Substitutivity, simple sentences, and belief reports

eftir Heimi Geirsson


If we have anti-substitution intuitions in both simple sentences and sentences that embed them, such as belief reports, then it seems (assuming semantic innocence) that we should look for the reasons for those intuitions in the simpler cases, which in this case are the simple sentences, and then see whether the solution for the simpler cases can be extended to the more complex cases.1

In what follows I will provide the basic elements of such an account within a Russellian framework. The story I tell will not evoke pragmatic implications and it will explain why it is not cogent to substitute coreferential names even when our listener is in the know about the relevant identity; i.e., our anti-substitution intuitions are not mistaken.

Our judgments about the truth/falsity of simple sentences, as well as belief reports, depends to an extent upon our audience. If we are talking to someone who has never heard of Bob Dylan but who knows Robert Zimmerman, then we could truly say that Robert Zimmerman has a beautiful voice instead of saying that Bob Dylan has a beautiful voice. And if we are reporting Diana’s belief about the beauty of Bob Dylan’s voice to someone who is in the know about both of Dylan’s names, then it appears that we can, in that particular case, truly report the belief with either

1. Diana believes that Bob Dylan has a beautiful voice.

2. Diana believes that Robert Zimmerman has a beautiful voice.

At the same time we need to be careful not to assume that we always have a choice of how we report beliefs to someone in the know about two names naming the same person. Unlike the Bob Dylan case, different names of the same person are often associated with different characters the person might portray, and what concerns us most is those characters that are not a normal development of a person, but rather when one intentionally takes on a different identity and by doing so creates a significantly different persona.2 Because of this kind of created persona we often use different names to indicate that we are talking about the person as one character rather than another.3 For example, it is misleading and in a sense wrong to say that Clark Kent saved a city from a meteor, for Clark Kent as Clark Kent, i.e., in the persona or guise of Clark Kent, never performs such feats. He only performs such feats in the guise of Superman. Nevertheless, the proposition expressed by the uttered sentence is not false. Further, if someone believes that Superman saved the city then we don’t report that Clark Kent saved the city for that would indicate that he had revealed his true identity and shown his superpowers in the guise of Clark Kent. At the same time it is not false to so report for the proposition expressed, namely that Clark Kent saved the city, is true. The information conveyed by the sentence is misleading and wrong while the information contained in the sentence, namely the proposition it expresses, is correct as the proposition expressed is true. If our intuition is working with the information conveyed by the sentence, then we correctly resist substitution. If our intuition is working with the informationcontained in the sentence, then we (Russellians, at least) don’t resist substitution. If we don’t make this distinction between the information conveyed by the sentence and the information contained in the sentence, then it is less likely that people who have conflicting intuitions about the substitutivity can reach a common ground about the nature of their disagreement.

If someone knew Superman’s true identity we should in general still refrain from using ‘Superman’ and ‘Clark Kent’ interchangeably when reporting events and when reporting beliefs. Supposing someone knew Superman’s true identity, it would still be misleading to report to her that Clark Kent saved the city, for that would suggest, misleadingly, that Clark Kent has shown his powers in the guise of, or in the persona of Clark Kent. And it would be outrageous to claim that if I tell someone that Kal El (Clark’s given name on Krypton) saved the city, my respondent never having heard the name ‘Kal El’, then I have thereby informed her that Superman saved the city, even though „Kal El saved the city“ and „Superman saved the city“ express the same proposition.4 Somehow we need to account for these scenarios and the naïve Russellian needs to do so without giving up singular propositions as being the objects of beliefs.


It is plausible to maintain that we file information away in our mind and that we organize the information about individuals so that different pieces of information about what we take to be the same individual are kept together in a web of information.5 I might, for example, collect information about Clark Kent into one web, and information about Superman into a different web, because I take Clark Kent and Superman to be different individuals, and I might collect information about Paderewski the Polish politician into one web and information about Paderewski the German politician into a different web, because I take them to be different individuals. Some beliefs are fairly central to the web and very important to the identity of the individual while other beliefs are less important and are often quickly forgotten and/or replaced. Central to a typical web of information about Clark Kent is that he is a journalist, wears glasses, is rather nerdy, and has a tendency to disappear when trouble is near. Information less central would include, e.g., some of the mundane events in Kent’s life, such as what he had for lunch, the color of his favorite suit, etc. Central to a typical web of information about Superman is, e.g., that he wears a strange looking body suit and a cape, that he can fly, that he is very strong and has powers that humans don’t possess, and that he has a knack for saving innocent people when they are facing grave danger.

