熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber....

31
熊本大学学術リポジトリ Kumamoto University Repository System Title Semantics and Pragmatics of Innovative Noun-Noun Compounds Author(s) Nishikawa, Morio Citation �, 23(�): 1-30 Issue date 1988-01-31 Type Departmental Bulletin Paper URL http://hdl.handle.net/2298/28407 Right

Transcript of 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber....

Page 1: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

熊本大学学術リポジトリ

Kumamoto University Repository System

Title Semantics and Pragmatics of Innovative Noun-Noun

Compounds

Author(s) Nishikawa, Morio

Citation 熊本大学教養部紀要, 23(外国語・外国文学編): 1-30

Issue date 1988-01-31

Type Departmental Bulletin Paper

URL http://hdl.handle.net/2298/28407

Right

Page 2: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

W&kKfc&M&M: WM.iff • IVWXWti %23>f : 1-30 (1988)

Semantics and Pragmatics ofInnovative Noun-Noun Compounds

Morio Nishikawa

[I] INTRODUCTION

[ 1 ] Preliminaries

Homo ludens. homo faber. homo loquens, homo mentiens. homo sapiens, all these 'definitions' ofhuman beings arc initially related with language. Language, as a means of cognition andcommunication, has been the strongest evidence for man to be an intellectual being.

Man has an existence of knowledge and action in terms of communication. He has beenexpanding the world of his own through his experience and knowledge. What he knows andunderstands is reciprocally informed and shared by a speaker and a hearer. What is unknown turnsout to be what is known as a result of speech or text communication. The more we obtain knowledge,the more we learn to acquire knowledge of what we do not yet know. Thus we are pragmatictravellers travelling from what is already known to what is newly known by way of calculating orinferring the former knowledge in actual use of language.

Our assumption is that what is already known can be shared, but what is unknown cannot beshared among the participants of the act of communication. It can only be asked about andquestioned. The goal of our comprehension task is to turn what is unknown to what is newly known.What is significant in the context of this paper is the assumption that comprehension orunderstanding takes place through the contextual knowledge which enables us to calculate or inferthe intended meaning of a given word. With a piece of already familiar knowledge as a point ofdeparture, we are expected to arrive at the goal of acquiring a new knowledge.

In the course of verbal communication, at least two people are involved: a speaker and ahearer, or a writer and a reader or (to use more general term) an addresser and an addressee (cf.Jakobson 1960 ! 353) . To avoid unnecessary confusion, we will accept the following descriptionprovided by Leech (1983 '. 13) from now on regarding the terminology of 'speaker' and 'hearer':

Following the practice of Searle and others, I shall refer to addressers and addresseesas a matter of convenience, as &• ('speaker') and h ('hearer') . These will be ashorthand for 'speaker (s) /writer (s)' and 'hearer (s) /reader (s)'. Thus the use ofabbreviations .v and h does not restrict pragmatics to the spoken language.

In fact, on the one hand, s provides a verbal message for h and produces a certain amount of writtentext and spoken utterance, and h comprehends or understands what is said (or written) andintended by s through a given expression. In order to produce utterances appropriately, s must havea set of the knowledge of grammar, a good amount of stored knowledge of the lexicon, and pragmaticknowledge : e.g. knowledge of the Cooperative Principle/Maxims (Grice 1968) and the Politeness

Page 3: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Morio Nishikawa

Principle/Maxims (Leech 1983 '. 81). And the selection and combination of linguistic constituents haveto take place appropriately for the verbal communication between s and h to be successful.

On the other hand, h is the counterpart of s in communication. He/She makes efforts tounderstand what is said and intended by s through a given expression. In this process, h is makingthe best use of the presupposed and shared knowledge of context which enables him/her to performthe task of comprehension.

When we mention noun-noun compounds (henceforth NNCs), many people will point out well-known and conventional NNCs such as the following :

( 1 ) The paddling pool was filled with lady birds.

— The Sunday Times 13/4 1986

( 2 ) Do you remember to leave a Christmas box for a dust man ?— Daily Mail 2/12 1985

( 3 ) They threw out 20 mailbags at three pre-arranged points along the railway lines.— Sunday Telegraph 2/2 1986

However, there are other kinds of NNCs which are not well-established. They cannot be found out inthe lexicon. We call them innovative NNCs. Notice the following :

( 4 ) Some 40 % of foam blanket is absorbed by trees and plants.— The Sunday Times 9/2 1986

( 5 ) It can handle only 40 planes, compared to 200 with MLS. and can be distracted byghost signals reflected from buildings or other large objects.

— The Sunday Times 3/8 1986

( 6 ) In the United States, she has four horses, and a bug-eyed tree animal which has acomplicated name but which is basically an over-sized bush baby.

— The Sunday Times 22/6 1986

They are not so well-established as the above institutionalized examples. We cannot understand what theseinnovative NNCs mean out of appropriate context. They are at most on the way to full acceptability. Andh is expected to understand or comprehend them by making the best use of the knowledge from thecontext where they are actually used. This interpretative process is based upon the assumption thatwhat is unknown can be understood by using the knowledge of what is already known.

The present paper is devoted both to semantic and pragmatic study of English NNCs. The basicview-point adopted in this paper is not transformational but is functional or contextual; that is tosay, we want to take account of a certain amount of knowledge derived from the context where agiven NNC is used. But we do not forget the heritage of structuralism which has been one of themain traditions of linguistics. It still provides us with an important understanding of the system orinner mechanism of human language. Structuralism, however, is not enough.

We also do not forget to learn from recent and current important works of English compounds.Some of them are works provided by transformationalists like Lees (1969) and Levi (1978), andlexicalists like Chomsky (1970). Jackendoff (1975), and Selkirk (1982). Being different in manyways, they are common in adopting a structuralist point of view, and their basic approach relies on a

Page 4: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds

context-free grammar of word-formation.

The data in this research are mainly taken from two sources. One is from journalistic writing,including newspapers and magazines, the other is from data provided by the native speakers of Englishthrough experiment. I have excluded literary works because they seem more or less to be consciouslyconstructing words which have primarily an aesthetic rather than communicative function. In theexperiment, the data provided by informants were obtained from an answer sheet. I asked 25 nativespeakers of English to pick up randomly two cards containing a noun, and to combine them. Amongthem, two were the flat neighbours from London and another two were the people working at theuniversity. The subjects were mostly the students of the University of Lancaster where I carried outthis research. The first noun was regarded as a head noun, and the second noun as a sub-categorizingword. The number of noun-noun combinations provided by one person was limited to three or less,because I had to make the data objective and varied. The nouns which I put on each card were allordinary and common nouns from the Oxford Picture Dictionary of American English.

[2] Production and Comprehension

Language is closely connected with the linguistic act of communication, which is notindependent from human knowledge of language. Halliday (1985 : xiii) claims: "all languages areorganized around two main kinds of meaning, the 'ideational' or reflective and the 'interpersonal' oractive." He also refers to 'the manifestation in the linguistic system of the two very generalpurposes.' They are " (i) to understand the environment (ideational), (ii) to act on the others on it(interpersonal) ." And thirdly, he points out a further meta-functional manifestation, that is, the'textual' function which breathes relevance into the other two. According to him. the linguistic act ofcommunication is nothing but an interaction between these three different phases.

Communication is a linguistic act of cooperation between s and h, and the intended meaning ofa word is realized by its use in context. The study of NNCs must not be imprisoned within theobscure area of mental process. In order to show the relevance of a functional model like Halliday's,it may not be insignificant to compare it with the more limited code model of communication. Fossand Hakes (1978 '. 11) elaborated the distinction between code and message, and they summarizedthe process of meaning transmission and translation as follows:

( a ) Language is essentially a means of relating two different kinds of patterns or formsof representation— sound and meaning. The representation of language sounds, thephonological system, is an external representation, the representation of meaning, thesemantic system, is an internal representation. The means of relating the two is alanguage's syntactic system.

( b ) We can think of the external and internal representations as involving differentcodes. The external, phonological code is the one in which language is transmittedfrom one person to another, from speaker to hearer.

( c ) The internal semantic code is the one in which speakers and hearers represent themeanings or messages of the sentences they produce and understand.

( d ) It is our knowledge of the language that allows us to translate the sounds thespeaker produced into a representation of the message he was trying to communicateto us.

Page 5: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Morio Nishikawa

This is a clear-cut explanation of the code model of linguistic communication. The concept of 'code' isclearly distinguished from that of 'message'. In this model, while the addresser is encoding amessage, the addressee is expected to decode it. To my understanding, the code is fundamentally thestored and conventional knowledge which is shared between the addresser and the addressee. It is

also the model which is abstracted from the message through its communal necessity or frequent usein actual communication. Thus the code model itself is a context-free model of comunication and it

does not work well out of context. Eventually, the code model does not work well when we comeacross innovative NNCs whose meaning can be calculated only contextually.

[ 3] Innovative NNCs

Foam blanket in ( 4 ) and ghost signals in ( 5 ) are the examples of innovative NNCs in English.They cannot be found out in the code model of the lexicon. That is to say. we cannot retrieve the'knowledge' of these words from any existing dictionary. We must infer or calculate the intendedmeaning by taking account of the knowledge from the context where these words are used. The codemodel does not work well in the process of understanding these innovative NNCs.

A compound is not a phrase but a word, consisting of two (or more words). The question as tohow a combination of existing words makes up another new word is the essential problem ofinvestigating compound formation not only in morphology but also in semantics and pragmatics.

