New paper on melodic predictability
30th August 2021
Check out David Quiroga-Martinez’s new paper on how our brains respond to melodic deviations while listening to simple melodies. He conducted the DCM analyses while he was visiting us at UCL.
You can read the paper here:
Quiroga-Martinez, D. R., Hansen, N. C., Højlund, A., Pearce, M., Brattico, E., Holmes, E., Friston, K., & Vuust, P. (2021). Musicianship and melodic predictability enhance neural gain in auditory cortex during pitch deviance detection. Human Brain Mapping. https://doi.org/10.1002/hbm.25638
Moved to SHaPS UCL
23rd August 2021
Today is my first day as a lecturer in the Department of Speech Hearing and Phonetic Sciences at UCL. I’ve just moved 10 minutes down the road from Queen’s Square to my new office in Chandler House. I’m looking forward to my new role, and building my group here.
30th July 2021
Podcasts are great. I rarely listened to podcasts before the pandemic, but I’ve discovered loads of good ones over the past year, which I didn’t know existed. I usually listen to them when I’m out on a walk or a run. It turns out there are even neuroscience podcasts too! I recently talked about our work on familiar voices with Wilf from Watercooler.FM. I highly recommend giving some of their episodes a listen, which cover a variety of topics in neuroscience. Ours is Series 2, Episode 30, in case you’re interested.
On the radio
4th June 2021
Tonight, I’ll be interviewed by Jess Brady about our new paper on voice training on Canadian radio station AM 980. If you’re not in Canada, you can listen online. Next week, I’ll be appearing on CBC (Canadian Broadcasting Corporation)—stay tuned for details!
Two new papers on voice familiarity
13th May 2021
This week, two of our papers on voice familiarity were published online. Both papers address the problem of how we understand what someone’s saying in noisy places, such as cafes, restaurants, and pubs. Our research shows that understanding what someone’s saying is difficult when a competing talker is present, but it’s much easier if we’re listening to someone who we’re familiar with—such as a friend or partner—than someone who’s unfamiliar.
In a study published in Neuroimage, we investigated the neural correlates of this benefit by comparing multivariate fMRI when the same sentence was heard with and without a competing sentence. When the target voice was familiar, the spatial pattern of brain activity for the sentence+masker more closely resembled the same sentence presented alone, than when the target voice was unfamiliar. This effect was strongest at intermediate levels of the auditory cortical hierarchy: left posterior STG, MTG, and planum temporale. We found no evidence of the familiar-voice benefit at lower levels of cortical processing (primary auditory cortex) or at higher levels, such as inferior frontal gyrus. Interestingly, the ‘neural benefit’ of a familiar voice correlated with the behavioural benefit to speech intelligibility, and not the degree of familiarity. This suggests the pattern of fMRI responses in these areas reflects the benefit to intelligibility rather than familiarity per se.
In a second paper, published in Psychological Science, we investigated how rapidly speech intelligibility improves when a novel voice is trained to become familiar. We trained participants with three different voices for different lengths of time. We found that intelligibility was better for a voice that had been trained for ~10 minutes than an unfamiliar voice, but was even better for a voice trained for ~60 minutes. Intelligibility of the voice trained for ~60 minutes was comparable to that of a naturally familiar friend/spouse (that we’ve reported in previous studies). Interestingly, recognition of all three of the trained voices was significantly above chance, and was no better for talkers that had been trained for longer. We found a dissociation between explicit recognition of a trained voice and its benefit to intelligibility. This fits with our previous work showing that a familiar voice doesn’t need to be recognisable to benefit intelligibility. Overall, these results highlight the great potential of such training for improving intelligibility in everyday settings. Training may particularly benefit older people and people with hearing loss, who experience particular difficulty listening in noisy settings, and people whose occupations require accurate speech perception in noisy surroundings, such as aircraft pilots.
The two papers are available here:
- Holmes, E. & Johnsrude, I. S. (2021). Speech-evoked brain activity is more robust to competing speech when it is spoken by someone familiar. Neuroimage. https://doi.org/10.1016/j.neuroimage.2021.118107
- Holmes, E., To, G., & Johnsrude, I. S. (2021). How long does it take for a voice to become familiar? Speech intelligibility and voice recognition are differentially sensitive to voice training. Psychological Science. https://doi.org/10.1177/0956797621991137
Simulating lesion-dependent functional recovery mechanisms
2nd April 2021
In our new paper, led by Noor Sajid, we examined different ways in which recovery could affect performance after brain damage—for example, following a stroke. The paper has now been published in Scientific Reports.
Noor simulated how different types of lesions affect word repetition performance (i.e., repeating heard words). This was based on a generative model of word repetition, which consisted of a default premorbid system and an alternative (less effective) system that could produce the same outcome—and active inference, which assumes that behaviour is Bayes optimal. The simulated lesions either damaged the premorbid system but left the alternative system intact, or partially damaged both systems. The results showed that complete damage to the premorbid system engaged the alternative system, which led to an initial drop in performance, but this recovered relatively quickly. Whereas, incomplete damage to both systems damaged performance less in the short-term, but more in the long-term, because performance didn’t recover as much after damage.
Overall, these simulations reveal that functional recovery after brain damage depends on the availability of computational and structural resources. While the model could develop resistance to insult, the effects of damage could not be overcome beyond a certain point, leading to persistent impairments.
The paper is available here:
Sajid, N., Holmes, E., Hope, T. M., Fountas, Z., Price, C. J., & Friston, K. J. (2021). Simulating lesion-dependent functional recovery mechanisms. Scientific Reports, 11, 7475. https://doi.org/10.1038/s41598-021-87005-4
International Women’s Day
8th March 2021
This International Women’s Day, I’m feeling grateful for all the amazing female mentors I’ve had, who have given me advice, confidence, and support over the years. Sometimes even seemingly small things—such as a nice comment after a presentation, or the knowledge that more senior colleagues care about your career—have a lasting impact. Today, I enjoyed reading these interviews with women at UCL who are based in the Faculty of Brain Sciences and within the Institute of Neurology:
I’m also grateful to have had the opporutnity to contribute to Athena SWAN initiatives in departments where I’ve worked. Even though we still have a long way to go, these small steps do make a difference. Personally, I’ve learnt that tracking gender statistics within a department (which is often not as comprehensive as we would hope) is crucial for clarifying the causes of underepresentation in a workplace. Going forwards, I hope that we can continue to strengthen our reporting, and use the knowledge learned to improve representation.
