PROTOTYPICAL MEANING VS. SEMANTIC CONSTRAINTS IN THE ANALYSIS OF ENGLISH GENITIVES

NAOKO HAYASE Osaka University

This paper deals with the meanings of English possessive genitives based on a prototype-theoretic approach combined with a semantic con- straint implemented here in the form of a schema. Prototype theory can account for the subtle, gradient differences in accessibility among the various readings of the possessive genitive construction, while a schema is better suited to capturing a variety of marked contrasts in acceptability among possessive data. The schema proves to be powerful in account- ing for cases which interact with contextualization. This suggests that the correct analysis of the meanings of possessive genitives becomes possible in terms of the combination of both the prototype-theoretic analysis and the schema-based approach.*

0. Introduction In this paper use the general framework of cognitive grammar to propose a two-part analysis of the semantics of the English possessive genitive construction. One part is based on the notion of prototype categories. In recent works of cognitively-based semantics (e. g., Lan- gacker (1987), Lakoff (1987), Taylor (1989a)), has been demon- strated that most words are inherently indeterminate in meaning and cannot be defined in terms of necessary and sufficient conditions for membership in the categories that designate. This amounts to saying that semantic categorization is not a matter of 'yes or no' but

* This is a paper based on and developed out of my presentation at the 64th Conference of the English Literature Society of Japan, held at Seinan Gakuin Uni- versity, May 23-24, 1992. am very grateful to many members for enlightening comments and discussion. I would like to thank Seisaku Kawakami, Yukio Oba for their constant encouragement and valuable suggestions during the writing of this paper. I also thank Michael T. Wescoat, and two anonymous EL reviewers for their insightful comments and suggestions on earlier versions of this work. Thanks are also due to M. T. Wescoat for invaluable, patient discussion and extensive sty- listic suggestions. Remaining inadequacies are of course my own.

English Linguistics 10 (1993) 133-159 -133- (C)1993 by the English LinguisticSociety of Japan 134 ENGLISH LINGUISTICS, VOLUME 10 (1993) one of 'more or less' and that the meaning of a word lies in a semantic prototype to which individual members may correspond in varying de- grees. A prototype category is an agglomeration of individuals or rela- tions that more or less resemble some prototype. It is convenient to conceive of a prototype category as a layered object of a kind that might be depicted by a set of concentric circles. At its very center lies the prototype, and in the layer immediately around the prototype are situated the members of the category that most closely resemble the prototype. As one moves outward, each successive layer contains members that resemble the prototype a little bit less than those mem- bers that reside in the previous layer. For any given level of the layered structure of a prototype category it is theoretically possible to provide an abstract description of all the members of the category that lie at that level or else at some level closer to the core, where the pro- totype resides. Such a description is called a schema, and it essentially picks out all the individuals or relations that resemble the prototype to a given degree. Schemata may be used to implement semantic con- straints and will provide the basis of the second part of our analysis. Various cognitive grammarians, e. g., Taylor (1989b), Durieux (1990), and Nikiforidou (1991), have expressed the view that the semantics of English possessive genitives can be adequately accounted for in terms of prototype categories. However, the principal claim of this paper is that the proper explanation of the meaning of the possessive genitive construction must be stated not only in terms of prototype categories but also with reference to schemata. The empirical prediction is that there is a threshold among degrees of resemblance to the prototype which must be met or surpassed in order for a given interpretation to be acceptable. Thus we predict two trends in empirical acceptability. On the one hand, we expect a smooth, gradient effect whereby some readings are relatively more accessible than others, and on the other hand we anticipate punctual drop-offs in acceptability if certain criteria to be defined below are not satisfied. Our goal in section 1 will be to demonstrate the utility of modeling the meaning of the possessive genitive construction through a prototype category. The members of this category will be relations that hold be- tween a possessor and a possessee, e. g., the 'ownership' relation or, perhaps less obviously, the 'parenthood' relation, etc. We shall moti- vate the approach by isolating the most generally accessible relation ex- pressible with the possessive genitive construction and positing it as the THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 135 prototype; we will then demonstrate that other relations conveyed by the construction form a descending hierarchy of accessible readings, in a manner consistent with the prototype-category model. As one progresses away from the prototype possessive relation, one eventually comes to a level in the prototype category where the mem- ber relations resemble the prototype so little that they cannot be con- sidered as felicitous readings for the possessive genitive construction. Therefore our task in section 2 will be to provide a schema that char- acterizes those members of the prototype category that resemble the prototype closely enough to be possible possessive-genitive readings. In section 3 we will examine a variety of infelicitous possessive forms to show that the relations between possessor and possessee do not adhere to the schema and therefore are not sufficiently close to the prototype relation to be acceptable. In section 4 we will examine classes of cases where context sharply affects the acceptability of possessive genitive forms. We can account for this context-sensitivity in the present model by demonstrating that contextual variations may cause a relation to satisfy the schema for pos- sessive genitives on one occasion, even though the same relation would not be admitted by the schema in a more usual, neutral context. These observations support the use of schema-based semantic con- straints, since an approach that employed only prototype categories would be hard pressed to account for the same data.

1. Possessive Genitives and Various Relations They Describe

In the present section, we will look at meaning relations expressible with possessive genitives and observe that they are more circumscribed than they are thought to be. In fact the relations between possessor and possessee seem to form a hierarchy wherein some readings are more readily accessible than others. This scalar effect is conducive to the prototype-category model. We will begin with a brief overview of the data.