In a compound, the second constituent, which is the head, is subcategorized by the first constituent.Due to the right-hand head principle in English compounds (cf. Williams 1981, Selkirk 1982 '. 20), thecategory of the second constituent is on a par with that of the whole compound. Consider the following :

( 7 — 1 ) water-bug( 7 — 2 ) water-colour( 7 - 3 ) water-level( 7 — 4 ) water-clock( 7 — 5 ) water-melon( 7 — 6 ) water-craft

In these examples, the varieties of meaning are illustrated in the combination 'water-X'. And, in ageneral sense, all the examples imply 'X has something to do with water'. Since they are characterizedby conventional or lexicalized meanings like in the examples ( 7 — 1 ) to ( 7 —6 ). we do not havemuch difficulty in understanding their meaning : e. g. water colour is " 1 . water-soluble pigment, appliedin transparent washes and without the admixture of white pigment in the lighter tones; 2 . a paintingdone in watercolours" (CED), and water clock is "any of various devices for measuring time that usethe escape of water as the motive force" (CED). However, if an given NNC is an innovative one likethe example below, we cannot retrieve its meaning from the lexicon. Consider the following :

( 8 ) water genaration

This is an example of an innovative NNC out of cantext. What interpretation can we provide to thisNNC ? And how can we specify one interpretation uniquely from among possible or candidateinterpretations ? Consider the following :

Page 6: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds

( 9 — 1 ) "today's children, who unlike their parents, have almost all learnt to swim

( 9 —2 ) "a group of children born under water under the supervision of a particularmedical expert advocating this method"

These are possible interpretations of this NNC provided by one of the informants, and they makegood sense. But, as a matter of fact, they cannot be applicable in (10).

(10) They are the water generation. Their ubiquitous little green bottles clutterrestaurant tables this festive time of year

— The Times 19/12 1985

This is an authentic example of the innovative NNC of ( 8 ) which is actually used in the context.Unlike the examples from (7 — 1) to (7 —6), this NNC does not inform us of any establishedmeaning. We have to infer or calculate what is meant by this NNC by taking account of theknowledge from the context. For example, water in water generation is semantically linked with "littlegreen bottles" which means non-alcoholic drinks. They are the generation who drink 'water' ratherthan drinking alcoholic drinks. They may be the generation allowed to drink beer or whisky, butthey do have their 'festive times' in restaurants, probably not in public house, and so forth.

We now move on to the next problem of communication: i. e. Is shared knowledge necessary forthe comprehension of the meaning of a given NNC? The answer in our position is yes. In otherwords, it is necessary for h to have a certain amount of shared knowledge in order to solve theproblem of comprehending the intended meaning of NNCs. Without any shared knowledge between sand h, h cannot find out the way to speaker's intended meaning. We summarize below a rough sketchof the conditions of verbal communication.

(A) There must exist a speaker (s) and a hearer (/i) in the course of verbalcommunication, even in the case of monologue where 5 himself/herself is h. S is anaddresser whose role is to produce the verbal message (or utterance) . // is anaddressee whose role is to comprehend or understand what is said and intended by s.

(B) They must be cooperative to make their communication successful. S is expected toprovide effectively the sufficient contextual knowledge for h to be able tocomprehend .v's intended meaning of the utterance. // is expected to do the task ofinferring or calculating appropriately the intended meaning by taking account of theshared and contextual knowledge between s and h.

(C) In the course of communication, there functions the shared knowledge between s andh. One type of shared knowledge is conventional. It is the socio-cultural knowledgewhich is shared among the membership of the community. We can acknowledge it inthe code model of the lexicon. The other type of shared knowledge is contextual. It isthe interpersonal knowledge available in the communicative process of productionand comprehension.

(D) For s, communication is an act of verbal production. He/She is expected to make agood contribution to the cooperative task of communication by providing effectivelythe sufficient information for h to be able to do the task of comprehension. For h

Page 7: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Morio Nishikawa

communication is an act of comprehension. He/She is expected to make a goodcontribution to the cooperative task of communication by retrieving the shared andconventional knowledge from the lexicon or by inferring what a given utterancemeans by way of taking account of the shared and contexual knowledge.

These conditions apply to the interpretation of NNCs as much as to other aspects of contextualizedinterpretation.

[ 4 ] Semantics and Pragmatics

A distinction is often made between two main linguistic components where meaning is dealtwith: semantics and pragmatics. They are often mixed and confused, which causes the study of'meaning' to be difficult (cf. Levinson 1983 I 17-35. Leech 1983 : 5-10).

The basic difference between the two is that semantics, on the one hand, is based upon thestructural and conventional aspects of meaning out of context. Pragmatics, on the other hand, isbased upon the functional and contextual aspects of meaning in terms of the study of meaning incontext. It also specifies uniquely one interpretation from among possible or candidateinterpretations with the help of contextual knowledge. In the course of communicative process ofcomprehension, h undertakes the problem-solving strategy of verbal communication. It is as follows :

(11) y= f (x)

That is, if a contextual knowledge V is given as a communicative value to an utterance such as

innovative NNC, then its contextual or intended meaning 'y' is determined. Therefore, h is expectedto find out a piece of contextual knowledge V in order to gain a sucess in comprehension. Considerthe following :

(12) worm queen

This is an innovative NNC obtained from the experiment. I will discuss this NNC in order to makethe problem explicit. Semantics provides us with the general knowledge of this NNC in such a way as" (a) queen which has something to do with worm". But what kind of queen is it? What kind ofinterpretation can we provide to this NNC ? It may or may not be "the most important worm in thereproductive cycle (cf. bees)." If the appropriate context is given now, h will be able to understandwhat it actually means, and we can see that the above ad hoc interpretation is not applicable.

(13) He threaded the bait onto the wonn queen and cast his line into the water.

Here works the contextual knowledge which enables us to infer what this NNC actually means. Itmeans, according to one informant, "a fishing hook—expensive—that is very good for holding wormswhen fishing." The contextual knowledge V in this example specifies the intended meaning of thisNNC. It could be, for example, "threaded the bait onto ..." or "cast his line into the water" or thecombination of the contextual knowledges. It has no semantic implication of "the most importantworm (which mate many times a year) ." Now we can hypothesize the function of contextualknowledge in the communicative process of comprehension as follows :

Page 8: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds

(14) The intended meaning 'y' of an innovative NNC will be specified if and only ifthe contextual knowledge V will be given appropriately.

As a sum, the study of meaning in semantics is static, and it may characterized by the one-niany relationship between an NNC and its interpretation. It is also structural in so far as it isconcerned with the context-free interpretation of a given utterance. The study of meaning inpragmatics, however, is dynamic and active because it is concerned with the actual use of language.It is also functional and contextual in so far as it is concerned with a certain kind of problem-solvingprocess of comprehension. Here works the assumption that the meaning of an innovative NNC shiftsfrom context to context.

Then, what is current situation regarding studies of compounds in English ? There are anumber of important concepts or ideas about the study of compounds in English. I would like tomove on to the next stage of discussing some of the recent studies of compounds.

[I] Recent Studies

[ 1 ] Preliminaries

In this chapter, we will make a broad survey of the research of compounding before examiningin more detail some studies which are of particular relevance to this paper.

A thorough study of 'nominal compounds' was first provided by Lees (1960) under the stronginfluence of early work on transformational grammar. Lees, dismissing the traditional descriptiveapproaches, proposed explicitly a transformational analysis of compounds. He claimed, for example, thatthe compound car thief is to be derived from the underlying sentence: 'The thief steals the car.' Theparaphrased sentence is hypothesized in his theory to be the deep structure of a given nominalcompound.

After a decade. Chomsky (1970) postulated a theoretical difference between the two positionsof 'transformationalists', who are liable to explain nominal compounds in terms of transformationaloperations like deletion or nominalizations, and 'lexicalists' who do not acknowledge any deepstructure approach in nominal compounds, and do not apply any transformational operation to them.He assigned theoretical superiority to the latter because transformations cannot explain adequatelythe difference between gerundive nominals like John' s refusing the offer and derived nominals likeJohn' s refusal of the offer.

Levi (1978) , as a generative scmanticist, provided a detailed study of the transformationalanalysis of what she called 'complex nominals (or CNs)' Along with Lees (1960, 1970). this was ahistoric experiment to discover how far a transformationalist can go in using only transformationaloperations to explain the derivation, form, and meaning of NNCs. She attempted to explore thesyntactic and semantic properties of complex nominals. She also attempted to construct atransformational model of the productive aspects of complex nominal formation. In order to carry itout. she proposed that all complex nominals must be derived from two syntactic processes(transformations) of predicate nominalization and predicate deletion. Then she postulated nine kinds ofwhat she calls Recoverably Deletcble Predicates (or RDPs) (e. g. CAUSE. HAVE, MAKE. USE, BE, IN.FOR, FROM, and ABOUT). Like that of Lees (1960), 'this approach is based upon the hypothesis that agiven 'complex nominal' is equivalent in meaning to its paraphrased underlying structure.

Jackendoff (1975), as a lexicalist, proposed a theory of the lexicon consistent with Chomsky' s'Remarks on Nominalizations' (1970). Jackendoff (1975 '. 655) enters 'phonological rules', 'morphological

Page 9: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Morio Nishikawa

rules', and 'semantic rules' of the lexion in the list of the redundancy rules in order to provide anadequate description of 'compound nouns'.

