Abstract
This chapter shows, using hidden Markov modeling, how the trial—to—trial variability of neural activity in gustatory cortex during taste processing can be described in terms of coherent sequences of states, with variability arising from differences in the timing of transitions between those states. Computer simulations of models, in which neural activity follows noise-induced transitions—or jumps—between attractor states, reproduce the observed neural dynamics in gustatory cortex. In other contexts such models can reproduce the observed behavioral variability in time estimation. Finally, decision-making can be improved if the initial “undecided” state remains an attractor and the decision is made when neural activity “jumps” via a stochastic fluctuation to a favorable attractor state. The requirement of sufficient fluctuations to produce such noise-induced decision-making is an example of stochastic resonance.