Back to Home
Kent Johnson: Research Statement

 


One of the oldest themes in the philosophy of language, going back to at least Plato, concerns language as a kind of window into the soul or mind. Most of my research carries on this tradition in a contemporary fashion, emphasizing current linguistic theorizing and its roles in a scientifically well-grounded psychology. Along these lines, I have developed views in three overlapping areas. A consistent theme emerging from them is that matters are considerably more complicated than they have often been thought to be.
    (i) Lexical semantics [e.g. 2002, 2003, 2004b, 2004c, 2006b, 2008b]. Word-meaning and lexical syntax is much more complex and counterintuitive than a great many philosophical views claim (given by Evans, Fodor, Peacocke, etc.). Few of us would pretheoretically identify the expression of “manner of motion”, “total affectedness”, etc. as semantic elements to be individuated as genuine parts of word-meaning. But our best linguistic theories strongly say otherwise. Similarly, human languages are too complicated to permit simple global relations between words and syntax, such as systematicity and reverse compositionality.
    (ii) Metaphysical/epistemological relation between current linguistic theories and broader psychological processes [e.g., 2004d, 2004e, 2006c, 2007b]. Many authors have claimed that our access to the meanings of our utterances is primitive, a priori, and immediate. But when made precise, such seemingly obvious claims are nonetheless empirical. Through a series of studies, I found that subjects had real difficulties identifying (orthographically and phonologically unrealized) semantic elements, despite receiving substantial assistance and priming. It seems that our relationship to language (in the sense of our higher cognitive processes’ abilities to consciously discern its workings and structure) is much more complicated and less thoroughgoing than the traditional views suggest.
    (iii) Methodological issues bearing on linguistic theorizing
[e.g., 2004a, 2007a, 2009a, 2009b, Forthcoming]. Perhaps more than any other discipline, linguistics has presented and defended itself as a “scientific” enterprise. I argue that many of methodological criticisms in the philosophical literature are simply unfounded, because the corresponding unquestionably scientific procedures are more subtle, and with good reason. Similarly, the simplified mathematical structure of e.g., Gold’s theorem adds no new constraints on the empirical facts of human language learning. (Some of these issues have led me to reflect on scientific and statistical methods more generally; e.g.2006a, 2008a, 2011.) However, there is still room for methodological improvement. Theoretical inferences in current linguistics are nearly always implicit, holistic  “expert judgments”. But decades of research in the decision sciences has shown that formal explicit methods are to be strongly preferred in such situations. As an extreme example of this, I considered the simple case of counting: how can we estimate the size of a data set used in linguistic theorizing? This turns out to be a difficult question. A theorem in several parts shows that many of the standard methods for performing such an estimation will not work, nor will anything like them.



Back to Home