCIs demonstrated measurable MAS thresholds <180° and on average larger than that from NH children. Eight of 11 children with BiCIs and all NH children had a MAS threshold <90°, requiring interaural differences only to gain the target intelligibility benefit; whereas the other 3 children with BiCIs had a MAS between 120° and 137°, requiring monaural head shadow for SRM. When target and maskers were separated at 180° on opposing hemifields, children with BiCIs demonstrated greater intelligibility benefits from head shadow and interaural differences than previous literature showed with a smaller separation. Children with BiCIs demonstrated individual differences in using auditory cues for SRM. From the MAS thresholds, more than half of the children with BiCIs demonstrated robust access to interaural differences without needing additional monaural head shadow for SRM. Both experiments led to the conclusion that individualized fitting strategies in the bilateral devices may be warranted to maximize spatial hearing for children with BiCIs in complex auditory environments.

Application of Immersive Virtual Reality to Pragmatics Data Collection Methods

This exploratory study investigated the usability of immersive virtual reality (VR) as a way of creating a role-play task to examine pragmatic competence, specifically the ability to produce the speech act of request. The study created a closed role-play task in two versions. One was a standard computer-based version in which participants read a written scenario displayed on the screen and produced the target speech act for the computer. The other one was a VR version in which participants put on a VR headset and produced the target speech act for the interlocutor in the virtual space. Five native and five non-native speakers of English completed both versions and participated in a follow-up interview. The purpose of the interview was to examine similarities and differences in participants' perceptions of the two role-play tasks in four areas: (1) thought processes (what they were thinking during the task), (2) recall (what they remembered about the task), (3) difficulty (what made the task difficult), and (4) enjoyment (whether the task was fun). Results revealed that the participants attended to various audio-visual cues in the VR scene and used them to guide their actions. The VR version also evoked greater emotional reactions from the participants.

UWSpeech: Speech to Speech Translation for Unwritten Languages

Existing speech to speech translation systems heavily rely on the text of target language: they usually translate source language either to target text and then synthesize target speech from text, or directly to target speech with target text for auxiliary training. However, those methods cannot be applied to unwritten target languages, which have no written text or phoneme available. In this paper, we develop a translation system for unwritten languages, named as UWSpeech, which converts target unwritten speech into discrete tokens with a converter, and then translates source-language speech into target discrete tokens with a translator, and finally synthesizes target speech from target discrete tokens with an inverter. We propose a method called XL-VAE, which enhances vector quantized variational autoencoder (VQ-VAE) with cross-lingual (XL) speech recognition, to train the converter and inverter of UWSpeech jointly. Experiments on Fisher Spanish-English conversation translation dataset show that UWSpeech outperforms direct translation and VQ-VAE baseline by about 16 and 10 BLEU points respectively, which demonstrate the advantages and potentials of UWSpeech.

Benefits of triple acoustic beamforming during speech-on-speech masking and sound localization for bilateral cochlear-implant users.

Bilateral cochlear-implant (CI) users struggle to understand speech in noisy environments despite receiving some spatial-hearing benefits. One potential solution is to provide acoustic beamforming. A headphone-based experiment was conducted to compare speech understanding under natural CI listening conditions and for two non-adaptive beamformers, one single beam and one binaural, called "triple beam," which provides an improved signal-to-noise ratio (beamforming benefit) and usable spatial cues by reintroducing interaural level differences. Speech reception thresholds (SRTs) for speech-on-speech masking were measured with target speech presented in front and two maskers in co-located or narrow/wide separations. Numerosity judgments and sound-localization performance also were measured. Natural spatial cues, single-beam, and triple-beam conditions were compared. For CI listeners, there was a negligible change in SRTs when comparing co-located to separated maskers for natural listening conditions. In contrast, there were 4.9- and 16.9-dB improvements in SRTs for the beamformer and 3.5- and 12.3-dB improvements for triple beam (narrow and wide separations). Similar results were found for normal-hearing listeners presented with vocoded stimuli. Single beam improved speech-on-speech masking performance but yielded poor sound localization. Triple beam improved speech-on-speech masking performance, albeit less than the single beam, and sound localization. Thus, triple beam was the most versatile across multiple spatial-hearing domains.

Cortical Tracking of a Background Speaker Modulates the Comprehension of a Foreground Speech Signal.

Understanding speech in background noise is a difficult task. The tracking of speech rhythms such as the rate of syllables and words by cortical activity has emerged as a key neural mechanism for speech-in-noise comprehension. In particular, recent investigations have used transcranial alternating current stimulation (tACS) with the envelope of a speech signal to influence the cortical speech tracking, demonstrating that this type of stimulation modulates comprehension and therefore providing evidence of a functional role of the cortical tracking in speech processing. Cortical activity has been found to track the rhythms of a background speaker as well, but the functional significance of this neural response remains unclear. Here we use a speech-comprehension task with a target speaker in the presence of a distractor voice to show that tACS with the speech envelope of the target voice as well as tACS with the envelope of the distractor speaker both modulate the comprehension of the target speech. Because the envelope of the distractor speech does not carry information about the target speech stream, the modulation of speech comprehension through tACS with this envelope provides evidence that the cortical tracking of the background speaker affects the comprehension of the foreground speech signal. The phase dependency of the resulting modulation of speech comprehension is, however, opposite to that obtained from tACS with the envelope of the target speech signal. This suggests that the cortical tracking of the ignored speech stream and that of the attended speech stream may compete for neural resources.SIGNIFICANCE STATEMENT Loud environments such as busy pubs or restaurants can make conversation difficult. However, they also allow us to eavesdrop into other conversations that occur in the background. In particular, we often notice when somebody else mentions our name, even if we have not been listening to that person. However, the neural mechanisms by which background speech is processed remain poorly understood. Here we use transcranial alternating current stimulation, a technique through which neural activity in the cerebral cortex can be influenced, to show that cortical responses to rhythms in the distractor speech modulate the comprehension of the target speaker. Our results provide evidence that the cortical tracking of background speech rhythms plays a functional role in speech processing.

