One of the main focuses of my research lies in the development and implementation of a new psycholinguistic approach, purely behavioral in nature, known as Auditory Classification Images (ACI). This approach draws inspiration from reverse correlation studies in vision psychophysics, which aimed to link random fluctuations in a stimulus to the corresponding perceptual response of a participant in a task, on a trial-by-trial basis. The objective of this project is to adapt and operationalize the reverse correlation technique for the auditory modality, with the ultimate goal of applying it to the exploration of speech perception, particularly phoneme comprehension. Auditory Classification Images enable us to visualize the « listening strategy » of a participant, revealing the information they extract from the auditory stimulus to perform the task.

The Matlab toolbox that we develop is available as a Github repository : https://github.com/aosses-tue/fastACI. It can be used to replicate previous auditory revcorr experiments or to design new ones.

With this toolbox you can run listening experiments as used in the studies below. You can also reproduce most of the figures contained in the mentioned references. The toolbox was also used to replicate Ahuma et al.’s seminal experiment (Ahumada, Marken & Sandusky, 1975). Experiments can be conducted on human participants through headphones, or on auditory models (Osses & Varnet, 2021; see also Osses & Varnet 2024)

REFERENCE	fastACI experiment name	TASK	Type of background noise	Target sounds
Ahumada, Marken & Sandusky, 1975	replication_ahumada1975	tone-in-noise detection	white	500-Hz pure tone vs. silence
Varnet & Lorenzi 2022	modulationACI	amplitude modulation detection	white	modulated or unmodulated tones
Varnet et al. 2013	speechACI_varnet2013	phoneme categorization	white	/aba/-/ada/, female speaker
Varnet et al. 2015, 2016a, 2016b	speechACI_varnet2015	phoneme categorization	white	/alda/-/alga/-/arda/-/arga/, male speaker
Osses & Varnet 2021	speechACI_varnet2013	phoneme categorization	speech shaped noise (SSN)	/aba/-/ada/, female speaker
Osses & Varnet 2024	speechACI_Logatome	phoneme categorization	white, bump, MPS	/aba/-/ada/, male speker from the OLLO database
Carranante et al., 2023, 2024.	speechACI_Logatome	phoneme categorization	bump	Pairs of contrasts using /aba/, /ada/, /aga/, /apa/, /ata/ from the same male speaker (S43M) in OLLO
Osses et al., 2023	segmentation	word segmentation	random prosody	Pairs of contrasts: /l’amie/-/la mie/, /l’appel/-/la pelle/, /l’accroche/-/la croche/, /l’alarme/-/la larme/

Administrative details

Collaborators: Alejandro Osses

Fundings:

NSCo Doctoral School, L. Varnet’s thesis scholarship (2012-2015)

Agence Nationale de la Recherche, projet ANR-20-CE28-0004 « Exploration des représentations phonétiques et de leur adaptabilité par la méthode des Images de Classification Auditive rapides » (2021-2023)

Selected publications and presentations:

English

Presentation of the auditory revcorr method

General overview of the toolbox

About the use of auditory models within the toolbox

Articles using the toolbox or reimplemented within the toolbox (see table above)

Français

Varnet, L. (2015). Identification des indices acoustiques utilisés dans la compréhension de la parole dégradée. Université Claude Bernard Lyon 1. (thèse)

Varnet, L. Présentation à la Fête de la Science (2020, ENS Paris): « Observer l’esprit humain sans neuroimagerie » (vidéo)

« Visualiser ce que nous écoutons pour comprendre les sons de parole » (vidéo)

Sur ce blog :
– L’image de classification auditive, partie 1 : Le cerveau comme boîte noire
– L’Image de Classification Auditive, partie 2 : À la recherche des indices acoustiques de la parole
– A visual compendium of auditory revcorr studies