Light and heavy reflexives 1 MARK SMITH Abstract Haiman (1983) has argued that the phonological weight of reflexives within many languages is determined by the type of verb that they are coupled with. Specifically, he has claimed that reflexives which occur with ‘‘introverted’’ verbs denoting ‘‘actions which one generally performs upon one’s self ’’ tend to be phonologically lighter than reflexives which occur with ‘‘extroverted’’ verbs denoting ‘‘actions which the subject usually per- forms towards others.’’ Contra Haiman, Kemmer (1993) has argued that light and heavy reflexive marking is determined by the degree to which the participants in the action denoted by a sentence are distinguishable. The current article defends Haiman against Kemmer’s critique but also advances many other criticisms of Haiman’s analysis. Thus, it is argued contra Haiman that introversion and extroversion can be properties not only of the actions denoted by verbs but of the actions denoted by sentences also. Furthermore, it is shown contra Haiman that notions of introversion and extroversion can di¤er across languages. Finally, it is demonstrated that although the distinction between introversion and extroversion can ac- count for more instances of the contrast between light and heavy reflexive marking than envisioned by Haiman, it still fails to explain every such in- stance. Problems with other theories which draw upon the distinction be- tween introversion and extroversion to motivate the contrast between light and heavy reflexive marking such as Reinhart and Reuland (1993) are also highlighted. Finally, it is shown that theories of reflexivization which do not appeal to the distinction between introversion and extroversion such as Chomsky (1986) or Pollard and Sag (1992) are unable to predict observed patterns of light and heavy reflexive marking. It is concluded that no satis- factory account of the contrast between light and heavy reflexive marking currently exists and that further data pertaining to issues such as diachronic evolution and cross-linguistic variation is required if such an account is to emerge. Linguistics 42–3 (2004), 573–615 0024–3949/04/0042–0573 6 Walter de Gruyter Brought to you by | University of Sussex (University of Sussex) Authenticated | 172.16.1.226 Download Date | 7/17/12 7:21 PM