1.1. Overview First of all, the possessive genitives we will deal with here serve to identify the head nouns that follow them:1

1 This paper is not concerned with descriptive or compound genitives, which have 136 ENGLISH LINGUISTICS, VOLUME 10 (1993)

(1) a. the funnel of the ship-the ship's funnel b. the funnel of a ship-a ship's funnel c. a funnel of the ship (Quirk et al. (1985: 1276)) (1a) and (1b) have felicitous possessive genitive counterparts, while (1c) does not. This indicates that a possessive genitive has a reading where it has some definite reference, even though the possessive itself may be definite or indefinite.2 Possessive genitives are most typically thought to express the relation of ownership, but many other relations are possible: (2) a. John's book b. John's father c. John's heart d. the book's owner e. the house's entrance f. last year's decision g. San Francisco's Golden Gate Bridge Kinship, whole-part, temporal and spatial relations are also expressed with the same construction. The relations depicted by the possessive construction are varied enough to make several linguists (e. g., Kemp- son (1977: 125)) claim that the meaning of the possessive genitive is in- determinate or even vague, and that any attempt to explain its semantic properties will be fruitless. In contrast, we shall argue that the relations which possessive con- structions can describe are in fact limited to a certain group. Whereas the 'semantic indeterminacy hypothesis' would entail the incorrect pre- diction that any application of a possessive genitive will do as long as the possessive identifies its head noun, the empirical data lead to an en- tirely different conclusion. (3) a. # the knife's fork b. # the hat's boy c. # the father's John d. # the collar's dog These examples indicate that the relation between possessive genitives and head nouns is more circumscribed.

been distinguished from possessive genitives syntactically and semantically. Posses- sive genitives have the syntactic structure [[NP's] N] ([[the child's] book]) and al- ways have definite reference, while descriptiveor compound genitivesdo not, and furthermore have the structure [N's N] ([a [children'sbook]]). 2 See Woisetschlaeger(1983: 147-149) for a relevant discussionon the matter of definitenessin NPs with genitiveNP determiners. presents severalexamples to show that nouns with genitiveshave only a single reading, namely, the definite one, and that they are not ambiguousbetween definite and indefinitereadings. THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 137

1.2. POSSESSION as a Prototype Category Next we shall demonstrate that the various meanings of the posses- sive genitive construction can be best accounted for in terms of a prototype category. Let us call this the prototype category of POS- SESSION. Toward that end, our first task will be to determine the prototype around which the category will be formed. While we have observed that possessive expressions are open to mul- tiple interpretations, the reading of ownership is the most prominent. A variety of data support this view. (4) John's book a. the book that John has b. the book that John wrote c. # the book that John kicked d. # the book that John tore If the meaning of were truly indeterminate, then all the re- lations in (4) would be acceptable as interpretations of John's book. Instead certain relations describable with possessive genitives stand out as prototypical: included among these relations are true ownership, as in (4a), and creatorship, as in (4b). Too much deviation from the norm results in difficulty in interpretation. In addition, the most favored interpretation of John's in (5) is that of 'possessor' rather than 'creator', because of the existence of the pre- positional phrase of himself. (5) John's picture of himself Though possible, it is unusual to take a picture of oneself. The most prototypical and the least eliminatable interpretation of the possessive genitive is that of possessor. Moreover, if we change the head noun to computer, for example, only the "possessor" reading is readily avail- able, especially when there is no special context. (6) a. Whose computer is this? b. It's John's. The most striking illustration of the prominent status of ownership is found in the predicative genitive construction. (7) a. This book is John's. b. # This father is John's. c. (#) The heart is John's. d. # The owner is the book's. e. # The entrance is the house's. f. # The decision is last year's. 138 ENGLISH LINGUISTICS, VOLUME 10 (1993)

g. # The Golden Gate Bridge is San Francisco's. As the examples in (7) show, relations other than ownership are ordi- narily excluded from this construction. These data suggest that the central, prototypical meaning invoked by possessive genitives is that of true ownership. Hence the relation of true ownership will constitute the prototype within the prototype category of POSSESSION. Let us now consider the next rank of near-prototypical relations, which are situated within the prototype category immediately adjacent to the core constituted by the true-ownership relation. Another con- struction involving genitives and allowing additional relations is the postmodifying genitive construction.3 Compare (3) with (8); (8) a. a book of John's b. That father of John's (, what a guy!) c. (#) That heart of John's (is always breaking.) d. # {the/an/that} owner of the book's e. # {the/an/that} entrance of the house's f. # {the/a/that} decision of last year's g. # the Golden Gate Bridge of San Francisco's (8a) can describe true ownership: it means a book John has, not a book about John. In addition, the kinship and body-part relations are acceptable, but other relations are excluded. This suggests that though true ownership is the prototypical core relation, kinship and body-part relations are to be situated close to the prototype within the prototype category of POSSESSION. Lastly, let us look at data where yet other interpretations close to this prototype are strongly preferred. (9) a. The invention of instant coffee in Japan was one of the most important events in economic history. b. The invention of instant coffee on the desk before us was one of the most important events in economic his- tory. Japan and desk in (9) are understood as locations or settings in which events take place. In (10), which corresponds to (9), possessive geni- tives are not accepted if they are understood as settings. (10) a. (#) Japan's invention of instant coffee (OK as Agent)

3 It is true that postmodifying genitives are syntactically distinct from possessive genitives, but their semantics are closely related. See Taylor (1989: 682). THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 139

b. # the desk (before us)'s invention of instant coffee As for (10a), it is much improved when Japan is interpreted not as a mere place but metonymically as a participant, i. e., the Japanese peo- ple in general. In this case, the relation described is regarded as simi- lar to the prototypical relation of POSSESSION: Japan is understood as a possessor of the event described by invention in that Japan brings it about. Conversely, in (10b) the difficulty of such a metonymical in- terpretation yields an awkward interpretation.4

(11) a. Japan→Japanese people b. the desk→#people at the desk In brief, the possibility of metonymical interpretation makes it easier to accept the genitive denoting place. More examples may be cited: (12) a. (#) Osaka's destruction of the theater (OK as Agent) b. (#) Western Japan's employment of woman executives (OK as Agent) c. # Moscow's attack of the army on the citizens The examples in (12) become acceptable if a metonymical reading of the genitive of place is possible, e. g., one in which the location is equated with an agency or institution. If this metonymical transition is a conceptually difficult task, as it is in (10b), then the only possible reading for the genitive is that of location, which results in low accepta- bility. Note that interpretation merely as a place is not enough to license a genitive. When there are several choices of interpretation for a geni- tive place-name, we tend to select the one closest to the prototypical relation and therefore to interpret the genitive of place as expressing the relation nearest to that of true ownership. Metonymical under- standing reflects an attempt to pursue the most nearly prototypical rela- tion possible from within the category of POSSESSION, i. e., the most appropriate relation representable with genitive expressions.

