When one considers the phrases eat bread and eat soup, one realizes that eat has a slightly different meaning in each phrase: in the latter expression, but not in the former, it covers the manipulation of a spoon. Continuing the procedure applied in KF to polysemous items such as ball and colorful, one would have to represent the dictionary entry for eat by a branching path, perhaps as in (20):

The selection restrictions at the end of each subpath would provide the information which makes possible the choice of the correct subpath in the contexts of bread and soup functioning as object Noun Phrases. But then the activity symbolized by eat is also different depending on whether things are eaten with a fork or with one’s hands; and even the hand-eating of apples and peanuts, or the fork-eating of peas and spaghetti, are recognizably different. It is apparent, therefore, that a KF-type dictionary is in danger of having to represent an unlimited differentiation of meanings.

Several escapes from this danger can be explored. The most direct one would prohibit branching of paths in a lexical entry except where they represent an experienced ambiguity in some non-ambiguous context. For example, if file can be understood as ambiguous (e.g. in the context I love to — things: ‘ 1. put away for storage; 2. abrade ’), the dictionary entry would represent the ambiguity by a branching of paths; on the other hand, if eat does not feel ambiguous in a general context such as I'd like to — something, the submeanings of eat would not be represented in the dictionary. But this will presuppose, as a primitive concept of the theory, an absolute distinction between true ambiguity and mere indefiniteness of reference. The difficulty of validating such a; distinction empirically makes its theoretical usefulness rather dubious, although it has been advocated, e.g., by Ziff (i960: 180 ff.).

A more elaborate solution, suggested by Kurylowicz (1955), could be stated as follows: a dictionary entry W will be shown to have two subpaths (submeanings), W₁ and W₂, if and only if there is in the language a subpath Z_i of some entry Z which is synonymous with W₁ and is not synonymous with W₂. According to Kurylowicz, the notions of polysemy (path branching) and synonymy are complementary, and neither is theoretically tenable without the other. Thus, the path for file would be shown to branch insofar as file₁ is synonymous With, put away, whereas file₂ is not. However, the condition would have to be strengthened to require the synonyms to be simplex, since it is always possible to have multi-word circumlocutions which are equivalent to indefinitely differentiated submeanings of single words (e.g. consume as a solid = eat₁; consume as a liquid = eat₂). On the notion of lexemic simplicity.

In any case, it is evident that some regard for the experience of previous semantic theorists could have saved KF from an unnecessary trap.