Problems with conceptual semantics
Jackendoff’s system could be criticized for precisely this feature: its highly abstract primitives. These may permit interesting connections to be made between apparently unrelated meanings, but how justified are we in believing that these connections are cognitively real? Clearly, the more abstract the conceptual primitives we propose, the greater the number of possible connections between domains we can make. This is a similar criticism to the criticism of arbitrariness earlier made against cognitive semantics (7.2.6). What guarantee do we have, for instance, that a conceptual feature like [PL] really exists? In its current early state, the theory seems somewhat arbitrary and unconstrained: the investigator simply looks for plausible underlying conceptual structures, but there are no clear procedures for determining when a primitive is justified.
Jackendoff has addressed this question in two ways. First, he has stated that it is simply too early to demand that the theory justify its primitives: as in any immature science, all we have to go on are hunches; only when we have a good description of the semantic phenomena can we begin to constrain the theory (1990: 4). Second, he adopts a holistic approach to the justification of his primitives:
In fact, an isolated primitive can never be justified: a primitive makes sense only in the context of the overall system of primitives in which it is embedded. With this proviso, however, I think a particular choice of primitives should be justified on the grounds of its capacity for expressing generalizations and explaining the distribution of the data. That is, a proposed system of primitives is subject to the usual scientific standards of evaluation. (1991: 12)
This reply might not satisfy everyone, since it does not tell us which generalizations are psychologically real and which are merely artefacts of the analysis. One possible way of constraining the generalizations is to look for ones for which there is some independent linguistic evidence. For example, Jackendoff’s proposal that verb iterativity as illustrated in (16) above and nominal plurality are varieties of a single conceptual feature, PL, might be supported by the fact that some languages instantiate this with an identical morphological category. In Siraya, for example (Austronesian, Taiwan; extinct), reduplication had these very functions (Adelaar 2000). Another might be to look for psychological or perhaps neurological evidence to support the analyses developed in Conceptual Semantics.
A different line of criticism would be to question how we know whether the primitives are actually primitive. Perhaps it will turn out that some of them can be decomposed into even smaller conceptual units. Jackendoff is not perturbed by this possibility. He has often stated that there is no way to know exactly how far decomposition will be possible; and, in fact, his own work has suggested further decompositions for what were originally presented as primitives, such as the proposed decomposition of Path and Place into dimensionality and directionality features. Semantics is no different in this way from the decomposition characteristic of physical sciences: particle physicists do not know how far their decomposition of matter will lead, but that does not deprive their current attempts of legitimacy.