If I take Clark Kent and Superman to be different individuals then I have one web of information about Clark Kent as Clark Kent and a second web of information about Clark Kent as Superman. I have, in effect, a Clark Kent web and a Superman web. I suggest that the web of information I have about an individual is typically accessed by a name we associate with the web, and that we then elicit information from the web. Consequently, when I hear a sentence containing ‘Clark Kent’ then information from the Clark Kent web is elicited, and similarly, when I hear a sentence containing ‘Superman’ then information from the Superman web is elicited. The information elicited determines how I represent the object in the proposition believed, and thus determines to a great extent how I believe the proposition. It is thus of great importance how a proposition is presented to me, i.e., whether a proposition containing Clark Kent as a constituent is presented to me with a sentence that contains ‘Clark Kent’ or ‘Superman’, for it depends on the name used in the sentence which web of information I draw from or, for that matter, form a new web as might happen if I take it that the name is of someone about whom I think I have no information.

The web of information I have on Clark Kent has to figure heavily in how I believe a singular proposition that has him as a constituent when the proposition is presented to me with a sentence containing ‘Clark Kent’, for the web of information I have about him as Clark Kent is what I draw from when I represent him as Clark Kent, and it thus determines largely the way in which I believe the proposition. Similarly, when I believe a singular proposition that has Clark Kent as a constituent and the proposition is presented to me with a sentence containing ‘Superman’, then the web of information I have about Clark Kent as Superman is what I draw from when I represent his as Superman.

The picture presented provides a simple explanation for why I might assent to, for example,

3. Clark Kent saved the city

and dissent from

4. Superman saved the city

while being fully rational. The reason is that I believe the proposition expressed in different ways. The names used to designate the object in the proposition expressed by the sentences elicit information from different webs of information. The reason that the names elicit information from different webs is not that I take the names to refer to different individuals, for even if I am in the know about Clark Kent being Superman the name ‘Clark Kent’ still typically elicits information from the Clark Kent web, and ‘Superman’ typically elicits information from the Superman web. The reason that the names elicit information from different webs is simply that the names are associated with different webs and so I typically access and elicit information from different webs when using the names. For example, if someone who has witnessed Superman saving the city tells me that Clark Kent saved the city, then that conveys misleading information, for the use of ‘Clark Kent’ prompts me to think of Clark Kent as Clark Kent saving the city and so indicates that he saved the city while in his Clark Kent persona; a highly unlikely scenario. I am in the know about the identity and I still resist substitution, the reason being that even though the two sentences express the same proposition, the information conveyed is very different. The information conveyed does not affect the truth value of the proposition expressed, but it does affect the cogency of free substitution.


An account of belief reports advances along similar lines.6 The choice of words in the belief report is important because it conveys some information about how the belief is held. It is reasonable to assume that those who know about Clark Kent have similar information at the center of their Clark Kent web and so there are some central information that come to most peoples’ minds when they think about Clark Kent. Same goes for Superman; it is reasonable to assume that those who know about him have similar information at the center of their Superman webs. When I report that Lois believes that Superman can fly, then I am indicating that Lois holds the belief in such a way that it is Clark Kent as Superman who can fly. If I were to report that Lois believes that Clark Kent can fly, then I wrongly indicate that Lois believes that Clark Kent, as Clark Kent, can fly. When I report that Lois believes that Superman can fly, then how I report the belief carries with it significant information, since I am assuming that my audience has similar key information about Superman as the rest of us.7 My report, in addition to indicating which proposition Lois believes, provides some information about how she believes the proposition. But while my report provides some information about how Lois believes the proposition, the information is not implicated pragmatically as there is no assumption of the speaker entertaining or believing any proposition(s) that might be pragmatically conveyed, and there is no assumption about the listener believing any proposition(s) that might be pragmatically conveyed. The information as to how Lois believes the proposition is instead conveyed by a choice of words used to express it.

