Neural Timing Models

We have recently proposed that the interval timing system is highly dependent upon functioning cortico-striatal circuits (Matell and Meck, 2000, 2002). A thorough evaluation of the anatomical and electrophysiological characteristics of these areas led us to propose that these neural circuits can provide the majority of the necessary computations for the clock, memory, and decision stages of the GTM. Briefly, each striatal spiny neuron receives 10,000 to 30,000 unique cortical inputs (Wilson et al., 1983). The synaptic strength of these inputs can be modulated by a dopamine reward signal, and through such modulation, the striatal neuron can be "trained" to respond to specific patterns of cortical activity (Beiser and Houk, 1998; Houk, 1995). Given the similarity between this recognition process and the one utilized by the beat frequency model of timing (Miall, 1989), we proposed that striatal spiny neurons detect beat frequencies of oscillating cortical neurons and, in so doing, are able to "time" intervals. Simulations of this striatal beat frequency (SBF) model, which involved adding variance to both the cortical oscillatory frequencies and striatal firing threshold and implementing a physiologically realistic two-state membrane potential in the striatal neuron, produced peak-shaped firing patterns similar to the behavioral data obtained in the peak-interval procedure (Matell and Meck, 2000, in press).

In contrast to an anatomically distinct interval timing system, which might be surmised from the GTM, almost all of the information-processing components of SBF lie within single striatal neurons. Although the SBF model proposes that the clock signal comes from cortical neurons, the clock stage integration process occurs through an alteration of the membrane potential of a striatal neuron. Similarly, temporal memories are stored as specific combinations of synaptic weights of the cortico-striatal synapses. When the proper combination of cortical neurons fires synchronously, the striatal neuron fires. This firing thereby indicates that the current clock value (cortical firing pattern) matches the temporal memory (synaptic strength template), thereby serving as a similarity function. As such, the firing of these spiny neurons over time can be viewed as the decision stage output.

Irrespective of whether this model is an accurate account of interval timing, some interesting processes occur within this model that we would like to elucidate. First, this model shows that it is feasible to build physiologically realistic interval timing models in which the information-processing components have very little anatomical separation. Secondly, close inspection of the SBF model shows that the processing of time in this model does not progress in a serial manner from clock stage through decision stage. Rather, the clock signal is filtered, prior to integration, by those synapses that represent the temporal memories. In other words, the clock stage is not upstream from the memory stage. Similarly, the anatomical circuitry of the cortico-striatal system suggests that the clock stage is not solely upstream from the decision stage: output from the striatum passes through the basal ganglia output nuclei (e.g., both segments of the globus pallidus, the substantia nigra pars reticulata, and the subthalamic nucleus), through the thalamus, and back to the cortex, thereby making up cortico-striato-thalamo-cortical loops. The output pathways of the stria-tum are believed to return to the same cortical regions that innervated the striatal neuron in question (Strick et al., 1995) and imply that the functioning of these regions is dynamic. In terms of the SBF model, decision stage activity in the striatal neuron produces changes in the cortical clock signal, which thereby alters subsequent striatal activity. The dynamic nature of these circuits or loops implies that unlike the GTM described above, the neural implementation of an interval timer is anything but independent and serial.

0 0

Post a comment