4 It depends on the context whether or not its genitive can be interpreted metonymicallyas an agent-like POSSESSOR. For example, the same word desk can behave differentlyaccording to the head noun or situation associatedwith it. (i) a. the desk's subscriptionto the periodical b. the desk's correctionof the proofs (at the press company) These examples allow for a metonymicalinterpretation of the desk as the people at that desk, and they are naturally acceptable. For detailed discussionof the issue, see Lakoff and Johnson (1980)and Yamanashi(1988). 140 ENGLISH LINGUISTICS, VOLUME 10 (1993)

To sum up, in this section we have attempted to show that certain re- lations expressed by possessive genitives are predominant. This state- of-affairs may be readily modeled by appealing to prototype categories. The most prevalent reading for possessives is that of true ownership, so we naturally propose this as the prototype around which the prototype category of POSSESSION will be formed. Other readings which are readily accessible with the possessive genitive, though not quite so readily as the true ownership relation, include kinship, the whole-part relation, and agency. The use of prototype categories affords us an effective model of scalar effects, since the latter relations may be in- cluded within the prototype category of POSSESSION in a location close to but distinct from that occupied by the relation of true owner- ship.

2. The Framework of Cognitive Grammar and Possessive Genitives Having proposed a prototype category to account for gradience in accessibility among relations expressible with possessive genitives, we now come to the second of our basic claims, i. e., that a relation must resemble the prototypical true-ownership relation to a given degree in order to be acceptable as a possible reading for the possessive genitive construction. Our analysis will characterize the required degree of re- semblance by means of a schema. Recall that a schema could be said to capture a degree of similarity to a prototype by characterizing all the members of the prototype category that resemble the prototype at least to the degree in question. Hence, in this section, we will show that there can be fairly sharp drop-offs in acceptability in addition to the gradual ones emphasized above, and we shall attempt to isolate a sche- ma that correctly characterizes all of the relations that may be express- ed by means of the possessive genitive construction. For that purpose, we shall introduce several concepts from cognitive grammar and show that these concepts allow us to provide a simple and yet interestingly general characterization of possessive genitives.

2.1. The Conception of Nouns Within the framework of cognitive grammar, it is claimed that the meanings of linguistic expressions are conceptual in nature and that they are characterized with respect to a background in the form of one or more cognitive domains. Langacker (1987: 147) claims that "[m]ost THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 141 concepts presuppose other concepts and cannot be adequately defined except by reference to them, be it implicit or explicit." The set of con- ceptual domains serving as a background for semantic characterization is called the base. A linguistic expression designates a substructure of the base called the profile. Most domains directly relevant to semantic characterization are of complex sorts called abstract domains, while some are basic in the sense that they cannot be reduced to more primi- tive ones. Examples of basic domains include time, three-dimensional space and so on.5 Langacker (1987: 183-184) gives one well-known example to illus- trate the relation between base and profile: the notion of [CIRCLE] is necessary to conceptualize [ARC] or [DIAMETER]. Divorced from the circle, an arc is merely a curve, and a diameter is nothing more than a straight line. In this sense, [CIRCLE] is conceptually a neces- sary component, a base required for [ARC]/[DIAMETER] to receive profile appropriately.6 Various kinds of domains serve as the base of nouns. January, Tuesday, hour, day and year are not characterized directly with respect to the basic domain of time but require abstract domains like a calen- drical system. This conceptual view on the meaning of nouns amounts to claiming that linguistic meanings appeal to general encyclopedic knowledge.

2.2. Schematic Characterization of Possessive Genitive Constructions Having introduced useful notions from cognitive grammar, we may in turn extract a schema which will characterize those members of the prototype category of POSSESSION that resemble the prototype close- ly enough to be possible possessive genitive readings. Toward that end, it will be claimed that the notion of domain, first utilized by Taylor (1989b), is useful in defining the schema; this concept will be supplemented with the notion of unique identifier, which plays a crucial role in making the possessive genitive construction acceptable.

5 Langacker(1987: 150, fn. 5) commentsthat the abstract domain discussedhere roughlycorresponds to what Lakoff calls an ICM (for idealized cognitive model) and what others call a frame. 6 For more detailed information and discussionon the matter, see Langacker (1987:Chs. 4-5). 142 ENGLISH LINGUISTICS, VOLUME 10 (1993)

2.2.1. The Notion of Domains as a Condition for Possessive Genitives Employing the cognitive grammar framework briefly explained above, Taylor (1989b) claims that possessive genitives must realize one participant within domains required in conceptualizing the head nouns with which they occur. Thus, possessive genitives are used to identify the referents of head nouns by referring explicitly to a certain partici- pant within the conceptual bases on which the interpretation of those head nouns depends. In the case of (13), for instance, [CIRCLE] is realized as a possessive genitive in order to identify its head noun dia- meter, since the former is necessary to conceptualize the latter. (13) a. the circle's diameter b. the circle's radius In conceptualizing [DIAMETER] or [RADIUS], as was explained above, [CIRCLE] is implicitly presupposed, i. e., it serves as a base do- main for diameter and radius. Though we do not believe that Taylor has provided a completely satisfactory characterization of the relations that may serve as readings for the possessive genitive construction, the domain-based approach is nonetheless extremely attractive and shall be adopted here as a part of the ultimate definition of the schema describ- ing appropriate possessive genitive relations. Let us reexamine the examples of prototypical and near-prototypical POSSESSION relations identified in the previous section from this viewpoint. What is crucial is that the possessive genitive must refer to a participant within the domain for the head noun or sometimes to the domain itself. For example, (14)-(16) are examples of the body/part and whole/part relations. (14) a. John's hands b. Mary's eyes c. that gentleman's arms d. Betty's intelligence (15) a. the cat's tail b. the bird's bill c. the dog's eyes d. the elephant's trunk (16) a. the ship's funnel b. the fence's new coat of paint c. the church's entrance d. the university's main library As shown in the case of [CIRCLE]/[DIAMETER], conceptualization of the part presupposes the concept of the whole. A part of a body like hand is understood with reference to the conceptual domain [ARM], which conjures up [BODY], which further evokes the whole person as the endpoint of the conceptual hierarchy (Langacker (1987: THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 143

