Presented by Zora Teofilovic
SUMMARY
OF “ADVERTISEMENT FOR A SEMANTICS
By Ned Block
PURPOSE: To advocate an approach to semantics relevant to cognitive psychology/science.
METHOD: Beginning with some desiderata, Block uses RTM, with an assumption that thoughts are structured entities.
Block argues that conceptual role semantics (CRS), specifically two-factor theory CRS, is the only approach that has the potential to satisfy all the desiderata listed below.
Block promotes a variation of the functionalism (i.e., the view that mental states are defined by their causes and effects).
______________________________________________________________________________
DESIDERATA
(D)
D1: Explain the relation between meaning and reference/truth.
D2: Explain what makes meaningful expressions meaningful. What makes a "cat" meaningful while "glurg" meaningless?
D3: Explain the relativity of meaning to representational system (this desideratum is a special case of D2).
D4: Explain compositionality. What exactly is the relation between the meaning of a sentence and words?
D5: Explain the relation between meaning and mind/brain. What is it to grasp or understand meaning? Can grasping meaning have a physical effect? Block says it has an effect on the brain.
Block's prime focus is on the brain. He adapts a form of materialism known as "token" identity thesis, which assumes that a particular mental event is a physical occurrence. Block states that while meanings are nonphysical abstract objects, there is still difference between a brain that grasps a certain meaning and a brain that does not -- The difference is in the causal properties of the brain that does. For this reason, we need an explanation for how a relation between a brain and a meaning can make a causal difference.
D6: Explain the relation between autonomous and inherited meaning.
Q: What is the difference between the mental representations and the non-mental representations (e.g. words on this page)?
A: The representations on this page to be understood require that they be translated into a language of thought (through reading or hearing), while the representations in the brain require no such translation.
Representations in our brain are referred to as autonomous meaning, while representations that require translations or transliteration are inherited meaning.
Autonomous meaning arises from the causal powers of the human brain (or any other object as long as it has equivalent powers). Q: What is the relation between autonomous and inherited meaning? And, how can a representation with autonomous meaning mean the same as a representation with inherited meaning?
D7: Explain the connection between knowing, learning, and using expression's meaning. It is clear that there is a close connection between the meaning of a word on the one hand, and what we know when we know or understand a word and what we learn when we learn a word, on the other hand.
D8: Explain why different aspects of meaning are relevant in different ways to the determination of reference and to psychological explanation. Two aspects of meaning that are relevant to psychological explanations are:
Indexical:
(1) "I am in danger of being run over."
(2) "Ned Block is in danger of being run over."
Block claims that believing and expressing (1) would have greater effect on him than (2). Why? Because he may not know that he is Ned Block. From this he concludes that there is an important difference between (1) and (2) with respect to causation of behaviour.
Note: Two different people may express utterances of (1) and both would have beliefs with the same content. Also, (1) and (2) express the same proposition according to a familiar way of individuating propositions (p. 619)
Narrow meaning is "in the head", it indicates supervenience on physical constitution. It also captures the semantic aspect of what is in common to utterances of (1) by different people.
Wide meaning, by contrast, depends on what individuals outside the head are referred to (i.e., wide meaning is not in the head). Wide meaning does not include narrow meanings. Utterance of (1) and (2) have the same wide meaning, but not the same narrow meaning.
How are narrow meaning/content and wide meaning/content relevant to psychological explanation? Well, narrow meaning (a narrow content) is better suited to predicting and explaining what someone decides or does, as long as the information about the external world is ignored (p. 620).
Narrow meaning also determines function from expressions and contexts of utterance onto referent and truth value. How? For example, when you and I say the word 'I' in (1), there is something we share, some semantic aspect of the word 'I' that in your context maps your token onto you and in my context maps my token onto me.
Block states that: (a) narrow meaning of "I" does not include one's conception of oneself; (b) shared semantic of 'I' does not suggest that these shared semantic aspect are exactly the same; (c) the narrow/wide distinction as Block describes it applies only to tokens and not types. (Will explain latter).