25th February 2021
The ARO meeting was held online for the first time this year. While I was disappointed to miss the opportunity to visit sunny Florida in February (a month that is noticeably less sunny in the UK), the schedule was in keeping with its usual high standard. I particularly enjoyed talks by Matthew Winn (“Speech perception with a cochlear implant: Percent-correct scores do not explain listening effort”) and Lori Holt (“Supervised error-driven learning in speech perception”) and a symposium on “Age-Related Plasticity Changes in the Central Auditory System”. I also greatly enjoyed one of the spARO events prior to the meeting, in which award-winning comedian Kathy Buckley talked about her experience of hearing loss and told some fascinating stories about people she’s met during her career. Her story is incredibly inspirational and I strongly recommend checking out her work.
Digital Engagement Project
24th January 2021
If you follow my research, you might remember that we recently developed some auditory figure-ground tests that measure fundamental auditory grouping processes. They assess people’s ability to group a pattern of tones against a background of random tones, which is relevant to understanding speech in noisy settings.
In The Hearing Brain Project—which was funded by an engagement grant from the Wellcome Centre for Human Neuroimaging—we co-designed some new animations, with input from a focus group, to help explain these sounds. I’ve created a video to describe the project and to showcase our new animations, which you can watch on YouTube at the following link: https://www.youtube.com/watch?v=91IwZFW613A
We envisage using these new animations to explain our task to patients and future research participants, and to disseminate our research more widely. We’re really grateful to the 5 volunteers who participated in the focus group, which took place over a series of 4 video calls, and to the professional designers, who helped bring our ideas to life. The volunteers reported that they find it difficult to listen in noisy places, and—as well as their help designing the animations—I learnt a lot about their experiences and preferences and really enjoyed getting to know them.
You can read a summary of the project on our centre’s website: https://engagement.fil.ion.ucl.ac.uk/grants/digital-engagement-grants-2020/
If you’d like to read more about the science behind our figure-ground tests, you can find some of our relevant papers here:
- Holmes, E., & Griffiths, T. D. (2019). ‘Normal’ hearing thresholds and fundamental auditory grouping processes predict difficulties with speech-in-noise perception. Nature Scientific Reports, 9, 16771. https://doi.org/10.1038/s41598-019-53353-5
- Holmes, E., Zeidman, P., Friston, K. J., & Griffiths, T. D. (2021). Difficulties with speech-in-noise perception related to fundamental grouping processes in auditory cortex. Cerebral Cortex. https://doi.org/10.1093/cercor/bhaa311
Patient case study in press at Cortex
7th December 2020
Recently, we had the rare opportunity to study a young patient who experienced a stroke, which affected the right hemisphere of her brain. After the stroke, she told us that she found it difficult to listen in environments containing multiple sounds, such as understanding speech in noisy places and picking out melodies in music.
We conducted a wide range of tests, including discrimination of basic acoustic features, more complex scene analysis, and tests involving overlapping sounds. She showed a general segregation deficit, displaying atypical performance for words-in-noise, music streaming, and figure-ground perception—despite intact peripheral function, working memory, and recognition of familiar melodies. In other words, she showed impaired segregation between objects when competing sounds were present, but preserved within-object analysis when we tested auditory features in isolation. Intriguingly, her deficits were most pronounced for sounds presented to her left ear.
The lesion affected the right inferior parietal lobule, posterior insula, and auditory cortex including planum temporale, but spared medial Heschl’s gyrus. We, therefore, think these impairments in auditory segregation are related to damage to non-primary auditory cortex including planum temporale, which has been implicated in auditory segregation in previous imaging studies. This unique case shows that these regions are critically involved in auditory segregation and reveal a new type of auditory agnosia, which is associated with similar symptoms as the visual symptom of simultaneous agnosia.
You can read the paper online here:
Holmes, E., Utoomprurkporn, N., Hoskote, C., Warren, J., Bamiou, D. E., & Griffiths, T. D. (2020). Simultaneous auditory agnosia: Systematic description of a new type of auditory segregation deficit following a right hemisphere lesion. Cortex. https://doi.org/10.1016/j.cortex.2020.10.023
21st October 2020
I’m looking forward to the first ever online APAN conference, which starts tomorrow. Even better, there’s no registration fee this year, which means that it’s a great time to get involved, no matter what your funding situation or where you’re located.
The conference agenda and a searchable list of abstracts are available online. All of the posters are available to view already, and most also have a 3-minute video digest, which I think is a great idea (even if I hated recording my own!).
I’ll be presenting my new work on modelling attention during cocktail party listening. If this sounds like something that might interest you, feel free to check out my poster (#94), 3-minute digest, or come and talk to me at the virtual poster session on Friday 22nd at 11:15–12:15 EDT (4:15–5:15pm UK time). Hope to see you there!
New paper accepted in Cerebral Cortex
23rd September 2020
In a previous study, we found that the ability to hear speech in noisy places varies widely among people, and relates to auditory figure-ground perception (Holmes & Griffiths, 2019). In other words, people who are better at figure-ground perception are better at speech-in-noise perception. In our new paper—which has just been accepted for publication in Cerebral Cortex—we examine the brain basis of the link between figure-ground and speech-in-noise perception using functional MRI and dynamic causal modelling (DCM).
We found common processes for figure-ground and speech-in-noise perception in auditory cortex: When the two tasks were more difficult, we found evidence for common disinhibition (implying an increase in gain) at the earliest stages of the auditory cortical hierarchy, including left primary auditory cortex. Ultimately, these results suggest a common cortical substrate that links perception of basic and natural sounds—and might explain why people who are worse at figure-ground perception are also worse at speech-in-noise perception.
The paper is available here:
Holmes, E., Zeidman, P., Friston, K. J., & Griffiths, T. D. (in press). Difficulties with speech-in-noise perception related to fundamental grouping processes in auditory cortex. Cerebral Cortex. https://doi.org/10.1093/cercor/bhaa311
How can hearing loss cause dementia?
31st August 2020
Hearing loss in midlife has been estimated to account for 9% of cases of dementia, and is thought to be the largest modifable risk factor. In our new perspective piece, we consider the directionality of this effect and possible brain bases for the link. We hope that this paper will be a useful starting point for researchers who wish to examine this link empirically.