In a very insightful paper Kent Bach explores a novel account of belief reports; an account that is motivated by the same concerns as the that has troubled Russellians, namely that belief reports do not seem to capture with great accuracy what one believes.8 He resorts to a view according to which belief reports do not report propositions believed; instead a belief report describes a belief. On the view he endorses it can be true that „Peter believes that Paderewski had musical talents“ and at the same time be true that „Peter disbelieves that Paderewski had musical talents,“ ‘Paderewski’ referring to the same person both times, because the ‘that’-clauses do not specify what it is that Peter believes; specifically, the two reports do not indicate that Peter believes and disbelieves the same proposition. The ‘that’-clause does not specify a belief as much as it describes the belief. Bach thus rejects what he call the specification assumption, which assumes that the ‘that’-clause in a belief report specifies the thing (proposition) that the believer must believe if the belief report is true.9 In fact, Bach claims that it is the specification assumption that creates the main problems for the accounts of belief reports that assume its truth.10

I prefer to avoid some difficulties that Bach’s view faces if we can retain his main insight together with the more traditional view that belief reports report propositions believed and, further, that they give some indication of how they are believed. Bach’s main reason for rejecting the specification assumption is that beliefs are finer grained than can be specified with a singular proposition. That, in turn, results in us ending up with belief reports that seem to run counter to our intuitions, such as when we substitute coreferential names in belief reports. But the view that one believes a proposition in a way accommodates the claim that exactly how one believes a proposition is not specified by the singular proposition believed. Further, while the singular proposition does not specify how one believes it, one’s choice of words when expressing the proposition can provide much needed clues as to how one believes the proposition. At the same time, the choice of words, while providing some help, cannot provide the details of how one believes the proposition expressed, for each person’s web of information associated with a name is likely to be somewhat different and the information elicited from the web upon hearing the name is likely to differ from one person to the next. The account presented to some extent captures Bach’s insight that a belief report describes a belief, for there is a sense in which the choice of words when one reports the belief describes how a belief is held, for using one name instead of another prompts one to elicit information from one web rather than another.


While I have so far concentrated on names, it is not only names that can prompt one to elicit information from one web of information rather than another. For example, a picture can work in the same way as a name in the sense that if one is shown a picture of Clark Kent as Superman, then that prompts one to elicit information from a different web than if one is shown a picture of Clark Kent as Clark Kent and, consequently, one may construct similar puzzles about assent/dissent, belief and their reports with the help of pictures instead of using names. In other cases a certain scent, just as a name, might prompt one to elicit from a web of information about someone, and a certain tune can prompt one to think of one certain individual in a particular way. While we should acknowledge that not only names can be used to create puzzles of the type we have been dealing with, the particular puzzles at hand have to do with singular propositions and we do tend to express them with the help of sentences.11

It is evident how the story told accounts for our intuition that someone might come to believe the same proposition all over again, without realizing it is the same proposition. I may, for example, take Paderewski the German politician to be different from Paderewski the Polish politician and so I have one two webs of information for Paderewski. Accordingly, I elicit information from one web when I take ‘Paderewski’ to refer to Paderewski the Polish politician, and I elicit information from a different web when I take it that ‘Paderewski’ is referring to Paderewski the German politician. Accordingly, I believe the relevant proposition in two different ways. The consistency problem, namely that one can assent to and dissent from the same proposition without detecting any inconsistency, is dealt with in a very similar way, for I don’t detect any inconsistency amongst my believes when the relevant propositions are believed in suitably different ways.

The translation problem and the no name problem can be dealt with once we acknowledge that the items in the web of information need not all be linguistic in nature. The best known and most widely accepted examples on non-linguistic representations are probably those that involve imagistic or pictorial representations, but we can probably represent objects in other ways as well, where some of those ways may depend on olfactory or tactile information that is not couched in words. Once we acknowledge this possibility, then the translation problem, where two people who don’t share a common natural language can nevertheless believe a proposition in the same way, can be solved. Two people who don’t share a language may nevertheless represent the object in a proposition in the same non-linguistic way, which allows them to believe the proposition in the same way. And once we acknowledge the possibility of non-linguistic representations being in the web of information then the no name problem is no longer an obstacle, as the person in question can have a web of information of non-linguistic items and she can access the web via non-linguistic clues, such as visual clues, smell or tactile information to name a few.