Q: What do considerations raised so far have to do with a semantics for psychology?
A: First, semantics for psychology should explain the distinction between narrow and wide meanings. Second, the theory ought to say why it is that narrow and wide meanings are distinctively relevant to the explanation and prediction of psychological facts (including behavior). Third, the theory ought to explain why and how narrow meaning determines a function from the context of utterance to reference and truth value (p. 621).
Block has so far been talking about sentences with indexical, but it can be extended to natural kinds of terms. He presents one of his own versions of the famous Twin Earth story (see p. 621-622 for details of the story). He asks us to consider Teen (of Earth) and her Twin Earth, Teente. The two individuals on the two worlds are identical in every respect except that they are distinct. Teen’s hero is Michael Jackson, while Teente hero is a distinct, but indistinguishable Michael Jacksonte. Teen and Teente each have thought with:
(3) Michael Jackson struts.
If they both would sincerely say “Michael Jackson has supernatural powers,” then, Block says they share the same delusion. This is narrow meaning and narrow content.
But we can regard the meaning and thought contents as distinct simply in virtue of the fact that Teen is referring to Michael Jackson and that Teen te is referring to Michael Jackson te. This is wide meaning and content.
There are two basic facts on which the narrow/wide distinction is based. One is that how you represent something that you refer to can affect your psychological states and behaviour. Furthermore, there is more to semantics than what is in the head. The contents of the head of a person who asserts (3) together with the fact that Michael Jackson struts, are not enough to determine whether (3) is true or false, since the truth value depends as well on whom "Michael Jackson" refers to.
Since the truth value of a sentence is determined by the totality of semantic facts, plus the relevant facts about the world, there is more to the totality of semantic facts about the sentence than is in the speaker's head. The "extra" semantic facts are about what the referring terms in the sentence refer to.
Thus, as in the indexical case, wide meaning and content are not well suited to explain change of mental state and behaviour. For example, the wide meaning of "water is wet" (in English - not Twin English) is the same as that of "H20 is wet" despite the possible difference effects of believing these sentences on mental state and behaviour (p.623).
Why has Block gone on about this desideratum (the narrow and wide distinction) at such length? Well, because the version of CRS that he is defending characterizes narrow meaning in terms of conceptual role. Harman's (1982) version of CRS does not deal with narrow content or meanings. Harman conceptual roles involve perceptual and behavioural interactions with objects in the world (so meaning is wide), while Block's stop at the skin. Block says that he prefers his own.
Consider Putnam's original Twin Earth story. There are substantial semantic differences between my twin's and my meanings and thought content because of the differences in physical and social environment. Nevertheless - there is some aspect of meaning in common to what I say and what he says. So it is the narrow meaning and content that is useful in stating nomological generalizations relating to thought, decisions, and actions.
CONCEPTUAL ROLE SEMANTIC
AND TWO-FACTOR THEORY
The idea of two-factor version is that there are two components to meaning: (1) conceptual role component that is entirely "in the head" (narrow meaning), and (2) external component that has to do with relations between the representations in the head (with their internal conceptual roles) and the referents and/or truth conditions of these representations in the world. This two-factor approach derives from Putnam’s argument that meaning could not both be "in the head” and also determine reference. The two-factor approach can be regarded as making conjunctive claim for each sentence- what its conceptual (or functional) role is and what its (say) truth conditions are (p. 627).
Q: What is the exact nature of the external and internal factors?
A:
The external factor is clarified by causal theory of reference or by
a theory of truth conditions, while internal factor (conceptual role) by the
causal role of the expression in reasoning and deliberating. A crucial
component of a sentence's conceptual role is a matter of how it participates in
inductive and deductive inference - eg. “Felix is a CAT” leads to infer that
Felix is an animal. Conceptual role is total causal role abstractly described
(p.627).
Important question for CRS: What counts as identity and difference of conceptual roles? For example, reasoning from Cat to that it is an animal - most people reach the same conclusion.