Aside from these generative linguists, compounding has been dealt with in some detail by anumber of notable scholars. Among them are Jespersen (1942), Marchand (1969), and Quirk et al.(1985). Jespersen gave a modest definition of compound : "A compound may perhaps be provisionallydefined as a combination of two or more words so as to function as one word, as a unit" (MEG :Chapter \l\, 1-1). He already predicted the necessity of taking account of contextual knowledge inorder to understand the meaning of a given compound.

Merchand (1969) made explicit the grammatical relation between the first and the secondconstituent words in the syntagmatic combination of compound. The definition of compound providedby Quirk et al. (1985 '. 1967, Appendix 1) was clear and fundamental. They say: "a compound is alexical unit consisting of more than one base and functioning both grammatically and semantically asa single word. In principle, any number or bases may be involved, but in English, except for arelatively minor class of items (normally abbreviated), compounds usually comprise two bases only,however internally complex each may be." In our understanding, however, we have to think of theinterrelation between the 'two bases' (i. e. lexical units) by taking account of a third element of realcontextual knowledge (see section [3] in chapter [I]).

Clark (1978) proposed a new pragmatically-oriented approach based upon the problem-solvingprocess of comprehension, which he calls 'intentional view' of meaning. He directs his attention to theprocess by which heares arrive at the interpretation which speakers intend them to understand. Thisview requires hearers to draw inferences that go well beyond the literal and direct meaning of thatexpression. He envisages an actual communicative situation between a- and /;, in which h is expectedto determine what s means by what s says.

Many things have been witten about compounds. Some are adequate but others do not takeaccount of the observation of real aspects of compounds. The fundamental question does not vary. Itis: "What kind of approach is more adequate to explain actual conpoimds ? " Now we have to discussand elaborate on some recent studies of compounding relevant to the present investigation.

[2] Marchand (1969)

Marchand (1969) begins his discussion with the definition of the word. The word for him is"the smallest independent, invisible, and meaningful unit of speech, susceptible of transposition insentences" (1.1.1). This phrase immediately reminds us of the meaningful unit of speech which hasbeen called the 'morpheme'. He claims:

A word, like any other morpheme, is a two facet sign which must be based on thesignificant/significate (signifiant/signifie) relationship posited by Saussurc. Synonymsof the terms which will also be used in this book are expression (signifiant) andcontent (signifie). (1.1.2)

In this passage, he assures himself that a word is a sign which designates something: there is aninterrelation between a sign (or a word) and what it designates. A word threfore is a designatorwhich denotes its designatum (henceforth D) . As for a compound, it is a word which can becharacterized by this designation process.

He defines a compound as follows :

Page 10: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds

( i ) The coining of new words proceeds by way of combining linguistic elements on thebasis of determinant/determinatum relationship called a syntagma. When two or morewords are combined into a morphological unit on the basis just stated, we speak of acompound. (2.1.1)

(ii) Combinations of all the kinds illustrated by steam boat and colour blind which containfree morphemes both for the determinant and the determinatum will be termedcompounds. (2.1.2)

In these passages, the essence of compounding is the determinant/determinatum distinction betweenthe two constituent words. The first word has the role of determinant, which subcategorizes thesecond word. And the second word has the role of determinatum, which functions as the head of the

whole compound. The distinction of determinant/determinatum is different from that ofdesignator/designatum in the sense that the former applies to the syntagmatic relation of the twoconstituents, but the latter applies to the relationship between a compound and what it designates.

According to Marchand, the fundamental idea of compounding lies in the function ofidentification performed by the second word and that of differentiation performed by the first word.He summarizes :

The principle of combining two words arises from the natural human tendency to see athing identical with another one already existing and at the same time different from it. Ifwe take steam boat, for instance, identity is expressed by the basis boat, the difference bythe word steam. (2.1.2)

He describes here the contrastive 'human tendency' to identify one thing by one word anddifferentiate it by another word. Thus, according to him. the second word (determinatum) in acompound plays the role of 'identification', and the first word (determinant) plays the role of'differentiation'. In compound formation, according to us, the process of 'identification' can beequated with that of categorization, and that of 'differentiation' with that of subcategorization.

Marchand introduces the concepts of expansion into the study of compounding. He describes itas follows :

Combinations of the types steam boat, colour blind, and rewrite which are mere morphologicalexpansions of the words boat, blind, ivrite respectively will be termed EXPANSIONS. Anexpansion will then be defined as a combination AB in which B is a free morpheme (word)and which is analyzable on the basis of the formula AB = B. (2.1.2)

Boathouse is an expansion of house, colour blind an expansion of blind, cryboy an expansion ofboy. (2.1.4.1)

However, I would prefer to say that the process of forming steam boat from boat is semantically asubcategorization of the determinatum boat. This is also explicated by Clark, E. (1985) in thefollowing description.

For adults, just as the well-established airman, horseman, and lawman pick out subcategoriesof the category of man, so the novel moon-flag and star-flag pick out subcategories of flag.

Page 11: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

10 Morio Nishikawa

We will accept here the term "subcategorization' in so far as our attention is directed to semanticand pragmatic aspects of NNCs.

[3] Warren (1978) and Leonard (1980)

Warren (1978) begins her "discussion with the following neat example:

(15) A snowflake is a flake consisting of snow, but a butter knife is not a knife made ofbutter.

She suggests in this passage that the first noun 'b' and the second noun 'a' are performing, as itwere, a division of labour: the second noun takes on the denotative function in a compound and thefirst noun takes on the subcategorizing function in a compound. Both nouns are integrated into athird noun of noun-noun compound. More importantly, she suggests that there are various kinds ofsemantic relation (or interpretation) between the two nouns. For example, it may be 'place' as instation master, or 'time' as in summer vacation, or 'material' as in marble statue and so forth.

Warren (1978 '. 49) makes a distinction between the semantic structure and informationstructure of NNCs. In semantic structure, she proposes that a topic/comment distinction coincideswith the determinant/determinatum distinction. According to her, the topic is that which we aretalking about and the comment is that which is said about the topic; the whole combination is apredication as in "pocket (comment) knife (topic) ". In information structure, she adopts therhematic/thematic distinction of Firbas (1964) and other Prague school linguists. She acceptsFirbas' s concept of the degree of communicative dynamism (CD) of rhematic or thematic elementsand applies it to NNCs. According to this theory, the thematic element is that which carries a lowdegree of CD. and the rhematic element is that which carries a high degree of CD. And she assignsdifferent degree of CD to compound elements. She claims : "The 'information structure' of acompound is that the rhematic element tends to precede the thematic element." However, we cannotdeny that there lies the problem of whether the concept of the degree of CD is appropriate or not.How can we measure 'the degree of CD' of the determinatum and that of the determinant ?

Warren' s main concern is with a corpus-based analysis of NNCs in terms of the semanticrelation between the constituent nouns. She sorts out the variety of the semantic relations of 'b' to 'a'based upon the prepositional paraphrases of a given NNC. Exampies of semantic relations with theappropriate prepositions are as follows :

( i ) A is something which fully constitutes B, or vice versa :

Source-Result (of) metal sheet

Result-Source (in) sheet metal

Copula Compounds girlfriend

(ii ) A is something of which B is a part or a feature or vice versa :

Whole-Part (of) egg shell

Part-Whole (with) arm chair

(iii) A is the location or origin of B in time or space :

Place-Object (in. at. on) coast road

Page 12: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds

Time-Object (in, at, on) Sunday schoolOrigin-Object (from) seafood

(iv) A indicates the purpose of B :

Purpose (for) f>ie tin

( v ) A indicates the activity or interest with which B is habitually concerned :

Activity-Object cow boy

(vi) A indicates something that B resembles :

Comparant-Compared (like) cherry bomb

These six major types of semantic relations are compact and suggestive, but they are still notsufficient. As she says, "the above is a condensed abstract of the semantic relations." And it is truethat there are many kinds of semantic relations in the combination of V to 'a'. What is important isthat one combination (NNC) is able to have a number of possible interpretations which cannot benecessarily categorized into one of the listed 'types' of semantic interpretation. Moreover, themeaning of one compound potentially shifts from context to context. This is especially the case ininnovative NNCs. In one of our informant-elicited examples, goose hammer means "hammar in amusical instrument (e. g. piano) covered in soft down-like material to produce a soft delicate note"or "a sacred instrument of the American Indians lavishly decorated with feathers used for strikingthe war drum" in addition to the interpretation : "hammar used on the farm to kill geese".

What is particularly useful in Warren' s research is that she gives a quantitative analysis ofthe frequency of semantic relations in English nominal compounds. All the data were taken from "theStandard Corpus of Present-Day American English" assembled at Brown University during 1963 and1964. She used 180 texts ; thus out of approximately 360,000 words, she found 4556 different NNCs.She excluded NNCs in which one noun is devcrbal, because the main aim of her investigation was tofind out what semantic relations between two nouns were at work. The statistical result is as follows :

Semantic Classes Number of Types

Source-Result (student group) 619Copula (girlfriend) 203Resemblance (clubfoot) 72Whole-Part (spoon handle) 921Part-Whole (arm chair) 341Size-Whole (3 day affair) 173Goal-OBJ (moon rocket) 14Place-OBJ (sea port) 483Time-OBJ (Sunday paper) 141Origin-OBJ (hay fever) 333Purpose (ball bat) 537Activity-Actor (cowboy) 157

3.994

Page 13: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

12 Morio Nishikawa

An additional 562 NNCs belonging to non-semantic classes are reported in her research: among themare dvandva compounds (e. g. aural-oral, Sino-Italian, etc.), idiosyncratic compounds (e. g. egg head,hunchback, etc.) , and unanalyzed compounds. This quantitative analysis tells us that the semanticrelations between elements of NNCs in the corpus are varied, and that those of Whole-Part, Source-Result, Purpose, Place-Object are used more frequently than others. In contrast, those of Goal-Objand Resemblance are very infrequently used.