You can read the paper here:
Griffiths, T. D., Lad, M., Kumar, S., Holmes, E., McMurray, B., Maguire, E. A., Billig, A. J., & Sedley, W. (2020). How can hearing loss cause dementia? Neuron. https://doi.org/10.1016/j.neuron.2020.08.003
Or, you can read coverage of the paper by Fox News here: Hearing loss may cause dementia, study finds
Working memory and speech-in-noise perception
19th August 2020
The ability to hear speech in noisy places varies widely among people (see Holmes & Griffiths, 2019). In our new paper, we tested the hypothesis that some of this variability relates to a general ability to hold sounds in mind, (i.e., auditory working memory) even when the sounds aren’t speech. This process might help us to combine foreground elements, like speech, over seconds to help separate them from the background of an auditory scene (e.g., noise or other people’s conversations).
We used a novel paradigm that tests auditory working memory for non-speech sounds that vary in frequency and amplitude modulation rate. We found that people who had better working memory for frequency were also better at understanding sentences in a noisy background (babble noise).
However, we found no correlations between speech-in-noise perception and working memory for amplitude modulation rate or phonological working memory (i.e., the ability to hold and manipulate phonemes or words in mind)—suggesting a special relationship between working memory for frequency and the ability to hear speech in noise.
The full paper is available here:
Lad, M., Holmes, E., Chu, A., and Griffiths, T. D. (2020). Speech-in-noise detection is related to auditory working memory precision for frequency. Scientific Reports, 10, 13997. https://doi.org/10.1038/s41598-020-70952-9
Generative models, linguistic communication and active inference
31st July 2020
We have another new modelling paper out this week. We describe a hierarchical model for linguistic communication, which produces theta-gamma coupling.
We illustrate linguistic communication by simulating a synthetic subject who can play the ‘20 Questions’ game. In this game, synthetic subjects take the role of the questioner or answerer, using the same hierarchical generative model.
We frame communication as an active Bayesian inference problem, which leads to forward-looking conversations. Crucially, this differs from previous approaches in the literature: it explicitly considers uncertainty, and the model can generate questions to resolve uncertainty. We show that a synthetic agent equipped with this generative model asks the minimum number of questions necessary to reach the correct answer in the ’20 Questions’ game. It doesn’t ask questions that have already been answered by it’s previous questions.
We simulate EEG responses using standard belief update schemes. Playing the ’20 Questions’ game induces belief updates at different levels of the model—producing a nesting of gamma responses at a theta rate. Empirically, this would be interpreted as theta-gamma coupling.
The paper includes several other simulations of communication between two agents. We show that, if the synthetic agent has precise beliefs about the nature of the scene, it can correctly answer another agent’s questions—and the beliefs of the two agents converge. We also simulated situations where, if the agent has very imprecise beliefs, it will acknowledge its own uncertainty by answering “I don’t know”. In addition, we illustrate a folie à deux: If two agents both start with imprecise beliefs about a visual scene, then their generative models will converge, even though neither agent knows the veridical state of a scene.
Overall, we envisage that this model will be a useful starting point for simulating more complex linguistic exchanges—that include metacognition, or which simulate language acquisition. You can read the full paper here:
Friston, K. J., Parr, T., Yufik, Y., Sajid, N., Price, C. J., & Holmes, E. (2020). Generative models, linguistic communication and active inference. Neuroscience & Biobehavioral Reviews. https://doi.org/10.1016/j.neubiorev.2020.07.005
How do people recognise speech?
20th May 2020
In our new paper, we introduce a novel model for synthesising and recognising speech.
Speech recognition is a complex problem. Our ears receive a continuous acoustic signal and, in order to understand the words that are spoken, we must parse the continuous signal into discrete words. But there is not a one-to-one mapping between words and the acoustic signal.
In our paper, we consider speech recognition as a Bayesian inference problem (based on active inference). Our generative model maps lexical, speaker, and prosodic factors to a continuous acoustic signal. The model can be inverted to recognise words, given the speech signal. To recognise sentences, the model must also segment speech. We conceptualise speech segmentation as an active process: Possible word boundaries are hypotheses, and the model infers the most likely combinations of words (which determines the segmentation). This enables the model to distinguish between alternative lexically-plausible segmentations, such as “Grade A” and “Grey day”. Segmentations are disambiguated via lexical priors. Thus, the model segments speech most consistent with its prior expectations (like humans do).
We also use the model to synthesise expected neuronal responses. These responses are based on belief updating under the generative model—which is the change in free energy as evidence is accumulated. We show that mismatch responses (e.g., MMN) emerge from this framework.
Overall, this framework generates quantitative (testable) predictions for behaviour and neural responses, using processes not specific to speech recognition. We anticipate this model could be useful for interpreting empirical data.
You can read the full paper here:
Paper featured on JASA landing page
1st June 2020
Ysi Domingo’s paper—Using spatial release from masking to estimate the magnitude of the familiar-voice intelligibility benefit—has been selected to feature on the landing page of the JASA website for the next 3 months. In this paper, we show that being familiar with someone’s voice provides a speech intelligibility benefit as large as spatially separating maskers by +/-15 degrees azimuth. You can read the full paper here:
Domingo, Y., Holmes, E., Macpherson, E., & Johnsrude, I. S. (2019). Using spatial release from masking to estimate the magnitude of the familiar-voice intelligibility benefit. The Journal of the Acoustical Society of America, 146(5), 3487–3494.
Audio code on GitHub
18th April 2020
I’ve recently created a new repository on my GitHub page, which contains some miscallaneous scripts that are relevant to auditory research. I’ve uploaded Praat scripts that I’ve used to analyse and manipulate the duration and pitch of speech recordings. I’ve also uploaded a MATLAB script for normalising the root-mean-square (RMS) amplitude of audio files within a directory. I found these types of scripts useful for my voice familiarity work, in which I manipulated the pitch of sentences spoken by a large number of talkers (e.g., Holmes et al., 2018, Psych Science; Holmes et al., 2020, JEP LMC). I’ve recently edited the scripts so that they’re more general, which will hopefully mean that they’re useful to other researchers.
Finally, I’ve uploaded a script for plotting the results of pure-tone audiograms. This script can be used for plotting either group results, or the results from an individual participant. I used this script to plot the group audiogram results from 97 participants in Holmes & Griffiths (2019) (Figure 2).