But CRS may face 'collaborative information' problem. Suppose you are prepared to infer from 'TIGER' to 'DANGEROUS' and I am not. Do our ‘Tigers’ have the same conceptual role or not? Even more significantly, what if we differ in inferring from 'TIGER' to 'ANIMAL' – Thus individuation of conceptual roles is needed (p.629).
I am not certain that Block is convinced that CRS theory is able to individuate conceptual roles—he definitely did not offer clear explanation on how to. However, he did mention that CRS cannot use analytic/synthetic distinctions because CRS cannot appeal to primitive (undefined) elements of language in terms which other elements are defined. In CRS, conceptual role is supposed to determine narrow meaning.
Role of English: Block only sometimes uses English to be the language of thought, but for the convenience he assumes that English is used only for communication and that all thought is in a language that does not overlap with English. It is in mentalese, for example, mentalese uses its language of thought symbols to develop a concept 'CAT'. This concept gets translated into language of communication. An analogy to this would be a computer. Decimal numbers are entered and displayed, but computation is done using binary number system - eg. 3 + 7 = 10 is in binary 011 + 111 = 1010. So we can speak of conceptual roles in English expressions even when we adapt the view that the internal computation is entirely in mentalese (p.633).
Clarification of the notion of standard association:
(1) English language is a social object - in speaking of the conceptual roles of English Block does not intend a theory of that social object. Conceptual role is meant to capture narrow meaning.
Since causal roles differ from person to person, CRS deals with idiolect narrow meaning rather than public language meaning.
(2) The existence of the mechanism that effects the standard association is empirical questions for "language mode" (p.633).
(3) The point of appeal to the language model is that it works without intervention of any intentional states. The language model works the same in lying and truth telling; the difference is to be found in the mentalese message (p.634).
(4) Intentionally left blank.
(5) There is nothing in CRS position that requires that sentence spoken have the same meaning as the sentence thought.
If language of thought hypothesis is true, two type of theories of meaning are possible - one for internal language, and one for external language. Block focuses of course on internal i.e. narrow meaning in idiolect. Final point of clarification: Though Block advocates CRS, he admits that he is far from a true believer.
In the past, Block (1978) argued against functionalism (explanation why he is embracing it now can be found on p.635).
Once again, the version of CRS Block has been talking about is a "two factor" version, in which conceptual role factor is meant to capture the aspect (or determinant) of meaning "inside the head", whereas the other is meant to capture the referential and social dimensions of meaning.
Block claims that Harman's version makes do with one factor CRS. How does he do without referential and social factors? By making his one factor reach out into the world of referents and into the practices of the linguistic community (p.636).
Here is Dretske's remark to Harman’s one factor:
"It sounds like magic: signifying something by multiplying sound and fury. Unless you put cream in you won't get ice cream out no matter how fast you turn the crank or how sophisticated the "processing". The cream, in the case of cognitive system, is representational role of those elements over which computations are performed. And the representational role of a structure is, I submit, a matter of how the elements of the system are related, not to one another, but to the external situations they express (p.638).
But the cream according to two-factor theory is conceptual role together with Dretske's representational role. Since CRS puts in Dretske's cream, plus more, there is no mystery how you get ice cream out of it.
Two Factors or One Factor
One might speak of Harman's conceptual role as "long armed" as opposed to the "short armed" conceptual roles of two-factor theory. Block’s objections to Harman is that he does not see how Harman can handle the phenomena one would ordinarily think of as being in the purview of a theory of reference without extending his explanation to the point where it is equivalent to the two-factor account. The point emerges as one looks at Harman's responses to problems that are dealt with by familiar theories of reference.
Here Block goes to some extent to show that Harman's "long armed" CRS is basically his "short arm." Harman should provide a good reason for thinking that the outside-the-body part of his long-arm conceptual roles differs from the referential factor of two-factor theory. One can easily transform a theory of the sort he advocates into a theory of the sort Block has been advocating. If you take Harman's long-arm conceptual roles and "chop off" the portion of these roles outside the skin, you are left with Block’s short-arm conceptual roles. If the outside-the-body part that is chopped off it adds up to some familiar sort of theory of reference, then the difference between Harman's one-factor theory and two-factor theory is merely verbal.
