What's Up can be Explained by Language Statistics

What’s Up can be Explained by Language Statistics Sterling Hutchinson (schtchns@memphis.edu) Department of Psychology/ Institute for Intelligent Systems, University of Memphis 365 Innovation Drive, Memphis, TN 38152 USA Max M. Louwerse (maxlouwerse@gmail.com) Department of Psychology/ Institute for Intelligent Systems, University of Memphis 365 Innovation Drive, Memphis, TN 38152 USA Tilburg Centre for Cognition and Communication (TiCC), Tilburg University PO Box 90153, 5000 LE, Tilburg, The Netherlands Abstract experimental tasks cue participants to refer to relevant perceptual representations, language processing is facilitated (Glenberg & Kaschak, 2002; Kaschak et al., 2005; Pecher, van Dantzig, Zwaan, & Zeelenberg, 2009). For example, Zwaan and Yaxley (2003) demonstrated that when word pairs appeared in their expected physical locations on a computer screen (e.g., ceiling presented at the top of the screen and floor presented at the bottom of the screen), comprehension was faster than when pairs appeared in unexpected physical locations on a computer screen (e.g., floor presented at the top of the screen while ceiling was presented at the bottom of the screen). That is, it is easier to process a word when the expected physical properties of the word match its actual physical properties. Accumulating research like this tends to suggest that individuals rely on perceptual representations, especially in everyday language comprehension. This embodied cognition account of semantic representations is often contrasted to an amodal (or symbolic) account of cognition, whereby language is represented amodally. A classical symbolic account of language representation argues that semantic information is seated in language and can be derived from relationships that exist between symbols instead of from the mental reenactment of biomechanical and perceptual experiences. In other words, meaning is represented in a linguistic structure within the brain encoded in a formal abstract language, and words are understood from their natural linguistic context instead of from their perceptual features. Recently, several studies have argued that an extreme symbolic or an extreme embodied cognition account is untenable, and that a more plausible cognitive model includes both perceptual and symbolic processes in language comprehension (Barsalou, Santos, Simmons, & Wilson, 2009; Louwerse, 2008; 2011a; Paivio, 1986). For instance, Louwerse (2008; 2011a) proposed the Symbol Interdependency Hypothesis. This hypothesis predicts that language encodes the perceptual information we tend to simulate. Consequently, language statistics allows for bootstrapping meaning with only minimal symbol grounding in perceptual experiences. Put differently, according to the idea of symbol interdependency embodied simulations and symbolic relationships are complementary in conceptual processes. Embodied cognition studies have demonstrated that when words found in high physical locations (e.g., bird) are positioned at the top of a screen they are processed faster than when they are positioned at the bottom of the screen. The reverse effect is obtained for words found in low physical locations (e.g., fish). This concept-location facilitation effect has been argued to demonstrate that cognitive processing is fundamentally perceptual in nature. However, questions can be raised with regards to the absolute or relative location of these concept-location words We investigated whether semantic judgments were made with respect to an absolute location on the screen (embodied explanation) or with respect to a relative location in comparison to other words included in the experimental session (statistical linguistic explanation). In a response time experiment we presented participants with physical-location words from existing studies at the top or bottom, top or center, and center or bottom of the screen. For animate words we found a concept location facilitation effect for words presented at the top of the screen, at the center of the screen, and at the bottom of the screen. In addition, however, language statistics explained RTs to center words. Findings indicated that participants made judgments relative to other words on the screen and not relative to their absolute location on the screen, lending support to a statistical linguistic explanation of the findings. Keywords: concepts; embodied cognition; symbolic cognition; concept-location facilitation; perceptual Introduction Embodied cognition theories state that language is understood through perceptual representations that are grounded in modality-specific somatosensory experience (Barsalou, 1999; Glenberg, 1997; Semin & Smith, 2008). Words become meaningful only after mentally reenacting external perceptions and experiences associated with that word. Thus, the patterns of neural activity that occur when comprehending a particular word would be similar to those patterns that occur when actually perceiving its referent (Hauk, Johnsrude, & Pulvermuller, 2004). In other words, according to embodied cognition theories mental representations are couched in the physical and perceptual experiences of the body. There is a wealth of evidence supporting the embodied cognition account, with evidence showing that when