Against monosemy, pt 2: Semantics & pragmatics

As I continue to prepare my paper on the semantics of compounding, I realized that I had a mostly finished review of Charles Ruhl’s (1989) On monosemy:  A study in linguistic semantics that I had never posted. This will be appearing in five parts over the next two weeks. I’ll provide a full bibliography in the final post, but hyperlinks for each citation appear at first reference. The goal here is to (1) present my thoughts about monosemy as a theoretical construct and (2) maintain a regular posting schedule while I am also devoting more of my time to getting my SBL paper finished on ὑδροποτέω and compounding. This is going to be a deep dive.

These posts are not written for the mere linguistically curious. This is fairly advanced material and assumes some knowledge of the field from the outset. Many might want to consider simply waiting until Part 5, where is going to be more practical, discussing the merits of using Ruhl for ancient language study. The truly determined are more than welcome to ask questions. I am always happy to answer.

And to get it clarified at the start:

  1. polysemy = many senses
  2. monosemy = one sense

Semantics & pragmatics

The difficulty with separating the semantics and pragmatics is that the latter, as an independent field of study, is effectively an accident of the false dichotomies created by generative grammar. The desire to make linguistics mathematically rigorous that drove mainstream generative linguistics from its beginnings resulted in a situation where domains of study were made compartmentalized and autonomous. Crudely speaking, within the mind there was the syntactic module and the lexicon. A syntactic tree was generated via syntactic rules; its lexical elements inserted properly, and the meaning was then interpreted from there. Lexical items were defined mathematically as well through componential analysis: a bachelor is [+male] and [+single].*

*The pope notwithstanding, of course.

Features were formalized with rules in the lexicon, termed selectional restrictions. These rules made sure that the brain’s “language organ” put the right words into the right places to prevent sentences like: The rock drank an entire gallon of pizza. The trouble was, however, that within this system there was no space for figurative language, metaphor, or really any creative language use. Within the structuralist and generative paradigm, rock is [-animate], but drank requires a grammatical subject that is [+animate]. Likewise, pizza is [+shape; -homogeneity], but both the verb drank, and the mass noun gallon requires a genitive (i.e. of N) with the opposite features: [-shape; +homogeneity]. The sentence is violating semantic well-formedness conditions and within generative formalisms would be expected to be preceded by the symbol: #.**

**Within generative syntax you will often find the asterisk (*) used for ungrammaticality and the pound sign (#) for semantically ill-formedness.

Perhaps all of this seems perfectly reasonable for dealing with meaning—at least until encountering Dwayne Johnson enjoying a gallon-sized pizza smoothie.

But surely such an interpretation of this sentence is the result of the shared knowledge, culture, and experience, yes? Surely this cannot be part of semantics, but pragmatics only. That’s precisely the position Ruhl (1989) takes in his proposal. And, granted, there is certainly a kind of logic to it. It is not particularly likely that someone would consider including a reference Dwayne Johnson in a lexicon entry for the word ‘rock’. And, indeed, we encounter Ruhl laying out just such an argument for the real-world differences between the phrases open doors and open presents. Quoting Lakoff (1987), he writes:

[from Lakoff (1987, 416)] Or take the word open. We open doors and open presents, and though the actions described by the words are very different, we would normally have to think twice to notice the difference. …

This certainly sounds familiar, and thus even reasonable. However, what warrants the claim that window, open, and run have more than one sense? Admittedly, the phrases open the door and open the present evoke quite different images, but why is the difference attributed to open, which doesn’t differ, and not also the? Why isn’t the difference located solely where there is difference: in the door-presents distinction and in the knowledge that people have about these two activities? (Ruhl 1989, x)***

*** Of course, anyone with a cursory knowledge of Lakoff (1987) should know the answer to that final question: The difference is precisely located the knowledge that the people have about these two activities—for Lakoff that’s where semantic variation and polysemy comes from—encyclopedic, cultural knowledge is the whole point. Ruhl clearly did not read Lakoff very closely in his haste to get a citation of it into his preface.

Ruhl continues on after this quote, suggesting that Lakoff’s mistake was locating the distinction between semantics and pragmatics in the wrong place. For Ruhl, meaning that is derived from physical and social environment must necessarily be pragmatic. As such, He positions himself solidly within the generative tradition that rejects the idea non-linguistic knowledge and cognitive systems are sources for language structure.  He also misses entirely the thrust of Lakoff’s point in the quote above. Lakoff (1987, 416), above, is not arguing that the various senses are ‘over’ are discrete. When Lakoff says, “very different,” he is talking about the folk linguistic knowledge of the man on the street. He is not making an assertion about the nature of meaning. Indeed, the quote begins Lakoff’s case study of the semantics of the preposition ‘over’, in which he is making that case that polysemy is organized and structured. While it may not be predictable from the central member, it is nevertheless non-arbitrary as a result of larger image schema and metaphorical models that encompasses the lexical item (Lakoff 1987, 460-1).****

**** That Ruhl only cites Lakoff (1987) in his preface and show no awareness of Langacker at all suggests Ruhl had already finished his entire book when Lakoff (1987) appeared in print and did not want to be bothered to actually engage with it in the body of his monograph.