OVERVIEW
The focus of the rest of the paper is to show how CRS satisfies the desiderata and to compare CRS with other semantic theories in this regard. Block will address two quite different (but compatible) kinds of semantic theories: reductionist and nonreductionist. A reductionist semantic theory is one that explains the meaning in nonsemantic term. A nonreductionist is the one that does not have reductionist aims. The nonreductionist theories Block will mention are possible - worlds semantics. (Davidsonian semantics and Kalzian semantics). The reductionist theories are CRS theories, by which Block means theories that explain the semantic in terms of the mental and what he calls "indicator" theories (metaphor for the semantic is the relation between a thermometer and the temperature it indicates, or the relation between the number of rings on the stump and the age of the tree when cut down). These theories regard the nomological relation between the indicator and what it indicates as the prime semantic relation (p. 639).
Note: The reductionist/nonreductionist undertakings are compatible.
Since CRS in the version Block is promoting is a two-factor theory, he says that it requires the partnership of a reductionist truth - conditional theory. An indicator semantics is a good candidate. Another candidate that is both truth - conditional and reductionist is Field’s interpretation (1972) of Tarski (p640).
The rest of the paper shows how CRS satisfies the desiderata sketched above and also will contrast CRS's treatment with treatments possible for other approaches.
What Is the Relation between Meaning and Reference/Truth?" (Desideratum1)
In other words, what is the relation between the two factors? Are the two factors independent? Do they fit together in a coherent way? Block argues that the conceptual role factor is primary because it determines the nature of the referential factor, but not vice versa.
To illustrate: What makes the causal theory of reference true? Facts about how our language works - specifically how it applies to counterfactual circumstances. It is possible that Moses did not do any of the things the Bible says he did. We use names such as "Moses” to refer to the person who carries the right causal relations to our uses of the name even if he does not fit the descriptions we associated with the name. This is a fact about the conceptual role of names one that can be discovered just by thinking about intuitions about counterfactual circumstances (p.643).
Briefly, what theory of reference is true is a fact about how referring terms function in our thought processes. This is an aspect of conceptual role. So it is the conceptual role of referring expressions that determines what theory of reference is true. Conclusion: the conceptual role factor determines the nature of referential factor (644).
Note the important differences between saying that the conceptual role factor determines the nature of the referential factor and saying that the conceptual role factor determines reference. Block holds the former, but not the latter. Two factor-theory is compatible with a variety of different mapping from a single conceptual role onto aspects of worlds. For example, a word with the conceptual role of our 'water' could map onto one substance here, another on Twin Earth, etc. What is in the head - conceptual role - determines the nature of reference without determining reference itself.
So it is the referential factor (as described in a theory of reference) that determines that 'water' picks out H20 on Earth, but XYZ on Twin Earth.
For example, on a causal theory of reference, this will be held to be a matter of causal relation to different liquids in the two contexts. But since the referential factor must take context into account in this way in order to impose reference, it will determine the function from context to reference (p.644).
What Is the Connection between the meaning of an expression and Knowing or Learning Its Meaning? (Desideratum2)
CRS says meaning is conceptual role. If someone uses a word (or a word functions in her brain) that has the conceptual role of 'dog', then the word in question means the same as 'dog'…For a word to have proper use is for it to function in a certain way; hence someone whose word 'dog' functions appropriately thereby knows the meaning of dog; hence evidence of function can be evidence of knowing the word. CRS allows us to see how knowing meaning is related to our ability to use language.
Let us return to the familiar claim that to know the meaning of a sentence is to know its truth conditions…the way the person represents the set of possible worlds may not capture its meaning (p.646).
