Olfactory network dynamics

Preparation

Using the Perusall system, closely read and annotate…

Discussion

The contrast between odour representations in the antennal lobe and in the mushroom body is enormous. An observer looking at activity in these two areas might never suspect that they are separated by only one synapse. Similarly, observing individual Kenyon cell responses alone does not hint at integrative properties that depend on presynaptic spike timing and on the coordination of input arrival. In other words, understanding the computations that take place in a circuit can be difficult if we fail to consider individual neurons as parts of a system in action. I would argue that our lexicon introduces subtle but real biases in our thinking about sensory processing. Our predisposition as sensory physiologists is to call ‘responses’ the spike patterns that follow a stimulus; we then use these responses to define ‘receptive fields’. In doing so, we forget that these terms are meant only to be operational. Our thinking about sensory integration seems to be much too linear and passive: stimulus a leads to a response in area x, which produces a response in area y, and so on. In reality, neural circuits are often massively interconnected and reciprocally connected. Similarly, our thinking generally ignores the fact that, with the exception of motor neurons, a given neuron is never an end-point or its ‘response’ an end-product. So, how a neuron behaves might be relevant not as a response per se (something to be analysed by us to estimate the information it contains about a stimulus, although this is, of course, useful knowledge), but as part of a transformation (possibly extremely complex and distributed) to help further processing (for example, optimization, storage, recognition and retrieval) in the area in which the neuron lies (for example, decorrelation in circuits of the olfactory bulb) or in ‘target’ circuits. Thinking about sensory integration in these active terms (considering ‘responses’ not only as products, but also as ongoing transformations towards some other goal) might be helpful as we try to understand some brain operations.

Laurent 2002