Get in touch if you’ve used these scripts and/or found them useful!
Action on Hearing Loss Future Leaders Day
19th March 2020
This time last month, I was invited to the Future Leaders Day at the Action on Hearing Loss head office in Highbury. I really enjoyed hearing more about the variety of ways in which Action on Hearing Loss are helping people with hearing loss, including initiatives to make workplaces deaf aware. I presented my research plan for my fellowship to staff and supporters, which led to some really engaging discussions. You can read more about the day on the Action on Hearing Loss website.
New paper accepted in Journal of Experimental Psychology
28th February 2020
In noisy places, it can be difficult to understand what someone’s saying. Yet, we can better understand words spoken by our friends and family members than the same words spoken by a stranger. In our latest paper—accepted for publication in the Journal of Experimental Psychology—we wanted to understand how familiar voices improve intelligibility. To that aim, we tested intelligibility of familiar and unfamiliar voices with a variety of maskers.
We recruited pairs of friends and romantic couples. During the experiment, they were asked to listen to listen to the sentence that began with a particular word, and tell us the words they heard in that sentence. Sometimes, the sentence they were listening to was spoken by their friend or partner, but other times it was spoken by someone they had never met. We always presented a competing stimulus at the same time: it was either a different talker speaking a sentence in the same language as the target (English), a different talker speaking a sentence in a language that was incomprehensible to the listener (Spanish or Russian), or unintelligible noise (constructed from the sentences presented in the other conditions).
We replicated the familiar-voice benefit to intelligibility that we have found in previous studies (see Johnsrude et al., 2013, Holmes et al., 2018, Domingo et al., 2019, and Domingo et al., 2019). Moreover, we found the largest benefit of a familiar voice when one other person was speaking a sentence from the same language at the same time. We found the benefit of a familiar voice was significantly smaller when the other person was speaking a different language that the listener could not comprehend. We found no familiar-voice benefit when the masker was unintelligible noise. These results suggest that familiar voices did not benefit intelligibility because they were more predictable or because they attracted greater attention than unfamiliar voices; rather, familiarity with a target voice reduces interference from maskers that are linguistically similar to the target.
Overall, our results demonstrate that the benefit to speech intelligibility from a naturally familiar (compared to unfamiliar) voice differs under different masking conditions. They suggest that familiarity with a target voice improves intelligibility by helping listeners to avoid interference from distractors that are linguistically similar to the target.
The paper is available here:
Holmes, E., & Johnsrude, I. S. (2020). Speech spoken by familiar people is more resistant to interference by linguistically similar speech. Journal of Experimental Psychology: Learning Memory and Cognition [Epub ahead of print]. doi:10.1037/xlm0000823
24th January 2020
I’m on my way to San Jose, CA, for the 2020 ARO Mid-Winter Meeting. Tomorrow, I’ll be talking about a new model for generating and recognising speech (“Active Listening”; PD 7), and on Sunday I’ll be presenting some fMRI work showing common neural substrates for figure-ground and speech-in-noise perception (PS 286). Looking forward to seeing many of you there. Come and say hello if you’re around
New year, new fellowship
23rd December 2019
I’m delighted to announce that I’ve been awarded a Pauline Ashley Fellowship from Action in Hearing Loss. I’m excited to get working on this project in the new year—which includes MEG and behavioural experiments and new modelling work. Watch this space!
Our research is featured in the Daily Mail
14th December 2019
Findings from our recent paper, which was published in Scientific Reports, has been featured in the Daily Mail, in a nice piece written by Victoria Allen. I enjoyed talking to Victoria about our research. Based on the comments, it appears that lots of people subjectively feel that they have ‘party deafness’. This may be particularly noticeable over the festive season, which is commonly filled with noisy parties and loud family gatherings. We hope that our tests might be useful in the future for assessing this type of difficulty, which is not well assessed by current clinical tests. We are currently working on adapting these tests so they are more suitable for the clinic, and our colleagues have been investigating these tests on children diagnosed with “auditory processing disorder” and people who have recently received cochlear implants. We are also considering training approaches that might help to improve listening in noisy places.
You can read the piece online at the Daily Mail—which hopefully provides a nice break from the U.K. election news! I’ve also put a couple of examples of our new tests on YouTube, so you can have a listen for yourself.
Why do some people find it difficult to understand speech in noisy places?
22nd November 2019
Lots of people find it really difficult to understand speech in noisy places. It’s been estimated that 5-15% of people who visit audiology clinics reporting hearing difficulty turn out to have “normal hearing”, as defined by current clinical standards. We currently don’t know why this is, or what causes this difficulty. In our new paper, published in Nature Scientific Reports, we show this difficulty can be caused by several factors. In other words, different people find it difficult to understand speech in noisy places for different reasons.
One reason is that there is large variability in clinical test results, even among people whose hearing is in the range that is widely considered to reflect “normal hearing”. The most widespread clinical test is to measure the quietest sounds that someone can hear at different frequencies. A cut-off value is used to determine whether someone has “hearing loss”. However, we found that—even if someone doesn’t have hearing worse than the cut-off—people who are closer to the cut-off are more likely to experience difficulty hearing in noisy places. Therefore, these clinical measures that are widely used are telling us more about someone’s hearing than we often assume.
The second reason is differences in grouping sounds. Sounds that reach our ears are continuous and overlapping. Our brains help us to group sounds we’re interested in and ignore other sounds. We found that people who are worse at grouping sounds are more likely to experience difficulty hearing in noisy places. Difficulty grouping sounds was a previously unknown factor affecting the ability to hear speech in noisy places, and we developed new tests to measure it. Our tests use simple tone patterns that are much simpler than language. We call these ‘figure-ground stimuli’. The ‘figure’ part consists of a pattern that—if grouped successfully—‘pops out’ from the background. I have created a YouTube video to demonstrate some examples; you can listen to these by clicking on the following link: https://youtu.be/Q19E8cOQWkU. Interestingly, we found that two different types of sounds help to tease apart different grouping processes that explain why people find it difficult to understand speech in noisy places. Some people find it difficult because they struggle to group sounds that remain the same frequency over time (first set of examples in the YouTube video), whereas other people find it difficult because they struggle to track sounds that change frequency (second set of examples in the YouTube video). People tend to do worse on these tests as they get older. However, our results show these tests are also useful for assessing individual differences among people who are the same age.