147)). The whole person constituting the base of hand is realized by the possessive genitive, John in (14), just as Taylor's analysis predicts. Additionally, this account explains the fact that relational nouns rep- resenting kinship also require domains for their conceptualizations. (17) a. John's father b. Mary's cousin c. Bill's children d. my aunt Nouns like father, uncle or cousin in (17) conceptually presuppose fun- damental notions from within a kinship network. The domain of [FATHER], for instance, would feature specifically two roles, a 'male parent' role and an 'off-spring' role, which could be filled by different real-world individuals on different occasions (Langacker (1987: 184- 185)). It is against these notions that the characterization of relational nouns of this sort are possible and plausible. Possessive genitives make reference to landmarks in this kinship network, making explicit the perspective from which the person designated by the head noun is a father, uncle, or cousin. Once again Taylor's domain-based analysis seems to provide the correct insights. Related to kinship terms are nouns like employer, occupant, etc., in (18). (18) a. John's employer b. the room's occupant c. the house's owner d. the child's guardian The relations in (18) seem to be somewhat far removed from the prototype discussed in section 1. Especially in (18c), the relation be- tween the possessor and the possessed is reversed; the literal possessor owner is realized as the head noun and the house, which is in fact pos- sessed, is the genitive possessor. However, the domain-based analysis still correctly admits the relations expressed in (18) as appropriate read- ings for the possessive genitive construction. The head nouns share a commonality with kinship terms; they also presuppose a relation in which at least two parties are involved. In the case of owner, the do- main is a relation of ownership in which participate the person who owns and the thing which is owned. The noun owner itself refers to the former, while the possessive genitives John's, the room's, and the house's explicitly focus on the latter. However the aim of the con- struction is the same as in the previous cases: to identify their head nouns, by appealing to relationships with the head noun's domain. Lastly, let us consider the remaining cases, which feature the true ownership relation, as seen in (19). (19) a. Mary's necklace b. John's book 144 ENGLISH LINGUISTICS, VOLUME 10 (1993)

c. Sally's car d. her computer Necklace, book and car differ from previous examples like circle, father, or owner in that the relationship expressed by the possessive genitive does not seem truly essential to the semantic characterization of the head noun. In what sense, then, is the relation of true owner- ship reflected in the domains of necklace, book, etc.? Recall that do- mains are complex structures which Langacker assumes to be arranged in hierarchies. In the interest of economy of representation, domains for particular words should surely share common substructures repre- senting more general concepts. For instance, all the head nouns in (19) would probably share a substructure constituting the shared do- main for all consumer goods, and that domain would contain a partici- pant corresponding to the product's owner. Hence, if this substructure within the domain of the head noun is employed, we discover that the relation of ownership is indeed available. Therefore, Taylor's domain- based approach once again yields the correct prediction. The notion of domain can readily explain the reason for the in- appropriateness of the forms in (3), reexamined here in (20). (20) a. the knife's {case/# fork} b. the hat's {owner/# boy} c. the father's {son/#John} We observe a sharp contrast in acceptability which is readily captured by the criterion of presence or absence in the domain of the head. For example, in # the hat's boy, the domain of [BOY] contains no no- tion corresponding to [HAT]. The same kind of explanation may be applied to the other cases. This state of affairs may be confirmed by examining such questions as the following:7 (21) a. Where is the {case/*fork} of this knife? b. Who is the {owner/*boy} of this hat? c. Where is {the son/*John} of this father? These examples underscore the utility of the notion of domain in char- acterizing the set of relations expressible with the possessive genitive construction. Before moving on, let us go back to the case of (19), because it also provides an argument for the prototype-category approach to modeling the meaning of possessive genitives. Suppose one relied only on the

7 The examplein (20b) is due to MichaelT. Wescoat. THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 145

domain-based schema in order to pick out relations expressible with possessive genitives. Since the true ownership relation is relevant to the head nouns in (19) only rather remotely through their identity as consumer goods, one might incorrectly expect the true-ownership read- ing to be somewhat obscure and inaccessible. For this reason we clearly need the notion of prototype category, which provides a place to specify that true ownership is the prototype among POSSESSION relations and therefore the most readily accessible. In sum, the sche- ma specifies the set of relations expressible with the possessive genitive, and the prototype category records their degree of accessibility as POS- SESSION relations.

2.2.2. The Need for a Unique Identifier within Domains Beyond Taylor's insightful domain-based characterization of relations expressible with the possessive genitive construction, there is another equally crucial constraint to the effect that, within the domain, unique identification is required. In other words, a possessive genitive should be an identifier, i. e., a participant role in a domain such that filling that role with a real-world individual would uniquely determine the identity of the individual that fills the role of the head noun. Observe the re- peated examples in (13) below and compare them with (22). (13) a. the circle's diameter b. the circle's radius (22) # the circle's arc The difference between (13a-b) and (22) is that the former succeed in uniquely identifying the head noun while the latter does not. With re- spect to a given circle there can be only one radius. On the other hand, arc in (22) is not uniquely identified, for there are many arcs in a single circle. The establishment of unique reference requires more in- formation to specify the target, as in (23). (23) The length of this circle's red arc is about 3 inches. Another instructive example is the following: (24) a. # the city's road b. the city's roads Ordinarily there are many roads in a city. This makes (24a) inade- quate, for we cannot tell which road is specifically referred to. In other words, the genitive does not serve as a unique identifier in (24a). In (24b), on the other hand, roads as a whole can have the desired kind of uniquely determined relation with city. This suggests that 146 ENGLISH LINGUISTICS, VOLUME 10 (1993) there should be some relation of unique identification between geni- tives and their head nouns. This conclusion is supported in a slightly different way in (25). (25) a. # the city's road (=(24a)) b. the city's best-paved road The superlative provides the head noun with a new, more restricted and detailed domain, thereby guaranteeing a unique correspondence with the possessor. Modification of the head noun supplements the domain, and once the domain is newly set for the head noun, the city can serve as an identifier for road. The crucial point here is that in (24) and (25), the possessor is constant, the city's, and the head noun is nearly constant, always some variant of road. Thus, if there is a con- trast among these examples, we must conclude that the 'city-to-road' relationship is insufficient on its own to license possessive genitives. The factor which saves the acceptable examples is invariably the unique identification imposed upon the head noun once the possessor is determined.8 We are now in a position to define a schema as a semantic principle which picks out the set of relations expressible by possessive genitives. (26) A possessive genitive is appropriate when the possessor may be interpreted as filling a unique identifier role within the domain associated with the head noun. This definition combines both Taylor's domain-based approach to pos- sessive licensing and the uniqueness condition just introduced. Where- as Taylor requires only that the possessor realize some participant with- in the domain of the head noun, we further require this participant to fill a unique identifier role. By the latter term, we designate a partici- pant role within the domain such that filling it would uniquely deter- mine the identity of the head noun. By insisting on unique identification as a condition on relations ex-

