Jonathan Peelle

Starting now in Crystal DE: Non-sensory processes in speech perception. Thread of talks below.

#AROMWM #ARO2023

Feb 14, 2023 at 7:04pmWeb
Show threadJonathan PeelleFeb 14, 2023

First: Jane Brown, "Selective Attention and Familiarity at the Musical Cocktail Party"

Looking at speech perception in the presence of background music (e.g. Lady Gaga with vocals+music, isolated vocals, isolated music). Both familiar and unfamiliar music.

(Side note I love seeing "just dance, it'll be OK, da da doo doo" on a slide.)

Familiar music reduces accuracy more than unfamiliar music, especially for full song.

Music (with vocals) reduced accuracy more than speech (I missed what kind of speech).

EEG: N1 amplitude bigger with vocals than without. Latency: only differ in unfamiliar (with > without).

Preliminary additional data looking at TRFs when listening to audiobooks+music. Attending to book: same TRF regardless of music familiarity. When speech is background, TRFs to unattended speech separate based on familiarity.

Show threadJonathan PeelleFeb 14, 2023

Next talk: Alina Kuimova "Attention Increases the Odds of Expectation-Induced Misperception: Behavioural and Pupillometric Evidence"

Starts with a nice demonstration of vocoded speech and the popout effect and how context can affect perception.

Segue to attention, and the inattention can result reduced intelligibility and performance for speech in noise. How to reconcile this with people's tendency to mind wander?

Does higher attentional engagement reduce the rate of expectation-induced misperception?

Vocoded words and written text: total mismatch, partial mismatch, total match. Importance varied using a monetary incentive. Task = same/different judgment.

Behaviorally, incentives specifically affected perception of more challenging word pairs (small effect).

Pupillometry (w/ GAMMs 👏): larger response for incentive trials.

Overall conclusion: increased attention not always beneficial during effortful listening.

Show threadJonathan PeelleFeb 14, 2023

Now Zilong Xie on "Lexical Bias in Phonemic Categorization: Effects of Spectral Degradation, Cognitive Load, and Aging"

Starts with a reminder of the Ganong effect where listeners' perception is biased by lexical characteristics. E.g., gift-kift, with an ambiguous g/k, more likely to hear "gift" (real word) and "g".

Lexical bias my increase when signal is degraded, or when listeners are inattentive. Current work: looking at spectral degradation, cognitive load, and age (adult aging).

Cognitive load manipulated using AV dual task, remembering visual images that are presented along with a phoneme that needs to be categorized.

High demands in the secondary task reduce accuracy. Vocoded speech did not show a systematic increase in lexical bias. Dual task also hard to see lexical bias, nor did adult aging differences (but small N). However, the vocodedxaging interaction may indeed affect lexical bias. More data to come!

Show threadJonathan PeelleFeb 14, 2023

Next talk is from Jusung Ham, "Effect of Feedback Reliability for Neurofeedback Training of Auditory Selective Attention"

Background: speech in noise is challenging, few treatment options. Can selective attention manipulations help? People whose EEG data seem to better disentangle attended vs unattended seem to perform better in noise.

Current work: train auditory attention using neurofeedback. If attention is trained, does it transfer to speech-in-noise performance?

Behavioral data suggest transfer from training to test (small improvement in accuracy, not seen in placebo). EEG data suggest differences in oscillatory activity in different frequency bands.

Future directions include improving the auditory attention decoder and moving to other populations (e.g., CI users).

Show threadJonathan PeelleFeb 14, 2023

Next up: Joseph Rovetti "Listening Effort: Separating the Subjective From the Objective".

Background: there are many definitions of "listening effort". Relationship of objective and subjective measures unclear - are they the same thing or not?

As speech goes from favorable to unfavorable, does one dimension increase sooner or differently than the other?

Online study using 30-second passages, dual task, gist-based question (yes/no) after. And effort ratings 1-7.

As SNR gets less favorable, subjective effort ratings increase fairly linearly. Secondary task RT is a bit noisier, but flat for a bit, and then rises at the end.

Many reasons for the disconnect between subjective and objective measures. But evidence for different dimensions of effort, even if interpretations still tricky.

Show threadJonathan PeelleFeb 14, 2023

Next speaker: Björn Herrmann (filling in for Mo Cui), "Listening Effort Reduces Eye-Movement Behaviors in Younger Adults"

Exp 1: sentences in multitalker babble at different SNRs, probe relatedness task. Comparing free-viewing vs. fixation. Less movement during fixation (as expected), but gaze dispersion seems to maybe track effort.

Exp 2: similar set up, but with moving object display (dot moves around the display to encourage eye movements). In this case fixation duration and gaze dispersion change with SNR.

Exp 3: Story listening. 10 minute stories, same babble. Moving object display (16 moving dots on the screen). Stories could be intact or scrambled. For intact stories, same effects of fixation and gaze dispersion.

Eye movements may thus be used as a measure of listening effort.

Show threadJonathan PeelleFeb 14, 2023

Last talk of the session: Jacie McHaney @soundbrainlab : "Deficits in Sensory Decision-Making Underlie Self-Perceived Hearing Difficulties"

Looked at self-perceived listening difficulties (SSQ). Do people with listening difficulty differ in their sensory decision processes? Had people do a phoneme categorization task (quiet or noise).

Drift diffusion model on RTs for the phoneme task, giving evidence accumulation rate and decision threshold. Decision threshold does not change as a function of SNR.

Generally people have lower accumulation rates at less favorable SNRs. Slope of this (across SNR) relates to self perceived difficulty.

EEG study looking at neural tracking of acoustic and linguistic information in continuous speech (audiobook of Alice in Wonderland). Those with more self perceived difficulties didn't differ on tracking of acoustics (acoustic model).

For language processing, split apart sublexical, word, and sentence information. No differences in tracking for sublexical and word-level...but listeners with more self-perceived difficulty showed increased tracking for sentence level information. Compensation?

I really liked this linking of self-reported difficulties to objective in-lab measures. 💯