This site and its associated videos are initially about truth and meaning, but ultimately are about how we understand our relation to the world. This is a bit of a heady claim - I will try to explain.

The first thing to appreciate is that saying what it is for some set of sounds or marks on a page to mean something in general, non-trivial terms is not an easy thing to do. It's fair to say there's no generally agreed theory. And since the things that are true are sentences, and a sentence's truth is related to its meaning, truth itself is similarly up in the air.

The problem, I believe, is that we're guided in our thinking about the subject by a picture of our relation to the world which is mistaken: utterly intuitive and apparently obvious, morally appealing even, but also false. I want to suggest that to get to a workable theory of truth and meaning, we have to question this picture in something like the way relativity theory and quantum physics taught us to question our familiar picture of the physical world. One surprising perk of this is that we get a nice solution to the problem of consciousness, down the road.

Suppose novice cook Fred is planning to roast a chicken - you say he should spatchcock it. He looks at you, bewildered. You explain in detail the whimsically-named technique. Later, when someone makes the same suggestion, Fred explains his reasons for preferring not to.

We have no problem seeing what's going on in this scenario. Initially Fred doesn't understand because he doesn't know the meaning of 'spatchcock'; following the explanation, he does understand, because he knows the word's meaning.

The scenario and our explanation of it suggest a theoretical challenge - to explain in general terms what meanings are and what's involved in grasping them, that they may lead to understanding. What, in short, is behind those 'because's? Alas, the fatal intuitive picture is already baked into this challenge.

The picture requires there to be three things in some relation: the person, the word, and the meaning. Learning a language in part is a matter of associating words to meanings, and understanding involves being able to put meanings together to represent how the world is or may be. It turns out there just are no things which can live up to this expectation.

The fatal step is taken right at the beginning, in thinking there is something substantial, something in need of theoretical explanation. Abandoning this thought requires turning a whole lot of familiar (comforting, genuinely appealing) ideas upside-down - or rather, right-side-up.

1. First, we promote that 'Ah-hah' (or maybe, 'Yeah, I get it') associated with hearing some sentences to first-class status, in our theoretical thinking about our life as reasoning agents. More accurately, we take as basic the sense that something feels right or wrong -assent or dissent. We give-up on the idea that there's any notion of 'grasping meaning' which explains it, in any substantive way.

2. Second, we show that this is enough to get us to truth, provided that we seek to understand it in the context of speech in a language community. The idea of focusing just on one speaker's relation to the word and the world is out the window. Ultimately we understand truth as the goal of a collective project by which each of us maximizes the number of sentences s/he knows which s/he agrees with (this turns out to be a very useful thing to do).

3. Third, we show that we are then able to understand word and sentence meaning solely in terms of truth.

(there are other things, too).

So what's the fuss, where's the promised Quantum Paradigm Shift?

Let's step back a bit. If you've done some philosophy or you're the sort of person who thinks about such things, you may be aware that questions about what grounds our beliefs about what's morally right and wrong are uniquely difficult. They seem always to terminate in some categorical principle ('You must...') which it's reasonable to doubt. In philosophers' jargon, these matters - matters relating to what's right or wrong and to what we ought or ought not to do- are 'normative' matters, and their logic and problems, the logic and problems of 'normativity'.

One of the central lessons of 20th century thinking about language and mind is that normativity is every bit as essential to thinking about rationality, language and meaning, and its problems every bit as pressing. Rationality may sound like a very lofty concept, but being rational really isn't a lot more than believing what you ought to, and not believing what you ought not to. Mostly that's the truth, but sometimes, if the preponderance of evidence is for something which in fact is false, it is in an important sense rational -right- to believe what's false. And similarly, meaning something by a word requires the possibility of using the word wrongly, mistakenly.

First, then, the big element - which I grant not all will immediately find so boggling - is that this understanding of meaning and truth requires shifting the 'locus' - the site, or origin - of normativity from the individual to the social. Because rationality is inherently normative, being at root about believing what it's right to believe, this has as a consequence that people are rational - I mean, literally, thinkers- only insofar as they can be understood as belonging to a community of language speakers. The picture -endemic to our theorizing about these topics- of the isolated thinker, struggling to justify her or his beliefs about a wider world and the people in it, fails to appreciate that its protagonist can't have got to where she is without these things' help. In metaphors, they are the tree on one of whose branches one sits, as one contemplates sawing - rather than a ladder one has climbed up, which one can kick away. (This may all sound a bit waffly and vague - the model of language use at the core of this project makes these ideas more precise).

The second implication is that our knowledge of the world originates in and is wholly mediated by language. In bald terms, unlike some philosophers, who have built world views taking what is immediately known or given to be 'sense-impressions', or others who have taken this to be the physical world, I am saying that what is given is language - sentences and our evaluations of them.

