Some neurons in the nucleus HVc of the songbird respond vigorously to sequences of syllables as they appear in the bird's own song (such as AB), but they respond weakly or not at all when the same syllables are played individually (A or B) or in a different order (BA). We have constructed a network model that replicates this temporal sequence selectivity. The model is based on recurrently connected networks that produce strong resonant responses when the pattern of excitation evoked by a stimulus matches the pattern of excitation generated internally within the network. In the model, syllable B does not generate such a resonant response by itself. However, if syllable A is presented to the network followed by syllable B, the activity generated by A modifies the effective connectivity of the network making it resonantly responsive to B. This produces a highly selective response to the sequence of syllables AB, but not to any other combination.
All Science Journal Classification (ASJC) codes
- Computer Science Applications
- Cognitive Neuroscience
- Artificial Intelligence