# Non-Doxastic Attitudes and Semantics of Natural Language

Project Overview: The project ‘Non-doxastic Attitudes and Semantics of Natural Language’ lies at the intersection of contemporary analytic philosophy, linguistics, and cognitive science. It addresses the problem of ascribing attitudes, primarily from the viewpoint of natural language semantics. When philosophers talk about ‘propositional attitudes’, they think of relations expressed by the operators such as ‘believes that’, ‘wants that’, ‘fears that’, ‘wonders if’, etc. These relations are usually conceived as holding between a  subject and a proposition, i.e., something that can be true or false, or at least sensibly stated after the word ‘that…’. To use an example, the sentence ‘Wojciech hopes that he will receive funding’ expresses a relation between Wojciech and the proposition that Wojciech receives funding.

The aim of the project is to propose a cognitively adequate and possibly uniform semantic theory of sentences reporting propositional attitudes, with a focus on the so-called ‘non-doxastic’ attitude reports. We use this term for the attitudes which either do not imply that the subject believes in the propositional content of her/his attitude (like ‘wonders’), or additionally include a certain non-doxastic component like emotion, evaluation, etc. (examples include: ‚fears’, ‘is glad’, ‘regrets’).

Research Problems:  It is well recognized that attitude reports in general pose fundamental problems for standard approaches to the semantics of natural language, taken by philosophers, linguists, or logicians. Roughly speaking, their truth conditions are very sensitive to simple modifications of the content expressed by the relative clause, including the modifications which are generally not semantically significant in other types of contexts. A basic illustration of that phenomenon concerns substitutions of co-referring terms. Already Frege and Russell have observed that a sentence such as (1) below can be true while the corresponding ascription (2) false at the same time, even though ‚the Evening Star’ is another name for Wenus, so anything which is true about the latter should be true about the former, as well:

(1) John believes that Wenus is a planet.
(2) John believes that the Evening Star is a planet.

Plainly, John may not be aware of the fact that the Evening Star is Wenus, and so the second ascription is inaccurate. Another persisting problem is whether attitude verbs support logical entailments. We can imagine that John actually believes in a proposition whose logical form is given by the scheme: $(p \land q) \to r$. At the same time, John may not believe in the proposition of the form $\neg r \to (q \to \neg p)$, as he does not recognize a logical entailment which holds between these two propositions.

However, non-doxastic attitude ascriptions are even more problematic. Consider the following pair of sentences:

(3) The murderer of Smith has been convicted.
(4) There is a person who murdered Smith.

The first sentence trivially entails or presupposes the second one, and so anyone would likely infer (4) from (3). For this reason, the ascription such as (5) below seems to entail (6):

(5) Anne believes that the murderer of Smith has been convicted.  (de dicto)
(6) Anne believes that there is a person who murdered Smith.

Yet, the analogical observation does not hold about the ascriptions of non-doxastic attitudes. Take ‘want’ as an example. We cannot infer the ascription such as (8) or (9) below from (7):

(7) Anne wants the murderer of Smith to be convicted.
(8) Anne wants there to be a person who murdered Smith.
(9) Anne wants there to be a person who murdered Smith and for him to be convicted.

The second and the third ascriptions attribute a perfidious desire to Anne, namely, that Smith would have been killed by someone. But ascription (7) does not have such a reading, plainly, it can be true when Anne does not have the perfidious desire about Smith. Rather, (7) presupposes that Smith has been already murdered or that Anne believes so. Anyway, (7) does not entail (8), nor (9).

We can illustrate the problem with another pair of attitude verbs. Consider first:

(10) Socrates knows that there are honest men.
(11) Socrates knows that there are men.

Granted that Socrates is rational, we should infer (11) from (10). Yet, we should not make a corresponding inference in the case of ‘wonder’:

(12) Socrates wonders whether there are honest men.
(13) Socrates wonders whether there are men.

Clearly, (12) does not entail (13). Moreover, the entailment may not hold, even though we substitute the embedded clause with one that is, in a sense, equivalent to it. Consider (14) and (15) below, where the latter sounds like a more wordy paraphrase of the former:

(14) The president has been assassinated.
(15) The president is dead and has been assassinated.

These two sentences seem to “say the same thing in different words”. However, we can get different ascriptions, if we placed these two sentences in the scope of a non-doxastic attitude verb, e.g.,

(16) John wonders whether the president has been assassinated.
(17) John wonders whether the president is dead and has been assassinated.

The ascription (16) may be true if, e.g., John already knows that the president is dead and is thinking about what the cause of the president’s death was. No such reading is available for (17): this ascription clearly entails that John wonders whether the president is dead and so it is false under the aforementioned assumption.

In light of the presented examples, the problem of determining the truth conditions of non-doxastic attitude ascriptions seems to be more challenging than for the ascriptions with ‘knows’ or ‘believes’. As we have seen, the non-doxastic attitude verbs do not, in principle, support these entailments which are generally taken to be valid in the case of the doxastic attitude verbs. Furthermore, it is not a lack of subject’s specific knowledge or his/her cognitive limitations that blocks the entailments illustrated by (7)-(17). In the case of non-doxastic attitudes, a rational subject may be perfectly aware that $p$ entails $q$, have an attitude towards $p$, and yet not have the same attitude towards $q$. The problem is a genuinely semantic one, namely, the second ascription in each pair considered above expresses a \textit{content essentially independent} from what the former ascription says, which is puzzling in light of entailment dependencies between their complement clauses. Hence, the non-doxastic attitude ascriptions provide a real challenge for standard semantic theories.

Project objectives: The main goal of the project will be to provide a solution to the presented problem concerning the logical and semantic properties of non-doxastic attitude ascriptions. In order to achieve this goal, we will address the following research questions:
– how acceptance/rejection of an attitude ascription in a context depends on the formulation of relative clause,
– which aspects of the relative clause (i.e., which components of their meaning or structure) affect the ascription evaluation (in particular, we will test the hypothesis that broadly conceived ‘focus/background’ distinction is a crucial factor in that respect);
– what inferences are employed (or accepted/rejected) by users of language when they perform reasoning about someone’s attitude;
– the proper analysis of ‘propositional content’ which serves as the argument for the attitude verbs.
Our investigations on the presented issues will be essentially based on experimental research on natural language. At the final stage of the project, we will offer concrete semantic analyses for selected non-doxastic attitude verbs (framed in a suitable formal setting) and consider a possibility of extending the proposed analyses to a uniform theory of attitude reports.

The project team is: Wojciech Rostworowski (principal investigator), Katarzyna Kuś,  Bartosz Maćkiewicz

Student participants include: Anna Jasińska, Grzegorz Płuciennik, Maciej Chlebny