Markus Werning - Neuroframes – A Neural Model of Situated Conceptualization

In the paper a philosophically grounded and empirically motivated theory is introduced that largely holds on to a representationalist view on concepts (Fodor, 1998; Werning, 2004a, 2005a) and at the same time views concepts as situated, i.e., based on senso-motoric schemata. Mental concepts, here, fulfil a twofold function: They constitute the meanings of linguistic expressions – words, phrases and sentences – and they provide the contents of intentional states, including beliefs, desires, the will, imaginations, recollections, expectations and perceptions.
   The approach of situated conceptualization (Barsalou, 2005) on the one hand opposes strictly modular accounts of semantics and perception (Fodor, 1983). According to modular approaches to semantics, the meanings of words and sentences are processed in an informationally largely encapsulated, autonomous and amodal way (Clifton & Ferreira, 1987). Candidates for cortical correlates of semantic processes are often localized in left temporal regions of the brain and may also involve the frontal cortex (Friederici, 2002). Cortical regions typically associated with either perceptual or motor processes in this paradigm are typically not regarded to be involved in the processing of semantics. Modular approaches towards perception, in turn, argue for informationally encapsulated, domain-specific and cognitively impenetrable modules for various perceptual tasks (Barrett & Kurzban, 2006, for review). Modularism with respect to semantics and perception would thus be hardly compatible with the view that the same mental concepts, respectively their neural correlates, are both meaning providers for linguistic expression and well as content providers for intentional states such as perceptions.
   On the other hand, the approach of situated conceptualization also dissociates itself from radical proponents in the embodied cognition movement (Brooks, 1991), who reject a representationalist model of the mind tout court (Keijzer, 2002).
   In our representationalist and situated approach the principles of internal and external compositionality (Fodor & Lepore, 2002; Werning, et. al. 2005) as well as the co-variation of internal concepts with external contents are crucial. Only thus can be explained how the semantics of languages can be learned in finite time, how communication about the external world is possible and, more generally, how finite systems are able to grasp a potentially infinite manifold of situations in a systematic way.
   Not to fall back on classical computational models of mental representation (Fodor, 1975; Pylyshyn, 1984), the paper recurs to the already partially developed theory of neuroframes (Werning & Maye, 2005, 2007). The aim here is to provide a neurobiologically plausible implementation of frames, which provides us with mechanisms of concept composition and decomposition and links concepts to syntactically structured linguistic expressions.
   Neuroframes are neurobiologically grounded in the theory of binding by synchrony (Singer, 1999, for review). Neurophysiological evidence for the role of neural synchronization in binding mechanisms pertains to the perceptual domain (Gray, et al., 1989; Engel et. al, 1997; Engel & Singer, 2001), expectation and attention (Engel et al., 2001; Fries et al. 2001, 2002), the domain of action control (Schnitzler et al. 2006), as well as the domain of language comprehension (Weiss et al. 2000, 2005). The theoretical analysis is supported by oscillatory neural network simulations (Maye, 2003; Maye & Werning, 2007).