|
Bota, Mihail The ETAM of egocentric navigation: the roles of the medial precentral cortex and the caudate/putamen Based on the distinction between map-based and taxon systems for navigation (O'Keefe & Nadel, 1978), we have previously proposed the World Graph (WG) model, which develops the hypothesis that cognitive and motivational states interact, and the Taxon-Affordances Model (TAM) for taxon determination of movement direction (Guazzelli et al., 1998). In the present model, the rodent egocentric navigational system is assumed to be implemented by a network of cortical and subcortical neural structures. These include the posterior parietal cortex (PPC), the medial precentral cortex (Fr2), and the caudate/putamen (CP). The present model extends the Taxon-Affordance Model (TAM) proposed by Guazzelli et al. (1998) for egocentric navigation. As in TAM, the extended TAM (ETAM) relates taxes to the notion of an affordance (possibilities for locomotion) and are extracted by the rodent PPC. The PPC module is also considered to encode distances and bearings to landmarks. In contrast to TAM, however, ETAM explicitly defines roles for Fr2 and the striatum. Based on neuroanatomical findings (Corwin & Reep, 1998) and lesion studies (King & Corwin, 1990), ETAM proposes that the local view and the representation of the environment are encoded in the caudal and rostral parts, respectively, of Fr2. The caudal Fr2 receives inputs from PPC and, in turn, projects to the rostral Fr2 (rFr2). The neural representation of the environment in rFr2 remains stable as long as drastic changes do not occur, due to excitatory recurrent connections with a proposed thalamic module (Fig.1); while the representation of the local view, encoded in the caudal Fr2 (cFr2), changes each time an action is taken. Lesion studies showed that stimulus-response (S-R) pairs are encoded in the rat CP (Packard & McGaugh, 1996). In this way, we consider that egocentric navigation can be seen as a chain of S-R pairs which are incrementally reinforced. In ETAM, S-R is implemented by the CP module (considered to be composed of two submodules, S1 and S2), which uses reinforcement learning to learn expectations for future reinforcements. S1 receives topographical connections from rFr2 and an action schema (AS) (Fig.2), and an S1 unit encodes a specific S-R pair. That is, for each representation given by the rFr2 and an action from AS, an S1 unit will be differentially activated. The S2 module is responsible for the implementation of the chain of S-R pairs. S2 receives excitatory inputs from S1 and a reward signal from a dopaminergic unit. S2 units are fully interconnected and the connections are modifiable using reinforcement learning. The connections from S2 to AS are also modifiable using reinforcement learning. A model for procedural learning that considers modifiable connections between striatal and cortical modules was previously proposed by Trappenberg et al. (1998). In contrast to the model proposed by Trappenberg et al., however, ETAM specifies the functional roles of the cortical modules that are modeled and considers that the striatum is involved in the implementation of the chain of S-R pairs. Moreover, at a given time, an action is selected on the basis of the available set of affordances and the signal from S2. Rats that were trained in a cross maze to find food in a specific arm and then injected with lidocaine in hippocampus or striatum, behaved as response or place learners, respectively, when released from the opposite arm to that initially used (Packard & McGaugh , 1996). The rats that entered the correct arm were considered to be place learners, while the animals that entered in the previously unbaited arm were considered to be response learners. ETAM replicates the results obtained by Packard and McGaugh for rats with lesions of the hippocampus, i.e., the simulated rats were able to learn a response learning strategy in a cross maze with one arm blocked, but were impaired in solving the maze when the starting point was changed from the south arm to the north arm, behaving as response learner animals. Besides assigning specific functional roles to PPC, Fr2 and the striatum, the present model predicts the existence of neurons in CP that encode for specific phases of spatial S-R tasks. The model also predicts the existence of specific firing patterns of neural ensembles in rFr2 for different egocentric views of the environment. Figures: Figure 1: http://rana.usc.edu:8376/~mbota/Joint_Symp_99/MBFig1.jpg Figure 2: http://rana.usc.edu:8376/~mbota/Joint_Symp_99/MBFigure2.jpg |