We don't stand in a knowledge relation to things in any theoretically interesting sense. All reasoning (all!) is conducted with language or extensions of it, and reasoning cannot breach its confines - il n'y a pas d'hors-texte (there is nothing outside of language). You say, 'Fido is in the living-room curled up on the couch', and picture him in your mind's eye there, sleeping. The natural thought is that the name 'Fido' is for you in some theoretically interesting way related to that wholly extra-linguistic mental image, not to mention the animal himself. This is what we are giving up on. Doubtless, Fido-related neural patterns are excited in your brain when you utter the sentence - this is not our concern.

To be clear, none of this precludes your usefully explaining your having said "Fido is on the couch." with the statement, "I just saw him there!" (let's assume you did). It is, though, to insist that there's nothing about this explanation to unpack (assuming weird circumstances don't require further elaboration), no mental event -your seeing of Fido- which it's the job of philosophers to elucidate. There are, admittedly, interesting questions about how there can be two independent, satisfactory explanations of what may appear to be the same fact (the scientific explanation, and the everyday one). Explaining this is one of the project's burdens.

 

The project introduced on this site explores a position in the space of theories of philosophical semantics which I do not find represented in discussions of the topic. It is similar in some ways to familiar theories, but not broadly the same as any one, so far as I can see.

Reduced to a slogan, it would be: In the beginning was the word (more properly, In the beginning was the sentence, but that kind of sucks all the allusiveness out of it. I'll work on it). The guiding thought is that our epistemic access to the world is through sentence valuations (so not, say, physical objects, retinal stimulations, or sense impressions). Alternately, the thought is that the informational relation between words and things ('Fido' and Fido) is relevant, if at all, only to neural network theorists, not philosophers of language. The routinely crushing burden of trying to situate information in semantic theory is wrongly assumed. Everything follows from this.

The reason for this position's neglect may be that it requires renouncing commitments endemic to mainstream thinking.

The Main Ideas

A satisfactory account of truth can be given solely in terms of speakers' sentence confidence. This account holds that

  • Token sentences are the primary bearers of truth.
  • Truth is 'dialectic', not idiolectic. That is, the definition of truth refers ineliminably to a community of speakers.
  • Being true is a matter of contributing to the maximization of aggregate sentence confidence.

  • Sameness of meaning is a matter of substitutability salva veritate in context of utterance.
    • Sameness of token phrase meaning can thus be defined solely in terms of truth.
  • Sameness of token phrase meaning is all we need for semantics - meanings are just sets of token phrases, all with the same meaning.
  • The usual apparatus of sense and reference has no role. The goal of being able systematically to explain peoples' sentence valuations in terms of the senses they ascribe to words, is renounced. Meanings here are in some ways like familiar referents, though -to note one difference- names of fictional entities like 'Sherlock Holmes' have a meaning.
  • Meanings are wholly derivative - they do no independent explanatory work.

  • Sameness of proposition can be defined in terms of phrase meaning.
  • Sameness of proposition is all we need for semantics - propositions are just sets of token sentences whose constituent phrases' meanings align.
  • Propositions are wholly derivative - they do no independent explanatory work.

  • The theory renounces as mistaken the familiar thought that the only way to explain a language speaker's ability to understand effectively infinitely many sentences given only a finite vocabulary and grammar is to attribute to her a grasp of something in the nature of a semantic theory.
    • Neural net theorists can explain the relevant classifications of sentences without semantic concepts.
  • Philosophers of language are simply mistaken in thinking they need to replicate in their terms the neural net theorists' achievement; that there is a residuum of mystery which intellectual responsibility compels philosophers to investigate.
  • Semantics' only job is to regiment our understanding of its terms, which does not require a compositional theory.* Compare theoretical linguistics of the last 70 years, whose goal was to give a scientific explanation of speakers' classifications of sentences into the grammatical and non-grammatical in terms of speakers' grasp of a generative grammar. Its final foundering on the revelations of neural machine-learning science makes vivid the pattern here. Linguistics, like plant taxonomy, is immensely valuable in providing a history and classification of it subject- matter, in its case the vast array of parts of human language, but it is relevant to cognitive psychology only as plant taxonomy is relevant to molcular or cellular biology (which is to say, a bit, but not centrally).

  • The normativity inherent in meaning derives entirely from the contract which binds language users in normal speech. No further basis is needed. Meaning and language being inherently normative, this entails that all language is public language.
  • Justification is in the first instance just the familiar activity of speaking sentences with the goal of changing other peoples' minds - in the model's terms, altering their token sentence valuations.
  • Rationality itself is seen to be a property of an individual only as considered (actually or potentially) as a member of a speech community. Put another way, the normativity inherent to rationality is inescapably a consequence of social interaction (this thought admittedly is hardly original to this project). A prevailing conception of individualistic rationality, epitomised in Descartes' Meditations, must be renounced as mistaken.

