Wierzbicka (1972:21) lists among object words sea, river, field, wood, cloud, mountain, wind, table, house, book, paper, bird, fish, in-sect, plant, animal, cat, apple, rose, birch, gold, salt, and so on —a very ‘open’ series, indeed, which reminds one of the open list of ‘natural kinds’ conjured up by the theories of ‘rigid designation’ (Kripke 1972; Putnam 1975). But, once one has decided to go on in this direction, the list of primitives cannot be a finite one. Moreover, the idea of a list of semantic primitives is devised in order to conceive of a dictionary-like competence free of any commitment to world knowledge, but, if one takes the option (b), then the dictionary competence is entirely depen-dent on the world knowledge.
(c) The primitives are Platonic ideas. This position is philosophically impeccable, but there is a historical (and therefore empirical) inconveni-ence: not even Plato succeeded in limiting in a satisfactory way the sys-tem of these universal and innate ideas. Either there is an idea for every ‘natural kind’, and the dictionary is not finite, or there are few very general ideas (One, Good, Multiplicity, and so on), and they do not succeed in distinguishing the meaning of any single expression.
One can, however, conceive of a fourth and more theoretical way. Suppose there is the possibility of establishing a system of primitives such that, by virtue of their systematic relationship, they must be finite in number. If one’s mind succeeds in doing this, this can be taken as proof that such a systematic arrangement in some way ‘mirrors’ the struc-ture of the human mind (and probably also the structure of the world).
Fortunately, we have a good example of such a system: it is repre-sented by a purely lexical system of hyponyms and hyperonyms organ-ized in the format of a tree such that every n-tuple of hyponyms postu-lates a single hyperonym, and every n-tuple of hyperonyms becomes an «-tuple of hyponyms of a higher single hyperonym, and so on, until the point where, irrespective of the number of hyponyms to be classified at the lower row of the tree, the tree necessarily tapers at a single upper-most node. Figure 2.2 represents a tree of this kind by simply reorganiz-ing the terms provided by Hjelmslev. One can say that ewe contains or comprises ‘sheep’ and (by a transitive property of this classification) con-tains or comprises ‘animal’. One can also say that this tree represents a system of meaning postulates in the sense of Carnap (1974). In fact, if the form of a meaning postulate is
(x) (Sx Ax),
the fact that x is a sheep postulates the fact that x is an animal and this is a sheep entails this is an animal.
animal
/ \
sheep human
| | | |ewe ram girl boy
FIGURE 2.2
A set of meaning postulates is, however, different from the system of Figure 2.2. Carnap’s formula holds even thought stands for raven and A stands for black. According to this meaning postulate, if this is a raven, this is black is an instance of analytic truth, and, if there was not a meaning postulate establishing that sheep are animals, this sheep is an animal would be an example of synthetic or factual truth. A set of meaning postulates is established on ‘pragmatic’ grounds (cf. Lyons 1977:204) and does not distinguish between encyclopedia and dictionary (see Carnap 1947)·
The system of Figure 2.2 represents, on the contrary, an ordered set of meaning postulates, because it is hierarchically structured; for this reason it must be finite. One can think that it can be so because the way in which a lexicon of a natural language establishes relationships of hypo/ hyperonomy reproduces some (as yet mysterious) structure of the human mind. Fortunately, one can disregard such a tremendous question. In any case, the system of Figure 2.2 (even were it ‘true’ or ‘natural’ or ‘universal’) is not an instance of a ‘powerful’ dictionary. Its inconveni-ences are the following: (a) it does not say what sheep or animal means (once again it does not explain the meaning of figurae); (b) it does not help one to distinguish a ram from a ewe, since both are sheep and animals; (c) it does account for such phenomena as hyperonorny hyponymy, meaningfulness and anomaly, redundancy, analytic truth’ contradictoriness, inconsistency, containment, and entailment, but ц does not account for such phenomena as synonymy, paraphrase, and semantic difference.
To conclude, the tree of Figure 2.2 cannot provide the means for giv-ing definitions. As Aristotle knew very well, there is a good definition when, in order to identify the essence of something, one selects attri-butes such that, although each of them has wider extension than the subject, all together they have not (Post. An. 2.963.35). There must be» full reciprocability between deftniendum and deftniens.
Supposing that /ram/ can be defined as the only «horned male sheep», then not only does this is a ram entail this is a homed male sheep » but also this is a homed male sheep entails this is a ram as well as this is not a ram entails this is not a homed male sheep, and vice versa. Deftniens and ‘ deftniendum can be substituted for each other in every context.
This cannot happen with the tree of Figure 2.2. Not only does this is an animal sheep not entail this is a ram, but also x is my preferite ram does not entail this is my preferite animal, all rams are homed does not entail all animals are homed, and, if one deletes the hyponym, one does not neces-sarily delete the hyperonym.
Thus we must now think of a different system that, while displaying the same ‘good’ characteristics of the tree of Figure 2.2, is also able to account for the phenomena that the latter leaves unsolved. Let us try, then, a second tree (Figure 2.3), which in some way reproduces the procedure used by naturalists in order to classify animal species.
It is certainly imprudent to equate linguistic inventories with taxonomies in natural sciences: Dupre (1981) has demonstrated not only that, where a layman identifies a species (for example, beetle), the entomologist identifies something like 290,000 species, but also that the lexical system of a natural language and scientific taxonomies overlap in a very ‘fuzzy’ way. We call tree both an elm and a pine tree, while a naturalist would say that the former is an ‘angiosperm’ and that the latter is not. There is no taxonomic equivalent of tree and no ordinary language equivalent of angiosperm.
However, Hjelmslev’s proposal can allow one to conceive of a sort of taxonomic tree as in Figure 2.3, designed to define without ambiguity and with the maximum economy a series of words, namely, dog, wolf, fox, cat, tiger, lynx, bachelor (as a seal), horse, ox, buffalo, sheep, mouflon, elephant and echidna. In such a linguistic (and natural) universe, one is not supposed to distinguish a horse from an ass or an elephant from a rhinoceros, and this explains why only certain lower disjunctions are called for. In this sense, the tree of Figure 2.3 overlaps only partially a current scientific taxonomy.
The tree of Figure 2.3 provides the picture of a very restricted uni-verse made up of natural kinds (of which the words in italic in the lower row provide names). We are obliged to consider this universe as a re-stricted one for the sake of our experiment: this universe is scarcely similar to the one of our actual experience, but when speaking of dic-tionaries one must conceive not only of very artificial languages but also of very artificial worlds. For instance, this universe takes into account neither ‘artificial’ kinds (such as a house or a chair) nor predicates, nor actions, nor social roles (such as ‘king’, ‘bachelor’, ‘pilot’). It has been remarked how it is difficult to account for all these problems at the same time (see Schwartz 1977:37-41). Aristotle (and Porphyry) dealt with natural kinds in the tree of substances, admitting that all the other phenomena should have been dealt with in one of the possible trees for the other nine categories (see 2.2.1). As for artificial kinds, Aristotle was convinced that one can deal with them as if they were substances, but according to some analogical procedures (see Metaphysics, lO43b21, 1070a- b).
It is not necessary to decide whether the linguistic terms in each node of the tree names classes included within larger classes or properties in some way contained or postulated by the terms naming the natural kinds listed in italic in the lowest row. One can say that any name of a subclass postulates its class or that every name of natural kinds postulates a hierarchical series of properties. In lexicographical representations of a system
