Ernst Cassirer's theory of language as a symbolic form, one of the richest and most insightful philosophies of language of the twentieth century, went virtually unnoticed in the mainstreams of modern linguistics. This was so for what seems to be a good metatheoretical reason: Cassirer insisted on the constitutive role of meaning in the explanation of linguistic phenomena, a position which was explicitly rejected by both American Structuralists and Chomskian Generativists. In the last decade, however, a new and promising linguistic framework has emerged — the framework of lexical semantics — which seems to bear close theoretical resemblance to Cassirer's theory. In this paper, I show how the empirical results accumulated within the framework of lexical semantics serve to validate Cassirer's most fundamental philosophical insights, and suggest that Cassirer's philosophy helps position these empirical results in their appropriate epistemological context. I discuss the following fundamental points, which, for me, constitute the backbone of both Cassirer's philosophy and the theory of lexical semantics: (i) natural language grammars constitute structural reflections of a deeply-rooted, highly structured level of semantic organization; (ii) the representational level of linguistic meaning, which is prior to experience in the Kantian sense, comprises a partial set of semantic notions, which language selects as centers of perceptual attention; (iii) this partial set is potentially different from the sets selected by other symbolic forms, such as myth, science, and art; and (iv) linguistic variability is to be explained in universalistic terms, thus allowing for specific patterns of variability within universally-constrained limits.