However, this is neither a conclusion nor a goal. Her conclusion must be a departure in thestudy of NNCs. That is to say, it may provide us with information about the frequency and variety ofsemantic relations in NNCs, but it cannot suggest to us how to understand and specify the intendedmeaning of an NNC in a given context. Furthermore, we cannot restrict ourselves to the goal oflisting or classifying the semantic relations between the constituents of NNCs. We must go further toexplain the meanings which are realized by use of NNCs in context.

Leonard (1980) provides a typological study of the semantics of noun sequence (NS) . Herstudy is also based upon a corpus analysis of NNCs and involved in using a computer. Irrespectiveof its large volume, what is told is rather simple and clear. Leonard's research gives a computationalanalysis of noun sequences based upon the "algorithm for the interpretation of English nounsequences". She divides noun sequences into several calculable semantic types, including (A)Sentence Type (hire car, bank manager, acrobatics team), (B) Locative Sentence Type (cove dweller,town property, (eatable, night gown), (C) Locative Type (carpet rod, mill girl, house wife), (D) AnnexType (hill side, mountain top, chapel window) , (E) Equative Type (boy friend, oak tree, sign board),(F) Material Type (stone lion, brick cottage), (G) Additive Type (Lamplon-Lufford report), and (H)Reduplicative Type (plop-plop).

A noun sequence (or NS) for her is an expression consisting of two or more nouns insequence, fulfilling the function of a noun without an overt syntactic link. In her research, 2,000noun sequences are used. The examples come from approximately 300,000 words of English fictionwritten between 1700 and the present age. And her semantic classification is performed by applyingthe interpretative algorithm to a large range of noun sequences.

Leonard's computer program for interpreting NSs achieved approximately 50% success inproviding reasonably acceptable paraphrases for NS considered out of context. This was aconsiderable achievement, but still suggests that there is much to be explained by the role of contextin the interpretation of compounds.

[4] Chomsky (1970) and Jackendoff (1975)

Chomsky (1970) raises some fundamental critiques of the transformational approach ofderived nominals. He argues that there is a crucial difference between derived nominals andgerundive nominals in a number of fundamental ways.

First, according to Chomsky (1970 ! 4), gerundive nominals, on the one hand, "can be formedfairly freely from propositions of subject-predicate form and the relation of meaning between thenominal and the preposition is quite regular." That is to say, the meaning of gerundive nominals hasa tendency to be straightforward and its productivity is much less restricted. On the other hand,derived nominals are much more restricted in their productivity, and are "quite varied andidiosyncratic."

I would like to discuss here the assumption that derived nominals cannot be explainedadequately by transformations because their meanings are "varied and idiosyncratic". For example,the noun government which is derived from the verb govern does not necessarily have the same

Page 14: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds 13

meaning as the gerundive nominal of governing. Government has an idiosyncratic meaning: "theexecutive policy-making body of a political unit, community, etc ; ministry or administration: thesystem or form by which a community, etc., is ruled" (CED). In this context, we come into conflictwith the requirement that transformations do not change the meaning of deep structure. That is,government has an idiosyncratic meaning in contrast to governing, although both nominals originatefrom the same verb govern, and the original meaning of govern is idiosyncratically changed into thenew meaning of government. Booij (1977 '. 11) calls it 'semantic irregularity' of derived nominals. InNNCs, the same thing happens. Words like seaside amusement and life insurance are far fromequivalent to seaside amusing and life insuring. For example, amusement has its own implication of"distraction or diversion of the attention from the point at issue ; beguiling, deception : thepleasurable occupation of the attention, or diversion of the mind" (OED). Amusement and insuranceare well-established derived nouns, but amusing and insuring are inflected gerundive nouns whichare far from well-established.

Second, the crucial difference between gerundive nouns (the underlined words in (17)) andderived nominals (the underlined words in (18)) are explicated by Chomsky (1970 ! 4-10) in termsof a restriction on the formation (or productivity) of derived nominals. Consider the following :

(16) ( i ) John is easy to please.(ii) John is certain to win the prize.(iii) John amused the children with his stories.

What Chomsky (1970) claims is that the structures in (16) can be transformed to the gerundivenominals of (17) but not to the derived nominals of (18). That is. the derived nominals in (18) areill-formed, although the gerundive nominals in (17) are well-formed :

(17) ( i ) John's being easy to please(ii) John's being certain to win the prize(iii) John's amusing the children with his stories

(18) ( i ) *John' s easiness to please(ii) *John' s certainty to win the prize(iii) *John' s amusement of the children with his stories

Here conies a restriction of the productivity of derived nominals.Third, derived nominals can become part of a noun phrase, whereas gerundive nominals

cannot. In other words, only the derived nominals can have the internal structure of noun phrases.Consider the examples below :

(19) ( i ) the proof of the theorem(ii) John' s unmotivated criticism of the book

(20) ( i ) *the proving of the theorem(ii) *John' s unmotivated criticizing the book

Fourth, derived nominals can be pluralized and occur with the full range of determiners like :

Page 15: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

14 Morio Nishikawa

(21) ( i ) John' s three proofs of the theorem(ii) several of John' s proofs of the theorem

Fifth, derived nominals can appear freely in the full range of noun phrase structures like :

(22) ( i ) John gave Bill advice.

(ii) Advice was given (to) Bill,(iii) Bill was given advice.

In fact, none of these criteria apply to the case of gerundive nominals. Furthermore, thetransformational approach to derived nominals cannot handle the case of a derived nominal withoutits base like agression (cf. * agress). And finally, affixes can be attached to derived nominals as ingovernment and governmental or confession and confessionism or confessionist. This cannot happen ingerundive nominals. Thus Chomsky (1968) raised several important different points betweenderived nominals and gerundive nominals. As for evidence for the lexicalist position. Jackendoff(1975 ! 640) claims :

The fundamental linguistic generalization that must be captured by any analysis of Englishis that words like decision are related to words like decide in their morphology, semantics,and syntactic patterning. For Lees 1963, it seemed very logical to express this relationshipby assuming that only the verb decide appears in the lexicon, and by creating the noundecision as part of a transformational process which derives the NP John's decision to go

lrom the S John decided to go. However, for reasons detailed in Chomskey 1970, thisapproach cannot be carried out consistently without expanding the descriptive power oftransformations to the point where their explanatory power is virtually nil.

Since the lexicalist hypothesis denies a transformational relationship between decide and decision,

their relation must be explained by a set of rules within the lexical component. Therefore, Jackendoff,as a lexicalist, follows the assumption that nominalizations cannot be dealt with transformationallyand that they can be treated independently in the lexicon. For Jackendoff (1978 '. 639) the essentialproblem is "to develop a notion of lexical redundancy rules which permits an adequate description ofthe partial relations and idiosyncrasy characteristic of the lexicon". Then he proposes to construct alexicalist theory of redundancy rules for word formation with respect to a list of phonological,syntactic, and semantic rules of existing words. Jackendoff (1975 '. 655) provides the followingexamples :

(23) a . garbage man, iceman, milkman, breadbasket, oil drum

b. snowman, gingerbread man, bread crumb, sand castle

c . bulldog, kettledrum, sandstone, tissue paper

And he claims : "Although the meanings of each compound is formed from the meanings of the twoconstituent nouns, the way in which the meaning is formed differs from line to line." The variety andunpredictability of the meaning of compound is taken into consideration in this passage. And hislexicalist theory of compounding comes into existence as follows (p. 655) :

Page 16: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds

In the lexicalist theory, we can dispense with exception features in the description ofcompounds. We simply give each actually occurring compound a fully specified lexicalentry, and in the list of redundancy rules we enter morphological rule 22 ((24) in thissection) and semantic rules 23 a, b. c ((25) a. b, c in this section) describing the data of 21a, b, c ((23) a. b, c in this section) respectively.

15

According to him. "the redundancy rules thus define the set of possible compounds of English, andthe lexicon lists the actually occurring compounds." And he provides the following schematicsamples of compounds (pp. 655-656) :

(24)

(25) a

["/ Ul M/l

r+N

|_ZtiTHAT CARRIES W]

\_Z MADE OF W]

\_Z LIKE AWj

+ N

Z

+ N

+ N.hi'

L + N_

rai

We now face the problem of semantic rules of each description in the data of ((23) a. b. c): i. e. theproblem of the arbitrariness of semantic interpretation. We cannot specify only one interpretation ofa given NNC out of context. In other words, there is no fixed rule of semantic interpretation out ofcontext. A garbage man may be "a man who collects garbage or rubbish", why, however, is it notpossible to assume that a garbage man is "a man who is considered to be rubbish or garbage" or "aman who has the word 'garbage' associated in some way with him (maybe he speaks 'garbage' veryoften) ? This is a frequent cause of misunderstanding between people whose backgroundassumptions are personally or collectively very different from each other. I remember the commenton mismatching terms or referents, provided by Levi (1978 ! 11) as follows :

For example, a foreigner asked to meet some American friends "at the ball park" couldunderstandably end up—alone—at a football field or outdoor handball court, unless he hadspecifically learned that Americans use ball park to refer to playing areas for base ball only.Similarly, a child asked to pick up an eggplant at the store might just as easily (andreasonably) come home with an onion or a mango if she had not yet learned the actualreferent of this term.

