Experiential Attitudes are Propositional

Abstract

Attitudinal propositionalism is the view that all mental attitude content is truth-evaluable. While attitudinal propositionalism is still silently assumed in large parts of analytic philosophy, recent work on objectual attitudes (i.e. attitudes like ‘fearing Moriarty’ and ‘imagining a unicorn’ that are reported through intensional transitive verbs with a direct object) has put attitudinal propositionalism under explanatory pressure. This paper defends propositionalism for a special subclass of objectual attitudes, viz. experiential attitudes. The latter are attitudes like seeing, remembering, and imagining whose grammatical objects intuitively denote (events or) scenes. I provide a propositional analysis of experiential attitudes that preserves the merits of propositionalism. This analysis uses the possibility of representing the target-scenes of experiential attitudes by the intersection of all propositions that are true in these scenes. I show that this analysis makes available the usual (Russellian) account of intensionality and the common (Boolean) logic for entailments.

1 Introduction

In the philosophy of language and mind, mental attitudes are typically taken to be propositional attitudes. The latter are attitudes like believing, claiming, and hoping that agents hold towards propositions (or towards proposition-like, or propositional content-bearing objects). The propositional nature of mental attitude contents is supported by the possibility of substituting the embedded that-clause CP in some (!)¹ attitude reports by a DP of the form ‘the proposition $[_{\textsc {cp}}~\,]$’ [in (1)] (see e.g. King, 2002; Forbes, 2018), of evaluating attitude contents for their truth or falsity [in (2)] (see Hintikka, 1969; Moltmann, 2020), and of generating ‘new’ attitude contents through the application of logical inference rules [in (3): existential generalization] (see Montague, 2007; Sinhababu, 2015).

The propositional nature of attitude content is captured by Hintikka’s (1969) classical relational account of mental attitudes (see Stalnaker, 1988; von Fintel and Heim, 2021) . This account identifies attitude content [$=$ a proposition] with the set of possible worlds in which this content is true. Attitudes are then analyzed through the inclusion of this set² [in (4): $\{w: {\text{the Earth is round in}} ~ w\}$] in Bill’s doxastic alternatives $\textsc {dox}_{{\text{Bill}} , @}$ [$=$ the set of possible worlds that are compatible with Bill’s beliefs in @; see (4b)]:

The observation that a large number of attitudes displays the behavior in (2)–(3) and allows for (some variant of) Hintikka’s account in (4) has recently given rise to attitudinal propositionalism (see e.g. Felappi, 2021; Forbes, 2006; Grzankowski, 2014; Grzankowski and Montague, 2018; Montague, 2007; Sinhababu, 2015; Zimmermann, 2016).³ The latter is the view that all mental states have truth-evaluable propositions as their semantic (information) content. Attitudinal propositionalism is a general thesis about the truth-evaluability of attitude contents that is independent of any particular theoretical implementation. Thus, it is compatible with different relational accounts, including possible worlds-based accounts (e.g. Hintikka, 1969; von Fintel and Heim, 2021), situation semantic accounts (e.g. Kratzer, 2002; Elbourne, 2013), impossible worlds-accounts (e.g. Zalta, 1997; Jago, 2014), and ‘primitive propositions’-accounts (e.g. Pollard, 2015). Since propositions still play a central role in some non-relational accounts (viz. as the contents of concrete, content-bearing objects), attitudinal propositionalism is also compatible with accounts that replace propositions with proposition-like objects (e.g. facts, possibilities; see Vendler, 1967; Parsons, 1993; Pryor, 2007) and with accounts that replace propositions with content individuals (e.g. beliefs, claims, or fears; see Kratzer, 2006; Moltmann, 2020; Moulton, 2015).

Attitudinal propositionalism is supported by the belief report in (1a) [see (4) and (5)]. It is even supported by a number of attitudes that prima facie do not have a propositional content. Such attitudes include desires, whose reports typically embed infinitives or direct objects⁴ rather than that-clauses [e.g. (6a)] (see Larson, 2002) resp. [(6b)] (see Quine, 1956; Montague, 1973). Below, the semantic content of an expression X in a context c is abbreviated ‘$\llbracket X\rrbracket ^{c}$’.