8 A reviewer has suggestedto me that John's daughter in John's daughter is in college, or my friend in I stayed at my friend's house does not necessarilyrefer to John's only daughter or my only friend. It is true, but my principal claim here is that a possessiveserves as a unique identifier of the referent of the head noun. It does not necessarilymean that there should be only one daughter or friend in the real-world. A participant role in the head noun can be filled by different real- world individualson different occasions. In the cases above, it is sufficient that one of John's daughters or my friends is uniquelyreferred to in that situation. THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 147 pressible with the possessive genitive construction, we are taking a different position from that advanced by Taylor. This is not to say, however, that the importance of unique identification has gone com- pletely unnoticed in Taylor's work. He mentions it in terms of exclu- sivity, but he seems to regard it basically as only one of six or more prototype conditions for the category of POSSESSION as an experien- tial gestalt. He does point out the prominent status of exclusivity among these conditions but does not argue for it by means of concrete examples. In the following sections, we shall show empirically that this characteristic actually plays a crucial and significant role in extend- ing the application of possessive genitives to non-prototypical meanings.

3. Cases Incompatible with the Schema Thus far we have proposed a schema with essentially two compo- nents: (i) the possessor must fill a participant role in the domain of the head noun, and (ii) that role must be of a sort that will uniquely deter- mine the identity of the head once filled. In the present section we will see how these two parts of the schema serve to rule out some sig- nificant classes of unacceptable possessive forms.

3.1. Inappropriate Application of Domains One major factor related to infelicity of possessive genitives is the choice of the cognitive domain used as the base for the head noun. Applying inappropriate domains to head nouns yields interpretation failure. For instance, let us consider the following examples of tem- poral possessive genitives. (27) a. # the year's woman b. # last year's singer c. # yesterday's building d. # tomorrow's Patricia (28) a. yesterday's work b. last year's news c. tomorrow's conference d. Monday's snowfall e. the morning's golf f. [The L. A. riot] could lose him [=ex-president Bush] November's election. (BBC Evening News 5/8/1992) The difference between (27) and (28) lies in the domains concerned: the concept of time is involved in the base of head nouns in (28), but not in those in (27). To explain the difference, let us examine the na- ture of entities and how they interact with the concept of time. Givon (1979) asserts that entities satisfy a 'time-stability criterion' 148 ENGLISH LINGUISTICS, VOLUME 10 (1993)

which is experiential in nature. The Time-Stability Criterion For Entities: An entity x is identical to itself if it is identical only to itself but not to any other entity (y) at time a and also at time b which directly follows time a. (Givon (1979: 320)) According to the criterion, nouns and verbs usually lie in a polar opposition: nouns typically denote entities (i. e., things), remain stable and do not change their identity throughout temporal transition; verbs refer to actions or events which "have most typically only existence in time (Givon (1979: 321))." The nature of time-stability mentioned above gives us a reason why things, typically expressed by nouns, cannot be identified through the notion of time. (29) a. the book on the shelf b. # the book at three o'clock c. the book I read yesterday As the examples in (29) illustrate, there is no significance in referring to a certain time in order to recognize things like a book, as in (29b), unless some temporal notion of actions or events is interposed, as in (29c). The same explanation can be applied to (27). The examples there are infelicitous, because on a context-free reading, head nouns like woman or singer do not typically evoke any relationship with an activity or event that would make temporal identification appropriate. Yet, the identification of things in terms of time necessarily presupposes the notion of events or activities where passage of time is involved. As for (28), notions of this kind are presupposed by the head nouns. We can easily associate typical event-related readings with the nouns in (28) based on encyclopedic knowledge: work is something we do; news is something which is reported or broadcast and which we hear or read; etc. These nouns have particular, culturally fixed domains needed for us to conceptualize their meanings.9 They evoke events, which are typi- cally expressible with verbs, as Givon (1979) states. This means these notions are identifiable in terms of time. Thanks to these relational notions involving temporal transition, identification of these nouns by

9 This approximatelycorresponds to Idealized Cognitive Models in the sense of Lakoff (1987:Ch. 4). See Lakofffor details. THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 149 time becomes feasible. According to Givon (1979), as we have seen, verbs can be charac- terized with respect to the temporal domain, while nouns are usually not. Then how about abstract nouns derived from verbs? (30) a. Last year's immigration of a group of boat people was a serious problem. b. This semester's decision to assign more difficult prob- lems led to disaster. c. The 1961 theme is the Dakota Territorial Centennial, with the pictures including the Lewis and Clark expedi- tion. [...], and today's construction of large Missouri River reservoirs. (Brown Corpus PC15 S35 (my emphasis)) As indicated in (30), abstract deverbal nouns can be identified in terms of time, for they denote events conceptualized as things but which are originally characterized by their temporal anchoring points. As for nouns derived from stative verbs, like resemblance, which are not re- lated to temporal passage, they are not so natural when identified in terms of time only. (31) # today's resemblance The choice of infelicitous domains thus gives rise to low acceptability in possessive genitive constructions. All the examples examined in this section are instances which do not satisfy the domain-oriented compo- nent of the schema extracted in the previous section. Next, let us con- sider some effects of the unique identity condition.

