On Propositional Indexing
I am interested in defending a conception of propositional content that has application to all intentional agents (i.e., agents with beliefs about the world). Since I hold that some non-linguistic animals have beliefs about the world, this requires an account of propositional content that has application to such animals. But how are we to go about attributing propositional content to the beliefs of non-linguistic animals? In answering this question, I draw on John Dilworth’s propositional indexing framework.
In his paper, “Semantics naturalised: propositional indexing plus interactive perception”, Dilworth advocates a propositional indexing view according to which cognitive states (understood as concrete causal occurrences) enjoy an isomorphic correspondence with propositions (understood as abstract truth-value bearing items). Maintaining such an isomorphism requires that we have some method for indexing a given proposition (such as the proposition “X is red”, with respect to some worldy object X) with a corresponding perceptual or cognitive state (such as an agent Z perceiving that X is red). This method, whatever it may be, may be identified with the specific epistemic conditions under which we would accept that the isomorphism in question holds. Moreover, Dilworth maintains that the epistemic conditions must be ones that can be met by individuals without any technical or specialised scientific knowledge. After all, it is part of our folk psychological practice to describe perceptual states in propositional terms. This imposes the constraint that the method for indexing propositions to cognitive states must be one which is available to everyone, including those lacking any specialised philosophical or technical expertise. Dilworth puts the point as follows:
Our understanding of propositional indexing is not intended to be restricted to specialised cognitive science procedure requiring technical expertise and detailed knowledge of such matters as the cognitive structures involved in perceptual functioning. Instead, the idea is that the everyday understanding by people in general, of when a particular proposition, such as “X is red” is true of a particular object X is to be correlated with a related understanding by such people of what kinds of behavioural evidence would justify a claim that the person had indeed correctly perceived the relevant fact. So the predominant epistemic issue is not the theoretical nature of propositional indexing as such, but rather the everyday conditions under which people in general would agree that it had successfully occurred.
Dilworth recommends that propositional indexing be unpacked in terms of what he calls “classification behaviour”:
For example, a paradigm kind of colour-related classification behaviour would be that of a person assigned a task of sorting some miscellaneous objects by their colour, and then putting object X into a box containing only red objects. This classification behaviour would provide evidence that the person P had perceived that object X was red. Consequently, if the person Q observing person P is considering the proposition that X is red, then Q would take person P’s classification as evidence that P had perceived that X is red, and hence that P’s relevant perceptual state S, whatever it may be, is indexed by the proposition ‘X is red’.
Significantly, Dilworth describes propositional indexing in third- rather than first-personal terms. This suggests a view according to which an agent’s cognitive state may be described as propositional just in case it is, in principle, possible for that cognitive state to be correlated with a proposition. However, the act of identifying the correlation need not be performed by the agent that is actually undergoing the cognitive state in order for the cognitive state to count as propositional. One upshot of this view is that the cognitive states of non-linguistic animals, which lack the ability to actually engage in propositional indexing, may nevertheless count as having propositional attitudes.
Dilworth’s observations about colour-related classification behaviour generalises to less overt types of classification behaviour. For example, a dog may be said to perceive a bone is edible just in case it is disposed to ingest the bone. Ingesting the bone, then, amounts to a type of classification behaviour (akin to the act of sorting bones into the set of edible items). Since the heuristic of classification behaviour is understood in dispositional terms, it is not necessary that the ingesting of the bone actually take place for the dog to count as perceiving that the bone is edible. Moreover, since a dog may engage in such classification behaviour without us having to attribute to the dog the concepts of “bone” or “edible”, the present account does not require concept possession as a prerequisite for having an agent’s perceptual state indexed by a proposition.
There is much that I find attractive in Dilworth’s framework. Specifically, I believe it provides the resources for an account of propositional attitudes that allows for such attitudes to be attributed to non-linguistic animals. (Admittedly, Dilworth may be reluctant to speak of propositional attitudes in this way, but that would only be because of his refusal to attribute semantic content to cognitive states in general and not because of any prejudice against such attributions in the case of non-linguistic animals. In short, both Dilworth and I agree in the even-handedness of our treatment of the cognitive states of linguistic and non-linguistic animals.) However, I want to conclude by highlighting a difficulty with Dilworth’s account; one that motivates a way in which my account differs crucially from his own.
Dilworth is committed to what he refers to as an “interactive theory of perception”, the considered version of which he puts as follows:
IP2: An organism Z perceives an item X to have the property of being F just in case X causes some sense-organ zi of Z to cause Z to acquire an X-related disposition D, such that D is an F-classification disposition.