Attitudinal propositionalism has been praised for its uniformity and ontological parsimony, and for its ability to capture natural language reasoning (more on this in Sect. 2). However, despite these merits, propositionalism has recently come under empirical pressure. This pressure stems from ‘objectual’ attitudes⁵ (see Forbes, 2000, 2006; Grzankowski, 2016; Montague, 2007). The latter are attitudes like liking, fearing, and worshipping that are reported through intensional transitive verbs with a direct object [e.g. (7a)]. Unlike the attitudes that are reported by (6a, b), objectual attitudes resist an analysis in terms of propositional attitudes. In particular, in contrast to (6b), (7a) is intuitively not equivalent to the result of supplementing its direct object, i.e. Mary, with a contextually given predicate (in (7b); see Zimmermann, 1993; Szabó, 2005; Grzankowski, 2016). This even holds for a supplementation with the lexically ‘poor’ predicate be there (or exist) [in (7c)] (see Forbes, 2006; Zimmermann, 2016; pace Parsons, 1997):

For example, John might not like Mary’s exemplifying any particular property (including her existence), but only Mary herself (see Grzankowski, 2016, p. 829; Szabó, 2005). In this scenario, (7b) and (7c) are false and, hence, not equivalent to (7a).

The unavailability of a propositionalist analysis for cases like (7a) has often been taken to suggest that all objectual attitudes that defy a syntactic analysis [see (6a)] or lexical paraphrase into a clause-embedding structure [see (6b)] are non-propositional. The present paper argues that this conclusion is too rash. In particular, the paper shows that a special subclass of objectual attitudes, viz. experiential attitudes (e.g. seeing, remembering, imagining), allows for a systematic propositional treatment. Experiential attitudes differ from other objectual attitudes in being relations to events or scenes, rather than to individuals [unlike (7a)]. Their common classification as ‘objectual’ attitudes is due to the fact that – like vanilla objectual attitudes—experiential attitudes can⁶ be ascribed through linguistic constructions with intensional transitive verbs and direct objects (compare (8) [experiential] with (7a) [non-experiential]):

My propositionalist account of experiential attitudes will start from the scenes that serve as the target of the experiential attitude and that have been personally experienced by the holder of this attitude (see Stephenson, 2010). The propositional content of these scenes is then identified with the intersection of all propositions that are true in these scenes. My account is supported by early findings from situation semantics (see Barwise and Perry, 1983; Higginbotham, 2003; Umbach et al., 2022). It is in line with Tulving’s (1972) notion of episodic memory (see e.g. Bernecker, 2010; Cheng et al., 2016), with the notion of experiential imagination (e.g. Dokic and Arcangeli, 2014; Peacocke, 1998), and with the observation from psychology that some cases of episodic memory cannot be distinguished from propositional memory on the basis of their information content alone (Tulving, 1985; Suddendorf and Busby, 2003; Klein, 2013; see my Sect. 6.2).

The paper is structured as follows: to support the propositionalist thesis about mental attitude content, I will first illustrate the merits of propositionalism (in Sect. 2). In Sect. 3, I then identify the defining property of experiential attitudes that is key to my propositionalist account, viz. their scene-directedness. I provide this account in Sect. 4 and illustrate its merits in Sect. 5. The paper closes with a conjecture about how the presented account could be generalized to non-singular scenes, and about the theoretical implications of this account (in Sect. 6).

2 Merits of Attitudinal Propositionalism

I have suggested above that the propositional nature of attitude contents has a number of merits. These include the availability of a uniform account of different intensionality phenomena and of a simple explanation of the multivaried behavior of natural language entailment. These merits of attitudinal propositionalism are generally acknowledged (see e.g. Felappi, 2021; Montague, 2007; Sinhababu, 2015). However, since they are typically accepted without formal argument—and since I will later need such argument to support my claim about the standardness and uniformity of my account –, I first present these merits:

2.1 Intensionality

Intensionality is a cluster of semantic properties comprising (i) referential opacity, (ii) non-specificity, and (iii) non-actuality (see Zimmermann, 2001, p. 516). Referential opacity is the ability of an expression (in a linguistic context) to resist the truth-preserving substitution by a co-referential or truth-conditionally equivalent expression (see Quine, 1956; cf. Frege, 1997). Non-specificity and non-actuality concern the ability of indefinite DPs [e.g. a unicorn in (9)] to allow for an unspecific reading (see Fodor, 1970) and to lack existential import (see Quine, 1956).