3.2. Referentiality and Possessives It is said that pronouns or proper names usually cannot be modified by possessive genitives, as shown in (32). (32) a. John's girl b. # John's Mary c. # John's Then what makes these expressions inappropriate? We will derive the answer to this question by means of the schema characterizing accept- able possessive-genitive relations formulated in (26). One of the relevant factors that make (32b, c) infelicitous may be a pragmatic one. As forms which uniquely identify head nouns, the ap- plication of possessive genitives to proper names and pronouns is generally quite redundant; the latter forms are referential expressions 150 ENGLISH LINGUISTICS, VOLUME 10 (1993) whose referents are already determined. Thus they need not be iden- tified with reference to anything else in a usual, neutral context. The following data further amplify this conclusion. (33) a. [A girl in the room] is playing the piano. b. # [Mary in the room] is playing the piano. c. # [She in the room] is playing the piano. Note that the referentiality of girl in (33a) is low by itself, so it is com- patible with a locative postmodifier.10 The exact contrary is true of the proper name and pronoun in (33b, c). Having confirmed that proper nouns and pronouns reject restrictive modification, let us reconsider the examples in (32) with respect to the schema in (26). Notice that the schema does not actually rule out such cases as this. Rather, proper names or pronouns with possessive geni- tive modifiers could be regarded as vacuously satisfying the schema, since unique identification is achieved even without the support of the possessor. We claim, therefore, that examples of this kind are infelici- tous because the addition of the possessive identifier is utterly redun- dant in respect of identification. Note that the infelicity due to redundancy disappears in some cases involving proper names, if such forms come to be ambiguous because two or more like-named people are being discussed. (34) A: Do mean the Mary who is Bill's girlfriend? B: No, I mean John's Mary. Here, Mary needs identification and (32b) may be used felicitously.11 Note that the foregoing observations about infelicities actually consti- tute the basis for an argument in favor of the unique identification con- dition on possessives. The fact that the application of possessive geni- tive to pronouns constitutes an utter redundancy is due entirely to the presence of the uniqueness condition in the schema in (26). Conse- quently the existence of this component of the definition is necessary to make the remainder of the analysis go through.

10 Needless to say, the referentiality of nouns is a matter of degree and has no clear boundary between high and low. 11 Example (34) is brought to my attention by MichaelT. Wescoat. THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 151

4. Extension from Prototype: Non-Prototypical Examples and the Role of Context Most cognitive grammarians who examine possessive genitives, e. g., Taylor (1989b), Durieux (1990), and Nikiforidou (1991), mainly discuss phrasal level phenomena. However, the living, dynamic aspects of the category of relations expressible with possessive genitives is noticeable only by paying attention to the sentence or paragraph level. This mac- roscopic perspective makes it clear that contextualization interacts with possessive genitive expressions to save various forms which otherwise would be inappropriate. In the present section we will take some of the seemingly inappropri- ate examples we have seen above and show that contextualization interacts with them so as to create a relation to satisfy the schema for possessive genitives. The purpose of this section is to show that a cer- tain number of rather sharp contrasts in acceptability arise from contex- tual variations and that these contrasts may be successfully modeled by appealing to the schema in (26), whereas the prototype-category approach would be powerless to handle them on its own. Three types of contexts will be introduced here; while not exhaus- tive, they do illustrate typical cases where contextualization markedly changes acceptability.

4.1. Modification with Superlatives as a Means of Modifying Domains We have already examined briefly one way to save possessives through contextualization: this involves supplementing some notion lacking within the domain of the head noun by modifying it with super- latives. Superlative restrictive expressions can facilitate access to the notion of time far more easily than can simple nouns, as shown in (35). (35) a. A: He is a singer. B: # # Of which year? b. A: He is (one of) the most popular singer(s). B: Of which year? The addition of a superlative to singer requires a new domain where the situation described by the superlative holds; hence, a temporal no- tion is selected and realized. In other words, modification of this kind helps to create a relation of unique identification between the posses- sive genitive and its head noun within the latter's domains, so that ex- amples which were originally unacceptable come to fit the schema de- 152 ENGLISH LINGUISTICS, VOLUME 10 (1993) scribed in (26). Compare the following examples with (35): (36) a. the year's # (most outstanding) woman (Taylor (1989b: 676)) b. last year's # (most popular) singer c. yesterday's # (most attractive) lady at the party d. the 18th century's # (greatest) concern While the schema in (26) readily makes the correct distinction between the examples in (36) with superlatives versus those without, it is very hard to see how any approach that employs only prototype categories could match this performance. The latter sort of analysis could predict the sharp contrasts in acceptability above only by positing a rather odd prototype category in which the relations between possessor and pos- sessed expressed by the examples with superlatives are close to the prototype of true ownership, while the relations expressed by the forms without superlatives are significantly farther removed from that pro- totype. However, positing a gap in 'prototypicality' between the ex- amples with superlatives and those without broad enough to account for the sharp contrast in acceptability seems totally unjustified. Re- call, after all, that the basic participants are the same in each case, so the difference between them should not be so terribly great. A similar argument may be constructed with expressions of location. (37) a. America's # (most exciting) city b. the country's # (best behaved) long-term prisoner c. She was said to be the world's # (most popular) singer. d. Made of the world's # (toughest) unbreakable plastic, Melamine dinnerware comes in almost 400 different patterns and dozens of colors. (Brown Corpus PE14 S30 (my emphasis)) e. In 1846 Matthew, B. Goodwin, jeweler and watchmak- er, became the town's # (first) telegrapher ... (Ibid. PJ56 S17 (my emphasis)) f. The bond issue will go to the state courts [...] and then the sales will begin and contracts let for repair work on Georgia's # (most heavily traveled) highway. (Ibid. PA01 S56 (my emphasis)) g. He became, by his own ability, Britain's # (best-known) man of property. For instance, the example in (37a) originally does not satisfy the sche- ma, in that America cannot serve as a unique identifier for only a sin- THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 153 gle city within it, and in (37b) there may be more than one long-term prisoner in a country. Both cases violate the unique-identity condition given in the schema. With superlatives, however, the domains of their head nouns are specified enough for their location to play the role of unique identifier. It is hard to see, here again, how to handle these data in a theory that employs only prototype categories. Let us note also that non-superlative modifiers may also produce similar, sharp contrasts in acceptability that support the argument above. (38) a. this country's # (only) university (Quirk et al. (1985: 325) (my emphasis)) b. Kenny Lane of Muskegon, Mich., world's # (seventh ranked) lightweight, had little trouble in taking a unani- mous decision over Rip Randall of Tyler, Tex., here Monday night. (Brown Corpus PA 13 S27 (my emphasis)) If the expressions in (36)-(38) are deprived of superlatives or related kinds of modifiers, the acceptability of all the expressions decreases dramatically in a manner that cannot be modeled plausibly with just a prototype category.

