4.2.2 Distributed network or state-dependent ("emergent?" models
Distributed networks model
The alternate view is that timing is distributed, meaning that many brain areas are capable of temporal processing and that the area or areas involved depend on the task and modality being used.
The above models [clock and spectral] represent top-down approaches where timing is addressed by inferring a computation and then implementing the computation with neurons. An alternative bottom-up approach is to start with biologically realistic assumptions and then to ask the extent to which temporal processing can be found as an emergent property. These models have no built-in temporal processing or selectivity with ad hoc assumptions. That is, they do not rely on explicitly setting oscillators, synaptic or current-time constants, or some other variable that, in effect, functions as a delay line.
In these network or state-dependent models, timing does not arise from clocks or even from brain systems specifically dedicated to temporal processing. Rather, the evidence from the cerebellum, for example, illustrates howtiming and performance on experimental tasks designed to study timing may be mediated by computations that include temporal processing but that are not accurately characterized as interval timers or clocks.
In contrast, models based on network dynamics may better generalize to the processing of more complex temporal patterns. In state-dependent network models (see above; Buonomano&Merzenich 1995, Buonomano 2000, Maass et al. 2002), the current state of the network is always dependent on the recent history of activity. Thus, in the above example, if the third input arrives at 200 ms, the network will be in a different state depending on whether the second pulse arrived at 50 or 150 ms. In these models, time- dependent properties, such as short- term synaptic plasticity, slow PSPs (e.g., GABAB or NMDA-dependent currents), or, potentially, slow conductance, function as state- dependent memory traces of the recent stimulus history. In contrast to single-cell models, these time-dependent properties are not tuned for any particular interval; rather these states are expressed as changes in the probability of different neurons becoming activated.
Given the inherent temporal nature of our sensory environment, and the continuous, real- time motor interaction with our environment,we favor the view that temporal and spatial information are generally processed together by the same circuits, and that there is no centralized clock for temporal processing on the scale of tens to hundreds of ms. Additionally, we propose that temporal processing does not rely on specialized mechanisms, such as oscillators or arrays of elements, as with a spectrum of different time constants. Rather, we believe that neural circuits are inherently capable of processing temporal information as a result of state-dependent changes in network dynamics.
One such class of models, state-dependent networks (SDNs), propose that neural circuits are inherently capable of temporal processing as a result of the natural complexity of cortical networks coupled with the presence of time-dependent neuronal properties (Buonomano and Merzenich, 1995; Buonomano, 2000; Maass et al., 2002). This framework, based on well characterized cellular and network properties, has been shown to be able to discriminate simple temporal intervals on the millisecond scale, as well as complex spatial-temporal patterns (Buonomano and Merzenich, 1995; Buonomano, 2000; Maass et al., 2002).
Here we examine the mechanisms and nature of the timing in this model and show that it encodes temporally patterned stimuli as single ‘‘temporal objects,’’ as opposed to the sum of the individual component intervals. This generates the counterintuitive prediction that we do not have access to the objective (absolute) time of a given interval if it was immediately preceded by another event.
In this model, there is no explicit or linear measure of time like the tics of an oscillator or a continuously ramping firing rate (see Discussion; Durstewitz, 2003). Instead, time is implicitly encoded in the state of the network— defined not only by which neurons are spiking, but also by the properties that influence cell firing, such as the membrane potential of each neuron and synaptic strengths at each point in time. Thus, even in the absence of ongoing activity, the recent stimulus history remains encoded in the network.
The standard model of temporal processing postulates a single centralized internal clock, which relies on an oscillator and an accumulator (counter) (Creelman, 1962; Treisman, 1963; Church, 1984; Grondin, 2001). The clock concept is generally taken to imply that the passage of time is counted in units that can be combined or compared linearly. In contrast, SDN models propose that for spans on the scale of tens to hundreds of milliseconds, time may be represented as specific states of a neural network. Within this framework, a 50 ms interval followed by a 100 ms interval is not encoded as the combination of the two. Instead, the earlier stimulus interacts with the processing of the 100 ms interval, resulting in the encoding of a distinct temporal object. Thus, temporal information is encoded in the context of the entire pattern, not as conjunctions of the component intervals.
We propose here that cortical networks can tell time as a result of time-dependent changes in synaptic and cellular properties, which influence the population response to sensory events in a history- dependent manner. This framework is applicable to the processing of simple intervals as well as more complex spatial-temporal patterns, and does not invoke any novel hypothetical mechanisms at the neural and synaptic level. Additionally, we propose that timing is not centralized, and can potentially occur locally at both early and late stages of cortical processing.