Chapter 1 the study of collocations

1.3.2  The Semantic Approach 
 
 
Collocation as a linguistic phenomenon associated with lexical semantics 
was described as early as 2,300 years ago.  Greek Stoic philosophers, according 
to Robins (1967), rejected the equation of "one word, one meaning" and shed 
light on an important aspect of the semantic structure of language: "word 
meanings do not exist in isolation, and they may differ according to the 
collocation in which they are used" (Robins 1967:21). 
 
In parallel to the lexical composition approach, where linguists 
recognised lexis as a level of analysis of language separate from grammar, in 
the semantic approach linguists attempted to investigate collocations on the 
basis of a semantic framework, also separate from grammar.  
 
142

---

 
Chomsky was among the first to suggest the treatment of collocations by 
semantics.  Even though Chomsky did not examine collocations, he 
distinguished between 'strict subcategorisation rules', i.e. rules that "analyze a 
symbol in terms of its categorical context", and 'selectional rules', i.e. rules 
which "analyze a symbol in terms of syntactic features of the frames in which it 
appears" (Chomsky 1965:95).  These rules assist the generation of grammatical 
strings.  The breaking of strict subcategorisation rules will result in strings such 
as e.g. 'John found sad' and 'John became Bill to leave', while failure to observe 
the selectional rules will give examples like 'Colorless green ideas sleep 
furiously' (Chomsky 1965:149).  He then finds that selectional rules play a 
marginal role in the grammar and suggests that they should be dropped from 
the syntax and be taken over by semantics. 
 
The Neo-Firthians' approach to the study of collocations was found 
inadequate by semanticists because it sorts lexical items into sets according to 
their collocations, but it does not explain why there are lexical items that 
collocate only with certain other lexical items.  In the lexical composition 
approach collocations and sets are studied as if the combinatorial processes of 
language were arbitrary (Lehrer 1974:176).  
 
Firth's theory of meaning was found to be insufficient for the study of 
collocations (Lyons 1966).  Lyons claims that Firth's definition of 'meaning' as a 
"complex of contextual relations" is puzzling, and he criticises the apparent lack 
of principles by means of which "lexical groups by association" can be 
established and "lexical sets" can be defined (Lyons 1966:289-297).  Overall, 
 
143

---

Lyons proposes an abandonment of Firth's theory of meaning, in which the 
statement of meaning by collocation was introduced, because it does not 
coincide with well-established theories of meaning and language description 
and furthermore there are other "more important meaning relations" which 
must be accounted for in a theory of meaning (Lyons 1966:295).  Even though 
Lyons seems to agree that 'collocations' restricted to "syntagms (or collocations) 
composed of a noun and a verb or a noun and an adjective" (Lyons 1977:261) 
are worthy of study by the semanticist, he does not believe that a separate 
collocational level has to be established.  Lyons also proposes that collocations 
should be studied only as part of the synchronic and diachronic analysis of 
language.  For the study of collocations Lyons proposes the notion of "lexical 
fields" founded upon "the relations of sense holding between pairs of 
syntagmatically connected lexemes" (Lyons 1977:261).  However, he advises 
against going to the extreme of "defining the meaning of a lexeme to be no 
more than the set of its collocations" (Lyons 1977:265-268).  He then proceeds to 
describe the principles of a strong version of field-theory as if the vocabulary of 
a language was a closed set of lexemes with each lexeme being a member of no 
more than one field.  However, the vocabulary of a language is an open system, 
and lexemes do belong to different fields due to their different meanings.  
Therefore, the study of vocabulary in a theory of lexical fields based on 
syntagmatic relations presents problems.  These problems led Lyons to suggest 
that descriptive semantics can get along well without syntagmatic relations 
(Lyons 1977:268).  Thus, Lyons decides to deal with the 'more important' 
 
144

---

paradigmatic relations of sense in his study of semantics, setting aside the 
study of syntagmatic relations altogether. 
 
