Indirect cognitive control through top-down activation ofperceptual symbols

摘要

In classical, "non-embodied" approaches to mental representation, human action is construed as the output of acentralized system that is heavily involved in forming an internal model of the world, one composed of "amodal" symbolsthat, for combinatorial and computational purposes (Pylyshyn, 2002), do not retain any of the properties of thesensorimotor states that gave rise to them (Arkin, 1998; Barsalou, 1999, 2008; Kosslyn, Thompson, & Ganis, 2006). Inthese frameworks (e.g., Burgess & Lund, 1997; Landauer & Dumais, 1997), knowledge is represented by truth tables,feature lists, and elements as arbitrary as those of computer memory (binary digits). In unpredictable environments, suchcentralized systems require tremendous computational power to continually update their amodal model of the world(Arkin, 1998). This error-prone and time-consuming translational process is unnecessary in modal systems (Arkin, 1998;Markman & Dietrich, 2000), for action-relevant information is already available in the environment or is embodied inrepresentations that maintain properties of the sensorimotor states that gave rise to them (e.g., perceptual symbols;Barsalou, 1999, 2008; Farah, 2000). As Pickering and Garrodvropose, this advantage of speed and accuracy rendersembodiment of form and meaning as mechanisms that may explain how the fast and intricate cognitive processes ofdialogue can be carried out so effortlessly.