In sum. an amalgamated list of grammatical information of a normal compound in terms of the

Page 17: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

16 Morio Nishikawa

redundancy rules, as provided by the lexicalist hypothesis, is not enough, in so far as there can beno fixed semantic rule of interpretation in compounding. Rather, an appropriate interpretation of agiven compound is likely to depend on its actual use in context.

[5] Levi (1978)

Levi attempts to explore the syntactic and semantic properties of complex nominals (for shortCNs) in English, and she also tries to incorporate into a grammar of English a model of productiveaspects of complex nominal formation. She proposes the concept of CNs which comprises nominalcompounds like the examples in (26 a), nominalizations like those in (26 b), and noun phrases withnonpredicating adjectives like those in (26 c) :

(26) a . apple cake, deficiency disease, time bomb autumn rams, doghouse,nicotine fit, windmill

b . Markovian solution, film producer, American attack, city planner.

Presidential refusal, dream analysis

c . electric clock, musical clock, electric shock, musical criticism,

electric engineering, musical interlude, electrical conductor, musical

comedy

Her basic approach is first and foremost transformational. Levi (1978 ! 50) claims :

Complex nominals are all derived from an underlying NP structure containing a headnoun and a full S in either a relative clause or NP complement construction ; on thesurface, however, the complex nominal is dominated by a node label of N.

This passage immediately reminds us of the transformational approach given by Lees (1960,1970).First, according to Levi, an NP (containing a head noun and a full S) in underlying structure

is changed into an N on the structure by way of a certain number of derivational transformations.She simply confuses here the two different levels of word and phrase. Due to the meaning-preservingprinciple of transformation, a 'N (oun) -P (hrase) ' in underlying structure and its corresponding 'N(oun) 'of compound must have the same meaning. However, this is not necessarily fulfilled between acompound (N) and its underlying structure (NP). For example, starvation bell is not necessarily "aswathe of land affected by starvation", but it may be "an assembly line in a factory where workersare badly paid."

Second, she proposes nine kinds of Recoverably Deletable Predicates (RDPs) which aresupposed to enable us to classify all kinds of CNs. She claims :

Complex nominals are all derived by just one of two syntactic processes: the deletionor the nominalization of the predicate in the underlying sentence.

For complex nominals derived by predicate deletion, a small set of RDPs can bespecified such that only its members, and no other predicates, may be deleted in theformation of CNs; the members of this set arc CAUSE. HAVE. MAKE. BE. USE, FOR,

IN. ABOUT, and FROM.

Page 18: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds 17

This idea may be convenient for us to think of the variety of NNCs. but there must be found acritical problem : the problem of arbitrariness or unspecifiability of meaning in NNCs out of context.For example, Levi (1978 '. 76-77) classifies gun boat into the predicate 'HAVE'. However, it is notirrelevant for us to think that gun boat may be "boat which MAKEs guns", or "boat FOR guns". Shealso classifies apple seed as an example of the predicate 'FROM'. However, is it impossible to considerthat apple seed is "seed FOR apples", or "seed which CAUSEs apples", or "seed IN apples" ? Weclaim that the meaning of a given NNC cannot be classified into one of her nine groups. In otherwords, it cannot be limited within a conventionally fixed concept of one of those 'predicates'. Therecan always be a potential different interpretation of a given NNC which shifts from context tocontext. We can only say at most that the more it is well-established, the less will be the number ofits possible interpretations. Conversely, the less it is established, the more will be the number ofpossible interpretations. Levi has a tendency to account for only well-established NNCs.Consequently, she did not pay much attention to the variety of possible interpretations of a givenNNC.

Unlike Levi (1978 ! 76-77) . we would like to propose evidence that one given NNC hasmultiple 'predicates' which are exemplified by the above mentioned cases of gun boat, apple seed, andan additional one, song bird :

(27) a . f-un boat : MAKE ("boat which makes guns")HAVE ("boat which has guns")FOR ("boat which is for guns")

b . apple seed : CAUSE ("seed causes apples")FOR ("seed for apples")FROM ("seed from apples")IN ("seed in apples")

<-' . song bird : MAKE ("bird which makes songs")CAUSE ("bird which causes songs")FOR ("bird which is for songs")

Thus we must acknowledge that there is the variety of possible interpretations of a given NNC (ornominal complex). And we will direct our attention to the assumption that not a limited number of'predicates' but an actual usage of NNC must come first.

So far we have not yet discussed whether the Levi' s term CN (complex nominal) isappropriate or not. But now we would like to criticize her usage of this term. Levi (1978 '. 147)claims :

This [Morphological Adjectivalization] is the transformation that is responsible forproducing nominal adjectives from prenominal nouns in CNs. as in marginal note frommargin note, thermal stress from heat stress, and musical clock from music clock. The changeeffected by this transformation is a highly superficial one; as noted earlier, the semanticand syntactic characteristics of CNs are the same, regardless of whether the prenominalmodifier is morphologically a noun or an adjective.

Our data suggest that this is not always true. First, the change effected by 'this transformation' is

Page 19: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

18 Morio Nishikawa

not 'highly superficial one' but a very fundamental one, because it is a very important linguisticprocedure which changes a word into a phrasal construction. Second, the semantic and syntacticcharacteristics of CNs are not the same. They are influenced to a great extent by whether theprenominal modifier is morphologically a noun or an adjective.

In the course of our experiment, we came across evidence of the difference, provided by aninformant, between the nominal compound margin note and the denominal adjective constructionmarginal note as follows :

(28) a . margin note: "note that is written in the margin of paper, etc."

b . marginal note: "note that is secondary or of slightly less importance thanmain body of subject"

We also have different kinds of data from journalistic writing as follows :

(29) Spending three months in and around the tropical rain forests of Borneo may beeveryone's idea of a perfect holiday.

St Helens Reporter 27/12 1985

(30) Cooper's collaborator, phychology lecturer. Dr. Marilyn Davidson, is one of thiscountry's leading researchers on women in management.

The Sunday Times 9/2 1986

In these examples, the underlined NNC, rain forests or psychology lecturer cannot be replaced by rainy

forests or psychological lecturer. Rain forests has the meaning: "dense forest found in tropical areas ofheavy rainfall" (CED), which is not the case for rainy forests. Psychology lecturer, on the one hand, is"lecturer in psychology", and it normally implies a combination of subject and occupation.Psychological lecturer, on the other hand, might suggest "a lecturer who employs psychological tacticsin a lecture to make his point." In fact, such examples are not rare. There are a number of pairs of'prenominal noun' and its related 'nominal adjective" whose meanings are different from each other.Consider the following :

(31) a . ( i ) health business( ii ) healthy business

b . ( i ) form letter(ii ) formal letter

In neither of these examples is ( i ) synonymous with (ii ) . Health business is. according to aninfant, "business or financial enterprise connected in some way with health; might have food or anysort of sports connection. Emphasis is very much on business: e. g. tourism, the leisure industry,etc." It usually appears with the definite article 'the'. But healthy business may be a "comment onsome activity people have participated in, e. g. running, jogging, walking, etc." or just astraightforward description of "the financial well-being of a business". It usually appears with theindefinite article 'a (or an) '. Form letter is "a single copy of a letter that has been mechanicallyreproduced in large numbers of circulation" (CED), but fonn aI letter may have something to do with'ritual style of writing' used in a certain kind of writing situation.

Page 20: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds 19

It may be true that, as Levi (1978 '. 38) says, the prenominal noun atom bomb and the nominaladjective atomic bomb are synonymous, but this cannot be generalized at all. Levi (1978 '. 38) furtherclaims that language skills and linguistic skills are 'totally synonymous, despite the fact that theprenominal modifiers are adjective in one case and noun in the other.' However, our experimentsuggests that this is not the case. Language skills means, skills of using language (e. g. speaking,listening, writing, or reading) '. linguistic skills, however, is likely to refer to more scientific andanalytical skills, e. g. the 'ability to scan texts, or to identify linguistic features. It is likely to bepracticed by a trained professional linguist.' That is to say. language in language skills and linguistic

in linguistic skills are not "totally synonymous". Our claim is that 'prenominal nouns' and theircompounding 'nominal adjectives' are different things, and therefore we cannot deal with these twodifferent constructions under the same and one term of complex nominals (CNs).

Ironically, the limitation of her syntactic-semantic analysis of complex nominals aredemonstrated by Levi herself. She excludes a predicate such as LIKE (or RESEMBLE) from herfixed set of Recoverably Deletable Predicates. This is simply because CNs with the interpretation ofLIKE have 'extended' readings which cannot be dealt with in her structural (or syntactic-semantic)model. Her premises do not accept that an NNC may have a metaphorically extended meaning. Butthis is irrelevant. We rather acknowledge that nominal compound can potentially have numerousextended readings in different contexts.

