Concepts and Characterizations

in this sense. For instance, predication can combine any predicate phrase (e.g. ‘is an F’ or ‘Fs’) with any noun phrase (e.g. ‘a’ or ‘The F’), regardless of what objects and properties those phrases denote; and the representational significance of that combination is just that the object denoted by the noun phrase possesses the property denoted by the predicate. By contrast, many non-linguistic systems employ combinatorial principles that make a much more robust representational contribution (Camp 2007). For instance, maps employ a spatial combinatorial principle: the spatial arrangement of vehicular items represents an isomorphic spatial structure among the corresponding entities in the world (up to a distance metric). And in turn, the fact that these non-linguistic systems employ representationally robust combinatorial principles significantly limits their expressive flexibility. For instance, because combining items on a map necessarily represents their referents as being arranged in an isomorphic spatial structure, maps are only capable of representing objects and properties as having spatial structures and locations. (Similarly, phylogenetic trees employ the spatial structure of branching lines, where those lines have the representational significance of differentiation with common descent, and this in turn means that phylogenetic trees can only represent objects that have ancestors and descendents.) In principle, the more neutral a representational system’s combinatorial principle is, the wider a range of concepts it can combine: the relative abstractness of its syntactic operation(s) permits correlatively greater compositional systematicity. As with semantic arbitrariness, it is natural to think of formal and natural languages as paradigms of abstractness. And indeed, the relative neutrality of predication, combined with semantic arbitrariness, does make language distinctively topic-neutral in comparison with other formats. This is obviously a desirable property in a representational system. Further, to the extent that a thinker’s conceptual abilities display a similarly high degree of topicneutrality, this suggests that it too employs something like a linguistic format. I believe that such an argument from topic-neutrality – along with the more quotidian fact that we talk so much – does support an inference to the conclusion that much of human thought is in language, or at least is language-like in a fairly strong sense of the term: that is, that it employs semantic and syntactic 4 For ease of exposition, I focus on predication; the point goes a fortiori for other combinatorial principles, such as functional application and Merge.