Aristotle’s third mode of simple relation in Categories makes an abrupt shift from things to words. While a homonym or synonym are defined and used as if they are relations between things, the only definition of paronym offered by Aristotle is that a paronym is a word for a thing defined by a change in ending (or termination, depending on the translation).

However, this might not be the only possibility. According the OED etymology, paronym is taken from a Greek word meaning “formed by a slight change of the word, derivative.” In fact, translator EM Edghill avoids using transliterations at all, calling a synonym equivocal, a homonym univocal, and a paronym derivative.

At the risk of rewriting Aristotle (not being skilled enough in Greek to know how accurate the translations really are), it seems to me that the idea that a thing can be derived from another thing is reasonably sensible. For example, corn (as in an ear of corn) is a product derived from corn or (maize, for non-Americans). They are homonyms in the sense that they sound the same but mean different things, and also paronyms because one is derived from the other. In another case (adjacent to Aristotle’s example) grammar and grammatical are derived from the same root, and hence are paronyms— but they can also be synonyms used in a sense like “the sentence is grammatical” and “the grammar of the sentence is good.” In this case, the affix does not effect the meaning of the term, only its syntactical position.

Hence, something may be a homonym and a paronym (same root, different meaning) or both a synonym and a paronym (same root, similar meaning). However, these aspects alone are not sufficient to establish a paronym as a unique category worthy of consideration—they suggest an overlapping zone between the two primary categories of synonymy and homonymy.

By definition a synonym cannot be a homonym. This makes it reasonably certain that they are exclusive categories. Is there any situation where a paronym is neither a homonym nor a synonym? This could be the case if we extend Aristotle’s relations back to things, the place he started from. Boards are derived from trees. They have a common “substance” but not a common root—in the derivational sense though, they might be said to be paronymous. They have neither a common sound, or quality in appearance—so they are not homonymous. Though one might say a tree has boards, one cannot freely interchange boards and trees when referring to these things, so they are not synonymous. If we allow for names which apply to the same substance but in different forms, then we can establish paronymy as a separate category of relation, not merely a subset of both homonymy and synonymy.

The word paronym literally means beside a name. A paronym is close to the related term, but not necessarily the same in either sound or meaning. It is a relationship of contiguity, like the homonym, but when applied to things a paronym is not always homonymous. It is a relation of meaning like the synonym, but it is not always synonymous. It is a well formed category.

Why is that important? I have elected to explore these fundamental relations using these terms because the more commonly used terms become confusing. Expressed or implied relations of similarity (synonymy) are usually characterized as metaphor. Expressed or implied relations of contiguity (homonymy) are usually characterized as metonymy. The majority of twentieth century linguistic theorists have reduced all relations of meaning to these two primary axes. Generally, synecdoche (substitution of the part for a whole, or the whole for a part) is taken as a sub-category of metonymy. A binary distinction of linguistic behavior is central to Saussure and the entire chain of linguistics and deconstruction that followed him—signs either work by substitution (synonymy) or combination (homonymy)—they are either continuous or discontinuous with their subject.

Other theories (primarily rhetorical, anthropological and historical rather than linguistic) claims that there are four “master tropes.” Descended from Ramus, through Vico, to Kenneth Burke, Levi-Strauss and Hayden White, the tropes are enumerated as metaphor, metonymy, synecdoche and irony. I have had a problem justifying this set of categories to myself—sure, they are useful, but why these four?

Why not add zeugma, erotema, or any of the other 185 tropes enumerated in various versions of classical rhetoric? By what criteria (other than utility) can we claim that these tropes are basic divisions of any semiotic process? The tropes of metaphor, metonymy and synecdoche are disruptions in the process of naming—in the case metaphor, it is a substitution of a different name which is assumed to be similar in some fashion. Metonymy names something by virtue of a shared quality. Synecdoche, substitutes a part for a whole or its converse. These figures can operate on the word level, not just at the proposition level—and hence, could be considered primary.

Revisiting Aristotle’s Categories has made it possible for me to justify to myself that paronymy is indeed a plausible category—it is closely analogous to synecdoche in that it is derivational and not necessarily bi-directional like the other tropes. By adding paronymy (or synecdoche) as a fundamental relation (or trope) rather than a sub-category, it seems possible to break out of the treacherous land of binaries. Irony, however, is a special case. A metaphor, a metonymy, or a synecdoche can be ironic. I still have not worked that one out.

Paul Ricoeur has demonstrated fairly conclusively (to me at least) that metaphor does not operate the same way at the proposition level as it does at the word level. Therefore, it is misleading to say that metaphor is a primary function at the level of simple relations— I think that synonymy is perhaps a better term. The constant misunderstanding of irony also makes it a frustrating term to apply to the fundamental relations of naming. There are major overlaps and blurred distinctions between all these terms. Centuries of writing on the tropes, particularly metaphor, have muddied most of the usable distinctions.

I made a tentative map for myself today, which might make the distinctions that I have argued here a bit clearer—I’m not completely comfortable with it, but at least it is a start.