BRIEF REPORT Words can slow down category learning

The uniquely human ability to use language not only allows for effective communication but may also provide necessary grounding for certain cognitive activities (e. Researchers have suggested that words may make abstract notions more tangible, allowing for the formation of concepts that are easier to interpret and manipulate (Clark, 2006), or that words provide additional processing power while task switching (Emerson & Miyake, 2003), reasoning about false beliefs (Newton & de Villiers, 2007), and learning new categories (Lupyan, Rakison, & McClelland, 2007). Lupyan et al. (2007) found that named categories are easier to learn than unnamed categories, even when the names are redundant. In their study, participants were shown a series of novel objects (" aliens "), asked to respond to them (either to approach or escape), and then given auditory feedback (a buzz or chime). In the word condition, the feedback was followed by the presentation of a redundant category name. Participants in the word condition outperformed those in the no-word condition. This study demonstrated that words can enhance category learning, but the question of just what words do remains. Two accounts have been proposed of how words may influence categorization. First, words may enhance a mental representation, providing an additional robust, perhaps symbolic, representation for the whole category of objects. For example, the word " cup " may allow us to think about the category cup more efficiently than the phrase " concave object with a handle " and more abstractly than a mental picture of a cup. In this sense, words provide " material symbols " that augment category representations (Clark, 2006; Lupyan et al., 2007). An alternative account is that the effects of words are attentional. Rather than augmenting representations, words may shift attention to the perceptual dimension relevant for the task at hand. In a categorization task, words may act as contextual cues, highlighting the features of an object that have proven to be predictive of category membership. In general, cues that have been probabilistically associated with some stimulus in the past enhance the detection, processing, and learning about that stimulus In addition, category learning has been shown to shift attention to diagnostic dimensions in the short term (e.g., Goldstone, Lippa, & Shiffrin, 2001), and possibly in the long term (e.g., Winawer et al., 2007). In the context of object categorization, shape is a property that is typically predictive of category membership ; every theory …