Proceedings of The Physiological Society
University College London 2006 (2006) Proc Physiol Soc 3, PC216
Distributed encoding of face identity and face emotion information by ensembles of neurons in sheep temporal cortex
Hanno Fischer1, Andrew J. Tate1, Keith M. Kendrick1
1. Cognitive Neuroscience, The Babraham Institute, Cambridge, United Kingdom.
The representation of face-based identity and emotion displays in the temporal cortex of sheep was investigated using behavioural test protocols in combination with bilateral 128 channel multi-electrode array (MEA) electrophysiology. Sheep were trained to discriminate between neutral and emotional facial displays of familiar vs. unfamiliar sheep. MEA arrays were inserted under halothane anaesthesia. ITC activity and response latency maps recorded during face emotion recognition (FER) tasks were analysed. During the tests, sheep showed a 92±2.8% (mean±SD) preference for sheep faces displaying a calm expression over faces displaying signs of stress/anxiety (i.e. enlarged protruding eyes, pupils showing the whites, flattened ears and flared nostrils) (N=5 animals, n=40 trials, t test, P<0.05). When the calm face in the pairs of familiar sheep was replaced by a calm face of an unfamiliar sheep, the animals maintained a significant preference for the calm face (N=5, n=40, P<0.05). These behavioural findings suggest that face emotion information can override face identity information in determining preference. MEA data comprised on average 236±15 neurons per hemisphere (N=3 animals), of which 12.3±4.2% responded significantly to face stimuli with either an increase or a decrease in firing rate (t test, P<0.05). The array response was calculated as the percentage of neurons changing their activity in response to a given face. The difference in array response across trials was 5.8±1.8% (N=3 animals, 10 trials/hemisphere) for a particular face or shape. However, the difference in array response between different faces or expressions was 16.8±2.8% (3 different faces, ANOVA, F=4.64, df=2, P<0.05). This suggests that population-encoding of a particular face identity/expression requires changes in a small proportion of neurons. In addition, our data show a 34.4±5.2% reduction in the number of ITC cells responding with an increased firing rate to familiar faces independent of the facial expression (sparsening), compared to unfamiliar faces (data from 60 trials in 3 individuals). Furthermore, the number of neurons with a reduced firing rate increased by 32.3±4.9% (N=3, n=60). For each hemisphere, the overall response latencies of the investigated populations of neurons were not significantly different in response to face identity (283±102ms, n=275) or facial expression (298±123ms, n=289, P<0.05, data pooled from 3 individuals). Between both hemispheres, the response latencies of neurons during familiarity tasks were compared to FaBER tasks. During tasks addressing familiarity alone, the average latency was 68.7±36.1ms shorter in the right hemisphere whereas during FER tasks, this latency was 35.4±43.2ms (N=3, n=70, P<0.05). This suggests that right hemisphere dominance may be less pronounced during FER than during face identity recognition. In summary, our data point towards functionally lateralised ensembles for encoding face identity and expression using sparse codes in combination with distributed representation.
Where applicable, experiments conform with Society ethical requirements