Based on this theory, Halliday (1966: 156) emphasizes specifying a linguistic item in collocational environment during the lexical analysis, in which the primary task is to demonstrate the tendencies of items that are frequently co-occurred with each other。 Afterwards Sinclair (1987, 2004) further pided the lexical tendency into two types: the terminological tendency (i。e。 distributing in an “open choice principle” condition) and the phraseological tendency (i。e。 distributing in an “idiom principle” condition)。 He affirmed that the meaning of a lexical item is defined by its distributional patterns in these two tendencies。 

Thanks to Firth and his followers like Halliday and Sinclair, who together verified collocational distribution is a key to lexical semantics in the 1950s and 1960s, and their eminent work has shaped contemporary lexical semantics research theoretically and opened the door to the promising experiments that establish a close relation between the meanings and distributions of lexical items (Miller& Charles 1991; Church et al。 1994)。

2。1。2 Colligation 

The word colligation originates from the Latin, referring to "tie together," and it seems as a type of collocation, however, it is much more abstract than the concept of collocation。 Colligation was defined as the interrelation of grammatical categories in syntactical structure by Firth (1957), who first proposed this concept in linguistics。 Colligation is based on lexical collocation, but it is on a syntactic level, while collocation is on a lexical level。 In short, the study of collocation focuses on the specific words in the habitual co-occurrence, and the study of colligation is on the abstract interrelation among grammatical categories。 They mutually influence and restrict each other。 Only when the two items are correlated with each other in the same grammatical structure can this structure be recognized as colligation。 

2。1。3 Semantic preference

Semantic preference is a kind of patterning which is based on collocation。 In this kind of patterning, a large amount of a word’s collocates have same or similar semantic features。 Stubbs (2001:65) defined semantic preference as a relation between a lemma and a group of words related semantically。 Sinclair (1996, 1998) identified it as one of four important types of semantic relations, whose principle depends on the relationship between a set of frequently occurring collocates and certain shared semantic property。 

Therefore it is a useful and practical way to provide a semantic profile of a word (i。e。 the semantic field that a word belongs to) and reveal the central senses it has。 Mutual Information (MI) score can better reflect semantic preference, which has been testified and wisely implied in corpus analysis of lexis nowadays。 MI test compares the joint probability and chance (observing two words independently) of the target two words。 The higher the MI score is, the two words are more likely to collocate with each other, i。e。 to demonstrate their strong semantic preference。 The MI score around 0 implies that the examined two words hardly collocate, while if it is about or higher than 3 suggests that the target two items are habitually collocate with each other。 

2。1。4 Semantic prosody

Semantic prosody, originating from “prosody” used by Firth, is coined by Sinclair (1991)。 It has become a new term in post- Firthian corpus linguistics, which is remarkably elaborated by Stubbs (1996) and Partington (1998)。 It describes the way in which the target apparently neutral words can be perceived with evaluative connotations through frequent occurrences with certain special collocates。 Stubbs (1996: 176) classified semantic prosody into three types: positive, neutral or negative。 If the collocates of the core word or phrase denote a positive meaning, it can be sorted to positive prosody。 When the core word or phrase possesses a mixed or intricate meaning, it will be classified into neutral one。 Accordingly examining the prosody of collocates the word or phrase frequently company with is a good way to find out the hidden attitudes and connotative implications。 论文网 基于语料库的同义动词的行为模式分析以imply,indicate,signify和suggest为例(5):http://www.youerw.com/yingyu/lunwen_195900.html