Even though Lyons (1977) provided only a criticism of the Firthian 
theory of meaning, there have been other semanticists who tried to put together 
a theory of lexical meaning based on the semantic properties of lexical units.  
This approach is the semantic approach to the study of collocations.  According 
to the semantic approach, the meaning of a lexical item is perceived as a 
combination of the semantic properties of that item.  It is the semantic 
properties of a lexical item that determine its collocates. 
 
Just as the Neo-Firthians tried to establish the lexis as different from 
grammar, the semanticists also tried to establish a semantic theory that is 
different from, but complementary to, grammar.  Katz and Fodor (1963) 
describe a semantic theory that would organise, systematise, and generalise 
facts about meaning (Katz & Fodor 1963:170).  They state that a semantic theory 
of a language would "take over the explanation of the speaker's ability to 
produce and understand new sentences at the point where grammar leaves off" 
(Katz & Fodor 1963:172-173).  They accept that one component of a semantic 
theory of a language is a dictionary of that language, and they proceed to 
describe the semantic markers for a few lexical entries of a model dictionary of 
English.  According to the semantic theory proposed by Katz and Fodor, each 
entry for a lexical item in the dictionary must contain a selection restriction, i.e. 
a condition for that particular lexical item to combine with others.  For example, 
the lexical item 'sleep' would require a subject with the feature [Animate], and 
 
145

---

the lexical item 'break' would require as object something that is a [Physical 
object] and [Rigid]. 
 
Due to the fact that under the semantic approach to the study of 
collocations each lexical item will be defined by semantic markers based on its 
meaning or meanings, Lehrer (1974) argues that the semantic approach is more 
likely to explain why certain words can be found together.  In his examination 
of syntagmatic meaning relations between lexical units, Cruise describes 
collocational restrictions as co-occurrence restrictions that are arbitrarily 
established (Cruse 1986:279).  For example, 'kick the bucket' can only be used 
with human beings, although its propositional meaning is simply 'die' and not 
'die in a characteristically human way'.  Similarly, 'blond' refers to hair, but 
describing a hairy animal or a fur coat as 'blond' would be unacceptable.  Cruse 
also distinguishes three kinds of collocational restrictions: systematic, semi-
systematic and idiosyncratic, according to whether, and if so to what degree, 
the semantic properties of a lexical item set up an expectation of a certain 
collocant.  Lexical units that belong to the category of systematic collocational 
restrictions are 'grill' and toast'.  Both verbs denote the same process from the 
point of view of the agent, but different patients: normally we 'grill' food that is 
raw, while we 'toast' food that is already cooked.  Semi-systematic are those 
collocational restrictions that still behave as presuppositions of the lexical item 
in question, but there can be certain exceptions to the general tendency.  For 
example, 'customers' obtain something material in exchange for money, while a 
'client' receives a less tangible professional or technical service.  So, butchers, 
 
146

---

bakers, and grocers have  'customers', but solicitors and architects have 'clients'. 
 However, banks seem to have 'customers' rather than 'clients' (Cruse 1986:281). 
 Finally, for lexical items that present idiosyncratic collocational restrictions, 
their collocational ranges can only be described by enumerating all their 
acceptable collocants (Cruse 1986:281).  For example, one can 'pay attention/a 
visit' but not ?'pay a greeting or welcome'.  Idiosyncratic collocational 
preferences, such as 'flawless performance' but not *'unblemished 
performance', do not give rise to presuppositions, according to the semantic 
approach, and Cruse wonders whether "idiosyncratic restrictions are a matter 
of semantics at all" (Cruse 1986:282).  A close study of what collocational 
restrictions can deliver to the sentence they are used in is totally justified, since 
they are not 'logically' necessary.  For example, 'die' and 'pass away' have the 
same meaning, but 'pass away' refers to human beings, so the use of 'pass 
away' in the sentence 'My grandfather passed away' adds semantic cohesion to 
it; if it is used to describe the death of a pet animal then it anthropomorphises 
the animal (Cruse 1986:280).  Due to the difficulty of the syntagmatic relations, 
Cruse (1986), like most lexical semanticists, finds that paradigmatic sense 
relations are "a richer vein to mine than relations of the syntagmatic variety" 
(Cruse 1986:86). 
 