Though it is not obvious that knowing truth conditions guarantees knowing meaning, the opposite is more plausible. As Harman has pointed out, CRS can explain this in the following way: normal users of language understand certain metalinguistic ideas, such as the disquotational use of 'true', and this is what gives them knowledge or truth conditions. The conceptual roles of 'true' and nonsemantic terms yield knowledge of biconditionals like "Snow is white" is true if and only if snow is white. But even if knowing a meaning involves knowing truth conditions, one should not conclude that knowing a meaning is knowing truth conditions. CRS account of learning can be illustrated by its solution to what might be called Fodor's Paradox, which goes like this.
1. Learning the meaning of a word is a matter of hypothesis formation and testing.
2. When we learn a new English term (e.g. chase) we can do so only be hypothesizing of definitions in terms already known (including terms of the language of thought).
3. The history of attempts to define English terms "decompositionally" (e.g. "try to catch") has been dismal failure. This suggests that most English terms cannot be defined.
4. Therefore, when term like 'chase' is learned by hypothesizing a definition in terms of a single term of the language of thought 'CHASE' which has the same meaning as 'chase'. In other words, the typical word-learning hypothesis has the form: 'chase' means 'CHASE'
5. Therefore for most terms of English we grasp them only because they correspond to (indeed, are standardly associated with) innate terms of mentalese.
Block calls the argument a paradox because the conclusion is unacceptable-did the nature foresee that we will invent enzyme, transistor or radars, for example. So, what premise must go? 1st one sounds plausible-example of that would be the types of error children make when they begin to speak.
CRS militates against 2. According to CRS, the way we learn a new English term need not be a matter of definitions at all. Rather, the term (or its newly formed mentalese standard associate) comes to have a certain function. To the extent that hypothesis are involved, they are hypotheses about how the term functions in thought, reasoning, problem solving, and so forth (647).
One learns a certain relations among the new terms themselves (e.g. in studying physics, the relation between force and mass nether which can be defined in old terms) and some relations between the new terms and old terms.
Note that CRS is not a psychological theory. In particular, though it can tell us that Fodor’s second premise need not be true, it is compatible with its actually being true. For it is compatible with the CRS account that the way one learns to use a new term correctly is by linking it to a term one already has that functions properly (p.648).
What
Makes Meaningful Expressions Meaningful?
According to CRS, what makes an expression meaningful is that it has a conceptual role of a certain type, one that we may call "appropriate". The difference between 'cat' and 'glurg' is that 'cat' has an appropriate conceptual role. While 'glurg' does not…the difference between meaningful expressions with different meanings ('cat' and 'dog') is a conceptual role difference within the category of appropriate conceptual roles.
What is the difference between sentences that have and sentences that lack conceptual roles? The answer is, according to CRS, certain causal properties. But why do sentences have the causal role properties they have? The answer is to be sought, so Block proposes, of the same sort that one would give to "Why genes have the causal properties they have?"
CRS aims for a reductionist account, i.e., a naturalistic-reductionist account, in proposing to explain a semantic property in terms of a naturalistic nonsemantic property (Causation). CRS’s reductionism and naturalism allow it to promise an answer to "What makes meaningful expression meaningful?” Nonreductionist approach cannot answer this questions (p.649).
Block argues that the most of the standard approaches have been primarily concerned with the relations among meanings not the nature of meaning itself- e.g. concerned with an aspect of compositionality: how meaning of large elements such as sentences are related to the meanings of smaller elements such as words. The main aim of most of the standard approaches to semantic has been to correlate meanings with certain objects, so that relations among meaning are mirrored by the formal relations among the corresponding objects. These approaches have often been concerned with a purely descriptive project, not with explaining the nature of meaning (p.650).
There are two competing families of approaches to semantics that are reductionist and therefore do have genuine answers to the questions posed in the desiderata above.
One of them is the approach of reducing meaning to mental content. Block calls this type "Gricean". This approach reduces speaker meaning to the content of speaker's intentions. An intention of a speaker's utterance would be to affect the propositional attitudes of hearers (listener) in certain way. This approach, however, has little to contribute to the project Block is discussing.
