Simulators

Simulators
Much work has shown that categories tend to have statistically correlated
features (e.g., Chin-Parker & Ross, 2000; McRae, de Sa, & Siedenberg,
1997; Rosch & Mervis, 1975). Thus, when multiple instances of the same
category are encountered, they tend to activate similar neural patterns in
feature maps (cf. Farah & McClelland, 1991; McRae & Cree, 2002). As a
result, similar populations of conjunctive neurons in convergence zones—
tuned to these specific conjunctions of features—tend to capture these
patterns (Damasio, 1989; Simmons & Barsalou, 2003b). Over time, this
population of conjunctive neurons integrates modality-specific features
across category instances and settings, establishing a multimodal represen-
tation of the category. Figure 2a provides a highly simplified and schematic
illustration of the resultant distributed system. Barsalou (1999b) referred to
these distributed systems as simulators. Conceptually, a simulator functions
as a type, integrating the content of a category across instances and
providing the ability to interpret later individuals as tokens of the type
(Barsalou, in press).
Consider the simulator for the social category face. Over time, visual
information about how faces look becomes integrated in the simulator,
along with auditory information for how they sound, somatosensory
information for how they feel, motor programs for interacting with them,
emotional responses to experiencing them, and so forth. The result is a
distributed system throughout the brain’s association and modality-specific
areas that establishes conceptual content for the general category of face.