Language as grist to the mill of cognition Alexandros Tillas ∗ Penultimate Draft – Non-Citable Abstract There is a growing consensus that natural language plays a significant role in our cognitive lives. However, this role of language is not adequately characterised. In this paper, I investigate the relationship between natural language and thinking and argue that thinking operates largely according to associationistic rules. Furthermore, I show that language is neither restricted to interfacing between a ‘Language of Thought’ and the conscious level, nor is it constitutively involved in thinking. Unlike available alternatives, the suggested view predicts and accommodates a large battery of empirical evidence. Furthermore, it avoids problems that associationistic views traditionally faced, e.g. problems of propositional thinking and compositionality of thought. Keywords: Language; Cognition; Associationism; Perceptual Categorisation; Linguistic Impairments; Aphasia. 0. Introduction There is a growing consensus that natural language plays a significant role in our cognitive lives. For instance, according to various thinkers, language greatly contributes to acquiring endogenous control over concepts – the building blocks of thoughts, (e.g. Barsalou 1999; Prinz 2002). Acquiring endogenous control over thoughts is to have thoughts produced by processes of thinking as opposed to processes of perceiving the appropriate stimulus, e.g. tokening or activating the concept TREE while thinking of a tree as opposed to looking at a tree. 1 Furthermore, natural language allows an agent to represent things that she could not have represented otherwise, (e.g. abstract concepts), as well as to exploit the advantages of logical structure (see below). But what exactly is the role of language in thinking? I start by briefly reviewing the main views on the relationship between natural language and thinking, and argue that none of them is satisfactory. In suggesting an alternative account, I start by examining the nature of thinking and then present an associationistic view about the relationship at hand, which avoids traditional problems ∗ Alexandros Tillas, atillas@phil.uni-duesseldorf.de, Institut für Philosophie, SFB 991, Project A03 ‘‘Grounded Cognition’’, Heinrich-Heine-Universität, Raum: 46.21.06.01, Kruppstraße 108, 40227 Düsseldorf, Germany. 1 In what follows, I will be using small caps for concepts. E.g. TREE for the concept ‘tree’.