Another option is Searle's reduction of intentionality to the brain or whatever has "equivalent causal powers". This approach allows Searle to avoid physiological reduction. Block does not like this one much. Another Gricean option is that championed by Schiffer and Loar … they connect the reduction of meaning to the mental with a functionalist reduction of the mental. A major difference between the functionalism-based Gricean theory and CRS is that the Gricean theory is not committed to any sort of representationalism (p.653).
Loar's argument is that a theory of meaning should not depend on speculative psychological claim such as representationalism. Block’s objection is this: if representationalism is false, CRS is certainly false. But if representationalism is true, Loar is stuck with an intention-based semantic for external language plus a conceptual role for internal language, while CRS makes do with the CR of semantics for both types of language.
In sum, the Griceans cannot claim that their account is to be preferred to CRS on the ground that their account has no empirical vulnerability, since both accounts have an element of empirical vulnerability (p.655).
Why
Is Meaning Relative to Representational System? (Desideratum3)
It is because the conceptual role of a symbol is a matter of how it functions in a representational system (this is the very reason why CRS is sometimes called functional role). How a representation functions in a system depends on the system. If meaning is function as CRS dictates, then the meaning is system relative (p661).
What
is Relation between Meaning and Mind/Brain? (Desideratum5)
How does brain grant meaning on its representations (recall that Block is ignoring mind and concentrating on brain)? Answer: For a person to grasp the meaning of a word is for the word (or its standard mentalese associate) to have a certain causal role in his or her brain
The difference between grasping a meaning and not grasping it is a difference in the causal role of entities in the person's brain and differences in such causal roles can make for differences in behavior and the rewards that are contingent on behavior (p.663).
What
is the Relation between Autonomous and Inherited Meaning? (Desideratum6)
Inherited meaning like those linguistic expressions on this page requires translations or transliterations into the language of thought of a reader or hearer for their understanding. Autonomous meaning, the kind of meaning of the elements of the language of thought itself, requires no reading or hearing, and thus no translation or transliteration in order to be understood. The questions that were raised are: What is autonomous meaning? What is inherited meaning? What is relation between autonomous and inherited meanings? Is one reducible to the other? Or are they both manifestations of a single type of meaning? Or are they unrelated phenomena with only superficial resemblance?
CRS answers to the first two questions are: autonomous meaning is conceptual role and so is inherited meaning…the conceptual roles of external language are inherited from those of internal language. So inherited meaning is inherited from autonomous meaning (p. 664). CRS explains the difference between autonomous and inherited meaning without giving up a unified account explanation of the two types of meaning. English written words and their utterances affect one another (via their effects on internal language)
If causal theories of reference can be made to work, they potentially have more to say about the relation between autonomous and inherited meaning than nonreductionist theories such as possible-worlds semantics, situation semantics, Davidsonian semantics, and other, because causal theories of reference can say something about the similarities and differences between the causal chains leading to "cat" and "CAT" that explains similarities and differences between the two representations. But causal theories of reference cannot capture the aspect of meaning inside the head.
For example, they cannot capture the aspect of sameness in meanings of the sentences of me and my twin on Twin Earth (despite the difference in our causal chains outside our heads). Further, the theory that Block is promoting can appropriate whatever successes causal theories of reference may have. Recall the version of CRS that Block favours is part of two-factor theory, the external factor of which can adopt aspects of a causal theory of reference account (explanations, justification). In sum, causal theories of reference cannot accomplish the task Block has set; and whatever they can accomplish can be appropriated by two-factor version of CRS (p.665).
Compositionality
(Desideratum 4)
Why is narrow meaning relevant to explanation of behaviour; and why is it relevant in the same way for me and my twin? Answer: Since my twin and I are physically identical, all of our representations have exactly the same internal causal roles, and therefore, the same narrow meanings. But why is narrow meaning relevant to the explanations of behavior in the first place? Answer: To have an internal representation with a certain narrow meaning is to have a representation with certain likely inferential antecedents and consequents. Hence, to ascribe a narrow meaning is to ascribe a syndrome of causes and effects, including in some cases, behavioral effects -or at least impulses on motor-output neurons (p.668).