Decisions, accumulators, and neurons: How secure a bridge?
The conjecture identifying spiking activity of certain neurons with a stochastic accumulator decision process has inspired a vigorous and productive research effort for 20 years. This effort has described decision accumulation processes in multiple brain regions even extending to noninvasive ERP and fMRI measurements. It has been buttressed by models of perceptual categorization, response inhibition, and visual search. Lately, though, several new findings have raised questions about the coherence and clarity of the mapping between neurons and decision accumulators. These include the first data showing how neurons identified with decision accumulators accomplish executive control and speed-accuracy tradeoffs and the first model of response time from ensembles of accumulators. The new results indicate that mapping between parameters of accumulator models and measurements of neural activity is not as transparent as originally presumed. The accumulator model framework will no doubt remain an effective means of quantifying performance and instantiating computations in various tasks. However, the construction of a more secure bridge between model and neural levels of description will require more assiduity in (1) accounting for multiple stages of processing each adding time and potential errors, (2) incorporating distinct neural processes from heterogeneous neurons in diverse neural structures, (3) articulating the transformations between spikes, ERPs, and BOLD, (4) specifying converging constraints to limit parameters in more complex models and (5) appreciating the logical and rhetorical scope of the mapping — true identity, quantitative analogy, or interesting metaphor.