Journal Club

Wednesday 9th October 2013

This week we covered this article: 

"Neural correlates of phonetic convergence and speech imitation'
In this article Garnier et al. address a facinating phomenon known as phonetic convergence. It is well documented that humans tend to imitate the actions of people they are interecting with. Phonetic convergence is the speech equivalent of this - people tend to take on various characteristics of the speech produced in conversation. Specifically they wanted to ask whether automatic imitation of speech characteristics (phonetic convergence) is the same process as deliverate imitation of speech. 
Subjects either heard a speech token (one of three vowels) and then had to i) produce that same speech token, ii) deliverately imitate that specific speech token or ii) produce that same speech token but deliverately try not to imitate the characteristics of the speaker but use their own habitual speech. 

My first query was the relationship between the task and phonetic convergence. The task used very short vowel production, and looked at how the precence of a single token influenced the subequent utterence. In my opinion this is really automatic imitation, rather than phonetic convergence, and so I will use that terminology from hereonin. They found that people showed a high degree of automatic imitation - when simply asked to produce the speech sound that they heard, the characteristics (f0) of the sound that they produced was significantly correlated with the characteristics (f0) of the heard sound. This was the case for both the automatic imitation (subjects were not asked to imitate the sound they heard) and the deliberate imitation condition (subjects were specifically asked to imitated the sound that they heard). Furthermore, they showed that subjects were able to 'inhibit' this imitation. 

Unfortuately it is hard to say from this data that people were inhibuting automatic imitation. It is clear from their data that when told to inhibit imitation subjects were able to modify their speech so that it did not sound like the heard token, but it is not clear how much like their nomal speech this inhibited imitation was. One subject for example, when instructed not to imitate the heard sound, produced sounds as different as possible to the heard sound indicating that they were using a strategy to not imitate, rather than using their normal speech, which is what we'd hope to see if they were truly inhibiting this automatic imitation. 

The authors then went on to use neuroimaging to look at how the brain dealt with automatic and deliberate imitation. They found no differences in how the brain dealt with these two processes. However when they looked at the correlation between how well subjects imitated and brain acivity they did see something interesting. They found a negative correlation between activity in auditory cortices and how well subjects imitated the sound that they heard. This means that the more activity in auditory cortex there was when listening to the stimulus, the worse the subject was at imitating that sound. It is not clear why this would be the case but it is interesting that better imitation appears to be associated with changes in neural activity during perception of the stimulus, rather than during execution of the movement. Overall, this was a really interesting research question but there were many things that are unanswered. Like most science in my mind, it has produced more questions than answers!