Formality and interpretation

« previous post | next post »

I've been reading Stanley Fish's 1989 collection of essays, Doing What Comes Naturally: Change, Rhetoric, and the Practice of Theory in Literary and Legal Studies. It's not yet clear to me what he's for, exactly — I'm reminded of the old joke about the post-modern gang leader who makes you an offer that you can't understand — but it's clear what he's against, namely the idea that texts have meanings:

The objective facts and rules of calculation that are to ground interpretation and render it principled are themselves interpretive products: they are, therefore, always and already contaminated by the interested judgments they claim to transcend. [Consequences]

This is not a small point, in his view:

It might seem that the thesis that there is no such thing as literal meaning is a limited one, of interest mainly to linguists and philosophers of language; but in fact it is thesis whose implications are almost boundless, for they extend to the very underpinnings of the universe as it is understood by persons of a certain cast of mind. [Introduction: Going Down The Anti-Formalist Road]

The "cast of mind" in question is, roughly, science and the idea that rational inquiry can lead towards truth — the whole Enlightenment project. But in the passage just quoted, he's discussing a much more specific argument, made by Ruth Kempson in her 1975 book Presupposition and the delimitation of semantics.

Fish uses an argument from this book as the rhetorical backbone of his 33-page lead-off essay "Introduction: Going Down the Anti-Formalist Road". And it's therefore embarrassing (for him) that he's misunderstood the context (and thus the content) of her work, in a way that makes her argument at best irrelevant to the point that he wants to use it to make.

Fish first deploys Kempson in the second sentence of his essay:

It is one of the theses of this book that many of the issues in interpretive theory can be reduced to a few basic questions in the philosophy of language. Consider, for example, the discussion of "presupposition" in Ruth Kempson's Presupposition and the Delimitation of Semantics.

As Fish explains, Kempson raised the question of whether presupposition is "a relation between statements (parallel to entailment, synonymy, etc.)" or "a property of the speaker's belief in uttering a sentence", and observed that "if presuppositions in terms of speaker-belief are considered to be a part of the semantic interpretation of sentences", then the meaning of sentences can't simply be defined "in terms of the relation between a symbol or set of symbols and the object or state described".

He takes Kempson to be making an argument with devastating consequences for rational inquiry:

Once you start down the anti-formalist road, there is no place to stop; remove the connection between observable features and the specification of meaning, and you also remove everything else that is supposedly independent of context: entailment, contradiction, grammaticality itself, all become as variable and contingent as presupposition.

And he depicts Kempson as drawing back in horror from the logical implications of her argument:

This, however, is not the conclusion Kempson reaches; and indeed it is the specter of reaching it that drives her in the opposite direction.

But the context of Kempson's 1973 PhD thesis, republished as her 1975 book, was not an argument about whether rational inquiry is possible, or even (mainly) an argument about whether meanings are things that sentences have, or things that people do. Rather, it was an argument among linguists and philosophers of language about whether proof theory or model theory is a better way to think about natural language semantics. More concretely, it was an argument about whether natural-language words and phrases should be interpreted directly, piece by piece, or whether they should first to be translated into some other representation, and interpreted only in that derived form.

In this context, the term "formal" usually had the meaning implied by the title of Richard Montague's famous 1970 essay "English as a formal language".

Montague argued that natural languages like English could be treated semantically as if they were "formal languages", in the sense that the meaning of phrases could be calculated recursively from the meaning of their parts, as denotations in a "model" universe. A simple analogy for this process is the recursive interpretation of subexpressions in a desk-calculator language — each atomic expression denotes a number, and so does every successively more complex expression, in a way that doesn't depend on the context that the expression appears in.

Ever since the work of Frege and Russell on the logic of quantifiers, it seemed obvious that natural languages can't work that way. The key to a coherent account of sentences with multiple quantifiers seemed to require abandoning the idea that noun phrases like "some people", much less words like all and some and many, are referring expressions at all. For Bertrand Russell and those who followed him, the way to make semantic sense of English was to figure out how to translate it into a different language — some kind of predicate calculus, for example — and then to work on the semantics of the translated form, whose sub-expressions are not in any sort of one-to-one correspondence to those of the natural-language original.

Montague showed that this wasn't necessarily so, at least not for the standard cases of multiple quantifiers and so on. His work was enormously influential.

