本文探討客語心理謂語的性質，特別著重於詞彙語意與句法的互動關係。根據句法表現，心理謂語可分為兩類：感事者主語（如英語動詞fear）以及起事者主語（如英語動詞frighten）。客語的「惜」（xiag4）、「愁」（seu5）、「惱」（nau1）、「驚」（giang1）、「譴」（kien2）為前者，而「嚇」（hag4）為後者。本文所有的語料皆取自客語語料庫，力求反映客語心理謂語的真實用法。感事者主語類心理謂語具有類似的句法分布，例如接受程度修飾，以及可以出現在結果或程度構式中。起事者主語類心理謂語以各種方式表達致使：詞彙手段如「嚇」，構詞手段如「V- 死」複合詞，句法手段如「分+NP+V」以及「得+ 人+V」（但後者已形成詞彙化之固定用法）。構詞性致使與句法性致使具有將感事者主語類心理謂語轉換成起事者主語類心理謂語之功能。因此兩者皆可視為一種對客語中起事者主語類心理謂語遠少於感事者主語類心理謂語的不平衡現象加以調整之機制。我們採納Goldberg（1995） 所主張的「構式語法」（ConstructionGrammar）理論。該理論認為語法的基本組成單位為構式（形式與意義的配對）。其優點為合理解釋客語心理謂語與本文中所討論的七種句式（其中三種為感事者主語類，四種為起事者主語類）的互動關係，而不需要去設立額外且特殊個案式的心理謂語意義。

This paper investigates the properties of Hakka psych predicates with afocus on the interaction of lexical semantics and syntax. Based on syntacticrealization, psych predicates in Hakka are divided into two types: experiencersubject(such as the verb fear), including xiag4 “to like,” seu5 “to worry,” nau1“to dislike,” giang1 “to fear,” and kien2 “to get angry,” and stimulus-subject(such as the verb frighten), including hag4 “to frighten.” All data in this paper arecorpus-based and reflect real-world usage of Hakka psych predicates.Psych predicates of the experiencer-subject type have similar syntacticdistributions such as allowing degree modification and occurring in result/extentconstructions. Psych predicates of the stimulus-subject type express causationin terms of lexical (e.g. hag4 “to frighten” ), morphological (e.g. V-xi2 compounds), or analytic (i.e. periphrastic, syntactic) means (e.g. [bun1+NP+V]and [ded4+ngin5+V], though the latter is fossilized, or lexicalized). Themorphological and analytic causatives have the function of converting psychpredicates of the experiencer-subject type into those of the stimulus-subject type.Thus they may be viewed as a kind of mechanism to counteract the asymmetrythat lexical psych predicates of the stimulus-subject type are outnumbered bythose of the experiencer-subject type in Hakka.We adopt the theory of Construction Grammar (Goldberg 1995) whichargues that the basic building blocks in grammar are constructions (formmeaningpairings). It has the advantage of explaining the interaction of the psychpredicates and the seven sentential constructions (three of the experiencer-subjecttype and four of the stimulus-subject type) discussed in this paper, withouthaving to resort to additional, ad hoc senses of the psych predicates.