The above properties are all exemplified in (9). In particular, on its fully intensional [= de dicto-] reading, this sentence neither commits Bill to the existence of unicorns [i.e. (9) does not entail (9c); see (iii)], nor does it attribute to him the belief that there is a particular unicorn which John is seeking [i.e. (9) does not entail (9b); see (ii)] or that John is seeking a griffin⁷ [i.e. (9) does not entail (9a); see (i)]:

To capture the referential opacity of attitude reports, contemporary semantics and the philosophy of language interpret attitude verbs as relations to (characteristic functions of) sets of possible worlds (Montague, 1970; Hintikka, 1979) [see (4)–(6)]. De dicto attitude reports then have the form in (10), where set abstraction over possible worlds (here: $\lambda w$) is highlighted in grey.⁸ The non-validity of the entailment from (9) to (9a) is then due to the existence of a world in which the set of unicorns and the set of griffins do not coincide (in the sense that, in this world, there is a unicorn that is not a griffin, or vice versa).

To emphasize the interpretation of a unicorn at the members of the set of worlds that serves as the semantic complement in (9), I use abstractors, $\lambda$, over world variables in syntax [along the lines of Percus, 2000; see (10a)]. In interpretations like (10b), the evaluation of a unicorn at the abstracted world variable w (i.e. $\textit{unicorn}_{w}$; see the grey-marked parts of the formula) captures this expression’s lack of existential import [see (iii)].

The non-specific interpretation of the DP a unicorn in (9) [see (ii)] is captured by giving the existential quantifier that is introduced by this DP narrow scope w.r.t. believe [see ‘$\textit{believe}_{@} (\ldots \exists x.\, \ldots )$’ in (10b)]. This scoping relation effects that (10) is already true in a scenario where John seeks a different unicorn in Bill’s different doxastic alternatives. Since (9b) is false in this scenario, the inference from (9) to (9b) is invalid.

The propositional interpretation of the infinitival and the DP complement in (6a-i) respectively in (6b-i) enables an account of the intensionality of (6a-i) and (6b-i) that is completely analogous to the account of the intensionality from (10). The propositional interpretation of (6b-i) is given in (11b), where bul$_{x, @}$ is the set of x’s buletic [= desire-]alternatives in @:

The above interpretation straightforwardly captures the referential opacity [see (12a)], non-actuality [see (12c)], and unspecificity of a unicorn in (6b-i) [see (12b)]:

Arguably, the intensional behavior of a unicorn in (6b-i) could also be modelled by interpreting this DP as an object of some other type, so long as this object is still intensional (see Percus, 2000) and can take scope below the attitude verb (see Russell, 1905). For example, the semantics from (Montague, 1970) [see (13)] (cf. Moltmann, 1997)—which interprets the direct objects of attitude verbs as intensional generalized quantifiers [= as functions from worlds to the set of properties P that are jointly exemplified at these worlds]—can likewise account for the intensional behavior of the direct object a unicorn in (6b-i):

However, since this interpretation does not treat the direct object as a proposition (s.t. this object cannot be a subset of John’s buletic alternatives in @), it blocks a Hintikka-style analysis along the lines of (4b) and (10b).

2.2 Entailment

The propositional interpretation of attitude complements [along the lines of (11)] is further preferred for reasons having to do with the modelling of natural language entailment (see Montague, 2007; Liefke and Werning, 2018; Theiler et al., 2018). In formal semantics, entailment is commonly modelled as (the functional counterpart of) set-theoretic inclusion. For two linguistic expressions X and Y, it thus holds that $X \Rightarrow Y$ iff $\llbracket X\rrbracket \subseteq \llbracket Y\rrbracket$ (see Keenan and Faltz, 1985; Kac, 1992). By the conditions on set-theoretic inclusion, entailment is restricted to pairs of expressions that denote objects of the same type. These include the designators of propositions and the designators of intensional generalized quantifiers. Admissible pairs of expressions then include the pairs in (14) to (16). They prima facie exclude the pair in (17):

In particular, since (14a) and (14b) each denote a set of possible worlds [viz. (18a) resp. (18b)], they are in the domain of this inclusion relation. The validity of (14) is then due to the set-theoretic inclusion of the set in (18a) in the set in (18b):

Assuming that ‘believe’ is upward monotonic in its complement position, an analogous observation holds for (15).

