Maintaining Intonation Contours in the Brain

The objects of linguistics are generally taken to be discrete, abstract objects; words or phonological features are examples.  These objects do not drift, spread, or degrade, and do not change except as the language mechanism functions. On the other hand, neuroscience shows that our working memory is dynamic; information is maintained in the patterns of firings of neurons.  But the neural processes behind short-term memory happen rapidly: neurons often fire 30 times per second, and can sometimes fire at more than 100 Hz.   Over the lifetime of a memory, the pattern of impulses is re-created  hundreds of times. How does the neural firing process maintain a pattern that represents a stable object for so many generations? Especially so, since that memory is not the only activity going on in the brain. But, complex dynamical systems can have stable behaviours.  These are known as "attractors".  An attractor is a special state of a system to which it returns, if subjected to a small disturbance. In this talk, I will describe the idea of attractors in more detail, and its application to linguistics, and how attractors relate to the (mostly discredited) idea of categorical perception. These ideas have experimental support for intonation contours in English, where subjects mimic a sentence's intonation. We see evidence to support the existence of discrete abstract objects, but not necessarily ones with properties that conventional phonology expects.