Complementary neural representations for faces and words: A computational exploration

Abstract

A key issue that continues to generate controversy concerns the nature of the psychological, computational, and neural mechanisms that support the visual recognition of objects such as faces and words. While some researchers claim that visual recognition is accomplished by category-specific modules dedicated to processing distinct object classes, other researchers have argued for a more distributed system with only partially specialized cortical regions. Considerable evidence from both functional neuroimaging and neuropsychology would seem to favour the modular view, and yet close examination of those data reveals rather graded patterns of specialization that support a more distributed account. This paper explores a theoretical middle ground in which the functional specialization of brain regions arises from general principles and constraints on neural representation and learning that operate throughout cortex but that nonetheless have distinct implications for different classes of stimuli. The account is supported by a computational simulation, in the form of an artificial neural network, that illustrates how cooperative and competitive interactions in the formation of neural representations for faces and words account for both their shared and distinctive properties. We set out a series of empirical predictions, which are also examined, and consider the further implications of this account.

Visual pathways from the perspective of cost functions and multi-task deep neural networks

“…When considering object based representations both distributed (Avidan & Behrmann, 2009;Haxby et al, 2001) and module-based representations Module-based representations, and theories stressing their importance, point to the existence of distinct cortical modules specialized for the recognition of particular classes such as words, faces and body parts. These modules encompass different cortical areas and, in case of the fusiform face area and visual word form area, even similar areas but in different hemispheres (Plaut & Behrmann, 2011). Conversely, distributed theories of object recognition point to the possibility to decode information from a multitude of classes from the patterns of activity present in a range of cortical regions (Avidan & Behrmann, 2009;Haxby et al, 2001).…”

Music and words in the visual cortex: The impact of musical expertise

“…Proverbio et al (2013) found that left fusiform activations to words were about one centimeter more anterior in musicians than in controls, without assessing the significance of this difference and using the spatially imprecise method of event-related potentials. Another example of interaction between categories of visual objects in defining the layout of cortical specialization was reported by Dehaene et al (2010), who showed that literacy was correlated with a shrinking of face-selective areas contiguous to the VWFA, and an increase in face-induced activation in the right FFA (see also Plaut and Behrmann, 2011;Pegado et al, 2014). Srihasam et al (2014) showed that when monkeys learned a second symbol set, activations induced by a first-learned set moved away slightly, again suggestive of competition between categories for cortical space.…”

Incremental learning of perceptual and conceptual representations and the puzzle of neural repetition suppression

“…Connectionist neural networks are composed of interconnected pools of computational elements referred to as "units," with simple rules of activity integration and propagation amongst the units via weighted connections, also referred to as "synaptic strengths." While such models have typically been abstracted away from the details of the brain, their underlying equations describing activity propagation are isomorphic to more biophysically derived population firing-rate models (e.g., Amit & Tsodyks, 1991;Gerstner, 1998;Wilson & Cowan, 1972;see Ermentrout, 1998, for review), and more recent versions of these models have explicitly included constraints from neuroscience, such as a bias for short-range connections, as well as including more anatomical constraints and constraints on mechanisms of plasticity (e.g., Braver, Barch, & Cohen, 1999;Gotts & Plaut, 2002;Hazy, Frank, & O'Reilly, 2007;Jacobs & Jordan, 1992;Ketz, Morkonda, & O'Reilly, 2013;Norman et al, 2006;O'Reilly, 2006;Plaut, 2002;Plaut & Behrmann, 2011;Usher et al, 1999).…”

“…In the current paper, I focus on a relatively simple but powerful notion that learning of this knowledge is incremental across experiences using Hebbian-like, activity-dependent plasticity at synaptic connections throughout the cortex, referred to hereafter as "incremental learning" (e.g., Becker et al, 1997;Friston, 2005;Friston & Kiebel, 2009;Jacobs, 1999;McClelland, McNaughton, & O'Reilly, 1995;McClelland & Rumelhart, 1985;McRae, de Sa, & Seidenberg, 1997;Oppenheim, Dell, & Schwartz, 2010;Plaut & Behrmann, 2011;Rogers & McClelland, 2004). A key feature of this idea is that different task demands can engage different cortical systems and cells within those systems, which, in turn, can qualitatively alter the nature of the neural representations that are acquired in order to support improved task performance (e.g., Farah & McClelland, 1991;Plaut, 2002;Rogers & McClelland, 2004;Tyler et al, 2000).…”