4.2. Contextual Supplementation of Domain Participants Let us now consider contextual enrichments that affect the inventory of participants in the domain of the head noun. Imagine, on an ab- stract level, any sort of discourse context that would cause the domain of a noun to feature a new participant that would have been absent on a context-free reading; in accord with the schema in (26) such a change would increase the potential range of acceptable possessors. Indeed, if a possessive genitive employed such a contextually-supplied domain participant as the possessor, we would predict a sharp contrast between the contextually enriched and context-free readings. Felicitously, empirical investigation reveals precisely this state of affairs. Recall that the awkwardness of # today's resemblance is due to the absence of time as a participant in the domain of the head noun. Con- sequently, this form is incompatible with the schema: the possessor corresponds to no unique identifier within the domain of the head noun. However, today's becomes appropriate when additional context is supplied, which shows, in this case, that the situation described is momentary and transient. 154 ENGLISH LINGUISTICS, VOLUME 10 (1993)

(39) a. # today's resemblance (=(31)) b. Today's (perceived) resemblance may fade as the two of them age. (39b) indicates that the present resemblance is only a temporary fea- ture, and the transitory aspect of the present situation receives focus of attention. It seems plausible, therefore, to assume that time has be- come a participant in the domain of resemblance. Furthermore, this sort of contextualization makes it relevant to identify the referent of the head noun by selecting one point on a temporal scale. Under the assumptions, the schema in (26) predicts the contrast in (39). In con- trast, an approach that dealt solely with prototype categories would have to posit a vast difference in the 'prototypicality' of the two possessor/possessee relations expressed in (39), in order to predict the sharp contrast in acceptability. However, such a wide divergence in degree of prototypicality seems totally unmotivated. Let us examine another example of the same kind. In (40a), on a context-free reading, the domain of Edinburgh contains no participant role that could be filled distributively by a group of people. However, the context in (40b) adds a new participant to the domain of Edin- burgh: the city becomes an abstraction in the mind of a person who conceptualizes it. (40) a. # other people's Edinburgh(s) b. [...] many times throughout her life Sandy knew with a shock, when speaking to people whose childhood had been in Edinburgh, that there were other people's Edin- burghs quite different from hers, and with which she held only the names of districts and streets and monu- ments in common. Similarly, there were other people's nineteen-thirties. (M. Spark, The Prime of Miss Jean Brodie (my emphasis)) Contextualization makes (40a) fitting for the schema by adding an ex- periencer participant role to the domain of the head nouns. It is hard to see how approach based only on prototypes could capture the con- trast in (40). Another example illustrates the same contextual effect. (41) a. (#)the room's teacher The room in (41a) is usually interpreted as a location, but typically the relation between the room and a teacher is an accidental one: the teacher happens to be in that location at that time, so the room cannot THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 155 serve as a unique identifier for teacher. Hence the schema in (26) is not satisfied. When context supplies unique-correspondence, however, (41a) becomes acceptable. (41) b. One day in Fall, in a small school in California, the SAT was being given, and all of the tiny classrooms were filled with nervous, college-bound teenagers. To each classroom was assigned a teacher who was to supervise the test. One room's teacher happened to be Mary Anderson, and it was on that day that her adven- ture began. The context revises the domain of teacher to reflect the information that on this occasion each teacher has a room assignment; thus, the do- main of teacher comes to have a participant role for a classroom. Since the classroom role is a unique identifier, the schema in (25) is satisfied, thanks to the effect of contextualization. While the schema- based approach adequately handles the contrast between (41a) and (41b), to attempt to account for the same data with only a prototype category would lead to the sort of problem encountered several times above.