Final remarks.

The sketch of an account presented here does not rely on implicatures, pragmatic or otherwise, nor does it rely on mistaken intuitions and evaluations. While using the metaphor of a web of information as an explanatory tool, doing so does not say much about the nature of beliefs. It remains a sketch, because how exactly we represent objects and propositions is for philosophers to speculate about and cognitive scientists to discover.


1. In „Partial Propositions and Cognitive Content,“ Journal of Philosophical Research, 21 (1996): 117-128, I argue that the simple cases are indeed the more basic cases, and that we should therefore seek to solve them first.

2. David Pitt has argued that we don’t name persons; instead we name personas. That view has some serious metaphysical implications that I can not go into here. I will simply assume the familiar view that names name objects, not personas.

3. I will not here try to provide precise conditions for when it is appropriate to use one name rather than another. As an example, though, while it does not seem appropriate to use only ‘Bob Dylan’ to refer to the singer, i.e., it is quite likely that those who knew him before he adopted the name ‘Bob Dylan’ can appropriate refer to him as ‘Robert Zimmerman’, the same is not true of David Bowie and the name he used for the persona he created, Ziggy Stardust.

4. Had I informed her of Superman saving the city, then she would be justified in believing that Superman saved the city. However, I have argued elsewhere that the justification for believing a singular proposition is crucially tied to how one believes it, and so given how the proposition is presented one can not assume that when one is justified in believing a proposition is one way, one is thereby justified in believing the proposition simpliciter. For more on this see „Justification and Relative Apriority,“ and „Justification and Ways of Believing.“

5. Several philosophers have used the analogy of a file, including my former self, suggesting that it provides a plausible explanatory account of how information is filed away. That analogy seems misleading for at least two reasons; first, it suggests that all pieces of information are equally important and, second, it doesn’t explain why it is that some pieces of information are more central than other, and thus more readily recalled when one thinks for the object. The web analogy, on the other hand, much like Quine’s web of belief, suggests that some pieces of information are more central than other, and that some are therefore more readily evoked than other.

6. In „Partial Propositions and Cognitive Content“ I argue that it is important to provide a cohesive account of beliefs and belief reports, and that the reason we resist substitutions in belief reports is found in how we believe propositions.

7. It certainly seems wrong, as Kent Bach writes, that „For us [who realize that Bruce Wayne is Batman], a belief which could be characterized as a that-Bruce-Wayne-is-a-wimp belief could equally be characterized as a that-Batman-is-a-wimp belief.“ „Do Belief Reports Report Beliefs?,“ Pacific Philosophical quarterly 78 (1997): 215-241, p., 229.

8. Brian Loar has also been a long time advocate of the view that propositions do not accurately capture beliefs.

9. See p. 221.

10. These would include, for example, the views of Salmon and Soames, Crimmins and Perry, Richard, and Schiffer. See M. Crimmins Talk About Beliefs (Cambridge: The MIT Press, 1992), M. Richard, Propositional Attitudes: An Essay on Thoughts and How We Ascribe Them (Cambridge: Cambridge University Press, 1990), S. Schiffer, „The ‘Fido’-Fido Theory of Belief,“ Philosophical Perspectives 1 (1989), pp. 455-480, „Belief Ascriptions,“ Journal of Philosophy 89 (1992), pp. 490-521.

11. Although we tend to focus on sentences as expressing propositions, it is possible to communicate the same proposition with and without the use of a name or an indexical. One might, for example, instead of using a name point at a person, or even flash a picture to indicate who we are talking about. Furthermore, if one flashes a picture, then it makes a huge difference, when talking about Clark Kent, whether the picture shows him as the journalist or as the superhero, just as it makes a huge difference whether one uses ‘Clark Kent’ or ‘Superman’.

« Til baka

Skildu eftir svar

Netfang þitt verður ekki birt. Nauðsynlegir reitir eru merktir *