Dilworth defines an F-classification dispositions as “a disposition, the manifestation of which is some F-classification behaviour. For example, on this account, to perceive that an object X is red is to acquire a disposition to classify object X in some red-related way.”
However, it is not clear that IP2 can accommodate cases in which an agent perceives things to be a certain way and yet fails to believe that things are that way. For example, suppose that I am led to believe (erroneously) that a room is equipped with special lighting that makes all the white objects in the room appear red. (However, there is no special lighting and all objects in the room are actually the colour they appear to be.) Due to my misinformation, I am disposed to sort all objects that appear red into the white box. According to IP2, it follows from my having this disposition that I do not perceive that the objects are red. But this gets things wrong. The thing to say is that I perceive the object as red, but I believe it is white. Consequently, IP2 is unable to accommodate cases in which the two—perceiving X is F and believing X is F—come apart.
Instead of IP2 I am committed to the following account of the indexing of propositional content to perceptual states:
(*) An agent has a perceptual appearance of X as having the property F just in case assenting to that appearance would dispose the agent to engage in F-classification behaviour with respect to X.
My account differs from Dilworth’s in two crucial respects. First, I shift from Dilworth’s focus on perceptual experiences of X as F (which Dilworth holds are factive both with respect to the existence of X and with respect to X possessing the property F) to a focus on perceptual appearances of X as F (which I take to be non-factive). Second, I see the propositional content of a perceptual appearance as parasitic upon the propositional content of the perceptual belief it would give rise to (even if it actually never gives rise to such a belief). Thus, I see beliefs as having inherent propositional content (directly indexed by the type of classification behaviour they would dispose an agent to engage in) and perceptual states as having derived propositional content (indexed by the propositional content of the belief they would generate if assented to).
The second point (which admittedly needs further unpacking) allows me to avoid the preceding difficulty facing Dilworth’s account. In the case of an agent who has a perceptual appearance of an object as red, but believes it is white, the agent’s perceptual state is indexed by the propositional content of the belief it would generate if it were assented to. Since assenting to the perceptual appearance would dispose the agent (in the preceding example) to engage in red-classification behaviour, the agent has a perceptual appearance of the object as red. This account of perception is anti-behaviourist (since there may be no behavioural indications of the content of an agent’s perceptual appearance) but still quasi-functionalist (since it sees the contentfulness of a perceptual appearance as partly dependent on the belief-forming function of perceptual appearances).
In his paper, “Semantics naturalised: propositional indexing plus interactive perception”, Dilworth advocates a propositional indexing view according to which cognitive states (understood as concrete causal occurrences) enjoy an isomorphic correspondence with propositions (understood as abstract truth-value bearing items). Maintaining such an isomorphism requires that we have some method for indexing a given proposition (such as the proposition “X is red”, with respect to some worldy object X) with a corresponding perceptual or cognitive state (such as an agent Z perceiving that X is red). This method, whatever it may be, may be identified with the specific epistemic conditions under which we would accept that the isomorphism in question holds. Moreover, Dilworth maintains that the epistemic conditions must be ones that can be met by individuals without any technical or specialised scientific knowledge. After all, it is part of our folk psychological practice to describe perceptual states in propositional terms. This imposes the constraint that the method for indexing propositions to cognitive states must be one which is available to everyone, including those lacking any specialised philosophical or technical expertise. Dilworth puts the point as follows:
Our understanding of propositional indexing is not intended to be restricted to specialised cognitive science procedure requiring technical expertise and detailed knowledge of such matters as the cognitive structures involved in perceptual functioning. Instead, the idea is that the everyday understanding by people in general, of when a particular proposition, such as “X is red” is true of a particular object X is to be correlated with a related understanding by such people of what kinds of behavioural evidence would justify a claim that the person had indeed correctly perceived the relevant fact. So the predominant epistemic issue is not the theoretical nature of propositional indexing as such, but rather the everyday conditions under which people in general would agree that it had successfully occurred.
Dilworth recommends that propositional indexing be unpacked in terms of what he calls “classification behaviour”:
For example, a paradigm kind of colour-related classification behaviour would be that of a person assigned a task of sorting some miscellaneous objects by their colour, and then putting object X into a box containing only red objects. This classification behaviour would provide evidence that the person P had perceived that object X was red. Consequently, if the person Q observing person P is considering the proposition that X is red, then Q would take person P’s classification as evidence that P had perceived that X is red, and hence that P’s relevant perceptual state S, whatever it may be, is indexed by the proposition ‘X is red’.