But there were always contrary voices. You could call them "anti-formalist", in the sense that they opposed the idea that English words and expressions should be assigned denotations as if they were the subexpressions of a "formal language". And the key issues that concerned them generally had to do with the role of context in interpretation.

Crucially, though, their contrary argument was not that a formal account of meaning is misguided and impossible — just that it needs to include a formal account of the role of context, by constructing context-dependent representations in terms of which grammaticality, entailment and so on can be calculated.

Ruth Kempson has recently written a very readable account of the past three decades of work along these lines: "The (re-)emergence of representationalism in semantics", to appear in Maienborn et al., Handbook of Semantics. Her abstract:

The major challenge posed to all those seeking to explain semantic properties of natural language is the pervasive dependence on context for the way in which expressions are understood. This chapter surveys the emergence of formal models of semantics against the background of the new emphasis in linguistics during the 1950's of formal modelling of grammars, and then details the way in which formal semanticists have increasingly turned to tools of proof theory for formal languages as they have grappled with this challenge. As we shall see, the starting point for this modelling of context dependence involved fierce disagreement between the protagonists as to whether or not explanations of natural-language content required some form of semantic representation, a view which those working within formal semantics paradigms initially so effectively ridiculed that this alternative was barely aired at all within semantics for the last decades of the last century. Yet as formal tools have been increasingly refined in order to express this pervasive context-dependence, the disagreement that led early on to the setting-aside of representationalism in semantics as a respectable methodology has, as we shall see, got so transformed that the initially conflicting sets of assumptions are now hard to distinguish.

Compare this to Fish's version of where the "anti-formalist" road should take us:

A formalist believes that words have clear meanings, and in order to believe that (or because he believes that) he must also believe (1) that minds see those clear meanings clearly; (2) that clarity is a condition that persists through changes in context; (3) that nothing in the self interferes with the perception of clarity, or, that if it does, it can be controlled by something else in the mind; (4) that meanings are a property of language; (5) that language is an abstract system that is prior to any occasion of use; (6) that occasions of use are underwritten by that system; (7) that the meanings words have in that system (as opposed to the meanings they acquire in situations) are or should be the basis of "general" discourses like the law; (8) that because they are general rather than local, such discourses can serve (in the form of rules or statues) as constraints on interpretive desires; (9) that interpretive desires must (and can) be set aside when there is serious public business afoot; (10) that the fashioning of a just political system requires such a setting aside, the submission of the individual will to impersonal and public norms (encoded in an impersonal and public language); (11) that this submission would be a rational act, chosen by the very will that would be held in check; (12) that rationality, like meaning, is an abstract system that stands apart from the contexts in which its standard is to be consulted; (13) that the standard of rationality is available for the settling of disputes between agents situated in different context; (14) that the mark of a civilized (lawful) community is the acknowledgment of that standard as a referee or judge; (15) that communities whose members fail to acknowledge that standard are, by definition, irrational; and (16) that irrationality is the state of being ruled by desire and force — that is, by persuasion — rather than by a norm that reflects the desires of no one, but protects the desires of everyone. All of these beliefs and more follow from and give support to the belief that words have clear meanings, and many of the essays in this book begin by challenging the linguistic thesis and end by challenging everything else.

The one positive conclusion from Fish's work that I believe I've grasped, so far, is the crucial role of what he calls "interpretive communities" in providing enough of a shared context — even if ephemeral and unfounded — for some minimal communication to take place. So it's ironic that he so completely fails to understand Kempson's work in the context of her native interpretive community.


  1. peter said,

    February 8, 2009 @ 11:03 am

    Locutions used for co-ordinating actions between different people (eg, requests, promises, commands, entreaties, etc) surely require, for their complete and accurate interpretation by a listener, some assumption by the listener of speaker-beliefs, -preferences and/or -intentions. Thus, it seems to me that, at least for dialogs over actions, accurate understanding of natural language expressions cannot be based only on relationships between syntactic components of statements. Is this not obvious?

  2. Mark Liberman said,

    February 8, 2009 @ 1:22 pm

    peter: Locutions used for co-ordinating actions between different people […] surely require, for their complete and accurate interpretation by a listener, some assumption by the listener of speaker-beliefs, -preferences and/or -intentions. […] Is this not obvious?

    Indeed. And it's no harder to treat all of this in a "formal" manner than it is to predict entailments in apparently impersonal statements.