One of the weaknesses of the semantic approach - the view that co-
occurrence of words is the result of their semantic properties - is that there is a 
large number of idiosyncratic co-occurrences or combinations that are 
arbitrarily restricted (see Cruse's examples above).  These constructions create 
 
147

---

problems for the study of collocations under a theory of lexical fields, and 
therefore they are left unexplained and marginal by semanticists.  To return to 
Halliday's example, since there is nothing in the meaning of 'tea' to explain why 
it collocates with 'strong' but not with 'powerful', according to the semantic 
approach, it will be listed as an idiom and as such it will be ignored in a study 
of lexical semantics.  Furthermore, as Lehrer (1974) points out, finding semantic 
features for each lexical item that would account for all its collocates is an 
extremely ambitious task (Lehrer 1974:178).  Fillmore (1978) also points out the 
difficulty of estimating the magnitude of collocational binding between lexical 
items, while he acknowledges the fact that a semantic theory must not accept 
the suggestion that all meanings must be described in the same terms.   
 
An example of how the semantic approach to the study of collocations 
can be best utilised was the compilation of a prototypical dictionary, the 
Explanatory Combinatorial Dictionary (ECD), of any language.  The ECD is 
related to the Meaning-Text theory which defines language as "a specific 
system of correspondences between an infinite set of meanings and an infinite 
set of texts" (Mel'cuk 1988:167).  As a core component of the Meaning-Text 
Model, the ECD, according to Mel'chuk (1988): 
 
 
"ensures the lexicalisation of the initial meaning (i.e., of semantic 
representation), uniting bundles of configurations of semantic elements 
into actual lexical units and supplying the enormous bulk of syntactic 
and lexical co-occurrence information that accrues from the individual 
lexical units of the language in question" (Mel'cuk 1988:167).   
 
 
148

---

 Each 
ECD entry is divided into three zones: a semantic zone, a syntactic 
zone and a lexical co-occurrence zone.  The latter comprises all the restricted 
lexical co-occurrences of the entry lexeme.  For this purpose, Mel'cuk and 
Zholkovsky, the ECD initiators, devised the concept of Lexical Functions that 
describe all the paradigmatic and syntagmatic relations that a lexeme can have 
with other lexemes (Mel'cuk & Zholkovsky 1988:42).  The above approach 
resulted in a large number of standard basic lexical functions - some of which 
had already been utilised in dictionaries for several decades (e.g. 'Syn' for 
synonyms) and others were new (e.g. 'Instr' preposition meaning 'by means of', 
and 'Propt' preposition meaning 'because of', 'as a result of') (see Table 1, 
above).  In the ECD version for French, Dictionnaire Explicatif et Combinatoire 
du Francais Contemporain, there are 53 lexical functions listed, and these are 
used together with the other semantic and syntactic information for the 
description of 50 lexical items.  Mel'cuk and Zholkovsky are considered 
pioneers in their lexicographic principles and the heuristic criteria they used for 
the compilation of the ECD.  The fact that only 50 lexical items were described 
in the French ECD underlines the extremely difficult task of listing all the 
semantic features of lexical items in an effort to account to all its collocates.  
Despite its limitations, the ECD could be used as "a central component of 
automatic text synthesis and analysis", as a "format" for the development of 
textbooks, pedagogically oriented dictionaries, and reference works, and also it 
can contribute to language theory (Mel'cuk & Zholkovsky 1988:66-67).   
 
149

---

 
Even though semanticists claimed that syntagmatic lexical relations 
should be studied under the scope of semantics, they did not proceed any 
further with the study of collocation and they did not make the phenomenon of 
'collocation' any more explicit.  Due to the irregularities and idiosyncrasies that 
collocations present, semanticists, who followed a similar role to grammarians 
(i.e. assigning semantic labels to sentence constituents and examining 
generalisable tendencies and regularities), preferred to study the more regular 
paradigmatic lexical relations, abandoning collocations to their rulelessness.