KF provides that to the terminal element of a path there shall be affixed a string consisting of syntactic or semantic markers, or both. The function of this string is to represent conditions on the non-anomalous employment of the word in the meaning represented by that path. For example, the suffix ((Aesthetic Object)) at the end of one of the paths for colorful would indicate that the adjective, in the sense corresponding to that path, is applicable as a modifier without anomaly only to head nouns which contain the marker (Aesthetic Object) in their paths.

This part of KF, too, is fraught with apparently insuperable difficulties. Consider the case of the adjective pretty. It seems to be applicable to inanimates and, among animates, to females. If its selection restriction were stated as ((Inanimate) V (Animate)-f (Female)),^b the normality of pretty girls as well as the anomaly of pretty boys would be accounted for, since girls has the marker (Female) in its path, while boys does not. But we can also say pretty children without anomaly, even though child does not contain (Female) in its path; in fact, English speakers will refer the sex of the neutral children from the attribute, and a theory concerned with ‘ the interpretative ability of speakers ’ must account for this inference. One way of doing so would be to reformulate the selection restriction more carefully, e.g. as ((Inanimate) v ~ (Male)), to be read: predicable of Inanimates and not predicable of Males. This would explain why pretty children is not anomalous, but would not yet show how we infer that the children are girls, since the projection rules only check on whether the conditions of selection restriction are satisfied, but transfer no information from the angle-bracketed position to the amalgamated path. Moreover, the case of Female = ~ Male is probably quite atypically favorable in that we have two values of a feature dichotomizing a large domain in a relevant way. If, on the other hand, addled were to be marked in its selection as restricted to eggs and brains, the restriction would be unlikely to be statable in terms of legitimate markers (without distinguishers); and again we would lack an explanation of how we know that in It’s addled, the referent of it is an egg or a brain. Here a restriction in terms of negatives, e.g. (~ [( ~ Eggs) V ( ~ Brains)]), would be intolerably ad-hoc.

Two alternatives may be considered. One would be to regard ‘Constructability with Z’ as an intrinsic feature of a dictionary entry W (i.e. of the ‘path’ of W), and not as a statement external to the path. The other is to adopt a more powerful conception of the semantic interpretation process, in which features of selection restriction of a word Z would be transferred into the path of another word, W, when it is constructed with Z. This is the solution adopted by Katz and Postal (1964: 83) for a special purpose [(c)], and it is the general solution [pp. 429 ff.]. But with respect to KF, it seems safe to conclude that the distinction between paths and selection restrictions is an untenable as its other specifications of the format of dictionary entries.

¹ If Female necessarily implies Animate, the notation could of course be simplified; see Katz and Postal (1964 16 ff.). On the theoretical implications of such simplifications.