Significantly, Dilworth describes propositional indexing in third- rather than first-personal terms. This suggests a view according to which an agent’s cognitive state may be described as propositional just in case it is, in principle, possible for that cognitive state to be correlated with a proposition. However, the act of identifying the correlation need not be performed by the agent that is actually undergoing the cognitive state in order for the cognitive state to count as propositional. One upshot of this view is that the cognitive states of non-linguistic animals, which lack the ability to actually engage in propositional indexing, may nevertheless count as having propositional attitudes.
Dilworth’s observations about colour-related classification behaviour generalises to less overt types of classification behaviour. For example, a dog may be said to perceive a bone is edible just in case it is disposed to ingest the bone. Ingesting the bone, then, amounts to a type of classification behaviour (akin to the act of sorting bones into the set of edible items). Since the heuristic of classification behaviour is understood in dispositional terms, it is not necessary that the ingesting of the bone actually take place for the dog to count as perceiving that the bone is edible. Moreover, since a dog may engage in such classification behaviour without us having to attribute to the dog the concepts of “bone” or “edible”, the present account does not require concept possession as a prerequisite for having an agent’s perceptual state indexed by a proposition.
There is much that I find attractive in Dilworth’s framework. Specifically, I believe it provides the resources for an account of propositional attitudes that allows for such attitudes to be attributed to non-linguistic animals. (Admittedly, Dilworth may be reluctant to speak of propositional attitudes in this way, but that would only be because of his refusal to attribute semantic content to cognitive states in general and not because of any prejudice against such attributions in the case of non-linguistic animals. In short, both Dilworth and I agree in the even-handedness of our treatment of the cognitive states of linguistic and non-linguistic animals.) However, I want to conclude by highlighting a difficulty with Dilworth’s account; one that motivates a way in which my account differs crucially from his own.
Dilworth is committed to what he refers to as an “interactive theory of perception”, the considered version of which he puts as follows:
IP2: An organism Z perceives an item X to have the property of being F just in case X causes some sense-organ zi of Z to cause Z to acquire an X-related disposition D, such that D is an F-classification disposition.
Dilworth defines an F-classification dispositions as “a disposition, the manifestation of which is some F-classification behaviour. For example, on this account, to perceive that an object X is red is to acquire a disposition to classify object X in some red-related way.”
However, it is not clear that IP2 can accommodate cases in which an agent perceives things to be a certain way and yet fails to believe that things are that way. For example, suppose that I am led to believe (erroneously) that a room is equipped with special lighting that makes all the white objects in the room appear red. (However, there is no special lighting and all objects in the room are actually the colour they appear to be.) Due to my misinformation, I am disposed to sort all objects that appear red into the white box. According to IP2, it follows from my having this disposition that I do not perceive that the objects are red. But this gets things wrong. The thing to say is that I perceive the object as red, but I believe it is white. Consequently, IP2 is unable to accommodate cases in which the two—perceiving X is F and believing X is F—come apart.
Instead of IP2 I am committed to the following account of the indexing of propositional content to perceptual states:
(*) An agent has a perceptual appearance of X as having the property F just in case assenting to that appearance would dispose the agent to engage in F-classification behaviour with respect to X.
My account differs from Dilworth’s in two crucial respects. First, I shift from Dilworth’s focus on perceptual experiences of X as F (which Dilworth holds are factive both with respect to the existence of X and with respect to X possessing the property F) to a focus on perceptual appearances of X as F (which I take to be non-factive). Second, I see the propositional content of a perceptual appearance as parasitic upon the propositional content of the perceptual belief it would give rise to (even if it actually never gives rise to such a belief). Thus, I see beliefs as having inherent propositional content (directly indexed by the type of classification behaviour they would dispose an agent to engage in) and perceptual states as having derived propositional content (indexed by the propositional content of the belief they would generate if assented to).
The second point (which admittedly needs further unpacking) allows me to avoid the preceding difficulty facing Dilworth’s account. In the case of an agent who has a perceptual appearance of an object as red, but believes it is white, the agent’s perceptual state is indexed by the propositional content of the belief it would generate if it were assented to. Since assenting to the perceptual appearance would dispose the agent (in the preceding example) to engage in red-classification behaviour, the agent has a perceptual appearance of the object as red. This account of perception is anti-behaviourist (since there may be no behavioural indications of the content of an agent’s perceptual appearance) but still quasi-functionalist (since it sees the contentfulness of a perceptual appearance as partly dependent on the belief-forming function of perceptual appearances).
Labels: Philosophy of Language