A higher-type analogue of the above holds for the generalized quantifier-interpretation of the DPs in (16):

Since DPs are typically interpreted in a different semantic type from embedded clause-CPs (viz. as generalized quantifiers vis-à-vis as propositions; see Montague, 1970; Moltmann, 1997), the complements of seek and try in (17) prima facie do not stand in a semantic inclusion relation. The propositional analysis of the complement in (17a) [see (11); based on (6b)] remedies the problematic type-difference. This analysis captures the entailment in (17) through the semantic inclusion of the set of worlds in which John finds a unicorn in the set of worlds in which John finds a mythical creature [in (20)]:

Admittedly, by itself, the modelling of natural language entailment is not yet unequivocal support for the propositional nature of mental attitude content: the entailments from (14) to (17) could also be modelled by assuming that attitude content takes the form of intensional generalized quantifiers. This assumption would straightforwardly capture the entailment relation in (16). The remaining relations could then be captured by interpreting linguistic expressions at face-value [as in (13)] and by lifting propositions to intensional quantifiers through a suitable type-shifter. However – apart from failing to account for the relation in (17) –,⁹ such a move would violate Partee’s principle of trying simplest types first (given that sets of worlds are simpler [= have a lower-rank type] than generalized quantifiers) and of only using type-shifting as a last resort (see Partee and Hendriks, 1997).

3 Experiential Attitudes

I have suggested above that experiential attitudes¹⁰ differ from ‘true’ [$=$ non-experiential] objectual attitudes in being relations to events or scenes (viz. a scene that saliently features some individual; e.g. a scene involving a [possibly non-specific] unicorn), rather than to individuals (e.g. the unicorn). To prepare my propositionalist account of experiential attitudes, I first provide support for the scene-directedness of experiential attitudes (in Sect. 3.1) and contrast ‘experiential’ with ‘propositional’ occurrences of experiential attitude verbs (in Sect. 3.2).

3.1 The Scene-Directedness of Experiential Attitudes

Like reports of ‘true’ objectual attitudes [see e.g. (7a)], experiential attitude reports [e.g. (21a), (22a)] resist a syntactic analysis or lexical paraphrase as a clause-embedding structure. However, they are equivalent to the result of supplementing their direct object with a contextually determined predicate (including the context-universal predicate exists; see (21b/c) and (22b/c), where ‘${\mathop {\equiv }\limits ^{c}}$’ indicates a contextual equivalence in the described scenario).¹¹ The examples below are inspired by (Forbes, 2006, ex. (14a)) and (Zimmermann, 2016, ex. (3)), respectively:

Assume that Mary imagined a unicorn which is/was cantering [see the scenario in (21)]. This attitudinal event can then be reported by (21a) as well as by (21b) or, in the described scenario, by (21c).¹²

The equivalences in (21) and (22) are reminiscent of the equivalence of objectual and scene-directed perception reports (see Barwise and Perry, 1983; van der Does, 1991; Falkenberg, 1989 [for German]):

As has been extensively discussed in situation semantics, (23a–c) are all adequately paraphrased by (24) in the described context:

Analogous paraphrases can be provided for (21a–c) [in (25)] and for (22a–c)¹³ [in (26)]:

The possibility of paraphrasing experiential imagination and depiction reports through explicit reference to scenes suggests that—like see—imagine and paint (and, as I will show below, remember) denote relations to events or scenes.¹⁴ The ‘scene-directedness’¹⁵ of these attitudes is further evidenced by the possibility of modifying the matrix verbs in reports of these attitudes by an experiential modifier like vividly or in vivid/perfect/lifelike detail [in (27), (28); see Stephenson, 2010]:

The scene-directedness of imagine- and paint-reports like (21a) and (22a) is further apparent from the possibility of replacing the direct object DPs in these reports by a scene- or event-denoting how-phrase [in (29), (31)] (see Umbach et al., 2022) and from these reports’ entailment to the truth of sentences that relate the agent’s personal (perceptual or mental) experience of the scene described by the complement [in (30), (32)] (see Stephenson, 2010).