  • The present conception of semantics can be got to through a kind-of thought experiment or model.
  • In Philosophical Investigations S.217, Wittgenstein remarks,
    How am I able to obey a rule?" -if this is not a question about causes, then it is about the justification for my following the rule in the way I do. If I have exhausted the justifications I have reached bedrock, and my spade is turned. Then I am inclined to say: "This is simply what I do."
    The model starts with the thought this suggests for semantics: that at bottom all we have is token sentences and our confidence-valuations of them.
  • Its elements are
    • speakers (many)
    • sentences = mere sound sequences (many)
    • for each speaker, a value function which maps token sentences to numeric values
      • valuations are made to depend in part on valuations of already-encountered sentences (beliefs).
    • A provisional, model-only pleasure property needed to motivate sentence exchange - what individuals seek to maximize by believing valuable sentences.
      • This pleasure property is cashed-out in terms of the practical benefits of believing truths.
  • It seeks to augment incrementally these bare resources to a point where familiar semantic concepts can be captured.
  • The model's interesting result is that it gives us the core semantic concepts (truth, meaning, proposition) without reference to the world.
  • The model sides with Wittgenstein against Quine and Davidson on the core matter of interpretation: valuations - and hence the semantics (or at least their antecedents) - of others' sentences are given. Far from being a central problem of semantics, they are for the model completely unproblematic.

  • The model makes plain Wittgenstein's point that all language is public language. The problem of consciousness is born of the mistaken thought that we can even get so far as to refer to the elusive objects it grasps at.
  • The model takes us beyond the Private Language argument: a direct analogue of the problem of consciousness naturally arises in it, despite no assumption having been made about the cognitive capacities or interior lives of the minimal speakers it postulates. The problem of consciousness is a consequence of the nature of language itself, properly understood.
  • Wittgenstein, now as a foil, from Philosophical Investigations Section 293:
    "If we construe the grammar of the expression of sensation on the model of 'object and designation' the object drops out of consideration as irrelevant."
    The suggestion is that the model of 'object and designation' is satisfactory for the semantics of some kinds of speech, just not sensation talk. The model tells us that this picture (as usually understood) is wholly misconceived. The beetle in the box is the world.
  • That is: the model -like familiar consciousness realism- puts pains (qualia) and middle sized dry-goods (as understood in familiar thinking about semantics) on the same footing. It departs from consciousness realism in denying any of these 'things' (as usually understood) a role to play in epistemology or semantics. Our access to carpets and avocados and smartphones is, so to speak, solely through language. However, language being inherently public, there is no way to recover qualia within the model, as we do these middle-sized dry goods.

The thinking behind the model may be made more vivid if I acknowledge a few of what I consider to be its counter-intuitive implications.

  • Democratic character of truth.

    The model holds a sentence true just in case it's an element of a set which would maximize value for all speakers of the language (and its absence would lower the value). This definition in a sense ranks all speakers' valuations equally. This may seem to imply that experts' valuations of sentences within their domain should carry a weight only equal to that of any social-media pontificator.

    This is not what the model implies. Truth in the model is an ideal - it is what would maximize aggregate value, were all speakers to have considered (per impossibile) all sentences. The thought is that in saying a token sentence is true, we are saying in effect that our all believing this sentence will maximize our collective sentence value (and by extension our pleasure) in the long term. The model allows that the expert, having considered more sentences in the domain than the pontificator, should be given more weight.

  • Language learning

    The model holds that an infant learning language simply begins to value some sentences but not others, without there being any systematic explanation to be had in terms of its learning the sentences' constituent words' meanings. The sound patterns of language being arbitrary, this may seem to require magical thinking.

    This objection is a reflection of the hold of our intuitive picture of language, that ordinary talk of learning of meanings is more or less a report of the workings of some underlying mechanism which it is the philosopher's or scientist's job to elucidate. Again though, there is no real problem here, once this picture is prised off what science-proper should be studying. Neural net researchers' models can explain the evolution of infants' reactions to language stimuli without recourse to meanings. And we of course can -must- retain our everyday explanations of one anothers' language undersanding in terms of meanings. A problem arises only if we look to consolidate these modes of explanation. There is, I should add, no obstacle to scientists explaining the predictive efficacy of ordinary meaning talk, so long as this doesn't aspire to the status of a reductive analysis. So long, that is, as the scientific explanation of our talk's predictive efficacy views language 'sideways-on', devoid of meaning.

  • Agents and people

    The model envisions speakers (called 'agents' in the model, solely because the abbreviating letter 's' was already taken by sentences) uttering sentences which hearers (other agents) value, or do not. Sentences can contain the names of people, and even speech reports. It is natural to take the names in sentences to refer to the agents to whom intuitively they correlate.

    This is incorrect. To take the names in sentences within the model to refer to the people whom they model would be to revert to the "Fido"-Fido model of semantics and break the theory. The model's elements are clear - there are in it, for example, no dogs or rugs, and similarly no people except as agents. The meaning -reference, if you prefer- of a name like 'Fido' of a dog is as specified above in the summary of meaning - language does not reach outside of itself, and the point applies equally to people. To take matters otherwise is in effect to conflate the object language of the model with the meta-language in which it is described.