Acknowledgement The authOrs would like to thank Muffy Siegel，who read an earlier draft of this article and offered many valuable comments and suggestions，and Lu Bingfu，who discussed sOme of the issues with us.we w。uld alsO like to thank the audience at IAcL-16 held in Beijing，2008，particularly Gu Yang，C-T.JameHuang，Jo-wang Lin，Liu Chensheng，and Lu Jianming.Needless tO say，all errors remain our responsibility Notes
In many but not all non-root clause context，SAs are fine.Huang and Li（2008）made a distinction between root clauses and some non-root clauses;this issue wnl be discussed in more detail in Section 5.
The distributional patterns with CAs are more complex than this.in- terested readers can refer to Huang（2006） for more specific discussions on subgroups 0f the complex adjectives，particularly with regard to the use of the so-caned predicate marker de 2 on some of them. Regarding clausal types, Huang（2006）only considers the predicate position of a root clause as the locus fOr this distributional pattern of SAs and CAs. See Sections 3.1 below for a discussion of small clauses in Huang（2006）and Huang and Li（2008）（cf·Gu （2008））.In Huang and Li（2008）we considered a few other clausal types such as conditionals，and also found SA in predicate positions in those clause tvpes.This issue will be discussed in Section 5，where we show that there arestill questions to be answered.
Huang（2006）points out that this constraint follows naturally from the coniunction analysis of modification as well as the Chierchian propertY theory. FOr ease of exposition， it is more convenient to refer tO type matching, a practice we will follow in this article. For further discussion of the literature on this point，see Zhu（1993）， Huang（2006，2008）and references cited therein.
Note that being 0f the same type is a necessary but not snmcient condition for modification.What the type matching constraint stipulates is that mlsmatch in semantic type will not result in good modification in syntax.
展开