The above behavior is displayed by all representational counterfactual attitude verbs (see Blumberg, 2019). These are verbs that “convey a mental picture” (see White et al., 2018) [representationality] and whose DP complements denote objects that may not exist at @ [counterfactuality]. They include—next to verbs of intentional physical and mental depiction (e.g. paint, imagine, visualize; see above)—verbs of non-intentional depiction (e.g. dream (of), hallucinate) [see (33a)]. The behavior from (24) to (32) is also observed for reports with direct object-taking occurrences of memory and attention verbs (e.g. remember, notice, observe) [see (34a)].

The remainder of this paper will defend propositionalism about the content of experiential attitudes. This defense will use the possibility of representing the target-scenes of experiential attitudes through sets of possible situations (see Sect. 4). To avoid wrong conclusions about the information content of the direct objects in experiential attitude reports, I precede my analysis with a distinction between experiential [= scene-directed] and propositional [= proposition- or fact-directed] occurrences of experiential attitude verbs (in Sect. 3.2). In contrast to the objects of experiential occurrences of these verbs, the objects of propositional occurrences have a straightforward—and widely acknowledged – propositional analysis.

3.2 Scene-Directed vis-à-vis Proposition-Directed Experiential Attitudes

Remarkably, many experiential attitude verbs also license finite that-clause complements. For the matrix verbs in (21a), (33a), (23a), and (34a), examples of such constructions are given below:

In contrast to their experiential counterparts, the that-clause reports in (35) to (38) intuitively do not describe an agent’s relation to a scene. Rather, they express the agent’s relation to a proposition that is true of this scene, or to a fact about this scene (see Barwise and Perry, 1983; Higginbotham, 2003; Stephenson, 2010). The propositional nature of (35) to (38) is evidenced by the observation that these reports resist a paraphrase in terms of scenes [for (38), this is shown in (39)]. In particular—in contrast to (34a) –, (38) is already true in a scenario in which Ron has only been told that a spider was hanging from the ceiling, but has not seen (or otherwise experienced) the spider.

The propositional nature of that-clause variants of experiential attitude reports is further supported by the observation that these reports fail the diagnostic tests for experiential attitude reports from Sect. 3.1.

The difference between experiential and propositional attitude reports reflects established distinctions in the philosophy of mind, psychology, and cognitive science. These include Dretske’s (1969) distinction between ‘non-epistemic’ [$\approx$ experiential] and ‘epistemic’ [$\approx$ propositional] perception, Tulving’s (1972) distinction between ‘episodic’ [$\approx$ experiential] and ‘semantic’ [$\approx$ propositional] memory, and Peacocke’s (1998) distinction between ‘experiential’ and ‘suppositional’ [$\approx$ propositional] imagining.

Following Hintikka’s relational account [see (4), (10)], the information content of the that-clause in (38) is identified with the set of worlds in which a spider is hanging from the ceiling (in (40a); see Ciardelli et al., 2018; Abusch, 2020):

Since the object that is denoted by the DP a spider in (34a) (viz. an event/a scene) has a different type from the object that is denoted by the that-clause in (38) (viz. a proposition/a fact), these objects are seemingly unable to stand in a semantic inclusion relation. I will show below that this is not the case.

4 ‘Propositionalizing’ the Content of Experiential Attitudes

To extract the propositional information content from the scenes that serve as the objects of experiential attitudes, I represent these scenes by sets of situations. The members of these sets are informationally incomplete parts of possible worlds that encode exactly the same qualitative [= non-spatio-temporal and -worldly] semantic information. The propositional representation of scenes uses the possibility of coding objects of a certain type (here: the general type for situations, events, and scenes) by sets of objects of this type (see e.g. Quine, 1946). My representation deviates from standard uses of this possibility in coding scenes by sets of (partial) situations, rather than by sets of (maximal, total) worlds. This deviation is motivated by the informational incompleteness of imagined and remembered scenes. The representation of informationally incomplete scenes by their maximally extending worlds would destroy this partiality (see Fine, 2017; Yablo, 2014).¹⁶

My strategy for the propositional representation of scenes is given in (41). This strategy identifies the propositional content of a scene s with a subset of the set, $\Sigma$, of situations whose members exactly verify (in the sense of truthmaker semantics, to be specified below) the conjunction, S, of all sentences that are true in s [see condition (i)] and—if applicable—whose members are located at the same spatio-temporal and world-coordinate, $l_{s}$ (for location of s), $t_{s}$ (for time of s), and $w_{s}$ (for world of s), as s [see condition (ii)]. In (41), $\Vdash$ is the truthmaker-semantic relation of exact verification (see Fine, 2017).¹⁷$s \Vdash S$ asserts that s is “wholly relevant for the truth of S”, such that s “does not include anything that fails to bear on the truth of S” (Moltmann, 2020, p. 167).