If we wanted to frame the entirety of the case study in Ruhlian terms: the image schema is itself the monosemic bias and the various context-dependent usages are the pragmatics that result from it. In a simple world, that would be the end of it: different ways of saying more or less the same thing. But Ruhl’s proposal is more complicated than that

On the whole, this pattern is consistent. Ruhl writes as someone on the brink of recognizing precisely the issue that led cognitive linguistics to its view of semantics but trying extremely hard to maintain a meaningful division between semantics and pragmatics in order to protect the autonomous system from the complex mess that is language in use.

Elsewhere he writes,

I am assuming a continuous scale of abstraction, from the most abstract, autonomous, mutually (i.e., strongly) systematic, and well-formed level (IDEAL) to the empirical, most concrete, non-autonomous, modularly (i.e., weakly) systematic, and ill-formed level. Roughly, this is the continuum from Language to Reality: from language as a (nearly) autonomous closed system to an experientially open system directly involved in and influenced by externalities. The different levels of abstraction are distinguished, essentially, by their degree of specific context.

To the conscious mind, the whole will always seem to be greater than the parts; the parts exist “in isolation” only because they have been detached from more concrete wholes. Consciousness typically puts the parts and the whole on the same level of abstraction, making the relation of part to whole problematic (emergence): the whole is reduced (but never successfully) to the compositional sum of the abstracted parts, and the abstracted parts are interpreted as if they were concrete, “basic” elements (138-40).

Ruhl seeks, in this section, to account for existence of both regularization and idiosyncrasy of language. Again, he is seeing the same difficulties and problems with traditional structuralist and generative theory that cognitive linguists were seeing and puts forward versions of the same critiques. Meaning is rarely successfully componential in nature. Language-in-use is complex, messy, and seemingly unsystematic. But he approaches these issues from an entirely different perspective than the functionalists of the same period. He creates a hard boundary between semantics (the lexicon) and pragmatics (language in use) and it does so specifically for reasons that functionalism rejects.

While Ruhl is very much aware of the problems and challenges in developing a theory of meaning, this awareness, paradoxically, seems do little to affect how he actually thinking about what a semantic theory should look like. This is especially clear in Ruhl’s evaluation of other linguistics, particularly Stockwell, Schachter and Partee’s (1973, Rulh’s SSP, below) analysis of pronoun semantics and Cohen’s (1979) study of the verb ‘insult’. Both these analyses are bread and butter structuralist semantics. And indeed, Ruhl praises both studies for what he views as their compelling conclusions—but on purely theoretical grounds:

We must note the strengths of this approach [that of Stockwell, Schachter and Partee (1973) and Cohen (1979)]. The “strongest” (most abstract and general) solution for such data would seem (other than the obviously wrong syntactic solution) to be the semantic one SSP and Cohen seek, because semantics, being more closed and abstract, imposes more constraints. A pragmatic solution seems an encouragement to “anything goes”; it is looser and vaguer than the systematic semantic features that Cohen so precisely specifies. A pragmatic solution is too easy and exemplifies a common dodge: discarding troublesome data to the relevant “garbage heap” (adverb, idiom, parole, performance) (Ruhl 1989, 147).

This quote just a few pages later, stands in start juxtaposition to Ruhl’s other statement provided earlier. Here Ruhl wears his generativity openly on his sleeve. While it may indeed exist as a continuum that ranges from language as autonomous closed system (Ruhl’s semantics) to language as experientially open system (Ruhl’s pragmatics), Ruhl has no interest in accounting for the existence of the continuum itself. It is certainly true that with his theory of monosemic bias, Ruhl is offering an alternative to the kind of analyses put forward by Cohen and Stockwell, Schachter and Partee. Yet, it is fundamentally important to understand that he agrees with them in terms of what makes an analysis a good analysis. His central fault for both is merely that they multiply sense unnecessarily, and thus inelegantly. Cohen (1979) and Stockwell, Schachter and Partee’s (1973) in the view of Ruhl wrong because they try to attribute too much semantic variation in their language data to the domain of semantics.  This is not a critique grounded in linguistic data. It is not empirical. Instead, the critique is very much within the standard generative domain: it is rationalist one. This is Cartesian Linguistics at its best (cf. Chomsky 2009): Use one’s faculty of reason to develop a theory and then build the language data around it.

So Ruhl is only interested in explanations that he views as existing within autonomous system. For him, the experientially open system is looser, vaguer. It is unsystematic, lends itself (apparently in Ruhl’s mind) to shoddy linguistic scholarship (i.e. the “discarding [of] troublesome data to the relevant ‘garbage heap’”[147]). In this context, his own solution is a maximally broad, single sense that covers fall usage. This is what his critique of Cohen and SSR is predicated on:

The need for us to recognize that words and sentences are much more remote from reality than we have believed. We have been continually guilty of Misplaced Concreteness. Any act of reference is much more complicated than simply putting a highly specified label on a concrete object or situation. That is the conclusion I have reached by assuming monosemy as my initial hypothesis (151).

What is going on here? Ruhl recognizes the continuum between regularity and idiosyncrasy; he is critical of other linguists who propose theories that get too deep into the experientially open side of the continuum. Ruhl’s monosemy is, thus, a sort of doubling down on linguistic autonomy and thus necessarily also doubling down on only describing linguistic competence at the expense of linguistic performance. In this context, it is not that polysemy does not exist in language, but simply that it is not part of the language faculty. It is not part of the autonomous linguistic system. Effectively this means that all functionalist and cognitivist accounts of semantics become less relevant to Rulh’s goals since they are necessarily focused on performance—language in usage—rather than linguistic competence, as we will discuss in part 3, coming in a few days.