Modeling the acquisition of cross-cultural differences in tonality with neural nets
Jamshed Jay Bharucha · The Journal of the Acoustical Society of America · 1988
The acquisition of scale schemas through cultural exposure was modeled with neural net simulations. Associative networks employing the back-propagation learning rule were exposed repeatedly to scale sets of one culture and were then tested for their ability to complete scale subsets of either the same culture or a different culture. The scales were taken from Western and North Indian music. Each network completed the scale subsets so as to most closely match the scales of its own culture. When the foreign scale subset was sufficiently different from any scale to which the network had been exposed, the network showed no clear preference for scale completion. In some cases, the foreign scale was interpreted as being similar to a native scale with a shifted tonic. It is argued that acoustical patterns that occur with considerable regularity in the environment become encoded through passive exposure and serve to generate expectations in context. Reaction time tasks were employed to study these expectancies empirically, for both Western and Indian music.