II.1 Taxonymy as a sub-division of hyponymy
II.2 On the transitivity of taxonymy
III. Meronymy – the part-whole relation
III.1 On the transitivity of the part-whole relation
IV.1 Differences between taxonomies and meronomies
IV.2 The borderline between the two types
Everyday we are confronted with lexical hierarchies in our correspondences, without actually thinking about it.
In this paper I will present two kinds of branching lexical hierarchies: taxonymy and meronymy, which both structure the vocabulary of a language.
We will further see that semantic relations, which are defined in terms of logical relations, underlie lexical hierarchies.
The problem is that it is not possible to give general, satisfying definitions of suitable test frames for the two kinds, but I will present different approaches that come close to final conclusions.
Another aspect I am interested in is the transitivity of taxonymy and meronymy – are they transitive, and if yes, in how far do we have to restrict transitivity?
In the final part of my paper I will draw a conclusion, containing a summary of the distinctions and similarities between taxonymy and meronymy.
If we talk about hyponymy, we take a closer look at the structure of lexical hierarchies.
If we want to refer to something, e.g. a dog, we have several possibilities to express this: We could say ‘spaniel’ (only, of course, if we talk about a spaniel), ‘dog’ or ‘animal’. It becomes clear that these lexical items are of “different levels of specificity” (Cruse 1975:153), and what we finally say depends on our point of view, whereas no one will disagree that ‘spaniel’ is more specific than ‘dog’, which itself is more specific than ‘animal’.
In ordinary dictionaries we will find an alphabetical structuring of the vocabulary of a language, but there are as well dictionaries that are conceptual, referring to lexical hierarchies as the one mentioned above. The first dictionary of this kind was Roget’s Thesaurus of English Words and Phrases, published in 1852, followed by other similar works like Dornseiff’s thesaurus for the German language (1933).
Various structuring principles can be found in all thesauri; one of the most important is taxonymy, a sub-division of hyponymy, which we will examine in I.1.
Without doubt all “semantic relations which underlie lexical hierarchies are defined in terms of logical relations” (Cruse 1975:26). If we take the example mentioned above we see that there is a logical relation between spaniel, dog and animal.
Lyons explains the relation between a specific, subordinate lexeme and a more general, superordinate lexeme with the help of class-inclusion (Lyons 1977:291): We can say that X includes Y, if X is the class of animals (flowers, …) and Y is the class of dogs/cats/horses/… (tulips/dahlias/roses/ …). In this case we talk about a relation of hyponymy. Furthermore he defines hyponymy on the basis of logical relations between sentences (Lyons 1977:292). He points out that syntactically identical sentences containing the items in parallel syntactic positions form good test frames for a relation of hyponymy.
Let us consider the following example: ‘It is a horse’ implies ‘It is an animal’; we can say that ‘horse’ is a hyponym of ‘animal’. Moreover, other hyponyms of ‘animal’ can be found: dog, cat, elephant,…. Those will be called co-hyponyms in the following.
Since co-hyponyms stand in parallel syntactic positions, they have to be of the same class of word (Lyons 1977:296).
To elucidate the relation of hyponyms, I will present some more examples:
text: drama, essay, novel, …
fruit: apple, banana, orange, grape, …
instrument: piano, trumpet, flute, drum, ...
All the examples work in the test:
‘Julie is reading a novel’ implies ‘Julie is reading a text’,
‘I ate an apple ’ implies ‘I ate a fruit’,
‘It is a flute’ implies ‘It is an instrument’.
The test does not only work for nouns, but as well for verbs (‘John punched Bill’ implies ‘John hit Bill’; ‘punch’ is a hyponym of ‘hit) and for adjectives (‘It is scarlet’ implies ‘It is red’; ‘scarlet’ is a hyponym of ‘red’).
Although this test seems to be a guarantee for hyponymous relations, there are two major problems:
1. It is difficult to formulate rules “determining whether, and in which direction, implication holds between two sentences that differ (…) only in respect of the specificity of a single lexical item” (Cruse 1975: 27). If we take the latter examples we find out that not all identical sentences will give proof of a relation of hyponymy:
‘Julie only likes novels’ does not imply ‘Julie only likes texts’,
‘I don’t like apples’ does not imply ‘I don’t like fruits’,
‘He doesn’t play the flute’ does not imply ‘He doesn’t play an instrument’.
Cruse says that “Lyons’ definition is useless as it stands” (Cruse 1975:27). In the definition we additionally have to mention that the logical structure of the sentence frame plays an important role.
2. The second problem is that “pairs of words that are not hyponymously related may give rise to implication (Cruse 1975:27):
’The scar is on John’s elbow’ implies ‘The scar is on John’s arm’, although ‘scar’ and ‘arm’ do not form a relation of hyponymy. This becomes obvious in another sentence:
‘It is an elbow’ does not imply ‘It is an arm’.
The cases mentioned in 1. and 2. must be filtered out, which is very difficult, since there are no satisfactory definitions of hyponymy.
It becomes clear that the simpler a test frame is in its structure, the better it works. Cruse for instance uses the following test frame to test relations of hyponymy (Cruse 1986:89):
X (be) Y.
All examples I’ve mentioned so far work in the test:
‘A dog is an animal’,
‘A tulip is a flower’,
‘A novel is a text’,
‘An apple is a fruit’, …
We can say that this is the only test frame that doesn’t leak.
 for more information on the different lexical levels see Cruse (1986:153ff)
 Lyons defines hyponymy and incompatibility as the most fundamental paradigmatic relations of sense in terms
of which the vocabulary is structured. (1968:463)
 more examples can be found in Cruse (1986:89ff)