    Similarly, the aspects of meaning called "functional sentence perspective" and "information structure" obviously have to do with relations among phrases, speakers, and listeners (or phases, writers, and readers), not the relations of phrases to communicatively deracinated states of affairs. And again, it's no harder to provide a "formal" account of such things than it is to give an account of (say) quantifier scope. (These problems are all *very* hard, in general, but that's another matter…)

    All of this is obvious enough, and has been obvious for long enough, that most linguists and philosophers of language will (I think) be puzzled about why Prof. Fish is getting so wrought up about it, and predicting such dire consequences for the entire project of rational investigation.

  3. Paul Kay said,

    February 8, 2009 @ 3:55 pm

    There was a faith healer from Deal
    Who maintained that all pain is unreal.
    "But, said he with a grin,
    When I sit on a pin
    I dislike what I FANCY I feel.

    — Anon.

  4. Stephen Jones said,

    February 9, 2009 @ 4:56 am

    I've spent about the last four weeks trying to read Lakoff's Women, Fire and Dangerous Things.

    It's rare that I admit a book is beyond my intellectual capabilities but this comes close.

    However I think I understand his demolition of the objectivist viewpoint and the idea that natural languages can be explained by logical languages.

    Never have I got the impression that Lakoff, or Puttnam, are suggesting 'objective investigation', is dead, any more than Heisenberg's uncertainty principle does.

    Anyway, isn't Fish something of a busted flush.

  5. Tadeusz said,

    February 9, 2009 @ 6:05 am

    What Fish is writing about should be seen on the background of literary studies, in which, I think, people did believe that there is one privileged interpretation and one meaning (granted, one per a generation of scholars). And I think that his opinions are still met with disbelief by other literary scholars. Actually, in what I have read by him, he does not suggest that there is no meaning in a text, but rather that the interpretation (i.e., the purported meaning) is relative to the interpretive community, especially when the reader forms an expectation about the function of the text (when you say "this is poetry" about a piece of nonsense the piece will be interpreted in a highly metaphoric way, etc.). This can be easily reconciled with what linguists and philosophers say. It is true, though, that he can produce a rambling text quite often.

  6. creole said,

    February 9, 2009 @ 9:02 am

    Sometimes we hear about "the Sapir-Whorf hypothesis", you don't know if this is a mere hypothesitation of certain intuitions. The same thing with the innatism but nobody feels like talking about the "Chomsky-Fodor or the innatist hypothesis" or something. The Montague's views don't deserve a special word like "innatism" nor a expression like "the Montague or the formal semantics hypothesis". There is a linguistic hidden dictionary here.

    [(myl) If you're saying that some ideas have widely-used names while some don't, and that some of these names come from the people who proposed the ideas and some don't, and that some of these ideas are crisply defined while some are more diffuse — well, yes. This is true in mathematics and political science and chemistry as well as in linguistics. ]

  7. Mark Liberman said,

    February 9, 2009 @ 9:34 am

    Tadeusz: What Fish is writing about should be seen on the background of literary studies, in which, I think, people did believe that there is one privileged interpretation and one meaning (granted, one per a generation of scholars).

    This is at best half true, it seems to me. (The idea that earlier generations of literary scholars believed in a single privileged interpretation strikes me as an overgeneralization or worse — that's certainly not what I.A. Richards thought, just for example — but never mind.) The subtitle of Fish's collection is "Change, Rhetoric, and the Practice of Theory in Literary and Legal Studies" (emphasis added), and he holds a faculty position in a law school. The essays in this collection take up legal issues at least as often as literary ones — the Anti-Formalist Road piece leads with a discussion of Cargill Comission Co. v. Swartwood, for example, and later takes up Arlen Specter's questioning of Robert Bork. So more is at stake than the interpretation of Paradise Lost.

    And one of the effects of Fish's arguments is to justify and reinforce the trend to omit all forms of linguistic analysis from the general curriculum. The current generation of lawyers, like the current generation of literary scholars, has (as a rule) no training in formal logic, no training in syntax, no training in the analysis of language structure and use at any level from phonetics to pragmatics. This seems to me to be a bizarre and historically aberrant mistake — but if every analytic technique, including logic — the "rules of calculation that are to ground interpretation" — is itself "always and already contaminated by the interested judgments [it claims] to transcend", then this state of affairs is nothing to worry about, and perhaps is even a good thing.