Ultimately, we hope that these new tests might be useful as a ‘cognitive audiogram’ that could be used alongside existing measures in the clinic. The tests may be particularly helpful for predicting hearing outcomes in populations who cannot be tested with language (e.g., young children, or people who are not fluent in the language that is spoken in their clinic). The tests might also be useful for predicting speech outcomes early after someone has received a cochlear implant, and we are currently testing this with our collaborators in Iowa.
You can visit the following link to read a media story about this research:
You can read the full paper here:
Holmes, E., & Griffiths, T. D. (2019). ‘Normal’ hearing thresholds and fundamental auditory grouping processes predict difficulties with speech-in-noise perception. Nature Scientific Reports, 9, 16771. doi:10.1038/s41598-019-53353-5
13th November 2019
We’ve created some new auditory figure-ground stimuli, which we tested in listeners with normal hearing. These stimuli consist of rapidly presented chords of pure tones. Participants had to detect a ‘pattern’ (the ‘figure’) in the stimuli, in which the tones stayed the same frequency at adjacent chords. Sometimes, the ‘figure’ had a gap in the middle, and other times it didn’t. I’ve created a demo video containing some example stimuli, which is available here: https://youtu.be/Q19E8cOQWkU
Two new familiar-voice papers
30th October 2019
Two of our new papers have recently come out, which were led by Ysabel Domingo.
The first—published in Journal of Experimental Psychology: Applied—looked at the familiar-voice benefit to speech intelligibility among people of different ages. We found that the benefit to intelligibility when a target talker is familiar is robust across settings. The magnitude of the benefit in young adults whose familiar voices were friends they had known for only half a year or longer was similar to the benefit in older adults whose familiar voices were partners they had lived with for more than 5 years. This implies that the familiar voice benefit develops relatively rapidly after we get to know someone as a friend and remains stable as we continue to know someone for longer periods of time. In both experiments, we found no benefit to intelligibility when the familiar voice was the masker.
In the second paper—published in JASA—we compared the familiar-voice benefit to intelligibility with the benefit due to spatial release from masking. These benefits are among the largest known intelligibility benefits, yet the former is cognitive and the latter is acoustic. We looked at the amount that two maskers needed to be symmetrically separated (in the azimuth plane) from a target voice for it to be as intelligible as a familiar voice with collocated maskers. We found that being familiar with a voice provides a speech intelligibility benefit as large as spatially separating maskers by +/-15 degrees
The papers are available here:
Domingo, Y., Holmes, E., & Johnsrude, I. S. (2019). The benefit to speech intelligibility of hearing a familiar voice. Journal of Experimental Psychology: Applied.
Domingo, Y., Holmes, E., Macpherson, E., & Johnsrude, I. S. (2019). Using spatial release from masking to estimate the magnitude of the familiar-voice intelligibility benefit. The Journal of the Acoustical Society of America, 146(5), 3487–3494.
New paper in Neuroimage
29th June 2019
Our new paper—Variational representational similarity analysis—is now available online at Neuroimage. This paper introduces variational RSA, a new multivariate approach.
Functional neuroimaging data are usually multivariate—for example, comprising measurements of brain activity at multiple (MRI) voxels or (MEG/EEG/ECoG) channels. Representational similarity analysis (RSA) is often used to quantify the similarity of multivariate activity between experimental conditions. Unlike classic RSA approaches, this paper describes a method for using standard variational inference procedures to quantify the contributions of particular patterns to the data. This enables us to use Parameteric Empirical Bayes to test for consistent effects across participants, and Bayesian Model Comparison to test hypotheses about condition-specific effects. Happily, this scheme allows one to test for mixtures of hypotheses using Bayesian Model Comparison.
You can read the paper online here:
Friston, K. J., Diedrichsen, J., Holmes, E., & Zeidman, P. (2019). Variational representational similarity analysis. Neuroimage
19th June 2019
I’m on my way to the University of Manchester to give a talk. I’m looking forward to presenting some new unpublished data, and catching up with colleagues.
18th June 2019
Tomorrow, I’ll be giving a talk at the Brain Meeting at the Wellcome Centre for Human Neuroimaging. It’s been great to look back over the years and bring together my current and past research—and appreciate the parallels between independent lines of research at different stages of my career.
11th February 2019
I’m currently at the ARO mid-winter meeting in Baltimore, which is a great meeting as usual! I enjoyed the symposium this afternoon on Hearing in Aging, featuring talks by two of my previous lab-mates (at different times)—Adele Goman and Björn Herrmann. Tomorrow, I’ll be co-chairing the Speech Perception session, which looks like it should be a really interesting session. At 2:45pm tomorrow, I’ll be talking about central and peripheral contributions to speech-in-noise variability among people with ‘normal’ hearing. Many thanks to Guarantors of Brain for funding my travel.
Speech in Noise Conference
11th January 2019
The annual Speech in Noise (SpiN) meeting seems to grow every year! Thanks to Sarah Verhulst for organising such an interesting meeting and for introducing us to the lovely (albeit a little rainy) city of Ghent!
Quentin Summerfield Festschrift
19th December 2018
Today marks 4 years since my PhD viva. It also happens to be the day that we’ll meet to celebrate the career of my PhD supervisor: Professor Quentin Summerfield. I’m grateful to have had such a wonderful PhD supervisor, who is renowned for applied research that has influenced health policy, improving the lives of people with hearing loss. Today, I’ll present some of the work we did together, and explain how it’s shaped my current thinking and research interests.
Upcoming talk at UCL Speech Sciences Forum
5th December 2018
Tomorrow, I’ll be giving a talk at the UCL Speech Sciences Forum. It’ll be held at Chandler House at 4pm—please feel free to come along. I’ll be talking about how cognition affects speech perception in people with normal and impaired hearing.
Two Different Ways Familiar Voice Information Can Be Used
29th November 2018
The Canadian Audiologist recently published our article, which summarises some of our key findings on familiar voices. Preparing this also reminded me about all the related work we’ve done on this topic, which should be available in the coming year.
You can find the article here:
Johnsrude, I., Holmes, E., & Deeks, J. (2018). Two Different Ways Familiar Voice Information Can Be Used. Canadian Audiologist, 5(6).
