Hierarchically structured mixture models are studied in the context of data analysis and inference on neural synaptic transmission characteristics in mammalian, and other, central nervous systems. Mixture structures arise due to uncertainties about the stochastic mechanisms governing the responses to electro-chemical stimulation of individual neurotransmitter release sites at nerve junctions. Models attempt to capture scientific features such as the sensitivity of individual synaptic transmission sites to electro-chemical stimuli, and the extent of their electro-chemical responses when stimulated. This is done via suitably structured classes of prior distributions for parameters describing these features. Such priors may be structured to permit assessment of currently topical scientific hypotheses about fundamental neural function. Posterior analysis is implemented via stochastic simulation. Several data analyses are described to illustrate the approach, with resulting neurophysiological insights in some recently generated experimental contexts. Further developments and open questions, both neurophysiological and statistical, are noted.

Research partially supported by the NSF under grants DMS-9024793, DMS-9305699 and DMS-9304250. This work represents part of a collaborative project with Dr Dennis A Turner, of Duke University Medical Center and Durham VA. Data was provided by Dr Turner and by Dr Howard V Wheal of Southampton University. A slightly revised version of this paper is published in the Journal of the American Statistical Association (vol 92, pp587-606), under the modified title Hierarchical Mixture Models in Neurological Transmission Analysis. The author is the recipient of the 1997 Mitchell Prize for ``the Bayesian analysis of a substantive and concrete problem'' based on the work reported in this paper.

The manuscript is available in either postscript or pdf