  8. Tadeusz said,

    February 9, 2009 @ 10:48 am

    thank you for putting Fish's ideas on a wider background. My selection of his papers was evidently one-sided. (Obviously I did overgeneralize in the two, three sentences). Yes, naturally, lawyers are supposed to know how to analyse a text or a statement, or at least to know what actually a dictionary contains (as they often use dictionaries in legal proceedings; I wonder whether they ever use grammars…). And it is alarming to think that the students do not study at least the rudiments of logic.
    But logically Fish should promote the use of analytic methods rather than reject them. How are you supposed to be able to tell how a meaning is "constructed" by the interpreter if you do not have the tools? Or is the tool the message for him?

  9. Mark Liberman said,

    February 9, 2009 @ 11:34 am

    Tadeusz: How are you supposed to be able to tell how a meaning is "constructed" by the interpreter if you do not have the tools?

    I believe that Fish's anti-foundationalism means that for him, there are no tools, at least not in the sense that you mean.

  10. Chris said,

    February 9, 2009 @ 11:44 am

    A formalist believes that words have clear meanings

    Well, my first impulse is to ask, based on a common meaning of "a": "Which formalist believes that?" Which I think makes some kind of point, but whether for Fish or against him I'm not entirely certain.

    Clearly my semi-serious question (contrasted with what native speakers will recognize as Fish's intended meaning) demonstrates that "a" doesn't have a clear meaning – at best, it has several, from which a selection must be made in interpretation. Can words of multiple letters be expected to do any better? And are we to believe that Fish's formalists miss this obvious point, or is he engaging in a bit of strawmanning?

    His following paragraph of non sequiturs doesn't improve the credibility of his argument, whatever his argument is. The existence of a phenomenon doesn't even imply that it *can* be perceived accurately, let alone that it *must* be (i.e. cannot be perceived inaccurately, I don't mean the imperative sense of "must").

  11. Adam said,

    February 9, 2009 @ 2:21 pm

    Count me in as a philosopher puzzled about why Fish thinks rational inquiry is doomed, as well as why he thinks formal methods in philosophy of language commit one to a caricatured Platonism.

  12. Bryn LaFollette said,

    February 9, 2009 @ 2:54 pm

    (9) that interpretive desires must (and can) be set aside when there is serious public business afoot; (10) that the fashioning of a just political system requires such a setting aside, the submission of the individual will to impersonal and public norms (encoded in an impersonal and public language);

    Wha? I must say I get completely lost by his line of "reasoning" at this point. I have the feeling that either his argument has just jumped the rails, or that it was not at all operating within any framework I was previously aware of (and interpretation is getting used in a novel, for me, technical application). I do however like the image of Semantic Interpretation getting set aside for "serious public business". What the hell is "serious public business", anyway, and when does one know it's "afoot"?

  13. bianca steele said,

    February 9, 2009 @ 4:16 pm

    When he discusses “interpretive communities,“ Fish is not talking (or not only talking) about communities in the social science sense. He means to include “English departments” as a type of “community,” and what he talks about sounds like what I might call a “methodology”; what is acceptable for academics in English departments, when they talk about literature, depends on what the members of English departments have decided to accept, and on nothing else. He lays out the argument pretty clearly in Professional Correctness.

    One presumes the theory applies to laypersons, as well. Therefore, for example, with regard to the interpretation of Julius Caesar, one expects that the way Americans have historically interpreted the Shakespeare play determines the range of acceptable readings of the play, at least in part. However, this isn’t Fish’s particular area of interest, and as I recall, his discussions of sites of interpretation range pretty much from the seminar room to the lecture hall.

  14. » OLDaily per Stephen Downes, 11 de febrer de 2009 TIC, E/A, PER…: said,

    February 11, 2009 @ 10:00 pm

    […] quelcom més que simplement recordar, i això és el que jo pense. Mark Liberman, Language Log. [L'enllaç] [etiquetes: aprenentatge, […]

  15. Robert F said,

    February 12, 2009 @ 5:55 am

    It's interesting that formal in mathematics means something slightly different from its everyday meaning. For instance, there are formal power series, which are infinite polynomials, manipulated according to the rules of algebra, but without regard to whether they converge to any value when a number is substituted for the variable.

RSS feed for comments on this post