Levi regards treatment of metaphorically extended readings of CNs as irrelevant in hersyntactic-semantic model because it is beyond her theoretical model. And her approach is essentiallyconcerned with the literal or "basic reading of CNs" (p. 108). This is the reason why Levi excludes ametaphorically extended reading of CNs with the predicate of 'LIKE'. Levi should have given asyntactic-semantic explanation of CN similes within her theoretical framework. But she could not doit. She only suggested how to account for CN similes in the area of pragmatics. Thus we are to takepragmatic aspects of NNCs into consideration more positively. In our understanding, the extendedreadings of NNCs are essential, because their meanings shift from context to context in actual use. Ametaphorically extended reading is no exception. In contrast, Levi (1978 '. 114) claims :

readings of particular lexical items or syntactic forms which extend beyond their literal (orcustomary) referents are so common as to make a structural analysis of such extensions(such as the analysis just discussed, based on a LIKE predicate) hopelessly inadequate tothe task. Rather it seems likely that a small set of pragmatic principles will provide the bestexplanation for the fact that we regularly use and interpret a great variety of linguisticforms as instances of semantic extension.

This suggestion of 'pragmatic principles' can be applied not only to metaphorically extended readingsof CNs but also to other kinds of readings of CNs which are classified in her theory in terms of oneof the fixed RDPs. As we already pointed out in this section. NNCs such as gun boat and apple seed

are capable of having multiple readings (or interpretations) . We cannot specify the intendedmeaning of these NNCs without knowledge of the context where they are actually used. What Levimentions about metaphorically extended readings of CNs can also be applied to other kinds ofnominal compounds which may have multiple readings in different contexts.

The transformational approach of NNCs provided by Levi (1978) thus contains severalunacceptable premises. In the course of discussing her approach, we acknowledge that there is alimitation of the context-free syntactic-semantic study of NNCs and that pragmatic approach must be

Page 21: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

20 Morio Nishikawa

inevitably taken into consideration in the study of multiple readings of NNCs.

[H] NNCs as Contextuals

There are roughly two kinds of NNCs. One is the conventional NNC whose meaning we canretrieve from the stored knowledge of the lexicon, and the other is the innovative NNC whoseintended meaning we arc expected to infer by calculating from the context where it is actually used.This chapter is devoted to a discussion of the schematization of conventional NNCs in terms of theshared knowledge of the lexicon, and of innovative NNCs as contextuals in terms of the sharedknowledge of the context.

[ 1 ] Schematization of NNCs

A conventional NNC like steam boat has a head 'a' (boat) which is subcategorized by 'b'(steam) . It denotes a designatum (D) which is grammatically categorized by 'a'. Thus it has theschematic description of the following :

(32) (D (categorized by a(subcategorized by b)))

Hence in steam boat :

(33) (STEAM BOAT (categorized by boat(subategorized by steam)))

where 'STEAM BOAT' is a designatum (D) of 'a' (boat) which is subcategorized by 'b' (steam).Furthermore, there must be a certain semantic interpretation between 'a' and 'b' in this NNC. Thisinterpretation can be considered to be a semantic relation of 'b' to 'a' which we will call I R { .Insteam boat, \ R \ turns out conventionally to be j powered by ( , because 'D' (steam boat) is a kind of'a' (boat) which is powered by 'b' (steam). Thus (32) and (33) must be elaborated as follows :

(34) (D (categorized by a((subcategorized by b)

in |R| )))

Hence in steam boat :

(35) (STEAM BOAT (categorized by boat((subcategorized by steam)

in 1 powered by ( )))

Thus in conventional NNCs, we are able to provide an appropriate interpretation of conventionalNNCs out of context by making the best use of the knowledge from the lexicon. That is to say, thesemantic relation of 'b' to 'a' (i. e. 1R } ) is specified conventionally. This kind of interpretation isapplicable only to well-established and conventional NNCs.

Page 22: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds 21

There is another kind of NNC which cannot be retrieved from the stored knowledge of thelexicon. Consider the following :

(36) snack war home pride water generationbaby bug bomb girl tennis can

night bird foam blanket tiger curtain

bottle man ironfist smoke hood

These are the examplcse of real innovative NNCs. They cannot be found in an English dictionary.They are neither well established nor institutionalized in any way. They are not stored in ourmemory and far from conventional but they make good sense. Moreover they unfailingly make a goodcontribution to the verbal communication between s and h.

As in (32) or (33), we can provide a schematic description of this kind of NNC such as bombgirl :

(37) (BOMB GIRL (categorized by girl(subcategorized by bomb)))

However, what is the semantic relation ( | R ( ) of V (bomb) to 'a' (girl) ? Can we specify it out ofcontext ? As a matter of fact, there are many possible or candidate interpretations of this NNC. Wecan only claim in this particular case that this girl has something to do with bomb, or this girl issubcategorized by bomb in terms of a certain semantic relation | R ( as follows :

(38) (BOMB GIRL (categorized by girl((subcategorized by bomb)

in \R\ )))

What is significant here is that I R } is ambiguous. In other words, there can be many j R ( 's if anappropriate context is not provided to this NNC. Briefly, for example, this bomb girl may be "a girlwho makes bombs", or "a girl who plants bombs", or "a girl who diffuses bombs", or "a figure of agirl marked on the side of a bomb as a means of identification", and so forth. In so far as there are a

number of j R } 's in innovative NNCs, there should be a number of D's (designata) which arecategorized by the head constituent 'a' which is subcategorized by 'b'. We now elaborate theambiguity of innovative NNCs out of context as follows :

(39)

[ 2 ] Inferring what is Meant by NNCs

Here is an NNC :

Page 23: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

22 Morio Nishikawa

(40) Two other climbers in the same party were airlifted to safety by a helicopterfrom a rock chimney in which they had been stranded overnight.

The Times 6/1 1986

What is a rock chimney out of context ? It may be a kind of chimney which has something to do withrock. It is in fact difficult to specify its intended meaning because it is contextual ; in other words, itis actualized by use in context. As in (35), we can only make a scheme as follows ;

(41) (ROCK CHIMNEY (categorized by chimney((subcategorized by rock)

in jK| )))

What is important is that 'R' cannot be specified without any knowledge from the context. We canonly provide possible or candidate interpretations to the given NNC. That is to say. it may be,according to an informant, ( 1 ) "a chimney stack made of rock or stone", ( 2 ) "a stack of rock inchimney formation which has been left standing when erosion has cleared surrounding rock", or ( 3 )"a narrow cleft or passage in a rock or mountain face encountered by rock climbers", etc. However,the problem crucially remains : i. e. how can we specify the intended meaning of the given NNC ? Inother words, how can we infer what is meant by this NNC? We now come to the fourth andcontextual element V which enables h to infer what is meant by this NNC in the communicativesituation between s and h. Thus there are four elements which are involved in NNC interpretation asfollows :

42) b : Determinant

a : Determinatum

I R } : Semantic relation of b to ax : Contextual knowledge

Suppose we name x for rock chimney, we can provide a process of inference :

(43) ( a ) X is something in which the climbers (the cohesive 'tie' or the referenceof "they") were stranded overnight.

( b ) From x a helicopter airlifted the climbers to safety.

And we presuppose that the 'climbers in the party' have something to do with mountain climbing,and they have nothing to do with any kind of real chimney on the roof of a house. This rock chimney

must be something like a chimney which is made of rock in the mountain where the party could bestranded overnight in an emergency. We now arrive at the intended meaning of rock chimney bymaking a contextual inference based upon the presupposed knowledge of the context where this NNCis used.

Thus the semantic relation }R { of 'b' to 'a' is specified by a certain amount of contextualknowledge 'x.' That is. the contextual knowledge V enables h to solve the problem of semanticambiguity in innovative NNCs.

However, when we say 1R | is specified by x, we must remember that 'a' or V is determinedby V. The ambiguity of | R ( is in fact that of the determinatum 'a' or the determinant 'b'. For

Page 24: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds 23

example, in hammer shelf, it may be "a shelf for a hammar" or "a geological outcrop of rockreminiscent of a hammer" ; or in computer corridor it may be "a corridor in a building where computerpeople work or have offices" or "a strip of land given over to computer production". These areexamples where the determinatum 'a' is ambiguous. On the other hand, in trunk bottle, it may be "abottle in the form of an elephant's trunk" or "a bottle not needed for immediate consumption or use,or to escape customs controls, and so packed in the trunk". And in zebra pot. it may be "a black andwhite striped pot" or "a pot shaped like a zebra/a pot with a zebra on". These are the exampleswhere the determinant 'b' is ambiguous.

In the former case, the ambiguous 'a' must be specified by the contextual knowledge V. and itsscheme is as follows :

(44) (D (categorized by ((a(specified by x))

((subcategorized by b)in JR| ))))

where the designatum 'D' is categorized by 'a' which is specified by a contextual knowledge V. andwhich is also subcategorized by 'b' in the semantic relation 1R { •

In the latter, 'b' is ambiguous, and it must be specified by V. And its scheme is as follows :

(45) (D (categorized by a(subcategorized by (b

(specified by x))in |R[ )))

where the designatum 'D' is categorized by 'a' which is subcategorized by 'b' which is specified byV in the semantic relation I R [ . Thus, to summarize :

( a ) An NNC is a third word which is realized by use in context where h can obtain acontextual or pragmatic knowledge 'x' to understand what is intended by sthrough a given NNC.

( b ) An NNC consists of a determinatum 'a' which categorizes a designatum 'D' and adeterminant 'b' which subcategorizes 'a'.

( c ) There is a case of NNC where 'a' which is subcategorized by 'b' is ambiguous andis specified by some contextual knowledge V; i. e. rock chimney, hammer shelf,

computer corridor, etc.

( d ) There is a case of NNC where 'b' which subcategorizes 'a' is ambiguous and isspecified by 'x'; i. e. trunk bottle, zebra pot, water generation, etc.