22nd November 2018
I’m very happy to visit Aston University today to talk about my research. I’ll present work demonstrating that listeners with moderate hearing loss have impaired selective attention, reasons why speech-in-noise perception varies substantially among people with ‘normal’ hearing, and how we can utilise knowledge about cognition to improve speech intelligibility.
Attention to Sound Meeting
14th November 2018
This Royal Society meeting was held at Chicheley Hall—a lovely location in the Buckinghamshire countryside. Most of all, I most enjoyed the relaxed discussion sessions, in which we debated topics related to auditory attention. Such an amazing programme of talks by several of my auditory heroes.
1st November 2018
I’m on my way to San Diego for my first Society for Neuroscience (SfN) meeting! I’ve been wanting to go to this conference for a while and am very grateful for the SfN Trainee Development award that has enabled me to attend this year. I’m also looking forward to the satellite meeting tomorrow on Advanced Perspectives in Auditory Neuroscience (APAN). The programs are available through the following links:
- SfN: http://www.sfn.org/Meetings/Neuroscience-2018/Sessions-and-Events
- APAN: https://www.med.upenn.edu/apan/assets/user-content/documents/apan2018/APAN_2018_talks_abstracts.pdf
1st October 2018
The following are press releases for our new Psych Science paper on voice familiarity:
- APS: https://www.psychologicalscience.org/news/releases/familiar-voices.html
- Western BrainsCAN: https://mediarelations.uwo.ca/2018/10/01/familiar-someones-voice-help-us-understand-theyre-saying/
BSA Basic Auditory Science Meeting
2nd September 2018
I’m on my way to The BSA Basic Auditory Science Meeting 2018 in Newcastle. This is my first BSA meeting since I returned to the UK. I’m looking forward to catching up with old colleagues and presenting a talk on my new work: ‘Normal’ hearing thresholds and figure-ground perception explain significant variability in speech-in-noise performance.
Familiar voices are more intelligible, even if they are not recognized as familiar
16th August 2018
Our paper in Psych Science is now available online! In this experiment, we were interested in why speech is more intelligible when it is spoken by familiar than unfamiliar people (see Johnsrude et al., 2013). In other words, what characteristics of someone’s voice do we rely on to better understand speech spoken by familiar people? We also asked whether these characteristics the same ones as those we use to recognise someone from their voice (for example, recognising that the person speaking to you on the phone is your mother).
We recruited pairs of friends who had known each other for longer than 6 months. Each participant heard sentences spoken by their friend and sentences spoken by the friends of other participants, who they did not know. We manipulated the recorded sentences, to simulate changes in vocal tract length (which determines the formants of speech) and/or voice pitch (which are determined by the frequencies). Large changes in voice pitch reduced participants’ ability to recognise their friend’s voice and reduced the intelligibility of familiar than unfamiliar voices, but did not completely eliminate the ability to perform either task (i.e., participants could still recognise the pitch-manipulated voice as familiar and they gained a small intelligibility benefit from that voice). By contrast, large changes in vocal tract length reduced the intelligibility of familiar voices, but completely eliminated the ability to recognise a friend’s voice as familiar. We found a significant interaction between voice manipulation condition (i.e., manipulations to voice pitch or vocal tract length) and task (i.e., recognition or intelligibility task), confirming that acoustic properties of a voice contribute differently to the ability to recognise a familiar person from their voice and the ability to understand the words that a familiar person is speaking. Moreover, most participants who were unable to recognise their friend’s voice as familiar when it was manipulated still received a speech intelligibility benefit from this same voice (i.e., participants were better at reporting words in the manipulated familiar voice than the same words in an unfamilar voice, who the participant had never met). These results demonstrate an interesting dissociation between the ability to recognise someone from their voice and the ability to understand the words that someone is speaking, suggesting that we use familiar-voice information differently in different contexts.
If you’d like to read more about this research, here’s a link to the paper:
Holmes, E., Domingo, Y., & Johnsrude (2018). Familiar voices are more intelligible, even if they are not recognized as familiar. Psychological Science. https://doi.org/10.1177/ 0956797618779083 [Epub ahead of print]
22nd May 2018
I’m on my way home from the 2nd Cambridge Representational Similarity Analysis and Advanced Computational Methods Workshop and it’s been a fantastic two days. Alex Wooglar gave a great talk on teasing apart hypotheses of attention using RSA for MEG, and she showed a novel Granger causality approach to RSA. I also enjoyed Cai Wingfield’s talk, where he described some studies combining automatic speech recognition approaches with M/EEG data recorded while participants listened to spoken words (for more information, see their paper). Johan Carlin’s talk was also very insightful, showing how differences in mean activity can influence the structure of multi-dimensional scaling output in surprising ways. Some interesting discussion followed about whether we can and should try to tease apart differences in ‘patterns’ of activity from differences in mean activity.
15th May 2018
I’ll be heading up to Newcastle tomorrow for a joint meeting of the Phonetics & Phonology and Auditory groups. I’m looking forward to talking about my new experiment (“Developing auditory figure ground tasks to better explain variation in speech-in-noise performance”) and to lots of interesting discussions about speech and language.
New paper accepted in Psych Science
5th May 2018
One of our papers on voice familiarity has recently been accepted in Psychological Science. In this paper, we show that familiar-voice information is used differently in different contexts—depending on whether the goal is to recognise someone from their voice or to understand the words that person is saying. I’ll post more details here when the paper goes live…
Preparing attention improves speech intelligibility
27th April 2018
Our new paper in Attention, Perception, & Psychophysics is now available online. In this paper, we show that preparing attention for an upcoming talker improves speech intelligibility, with greater benefits when we have longer time to prepare. This finding supports a mechanism by which partially preparing attention improves performance, but performance does not peak until we are in a fully prepared state.
In this experiment, we presented an visual cue that instructed participants to attend to a talker who was at a target location (left/right) or who was of a target gender (male/female). We varied the cue-target interval between the time at which the visual cue was revealed and the time at which the target talker and two other distracting talkers started to speak. Response times for reporting words spoken by the target talker became significantly shorter as the duration of the cue-target interval increased from 0 to 2 seconds. Most previous experiments investigating endogenous visual and/or auditory attention have not tested cue-target intervals as long as 2 seconds, yet we found a significant improvement in performance between cue-target intervals of 1 and 2 seconds. Thus, our findings suggest that future experiments should include intervals up to and beyond 2 seconds to maximise benefits of advance cueing on task performance.
