Document Type

Article

Publication Date

5-15-2018

Publication Title

Frontiers in Neuroscience

Department

Department of Psychological and Brain Sciences

Additional Department

Geisel School of Medicine

Abstract

The human brain rapidly deploys semantic information during perception to facilitate our interaction with the world. These semantic representations are encoded in the activity of distributed populations of neurons (Haxby et al., 2001; McClelland and Rogers, 2003; Kriegeskorte et al., 2008b) and command widespread cortical real estate (Binder et al., 2009; Huth et al., 2012). The neural representation of a stimulus can be described as a location (i.e., response vector) in a high-dimensional neural representational space (Kriegeskorte and Kievit, 2013; Haxby et al., 2014). This resonates with behavioral and theoretical work describing mental representations of objects and actions as being organized in a multidimensional psychological space (Attneave, 1950; Shepard, 1958, 1987; Edelman, 1998; Gärdenfors and Warglien, 2012). Current applications of this framework to neural representation (e.g., Kriegeskorte et al., 2008b) often implicitly assume that these neural representational spaces are relatively fixed and context-invariant. In contrast, earlier work emphasized the importance of attention and task demands in actively reshaping representational space (Shepard, 1964; Tversky, 1977; Nosofsky, 1986; Kruschke, 1992). A growing body of work in both electrophysiology (e.g., Sigala and Logothetis, 2002; Sigala, 2004; Cohen and Maunsell, 2009; Reynolds and Heeger, 2009) and human neuroimaging (e.g., Hon et al., 2009; Jehee et al., 2011; Brouwer and Heeger, 2013; Çukur et al., 2013; Sprague and Serences, 2013; Harel et al., 2014; Erez and Duncan, 2015; Nastase et al., 2017) has suggested mechanisms by which behavioral goals dynamically alter neural representation.

DOI

10.3389/fnins.2018.00316

Comments

Open access publishing fee supported by the Dartmouth Open Access Publishing fund.

COinS