[ 3 ] The concept of 'context'

So far a contextual knowledge V has been found in the given context. It has been, in thissense, intra-linguistic knowledge. The It's process of understanding a given innovative NNC has beenbased upon the knowledge from the linguistic context. However, we must take into consideration a

Page 25: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

24 Morio Nishikawa

different kind of contextual knowledge which is socio-cultural or extra-linguistic. It cannot be found

out explicitly in the given linguistic context, but it should be known and shared socio-culturallyamong people as implicit knowledge.

Here is an innovative NNC which can be understood extralinguistically.

(46) Aids farm plea by Nobel Nun.Daily Mail 4/1 1986

When we first come across this NNC. we can only guess that this 'Nun' must have something to dowith 'Nobel'. Or we can only provide the following schematic interpretation :

(47) (NUN (categorized by Nun((subcategorized by Nobel)

in |Rt )))

In this case, we can determine the intended meaning if and only if we know the fact that there is anun who had got the Nobel prize. Otherwise, we cannot specify who and what this nun is. Thepresupposed knowledge about this NNC does not come from any cohesive knowledge in the text, butit comes from the socio-cultural or extra-linguistic knowledge which is shared among people whoknow something crucial about the designatum of this NNC.

Here is a case where extra-linguistic contextual knowledge works to a great extent.

(48) Iron lady meets iron man, Omar khalifa, yesterday.The Sunday Times 25/5 1986

Generally speaking, we know who this iron lady is if we have a certain amount of backgroundknowledge about this person. Here also works the socio-cultural or extra-linguistic knowledge for h

to understand what is meant by this NNC. This lady has something to do with iron. It is no doubt thenick-name of a British lady politician who has a strong volition to decide and act in difficultsituations. Her personality may be characterized by the inherent semantic feature of iron. Thus wecan make a guess about the designatum of this NNC, although we do not have any presupposedlinguistic or cohesive knowledge.

We now come to the summary of the concept of 'context'. There are roughly two kinds ofcontextual knowledge which enables h to comprehend what is meant by a given NNC. One is cohesiveand linguistic knowledge which is shared in the communicative situation between s and h. The other

is socio-cultural or extra-linguistic knowledge which is shared among people (including s and h) whobelong to the same linguistic community.

[ 4 ] Constraints of NNC Formation

The combination of the determinant 'b' and the determinatum 'a' in the formation of NNC can

be varied. But all combinations are not always possible. They undergo some constraints on NNCformation. For example, we know that enemy country is a common NNC, but the contrastive NNC'friend' country is not so. We now remember the word ally which is to block friend country. Friend

country may be potential, but the existing synonymous word 'ally' blocks the instantiation of thisnewly coined NNC. In the same way, sauce pan is just a common noun which means "a metal

Page 26: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds 25

container of moderate depth, usually having a long handle and sometimes cover, for stewing, etc"(ACD). And it cannot be replaced by 'stove pot' which cannot be a real NNC because of the blockingof the existing word. Furthermore, we have the NNC sea salt and its contrastive NNC is rock salt

which cannot be replaced by any other tentative NNC such as land salt or earth salt. Blocking,according to Aronoff (1981 '. 43), is the name given to the phenomenon of "the non-occurence of oneform due to the simple existence of another."

Sister city may have the contrastive NNC brother city. But brother city is not in fact possible,because sister city itself blocks brothercity. And in addition, city is diachronically compatible with thefemale feature of sister, but not with the male feature of brother. Thus the feature compatibility orincompatibility between the two constituent words works in the acceptability or non-acceptability ofthese two NNCs.

As for the incompatibility of the semantic feature between the two constituents, we have adifferent kind of combination : i. e. ox cow, stallion mare, cock hen (or rooster hen) , etc. Here is thefeature contradiction of sex between the two constituents. However, the NNC formation is not

unacceptable if it is characterized by metaphorical relevance. Frogman, lady bird, snake river, etc. areNNCs which are incompatible or contradictory in semantic feature, but they make good sense as NNCs.

Now we can hypothesize the principle that a newly coined NNC must be blocked by the priorexistence of an established word, and that the incompatibility of the semantic feature between theconstituent words causes a given NNC to be unacceptable, if it is not metaphorical.

There should be another hypothetical constraint which we will call the principle of humanrelevance. We have the NNCs such as life insurance, death mask, health center, etc., but we usually donot have contrastive NNCs such as death insurance, life mask, disease center, etc., because they areirrelevant in the realistic sense of human beings. We don' t have prohibition ticket in contrast withpermission ticket. These are the examples which are irrelevant in the ordinary socio-cultural context.

To avoid redundancy is a general principle of verbal communication along with the 'Maxims ofManner' of Grice (1975) :

( i ) Avoid obscurity of expression.(ii) Avoid ambiguity.(iii) Be brief,(iv) Be orderly.

These maxims are also applicable to NNCs. In fact, the rich variety of NNCs in English is supportedby these maxims. So the combination of synonymous words is not common in English. For example,infant child, street road, girl lass, etc., are not allowed to come to be existent. In the same way. pony

colt, weather climate, treaty pact, etc., are not realistic. Furthermore, flower tulip, animal horse, fish trout,

etc., are not relevant because the second constituent word already implies the feature of the firstword. That is to say, tulip is a kind of flower, and horse is a kind of animal, and so on. These arecases of redundancy in NNCs, and this kind of redundancy must be checked by the principle of non-redundancy in NNC formation. To summarize the hypothetical constituents of NNC formation :

( i ) The NNC formation must be checked by blocking (the prior existence of anestablished word).

(ii ) The NNC formation must be checked by the crucial incompatibility of the semanticfeature between the two constituent words, if it is not metaphorical.

Page 27: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

26 Morio Nishikawa

(iii) The NNC fomation must be checked by lack of realistic relevance to human life.

(iv) The NNC formation must be checked by the redundancy of semantic featurebetween the constituents.

[VI] Conclusion

Innovative NNCs arc ubiquitous in English. To combine one noun with another is a commonway of coining a new NNC. We claimed that there are two kinds of NNCs : conventional andinnovative. In conventional NNCs such as light house and gas light, we are able to retrieve theirmeaning from the stored knowledge of the lexicon which is shared by s and h. In innovative NNCssuch as watergeneration i\m\foam blanket, we are required to infer their meaning by taking account ofthe knowledge drawn from the context where they are used. Here semantics and pragmatics interact.That is to say, semantics, on the one hand, provides us with possible or candidate interpretations ofa given NNC based upon the code model of the shared knowledge of the lexicon. Pragmatics, on theother hand, works to specify from among them one meaning by taking account of the contextualknowledge. Here we have the inferential processing of meaning in order for h to arrive at s'sintended meaning of a given NNC.

There are two sub-concepts of the knowledge of context. One is cohesive or linguisticknowledge and the other is socio-cultural or extra-linguistic knowledge. In the former, h is to obtainthe crucial knowledge of context from somewhere in the utterance or text. In the latter, however, h isto obtain it from the socio-cultural or extra-linguistic knowledge which is shared among people whobelong to the same linguistic community. We now summarize as follows :

( 1 ) An NNC is a linguistic product of the process of naming to an unknown thing oridea. It has therefore a designatum (D) which is categorized by the head of NNC.It is characterized by a denotative correspondence between an NNC and itsdesignatum.

(2) An NNC consists of two elements. One is the determinatum 'a' which issubcategorized by the other which is the determinant 'b' in the syntagma of#b+a#.

( 3 ) The syntagma of the elements 'a' and 'b' implies a semantic relation 1R [ of 'b' to'a', which is a conceptual frame of our mental process of interpretation.

( 4 ) Potentially. 1R { is not limited in number out of context.

( 5 ) The well-established meaning of conventional NNC can be retrieved from thestored knowledge of the lexicon.

( 6 ) | R | is uniquely specified by the crucial knowledge from the context where aninnovative NNC is actually used.

( 7 ) Thus the context-free grammar of NNCs is not adequate.

( 8) Only the semantic approach is not enough. It can only provide a number ofpossible or candidate interpretations of an NNC out of context.

(9 ) Pragmatic approach aims at specifying one unique interpretation from among

Page 28: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds 27

possible or candidate interpretations of an NNC which come from semanticapproach.

(10) Here works the concept of relevance which is the shared knowledge between aspeaker and a hearer to determine whether or not a given semantic relation \ R (is acceptable.

The process of understanding or comprehension goes from the level of what is already known

('given' information) to that of what is newly known ('new' information). A theory of understandinginnovative NNCs can be applied to this process. 'Given' information of shared knowledge enables h

to understand the meaning of conventional NNCs by retrieving it from the lexicon. 'New' information,on the other hand, must be processed or calculated by taking account of the contextual knowledge.

In chapter [ I ] . we discussed some fundamental problems of NNCs. 'Data and method' wereintroduced at first. And then the communication model of 'production and comprehension' wasdiscussed, focusing upon the process of inferring or calculating what is said and intended by aspeaker. In chapter [ II ] , we discussed the historical background of the study of Englishcompounds, referring to Marchand (1969) , transformationalists like Levi (1978) , lexicalists likeChomsky (1970), and other important grammarians. In chapter [III] , the concepts fundamental toNNCs were discussed, including the definition of NNC. The categorization and subcategorizationprocesses were investigated and schematized based upon the inferential process of understanding agiven NNC. There we focused on the functional or contextual analysis of innovative NNCs in termsof the position that the contextual knowledge V is indispensable for // to understand s' s intendedmeaning of a given innovative NNC. This position raises the crucial problem of the interactionbetween the semantic and pragmatic approaches in the study of NNCs.