If you want to read more about the experiment, here’s the paper:
Holmes, E., Kitterick, P. T. & Summerfield, A. Q. (2018). Cueing listeners to attend to a target talker progressively improves word report as the duration of the cue-target interval lengthens to 2000 ms. Attention, Perception, & Psychophysics. [Epub ahead of print]
Talk at Royal Holloway this week
19th February 2018
On Wednesday (21st February), I’ll be visiting Royal Holloway and will be giving a talk on our recent work on familiar voices. If you’re around, come along to find out (1) how being familiar with someone’s voice helps you to better understand what they’re saying, (2) how long it takes to become familiar with a new voice, and (3) how patterns of brain activity elicited by a voice depend on its familiarity.
Semantic context reduces listening effort for hearing-aid users
17th February 2018
Our new paper has just been published online in the International Journal of Audiology. The paper arose from a collaboration with the National Centre of Audiology at Western University, with Susan Scollie and Paula Folkeard.
People with hearing impairment are known to find listening to speech particularly difficult and effortful when reverberant noise is present. Here, we measured speech intelligibility and subjective ratings of listening effort when people with hearing impairment listened to speech in the presence of reverberant cafeteria noise. When participants listened with a hearing-aid setting designed for reverberant environments, they reported more words correctly and reported lower listening effort than when they used a standard omnidirectional hearing-aid setting. Participants received a greater improvement in speech inteligibility, and similar reduction in listening effort, when they listened to sentences preceeded by a same-topic than different-topic sentence.
Semantic context and hearing-aid setting had additive effects on speech intelligibility and listening effort, demonstrating that the benefit of semantic context can extend above and beyond the improvement gained from changing hearing-aid settings. These results highlight the enormous potential of cognitive factors for improving speech intelligibility and reducing perceived listening effort in noisy acoustic environments.
You can read the paper here:
Holmes, E., Folkeard, P., Johnsrude, I. S., & Scollie, S. (2018). Semantic context reduces sentence-by-sentence listening effort for listeners with hearing impairment. International Journal of Audiology.
8th February 2018
I’m on my way to the 41st ARO MidWinter meeting and half the plane is filled with auditory researchers! Alex Billig and I organised a Young Investigator Symposium on ‘Non-acoustic influences on speech perception in normal and impaired hearing’, which will take place on Tuesday. The symposium highlights the wide variety of factors that influence speech perception and will feature talks by Tessa Bent, Ross Maddox, Alex Billig, Elin Roverud, Stefanie Kuchinsky, Pranesh Bhargava, Malte Wöstmann, and myself. You can find the abstract book here: ARO 2018 MidWinter Meeting Abstracts
Two new papers accepted
18th January 2018
A great start to the new year—2 papers accepted within the past week! Watch this space for details…
Moved to UCL
2nd January 2018
Recently, I made the transatlantic move from London Ontario to London England to start a new position at UCL. I’ll be based at the Wellcome Trust Centre for Neuroimaging (known locally as the FIL) at 12 Queen Square. I’m looking forward to making use of the 3 fMRI scanners and the MEG scanner here.
14th November 2017
I’ve just got back from the 2017 Annual Meeting of the Psychonomic Society. It was my first time at Psychonomics and I presented some of my new work on familiar voices. I’m very grateful to have received a Women in Cognitive Science Travel and Networking Award to attend.
New paper out in JARO
5th October 2017
Our new paper has just been published online in JARO. The paper describes two experiments that investigated whether envelope following responses (EFRs) are affected by attention.
We found that EFRs were modulated by frequency-specific attention when we used stimuli with lower amplitude modulation rates (93–109 Hz), but not when we used stimuli with higher amplitude modulation rates (217–233 Hz). The results demonstrate that amplitude modulation rate can dramatically affect the conclusions of EFR studies—implying that different neuronal populations contribute to EFRs at these lower and higher rates.
Auditory Cortex meeting
8th September 2017
Tomorrow, I’ll be flying out to the International Conference on Auditory Cortex. I’m looking forward to presenting my latest fMRI work on ‘How are familiar voices represented in auditory cortex?’, catching up with colleagues, and enjoying the beautiful scenery in Banff!
Speech recording videos
21st August 2017
Yesterday, I shared some code that I wrote to generate videos, which I have used to standardise speech recordings. When making sentence recordings, these videos have been helpful in ensuring that word onsets are aligned between different talkers, who might otherwise speak at very different speeds. The code is quite flexible and has been used to generate sentence videos in 4 different languages for 10 research projects in the lab. You can find the code on my Github profile.
4th August 2017
I’ve recently started using Github and have created a profile to share some of my code. For example, I’ve made available the code I wrote to calculate Phase Coherence and an analysis method I developed for estimating the dissimilarity in source locations between two conditions (termed ‘Source Dissimilarity Index‘). I have programmed both of these analyses in MATLAB and Python.
You can view my Github profile here. I would be interested to hear from you if you’re using the code or have found it useful.
New paper on attention in hearing-impaired children
22nd May 2017
Another paper from my PhD has recently been published. The paper provides evidence that children with sensorineural hearing loss have a reduced ability to direct selective attention to a talker of interest compared to children of a similar age who have normal hearing.
People with hearing impairment often experience particular difficulty when listening to speech in the presence of other competing speech. Since auditory attention is crucial for separating simultaneous speech, we tested the hypothesis that auditory attention is atypical in hearing-impaired children.
Participants were cued to attend to a talker (defined by their location or gender) who spoke in a mixture of three talkers. To isolate attentional processing from peripheral acoustic processing, we measured electro-encephalography (EEG) when participants prepared their attention for an upcoming talker. Hearing-impaired children showed a reduced ability to prepare spatial attention compared to normally-hearing children. Preparatory EEG activity was not restored when the hearing-impaired children listened using their acoustic hearing aids. These findings demonstrate that sensorineural hearing loss has consequences for central processing, providing a shift in thinking from the traditional view of hearing loss as purely a peripheral phenomenon.
