Blake Lemoine
 

            The grammaticalization of ungrammatical sentences is a phenomenon with which all syntacticians are intimately familiar.  Many ungrammatical sentences become more grammatical with repeated exposure.  This paper presents two potential mechanisms to account for how grammaticalization happens in an individual.  The first of these is ad hoc creation of functional elements.  The sentence, “What did who do to whom?” can be made grammatical by stipulating a question that asks for non-animate unknowns (i.e. C[Q+, Wh+, Anim-]).  A second type of sentence such as “Who is there a picture of on your desk?” can be made grammatical by altering the mental parsing algorithm to allow for movement out of subject and adjunct positions.  If this hypothesis is correct then there are several empirically testable phenomena that should differentiate these types of grammaticalizations.  After a new question type is created, lexical lookup will occur quickly.  The algorithmic grammaticalization would not enjoy this same repeated use efficiency.  If these are the primary methods for grammaticalization then there exist sentences that are more difficult, if not impossible, to make grammatical.  One such example is “What did what do what to?”  Early test results indicate that these empirical phenomena are in fact present.  Lab studies will not be completed until early February.  The experiment consists of the subjects reading a grammatical sentence and answering a simple question about it.  The prompt sentences are simple and grammatical, but the grammaticality of the questions varies in such a way as to test the hypothesis.  The primary indicator for speed of grammaticalization is latency, but correctness of answer is used as a secondary indicator.  Examining how a person makes an ungrammatical sentence grammatical will provide insight into the nature of the human parsing algorithm as well as the nature of the lexicon’s role in grammaticality judgments.