Hyponymy
Hyponymy (Greek hypo- ‘under’) is the lexical relation described in English by the phrase kind/type/sort of. A chain of hyponyms defines a hierarchy of elements: sports car is a hyponym of car since a sports car is a kind of car, and car, in turn, is a hyponym of vehicle since a car is a kind of vehicle. Other examples of hyponym hierarchies include
• blues – jazz – music,
• ski-parka – parka – jacket,
• commando – soldier – member of armed forces,
• martini – cocktail – drink and
• paperback – book.
A standard identification procedure for hyponymy is based on the notion of class-inclusion: A is a hyponym of B if every A is necessarily a B, but not every B is necessarily an A. For example, every car is a vehicle, but not every vehicle is a car, since there are also buses, motorbikes and trucks. Hence, car is a hyponym of vehicle. Furthermore, hyponymy is usually taken to be transitive: if A is a hyponym of B, and B of C, then A is a (more remote) hyponym of C.
As we will see, hyponymy is a major semantic relation in the grammar of many languages. Furthermore, a particular type of hyponymy, taxonomy, discussed in the next section, is an important aspect of the way we talk about the natural world.
Hyponymy also has a crucial communicative function. It often happens that we are unable to retrieve the most accurate, precise term for the referent we have in mind. At other times, mention of the most precise term would be needlessly informative and thus violate one of the pragmatic constraints which often seem to be operative in communication (see 4.4). In cases like these, the existence of a term (referred to as a hyperonym) further up the hyponymic hierarchy allows reference to be accomplished. Thus, wanting to mention the fact that my brother has started learning to play the sackbut, but momentarily unsure of the name of this instrument (or worried that my interlocutor will not know what I’m talking about), I can simply say my brother is learning a weird musical instrument, using the hyperonym musical instrument to refer to its hyponym sackbut. The possibility of referring at a number of different hierarchical levels is also crucial for cross-cultural communication. At specific levels of categorization, languages often lack exactly corresponding terms: Japanese wasabi, for example, isn’t accurately translated into English by any of the choices mustard, chutney, vinaigrette, etc. But in order to explain what it is, a combi nation of modifier and hyperonym can always be found: thus, wasabi can be referred to as a horseradish condiment. Similarly, the names of the various female outer garments often worn in Muslim countries lack precise English equivalents. But by adding modifying adjectives to appropriate superordinate terms, translations can be given: khimar ‘long veil’, chador ‘full-body cloak’.
The concept of hyponymy can be made intuitively clear on the basis of examples like those given above, and hyponyms in other languages are often easy to identify: in Tzeltal (Mayan, Mexico), for example, chenek’ ‘beans’, ixim ‘corn’, ti’bal ‘meat’ and wale’ ‘sugarcane’ are among the obvious hyponyms of we’lil uch’balil ‘food’ (Berlin 1992: 186). But as soon as one tries to make the notion of hyponymy explicit various problems are encountered. The definition of hyponymy as class-inclusion, for example, seems to be too powerful, since there are many cases which fi t the class inclusion definition which could not be described with the formula kind/ type/sort (Cruse 1986). For example, as noted by Wierzbicka (1984), every (male) policeman is necessarily someone’s son, and not every member of the category ‘someone’s son’ is a policeman, but this doesn’t mean that a male policeman is a ‘kind of son’, and we would not want to describe the relation between male policeman and someone’s son as an example of hyponymy.
Even the linguistic definition of hyponymy as the kind/sort/type relation admits instances which seem remote from the standard exemplars of hyponymy because they do not defi ne a hierarchy. In English, for instance, one might very well utter the sentences in (13), for example in the context of an explanation to someone unfamiliar with the word involved:
In none of these cases, however, would we wish to claim that the nouns related by the phrase a kind of are hyponyms. Kind of, in other words, seems to have a variety of values in English, not all of which correspond to the strict class-inclusion model: in (13), kind of serves to establish a comparison between two terms without introducing any claim of class-inclusion of the sort which could defi ne a hierarchy. This isn’t such a problem for determining hyponymy in our native language, but it poses a particular challenge when the lexical structure of an unfamiliar language is under investigation. If English kind of seems ambiguous between a ‘strict hyponymy’ reading and a looser, comparison reading, how can we decide whether the equivalent of kind of in an unfamiliar language is being used in a strict or a loose sense? In Tok Pisin (English-based Creole; Papua New Guinea), for example, we find the translation equivalent of kind of, kain, used in the following definitions:
Judging from the translated definitions, the words concerned are the Tok Pisin translations of ‘grass hut’, ‘nest’ and ‘tent’. Are they, however, hyponyms of TP haus? Without an appreciation of the range of uses of kain in TP, we are unable to tell. (The mere fact that the TP definienda contain the word haus is no evidence: in English, a publishing house, a doll’s house and a Royal house are not kinds of houses: the first is a kind of company, the second a kind of toy, the third a kind of family.)
Hyponymy is often exploited by languages with classifier systems (Allan 1977; Aikhenvald 2000). In noun-classifying languages, the noun phrase obligatorily contains a morphological element (the classifier) whose choice is determined by semantic features of the referent of the head noun. Often, the semantic basis of this classification is implicitly hyponymic, with a given classifier naming a superordinate class of which the head noun is a particular kind. Thus, noun phrases in Jacaltec (Mayan; central America: Aikhenvald 2000: 285) contain a classifier morpheme which assimilates the noun to a broader set of superordinate kinds or classes. For instance, the person ‘John’ and the animal ‘snake’ are implicitly represented in (15) as hyponyms of the classes ‘person’ and ‘animal’ through the use of the classifiers naj, which classifies the noun as a human, and no7, which classifies it as an animal (Aikhenvald 2000: 82):
The number of classifiers may often be quite high: a non-human noun, for example, will be accompanied by one of the eleven following classifiers (Aikhenvald 2000: 285), depending on the semantic kind of which it is a hyponym:
In Burmese (Tibeto-Burman, Myanmar), classifiers are based on the function which the noun fulfils:
Sometimes it is the verb which takes the classifier. This is the case in Ojibway and Cree (Algonquian, Canada), for instance, where verb classifiers categorize the referent of the verbal argument in terms of its shape, rigidity, size, structure, position and animacy, as in (18):
This classification relies on implicit hierarchies of long, one-dimensional and flexible things, flat and round things, etc. Implicit hyponymic structure is therefore an important principle in the grammatical structure of classifier languages. Elsewhere, however, it may be the case that hyponymic structure is minimal or even absent for certain lexemes. As Jackendoff (2002: 345) points out, the hierarchical connections of junk and puddle would not seem to be an important part of their meaning.
