Laboratory: Laboratoire de Psychologie Cognitive
Team name: Cognition comparée
Team leader(s): Joel Fagot
First name and last name of the proposed
PhD or Post-Doc supervisor(s):
Florence Gaunet (co-supervisor)
Anne-Lise Giraud (co-supervisor). Imera
position from Feb to June 2019 (LPC and LPL).
The student
will be 100% based in L.P.C./Marseille.
Collaboration
with T. Legou/L.P.L. Research conducted in the Labex ILCB framework (Institute
of Language, Communication and the Brain).
Title of the proposed
PhD or 2-year Post-doc research project:
Contour,
rhythm or content? What does dogs brain grasp from human speech?
Summary of the
proposed research project
o State of the art: Dogs have so closely
coevolved with humans that they are able to make use of short word sequences to
execute simple orders (e.g. “Titus, bring the ball”) (Kaminski et al., 2004). The
dog speech perception capacity is particularly interesting to explore from a
language evolution viewpoint because, while dogs cannot speak (Riede &
Fitch, 1999), they show a strong motivation to communicate with us (Gaunet
& El Massioui 2014). In humans, understanding connected speech requires
segmentation of the acoustic flow. This capital pre-processing step involves
slow neural oscillations, which naturally occur in the auditory cortex at a
rate close to the average human syllabic rate (4-7Hz) (Giraud and Poeppel 2012).
Since the syllabic structure is constrained by the motor system, its
contribution to the speech perception is intensely debated (Park et al. 2015).
o Objectives: This project proposes
to explore the neural bases (including laterality) of dog speech processing to
address the contribution of the motor system to speech perception. We expect to
determine whether dogs perceive speech on the basis of acoustic cues they can
themselves produce, i.e. short “syllable-like” intonated sounds, or whether alternatively,
they can be sensitive to cues that they cannot produce at all, phoneme-level
cues.
o
Methods:
Project
1: acoustic analysis. We will first record single and polysyllabic animal vocalizations.
Polysyllabic vocalizations can be obtained in dogs by mimetism with human
speech (Gaunet & Legou In prep; Legou & Gaunet In prep). We will use these recordings to perform a comparative
acoustic analysis of dog and man vocalizations. We will compute
modulation power spectrum
(MPS) as in (Arnal
et al. 2015), for both types of signal and will determine the overlap of dog
and man modulation acoustic landscape. This analysis will serve to interpret
the neural responses and to determine how much of the human
communication MPS area is also
used by dogs. A preliminary study analyzing dogs’ vocal productions, using
internet videos posted by dog owners, is in
preparation by Legou and Gaunet.
Project 2: non-‐invasive electrophysiology. We plan to record
EEG in 5 dogs (Kis et al.
2014, 2017; Bunford et al. 2017)
to explore the ability of auditory cortical
rhythms to track syllable boundaries in human speech in: 1/ natural continuous speech of an
unknown human intended to the dog, 2/ natural continuous speech of the dog’s
owner intended to the dog, 3/ a simple sentence usually understood by the dog pronounced by the dog’s owner with
prosody, 4/ the same condition without prosody, 5/ aprosodic single words, 6/
single words with intonation.
The data will be analyzed individually
using state of the art methods, involving standard analyses of evoked
potentials for single words, and more advanced methods of speech/brain cross-‐correlations (cross
spectral density etc.) and EEG time-‐frequency
analyses
for continuous speech, as we recently described (Pefkou et
al. 2017). We will also use detrended fluctuation analysis to evaluate long
range temporal correlations in dogs EEG at rest and during
speech (Borges et al. 2017).
o
Expected results
We expect to determine
whether the dog’s auditory cortical system spontaneously track syllables in
speech, how much this tracking depends on the owner’s voice and/or the owner’s
prosody, whether word content can be accessed in dogs without prosodic cues, and
finally whether their integration capacity fits with long-range autocorrelation
properties of their neural activity.
o Expected candidate profile: Basic knowledge in
EEG (e.g. master training or more), interest in animal cognition, good contact
with animals, very good French level.