The paper can be accessed here:
Holmes, E., Kitterick, P. T. & Summerfield, A. Q. (2017). Peripheral hearing loss reduces the ability of children to direct selective attention during multi-talker listening. Hearing Research, 350, 160–172. doi:10.1016/j.heares.2017.05.005
Pint of Science
11th May 2017
On Monday, I’ll be giving a talk at the London Ontario Pint of Science festival. The Pint of Science festival takes place every year at locations all around the world, but this is the first year that the festival’s being held in London Ontario. The idea is to convey the latest scientific research to the public in an interesting and accessible way. The festival comprises a series of talks and discussions, which are held in local pubs. You can attend any of the events for free by registering on the website: https://www.pintofscience.ca/london-2017. My talk, “Can you hear me?”, is part of the Beautiful Minds session on May 15th.
Paper accepted in The Journal of Neuroscience
21st April 2017
Björn Herrmann and I recently received notification that our paper—entitled “Revisiting the contribution of auditory cortex to frequency-following responses”—has been accepted for publication in The Journal of Neuroscience. The paper highlights accumulating evidence that auditory cortex contributes to FFRs at frequencies traditionally thought to predominantly reflect brainstem sources. This idea is becoming increasingly important, because researchers have often assumed that differences in FFRs at these frequencies must reflect differences in brainstem processing. Nevertheless, the relative magnitudes of cortical and brainstem contributions to FFRs at different frequencies remain unclear. In the paper, we suggest avenues for future research that could help to improve understanding of the neural generators of FFRs and of brainstem processing.
Holmes, E., & Herrmann, B. (2017). Revisiting the contribution of auditory cortex to frequency-following responses. The Journal of Neuroscience, 37(21), 5218–5220.
Two talks at ARO meeting
16th February 2017
The Association for Research in Otolaryngology (ARO) Mid-Winter meeting was held in Baltimore MD this year. As usual, the conference was packed with interesting research and it was difficult to see everything in the time available! Nevertheless, I always enjoy ARO for finding out about the latest research and having great discussions.
This year, I gave two talks and co-moderated a podium session, so the conference was even busier than usual! The Speech Perception session that I was moderating—on the first afternoon of the conference—included lots of great talks. In the same session, I talked about some of my recent work on the voice cues that enable us to understand and recognise speech spoken by familiar talkers (such as a good friend or partner).
A few days later, I gave a talk in the Brainstem session. I presented the results of my recent experiment showing that attending to sounds of particular frequencies affects envelope following responses (EFRs) at lower (93–109 Hz) but not higher (217–233 Hz) frequencies. There was an excellent turnout and I got lots of great feedback.
Collaborative Research Funding
3rd February 2017
I’ve recently had the opportunity to be involved in the funding process for the Brain and Mind Institute Postdoctoral Collaborative Research Grants. These grants are a new initiative by the Brain and Mind Institute to promote collaboration between different labs and give postdocs the opportunity to carry out independent research projects. I’ve been fortunate enough to see the funding process through from start to finish and gain insight into the entire process. I formed a committee to review the applications and was also involved in reviewing the applications myself. Overall, it was a very interesting and rewarding experience.
Listen up: Our brains prepare to ‘tune in’ to a voice of interest
6th December 2016
A lay summary of my 2016 Hearing Research paper was published yesterday on the Atlas of Science website, which aims to promote the dissemination of recent scientific papers to a broader audience. All of the articles are written by authors. The website includes lots of other interesting research summaries on a variety of topics.
Talk at FFR Workshop
20th May 2016
On Wednesday 18th May, I traveled to Boston MA, for the 2nd Frequency Following Response (FFR) workshop. My talk, entitled “Effect of attention on envelope following responses”, was on the second day of the workshop. There was an exciting program of talks and lots of great opportunities for discussion!
New paper published in Hearing Research
10th May 2016
A paper from my PhD has been accepted in Hearing Research and is now available online. The paper reports three EEG experiments investigating the time course of preparatory attention, when adults (18-27 years) and children (7-13 years) were cued to attend to one of two talkers during multi-talker listening.
Adults and children were able to use visual instructive cues for spatial location (left/right) and gender (male/female) to select words spoken by the target talker. Both adults and children showed preparatory EEG activity very soon (< 50 ms) after they were cued to the spatial location of a target talker. Adults, but not children, showed preparatory EEG activity when they were cued to gender. Although, preparatory EEG activity for gender was only apparent when the cue for gender predicted the specific identity of the male or female talkers and was not present when the gender cue could refer to one of three different male or female talkers.
Overall, the results provide converging evidence that listeners evoke consistent preparatory brain activity for selecting a talker by their location (regardless of their gender or identity), but not by their gender alone.
The paper can be accessed here:
Holmes, E., Kitterick, P.T., Summerfield, A.Q. (2016). EEG activity evoked in preparation for multi-talker listening by adults and children. Hearing Research, 336, 83–100. doi:10.1016/j.heares.2016.04.007
Does instrument familiarity affect the ability to separate multiple competing melodies?
9th May 2016
One of my ongoing research projects was featured in a recent newsletter of the Music Learning Across the Lifespan (MLAL) group of Western University. The project is a collaboration between researchers at the Brain and Mind Institute and at the Don Wright Faculty of Music.
We were interested in whether musical instrument learning affects the perception of melodies played by a trained instrument, compared to instruments on which participants had never been trained. Specifically, we aimed to investigate whether familiarity with an instrument helps people to perceptually separate (or “pull apart”) melodies played simultaneously by different instruments, if one of the instruments is familiar.
You can read about this project and other exciting projects of the MLAL group online in the Spring 2016 newsletter.
The MLAL group has been very successful in promoting collaboration between different departments at Western University. The group has funded collaborative pilot studies and I was very grateful to recieve funding as a Prinicipal Investigator to carry out a project in collaboration with the Don Wright Faculty of Music at the end of last year. The project will investigate whether musical instrument learning affects brain responses in a timbre-specific manner.
Three Minute Research Competition
28th April 2016
Today, I entered the Three Minute Research competition for postdocs at Western University. The competition was modeled on the increasingly popular Three Minute Thesis competition, which was first held at the University of Queensland in 2008. The aim of the computition is to convey your research in less than three minutes and engage an audience of non-specialists—with only one PowerPoint slide allowed! A panel of judges evaluated our presentations and I was awarded a prize, which includes funding for travel to a future conference. Overall, the competition provided an interesting opportunity to think in more depth about the broader implications of my research and was a great exercise in science communication. I recommend that everyone has a go at this in the future!