Brain shape predicts language learning success

Dr Narly Golestani, of the UCL Institute of Cognitive Neuroscience, said: "The bigger picture is that we're starting to understand that brain shape and structure can be informative about people's abilities or pathologies. Why people are good at some things and not others is evident from these scans, the differences can sometimes be seen just by looking. This might one day allow us to screen for language learning or for language disorders."

Sixty participants undertook ten blocks of twenty trials of training to learn to distinguish two different sounds - the French 'dental' 'da' sound and the Hindi 'retroflex' 'da' sound. Faster language learners recognised the differences in the 'da' sounds within a few minutes. Slower learners struggled even after the maximum twenty minutes of learning.

The team took anatomical magnetic resonance imaging (aMRI) scans of the brain structure of the 11 fastest and 10 slowest learners and found a number of anatomical differences between them. This was followed up with diffusion tensor imaging (DTI) scans, which allow finer measurements of white matter brain structure.

The differences in language learning ability were not linked to grey matter but to white matter, which is involved in connectivity - linking parts of the brain. The scans show a greater white matter volume in the left auditory regions (Heschl's gyrus) in faster learners, and a difference in the position of Heschl's gyrus in the right hemisphere between faster and slower learners. There is also a greater asymmetry in the volume of the parietal lobes (which in the left hemisphere is thought to be involved in speech sound processing), in faster compared to slower learners, with faster learners having a more pronounced left-greater-than-right volume difference.

The sounds that subjects were trained to distinguish were difficult to tell apart - the French sound is very similar to the Alveolar 'd' sound (an English consonant sound pronounced with the tip of the tongue near the upper teeth) i.e. dad. The Hindi sound is pronounced with the tongue curled upwards onto the roof of the mouth i.e. ardent. Both synthetic sounds were followed by the vowel 'a', so the participants heard 'da'. Subjects had to decide which of the two 'da' sounds they had heard and were told whether or not their response was correct.

The training got harder as it went on. At first the most distinct versions of these two 'da' sounds were heard; with the Hindi 'da' at one end of the synthetic acoustic range and the French 'da' at the other. If a person identified 80 % or more of the sounds correctly, the next training block was made to be more difficult by presenting sounds that were more acoustically alike.

Some of the faster learners learned to distinguish the most acoustically similar sounds within 3 or 4 blocks - this lasted about 6-8 minutes. The slowest learners were still guessing randomly with the most acoustically different sounds even after 10 blocks of training - around 20 minutes later.

Dr Golestani said: "What is special about sounds such as the dental versus retroflex consonants is that the acoustic difference between them is only in the first 40 milliseconds, in other words, your brain needs to work quickly. We speculate that we find differences in white matter volume between faster and slower learners because more white matter (resulting either from more white matter fibres or from thicker fibres) would support more efficient transmission of neural signals between brain regions. Given that we found differences both in the auditory regions and in the parietal cortex of the left hemisphere, there may be anatomical connectivity differences in these regions in the left hemisphere between faster and slower phonetic learners too. I am currently doing analyses on the DTI scans to test for those differences. This body of research indicates how much we can understand about people's behavioural differences just by looking at differences in brain structure. This type of work is very relevant to an understanding of the anatomical characteristics which underlie and support neural function."

Notes for Editors:

1. The paper 'Brain structure predicts the learning of foreign speech sounds' appears in Cerebral Cortex on the 3^rd April 2006. GET TH E LINK

2. The authors are: Narly Golestani (UCL Institute of Neuroscience) and Nicolas Molko, Stanislas Dehaene, Denis LeBihan and Christophe Pallier (Service Hospitalier Frédéric Joliot, France)

3. Listen to the test sounds here: http://www.zlab.mcgill.ca/home.html Go to 'publications', and choose 2001-2005. Go to 'supplementary material' under 'Golestani, N., Paus, T., and Zatorre, R.J. (2002) Anatomical correlates of learning novel speech sounds. Neuron, 35, 997-1010' Click on: CV1 CV2 CV3 CV4 CV5 CV6 CV7 under 'Hindi consonant-vowel syllables' to play the 7 sounds.

4. The French Ministry of Research supported the project.