The effect of reverberation time on sound masking method to improve speech privacy in open-plan offices

Lack of speech privacy is a significant problem in open-plan offices. Sound masking, which essentially covers the target speech with the competing speech simultaneously (masker), can increase speech privacy. However, a conversation takes place in a reverberant room at which will reduce speech intelligibility. This study observed the effectiveness of reverberation time on the sound masking to the target. In this early study, three conditions of stimuli were used, namely clean, 0.3 s and 0.6 s reverberation time added to both target and masker, respectively. The STI and subjective measure were used with twenty-five volunteers. As expected, the results suggest that the longer the reverberation time, the lower the effectiveness of the masker strength. The STI agrees with the subjective measure.

Speech intelligibility in a realistic virtual sound environment.

In the present study, speech intelligibility was evaluated in realistic, controlled conditions. "Critical sound scenarios" were defined as acoustic scenes that hearing aid users considered important, difficult, and common through ecological momentary assessment. These sound scenarios were acquired in the real world using a spherical microphone array and reproduced inside a loudspeaker-based virtual sound environment (VSE) using Ambisonics. Speech reception thresholds (SRT) were measured for normal-hearing (NH) and hearing-impaired (HI) listeners, using sentences from the Danish hearing in noise test, spatially embedded in the acoustic background of an office meeting sound scenario. In addition, speech recognition scores (SRS) were obtained at a fixed signal-to-noise ratio (SNR) of -2.5 dB, corresponding to the median conversational SNR in the office meeting. SRTs measured in the realistic VSE-reproduced background were significantly higher for NH and HI listeners than those obtained with artificial noise presented over headphones, presumably due to an increased amount of modulation masking and a larger cognitive effort required to separate the target speech from the intelligible interferers in the realistic background. SRSs obtained at the fixed SNR in the realistic background could be used to relate the listeners' SI to the potential challenges they experience in the real world.

Impact of non-individualised head related transfer functions on speech-in-noise performances within a synthesised virtual environment.

When performing binaural spatialisation, it is widely accepted that the choice of the head related transfer functions (HRTFs), and in particular the use of individually measured ones, can have an impact on localisation accuracy, externalization, and overall realism. Yet the impact of HRTF choices on speech-in-noise performances in cocktail party-like scenarios has not been investigated in depth. This paper introduces a study where 22 participants were presented with a frontal speech target and two lateral maskers, spatialised using a set of non-individual HRTFs. Speech reception threshold (SRT) was measured for each HRTF. Furthermore, using the SRT predicted by an existing speech perception model, the measured values were compensated in the attempt to remove overall HRTF-specific benefits. Results show significant overall differences among the SRTs measured using different HRTFs, consistently with the results predicted by the model. Individual differences between participants related to their SRT performances using different HRTFs could also be found, but their significance was reduced after the compensation. The implications of these findings are relevant to several research areas related to spatial hearing and speech perception, suggesting that when testing speech-in-noise performances within binaurally rendered virtual environments, the choice of the HRTF for each individual should be carefully considered.

Selective attention modulates neural envelope tracking of informationally masked speech in healthy older adults.

Speech understanding in noisy situations is compromised in old age. This study investigated the energetic and informational masking components of multi‐talker babble noise and their influence on neural tracking of the speech envelope in a sample of healthy older adults. Twenty‐three older adults (age range 65–80 years) listened to an audiobook embedded in noise while their electroencephalogram (EEG) was recorded. Energetic masking was manipulated by varying the signal‐to‐noise ratio (SNR) between target speech and background talkers and informational masking was manipulated by varying the number of background talkers. Neural envelope tracking was measured by calculating temporal response functions (TRFs) between speech envelope and EEG. Number of background talkers, but not SNR modulated the amplitude of an earlier (around 50 ms time lag) and a later (around 300 ms time lag) peak in the TRFs. Selective attention, but not working memory or peripheral hearing additionally modulated the amplitude of the later TRF peak. Finally, amplitude of the later TRF peak was positively related to accuracy in the comprehension task. The results suggest that stronger envelope tracking is beneficial for speech‐in‐noise understanding and that selective attention is an important ability supporting speech‐in‐noise understanding in multi‐talker scenes.

Effects of speech-rhythm disruption on selective listening with a single background talker.

Recent work by McAuley et al. (Attention, Perception, & Psychophysics, 82, 3222-3233, 2020) using the Coordinate Response Measure (CRM) paradigm with a multitalker background revealed that altering the natural rhythm of target speech amidst background speech worsens target recognition (a target-rhythm effect), while altering background speech rhythm improves target recognition (a background-rhythm effect). Here, we used a single-talker background to examine the role of specific properties of target and background sound patterns on selective listening without the complexity of multiple background stimuli. Experiment 1 manipulated the sex of the background talker, presented with a male target talker, to assess target and background-rhythm effects with and without a strong pitch cue to aid perceptual segregation. Experiment 2 used a vocoded single-talker background to examine target and background-rhythm effects with envelope-based speech rhythms preserved, but without semantic content or temporal fine structure. While a target-rhythm effect was present with all backgrounds, the background-rhythm effect was only observed for the same-sex background condition. Results provide additional support for a selective entrainment hypothesis, while also showing that the background-rhythm effect is not driven by envelope-based speech rhythm alone, and may be reduced or eliminated when pitch or other acoustic differences provide a strong basis for selective listening.