Acknowledgements

I must offer specific thanks to Prof. Geoffrey N. Leech who supervised and encouraged me tocontinue this research at Lancaster University. I am also grateful to Dr. Ted Briscoe who advised meand provided many valuable suggestions in the course of my research. Nick Campbell, ChrisFeudtner, and Phil Elenweight were my colleagues at Lancaster who were kind enough to share theirtime for discussion of the problems of English word-formation. Dr. Alan Rosen, my colleague atKumamoto University, kindly read the manuscript and gave me stimulating suggestions. And finally,my thanks must also go to Mr. Peter Buckley who supported my academic life in Lancaster as acounsellor of my scholarship.

Dictionaries

Barnhart, C. L. (1947) The American College Dictionary [ACD] , Random House, New YorkCoulson. J. et al. (1978) The New Oxford Illustrated Dictionary, Oxford Univ. Press, OxfordHanks. P. (1979) Co//i'ns English Dictionary [CED] . Collins. London and GlasgowHornby. A. S.(1948) The Oxford Advanced Learner's Dictionary of Current English, Oxford Univ. Press.

Oxford

Murray. J. A. H. et al. (1884, 1928) The Oxford English Dictionary [OED] , Oxford Univ. Press.Oxford

Parnwell, E. C. (1980) Oxford Picture Dictionary ofAmerican English. Oxford Univ. Press. OxfordStein. J. (1966) The Random House Dictionary of The English Language, Random House. New York

Page 29: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

28 Morio Nishikawa

Bibliography

Adams, V. (1973) An Introduction to Modern English Word-Formation, Longman, LondonAronoff, M. (1981) Word Formation in Generative Grammar (second printing), The MIT Press, Cam

bridge, MassachusettsAustin, J. L. (1962) How to Do Things with Words. Oxford Univ. Press, OxfordBates, E. (1976) Language and Context-The y\cquisition of Pragmatics, Academic Press, New YorkBauer, L. (1983) English Word-Formation, Cambridge Univ, Press, CambridgeBierwisch, M. and K. E. Heidolf (1970) Progress in Linguistics, Mouton, The HagueBolinger, D. (1975) Aspects of Language, 2nd edn., Hartcourt Brace Jovanovich, New YorkBooij, G. (1977) Dutch Morphology, The Peter de Ridder Press. AmsterdamChafe, W. (1970) Meaning and the Structure of Language, Univ. of Chicago Press, ChicagoChomsky, N. (1965) Aspects of the Theory of Syntax, MIT Press, Cambridge, Massachusetts

(1970) "Remarks on Nominalization*. Readings in Transformational Grammar (eds.Jacobs. R and P. Rosembaum), Blaisdell, Waltham. Massachusetts

(1986) Knowledge of Language-Its Nature, Origin, and Use, ((Covergence) -A SeriesFounded by R.N. Anshen). Praeger Publishers, New York

Clark, E. V. (1981) 'Lexical Innovations : How children learn to create new words'. The Child's Con

struction of Language, Academic Press, London and New YorkClark, E. V., S. A. Gelman, and N. M. Lane (1985) 'Compound Nouns and Category Structure in

Young Children', Child Development No.56. 84-94Clark, H. (1978) 'Inferring what is Meant', in Levelt and Flores d' Arcais 1978 ;295~322

(1981) 'Making Sense of Nonce Sense', The Process of Language Understanding (eds. G. B.Flores d' Arcais et al.) . Wiley, New York

Clark, II. and T. B. Carlson (1981) 'Context for Comprehension', Attention and Performance IX (eds.Long and Baddelay). 313-330, Laurence Erlbaum Associates Publishers,

Clark, H. and R. J. Gerigg (1983) 'Understanding Old Words with New Meanings', Journal of VerbalLearning and Verbal Behavior vol. 22, 591-608

Dijk, T. A. van, (1977) Text and Context, Longman, LondonDik, S. C. (1978) Functional Grammar, North-Holland, AmsterdamDowning, P. (1977) 'On the Creation and Use of English Compound Nouns', Language 53, 810-842Fillmore. C. (1977) 'Topics in Lexical Sematics'. Current Issues in Linguistic Theory, (cd.) R. W. Cole,

Bloomington. Indiana

(1981) 'Pragmatics and the Description of Discourse', Radical Pragmatics (ed.) P. Cole,Academic Press, New York

Foss, D. J. and D. T. Hakes (1978) Psycholinguistics : An Introduction to the Psychology of Language,Englevvood Cliffs. Prentice Hall, New Jersey

Fudge, E. (1984) English Word Stress, George Allen & Unwin, LondonGarnham. A. (1985) Psycholinguistics, Methuen, London and New YorkGrice, H. P. (1968) 'Utterer's Meaning, Sentence-Meaning, and Word-Meaning', Foundations of Lan

guage vol. 4, 225-242(1975) 'Logic and Conversation' in Cole and Morgan (1975). 41-58(1981) 'Presupposition and Coversational Implicative', Radical Pragmatics, Academic

Press, New York

Halle, M. (1973) 'Prolegomena to a Theory of Word Formation', Linguistic Inquiry, 3-16Halliday, M. A. K. (1978) Language as Social Semiotic, Edward Arnold, London

Page 30: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

Semantics and Pragmatics of Innovative Noun-Noun Compounds 29

(1985) An Introduction to Functional Grammar, Edward Arnold, LondonHalliday, M. A. K. and Hasan, R. (1976) Cohesion in English, Longman. LondonJakobson. R. (1960) 'Linguistics and Poetics'. Style in Language, (ed.) T. A. Sebeok, 350-377, The

MIT Press, Cambridge, MassachusettsJackendoff, R. (1975) 'Morphological and Semantic Regularities in the Lexicon',Language 51. 639-671Jaspersen, O. (1942) A Modem English Grammar on Historical Principles (Vol. VI), George Allen and

Unwin, London(1965) The Philosophy of Grammar, W. W. Norton &Company, New york

Jones. K. S. (1983) 'Compound Noun Interpretation Problems', (Science and Engineering ResearchCouncil), Computer Laboratory, University of Cambridge, Cambridge

Kagcyama. T. (1982)'Word Formation in Japanese', Lingua Kol. 57, 215-258Leech, G. N. (1980) Explorations in Semantics and Pragmatics. John Benjamins, Amsterdam

(1981) Semantics (2 nd edition), Harmondsworth, penguin Books(1983) Principles of Pragmatics, Longman, London

Lees. R. B. (1960) The Grammar of English Nominalizations, Indiana University Press. Bloomington.Indiana

(1970) Problems in the Grammatical Analysis of English Nominal Compounds', Progress inLinguistics, 174-186, Mouton, The Hague,

Leonard. R. (1981) An Algorithm for the Interpretation of English Noun Sequences by Computer, Ph. D.Thesis (University of Lancaster), Lancaster

Levinson, S. C. (1983) Pragmatics, Cambridge Univ. Press, CambridgeLicber, R. (1983) 'Argument Linking and Compounds in English', Linguistic Inquiry vol. 14, no.

2.251-285

Lipka. L. (1972) Semantic Structure and Word-Formation, Fink, MunichMacLaran, R. (1983) 'On the Interaction of Semantics and Pragmatics', CLCS Occasional Papers,

Dublin

Marchand, H. (1969) The Categories and Types of.Present Day English Word Formation, C. H. Beck'sVerlagsbuchhandlung, Munchen

Miller, G. A. (1978) 'Semantic Relations among Words', Linguistic Theory and Psychological Reality,(eds M. Halle et al.), 60-118. The MIT Press. Cambridge. Massachusetts

Prince. E. F. (1981) 'Towards a Taxonomy of Given-New Information', in Cole (1981) ; 223-256Quirk, R, et al. (1985) A Comprehensive Grammar of the English Language, Longman, London and New

York

Rappaport, M. (1983) 'On the Nature of Derived Nominals', Papers in Lexical-Functional Grammar(eds. L. Levin et al.). Reproduced by Indiana Univ. Linguistics Club, Indiana

Reinhart, T. (1981) 'Pragmatics and Linguistics : an analysis of sentence topic', Philosophica vol. 27,53-94

Roepcr. T and E. A. Siegel (1978)'A Lexical Transformation for Verbal Compounds'. Linguistic Inquiry vol. 9. no. 2. 199-260

Sag. I. A. (1981)'Formal Semantics and Extralinguistic Context'. Radical Pragmatics, (ed.) P. Cole,Academic Press, New York

Searle. J. R. (1969) Speech Acts, Cambridge Univ. Press, Cambridge(1979) Expression and Meaning, Cambridge Univ. Press, Cambridge(1983) Intentionality,Cambridge Univ. Press, Cambridge

Selkirk, E. (1982) The Syntax of Words, The MIT Press, Cambridge. Massachusetts

Page 31: 熊本大学学術リポジトリ Kumamoto University Repository System · Homo ludens.homo faber. homo loquens,homo mentiens.homo sapiens,all these'definitions'of human beings arc

30 Morio Nishikawa

Sperber, D. and D.Wilson (1986) Relevance-Communication and Cognition, Basil Blackwell, OxfordWarren, B (1978) Semantic Patterns of Noun-Noun Compounds, Gothenburg Studies in English 41,

Goteborg, Acta Universtatis GothoburgensisWilliams. E. (1981) 'On the Notions "Lexically Related" and "Head of a Word"' Linguistic Inquiry vol.

12. no. 2. 245-274