4.3. Contrastive Context as a Means of Modifying Domains The most powerful means of modifying domains is to make use of contrastive context. It explicitly introduces a new participant role to a domain. Recall the example in (34) in 3.2., where John's Mary be- comes completely acceptable. (34) A: Do you mean the Mary who is Bill's girlfriend? B: No, I mean John's Mary. Upon saying the Mary who is Bill's girlfriend, A alters the domain of Mary; subsequently the role of boyfriend (i. e., 'the person for whom Mary is a girlfriend') is included in the domain, though it surely needn't have been there before. Thus, when B utters John's Mary there is a participant role available for John to fill and thereby identify Mary. The need to differentiate two like-named people thus leads to domain expansion in order to find distinguishing criteria, and once this is accomplished the previously infelicitous form John's Mary is in con- formity with the schema in (26). In contrast an analysis based purely on a prototype category would be forced to the implausible conclusion that the relation that holds between someone named Mary and her boyfriend is vastly closer to the prototypical POSSESSION relation on 156 ENGLISH LINGUISTICS, VOLUME 10 (1993) an occasion when like-named people are being discussed than it is otherwise. Let us look at another effect of this kind of contrastive use. We have seen that, because of its high time-stability, there is no participant role concerning time within the domain of a noun like girlfriend. However, the contrastive use alters the acceptability of temporal pos- sessors. (42) a. # today's girlfriend b. He always brings his girlfriend to football games, but he is quite a playboy, so today's girlfriend may be different from yesterday's. With contrastive use, a temporally transitory aspect with respect to girl- friend receives focus of attention: the domain of girlfriend is modified so as to include a participant role for the time at which the relation obtains. Accounting for the contrast in (42) would once again push an analysis based on prototypes alone to an implausible extreme. Similar examples include the following: (43) a. Today, children are growing up physically far earlier than their parents did. [...] Today's parents cannot seem to accept that the girl who starts menstruating at eleven is not super-advanced. (LOB SF17 P21-P25 (my emphasis)) b. Many say it is because today's child is much better fed than her ancestors. (Ibid. SF17 P37 (my emphasis)) Contrastive contextualization emphasizes temporal transition, which adds participant roles corresponding to time to the domains of parents and child. Notice, however, that what the contrastive context does is mainly to modify domains; in order for a given expression to be perfectly accept- able, the unique-identity condition must be observed as well. (44) a. ?Maintenance on the city's road must be just as difficult and expensive as that on the county's. b. Since the two roads run through the same type of mar- shy terrain, maintenance on the city's road must be just as difficult and expensive as that on the county's. In the contrastive use of city's road and county's road, city and county come to be newly added participants in the domain, but this is not enough for them to be unique identifiers. Owing to the information, contextually supplied in advance, that exactly two roads are under dis- THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 157 cussion, it is guaranteed that city and county will be unique-identifiers, since the multiplicity of other roads that belong to the city or county have been eliminated from the domain of roads. Here, the schema- based analysis correctly predicts the acceptability of these cases, for which an approach based solely on prototype categories could not pro- vide a plausible analysis. Lastly, let us look at some additional cases with high referentiality. Generally, the objects contrasted or compared should be basically under the same category. If we follow this principle, then even nouns with high referentiality, when situated within a contrastive context, can take a possessive genitive form. Recall that proper names, like the Statue of Liberty or the Bay Bridge, have high referentiality all by themselves. It is redundant, therefore, to identify them with genitives of location. Contrastive contextualization, however, can make their possessive expressions more natural, as in the (b) examples of (45)- (46). (45) a. (?) America's Statue of Liberty b. America's Statue of Liberty is not so famous as France's Eiffel Tower. (46) a. (?) San Francisco's Bay Bridge b. The first superhighways-New York's Henry Hudson and Chicago's Lake Shore, San Francisco's Bay Bridge and its approaches, a good slice of the Penn- sylvania Turnpike-were built as part of the federal works program which was going to cure the depres- sion. (Brown Corpus PJ60 S71 (my emphasis)) In these cases the contrasted head nouns are different, but they can be categorized as the same: in (45) the contrasted head nouns have in common that both are famous architectural monuments specifically named the Statue of Liberty and the Eiffel Tower; in (46) what is com- mon among the proper names is that they denote what are categorized as superhighways. Under the present contrastive contexts, the domain for each monument or superhighway comes to contain a location role. To summarize the discussion so far, contrastive contextualization modifies domains by explicitly supplying and focusing a new participant role which may support a possessive genitive, in the sense of making it possible for that possessive to satisfy the schema in (26). Conversely an analysis depending only on a prototype category cannot handle these contrasts in any satisfactory way. 158 ENGLISH LINGUISTICS, VOLUME 10 (1993)

5. Conclusion

In this paper we have attempted to make use of the insights afforded by the prototype-theoretic approach to the analysis of possessive geni- tives, while adding to that a new mechanism to remedy certain empiri- cal inadequacies. The prototype-based model is capable of making va- rious useful predictions about subtle differences in the accessibility of alternative readings for possessive genitives. However, a semantic constraint implemented in the form of a schema is necessary to model sharp contrasts in acceptability that may arise between examples even though they seem to express relations that resemble the prototype of true ownership to more or less the same degree. We have seen many examples of the latter type here, and probably the most interesting and convincing among these were the data presented from a macroscopic point of view in section 4; these change dramatically in acceptability with simple shifts. in discourse context. While the approach that takes advantage of a schema-based constraint could readily handle these cases of living, dynamic interaction, any analysis that uses only a prototype category would be forced to posit implausibly wide gaps in prototypicality between relatively similar examples in order to account for the sharp contrasts in acceptability. Hopefully future research will test whether a distribution of labor between prototype categories and semantic constraints will be equally applicable to areas of meaning beyond the possessive genitive construction.

REFERENCES

Deane, Paul (1987)“English Possessives, Topicality, and the Silverstein Hierar-

chy,”BLS 13, 65-76. Durieux, Frank (1990) The Meanings of Specifying Genitive in English, Antwerp Papers in Linguistics 66. Givon, Talmy (1979) On Understanding Grammar, Academic Press, New York. Grice, H. Paul (1975)“Logic and Conversation,”Syntax and Semantics 3, ed. by Peter Cole and Jerry Morgan, 41-58, Academic Press, New York.

Hawkins, Roger (1981)“Towards an Account of the Possessive Constructions:

NP's N and the N of NP,”Journal of Linguistics 17, 247-269. Kempson, Ruth (1977) Semantic Theory, Cambridge University Press, Cam- bridge. Lakoff, George (1987) Women, Fire, and Dangerous Things, University of Chi- THE ANALYSIS OF ENGLISH POSSESSIVE GENITIVES 159

cago Press, Chicago. Lakoff, George and Mark Johnson (1980) Metaphors We Live by, University of Chicago Press, Chicago. Langacker, Ronald (1987) Foundations of Cognitive Grammar 1, Stanford Uni- versity Press, Stanford. Nikiforidou, Kiki (1991)“The Meanings of the Genitive: A Case Study in Semantic Structure and Semantic Change,”Cognitive Linguistics 2, 149- 205. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik (1985) A Comprehensive Grammar of the English Language, Longman, London. Taylor, John R. (1989a) Linguistic Categorization: Prototypes in Linguistic Theory, Clarendon Press, Oxford. Taylor, John R. (1989b)“Possessive Genitives in English,”Linguistics 27, 663-

686.

Woisetschlaeger, Erich (1983)“On the Question of Definiteness in‘An Old

Man's Book’,”Linguistic Inquiry 14, 137-154. Yamanashi, Masaaki (1988) Hiyu to Rikai (Metaphor and Understanding), Tokyo Daigaku Syuppan-kai, Tokyo.

2-9-23 Hiyoshidai Takatsuki-shi Osaka 569