The Effects of Quantized Re-timing of Speech Rhythm on Spoken Language

Overview
What's going to happen next? is a question the brain constantly seeks to answer. In
neuroscience, prediction is increasingly seen as the key to language comprehension, but less
is known about how listeners predict when linguistic events (sounds, syllables) will occur,
than about what is coming. Current theories of neural speech processing suggest a key role
for rhythm, yet neuroscience and the speech sciences lack a shared understanding of the
basics of rhythm. This project seeks to address this gap by asking: Can speech be re-timed so
as to maximise its comprehensibility? The research will use phonetic and musical
approaches to inform the re-timing of speech and the effects on comprehension will be
tested using innovative eye-tracking and EEG paradigms.
Context
A crucial aspect of speech comprehension is the ability to anticipate lexical or grammatical
units. This has been demonstrated by extensive research showing characteristic neural
responses to violations of expectation in speech (e.g. Hagoort et al., 2004). According to
recent theories in cognitive neuroscience, listeners achieve such anticipation in part by
tracking regularities in speech rhythm, at timescales roughly corresponding to individual
vowels and consonants, and to syllables (e.g. Giraud and Poeppel, 2012; Gross et al., 2013).
However, this picture does not match well with phonetic or behavioural reality. First, there
is little evidence of rhythmic regularity in the production of syllables in naturally occurring
speech (Cummins, 2012). Second, the rhythm-tracking approach suggests that making
speech more regular would help comprehension. But, while presenting isolated syllables at
regular time intervals can yield processing benefits (Otterbein et al., 2012), regularising the
intervals between syllables in connected speech instead has a processing cost (Aubanel et
al., 2016). In sum, regularity can be beneficial, but is not always: the relationships have to be
right for the speech being presented, in ways that are poorly understood.
A speech signal does not necessarily need to be regular to be rhythmic. In response to
musical rhythms, neural correlates of the pulse (or beat) can be found when listeners are
presented with complex polyrhythms where the pulse is not physically present in the signal
(Tal et al., 2017). Similar principles might apply in speech, but the relevant empirical work
has not been done, and speech research has a much reduced lexicon of terminology for
rhythm compared to music (Smith et al, 2014). In order to understand how the temporal
characteristics of speech facilitate comprehension, it may be necessary to consider a more
nuanced approach to rhythm, informed by music.
This project seeks to improve understanding of the role of rhythm in speech comprehension
by answering two Research questions:
1. Does re-timing speech to increase its rhythmicity help speech comprehension, when a
more sophisticated approach than equalising syllable durations is used?
2. If so, on which properties of the speech signal should re-timing ideally be based?