Condition (i) ensures that the propositional representation of the target scene is a set of isomorphic [= qualitatively identical] situations (in the sense of Kratzer, 2002, p. 667; see Fine, 1977, p. 136). In combination with this condition, Condition (ii) ensures that specific scenes, i.e. scenes that are located (or ‘anchored’) in a particular spatio-temporal and world-coordinate (e.g. Ron’s remembered scene from (34), in which a spider is hanging from the ceiling)¹⁸ are represented by a singleton sets of situations. Non-specific scenes, i.e. scenes without a particular spatio-temporal and/or world-coordinate (e.g. Mary’s imagined scene from (21a), in which some (!) unicorn is cantering), are represented by a (typically non-singleton) set of situations.

The possibility of representing non-specific scenes by sets of situations allows us to propositionally model the contents of representational counterfactual attitudes like imagining. To see how this works, consider the scenario in Fig. 1.¹⁹ In this scenario, Mary is imagining a white unicorn with a peach belly and a rainbow-colored mane and tail that is cantering. However—as is often the case in counterfactual attitudes –, the featured unicorn is not a specific individual that inhabits a particular possible world.

Fig. 1

Mary imagining a unicorn

Following the strategy from (41), we represent Mary’s imagined scene by the set of situations that serve as exact truthmakers for the conjunctive sentence in (42) [see (43)]. The members of this set encode exactly the same qualitative information. They differ in being parts of different possible worlds. Since, in some worlds, there may be multiple unicorns that each exemplify the properties in (42), this set may even contain situations with different spatial and/or temporal coordinates that are part of the same world.

To illustrate the propositional information content of Mary’s imagined scene in Fig. 1, consider the representation of this scene in the universe from Fig. 2: this universe consists of three worlds, viz. $@,\,w_{1}$, and $w_{2}$. These worlds each contain three situations with distinct²⁰ spatio-temporal coordinates that may, however, overlap (see the informational inclusion of $\sigma _{1}^{ii}$ in $\sigma _{1}^{iii}$). The semantic information that is true in these situations is abbreviated as follows:

A::

a white unicorn with a peach belly and a rainbow-colored mane and tail is cantering

B::

a white unicorn with a peach belly and a silver mane is basking in the sun

C::

a spider is hanging from the ceiling

D::

Samuel is eating a fish

Fig. 2

The propositional representation of Mary’s imagined scene from Fig. 1 [= the set of ‘white’ situations]

The information from (42) is exactly true in the ‘white’ situations, viz. in $\sigma _{1}^{i}$, $\sigma _{1}^{ii}$, and $\sigma _{2}^{iii}$. As a result, in the context from Fig. 1, the information content of a unicorn in (21a) can be represented by the set $\{\sigma _{1}^{i}, \sigma _{1}^{ii}, \sigma _{2}^{iii}\}$.

Note that, because of the non-specificity of Mary’s imagined scene, (21a) [copied in (44a)] is intuitively equivalent to the that-clause report from (44b) in the described context:

By replacing the familiar semantic values of sentences and clauses (viz. sets of possible worlds in which the clause is true; see Sect. 1) by sets of their exact verifiers/situations,²¹ we can straightforwardly capture the above equivalence.

Notably, contextual equivalences like (44) do not hold for perception and memory reports. In particular, the DP a spider in (34a) does not, in any context, have the same propositional information content as the CP that a spider was hanging from the ceiling [see (46)]. This non-identity even holds if the fact that a spider was hanging from the ceiling was the only thing that Ron saw in the remembered scene.²² This is so since the semantic complement in (34a) (i.e. Ron’s formerly perceived – and now remembered—scene that features a particular spider) is a specific spatio-temporal part of the actual world. This differs from the semantic complement in (38), which is a spatially unspecific part of some world.

To capture the specificity of the remembered scene/situation in (34a) (hereafter called ‘$\sigma$’), I represent this scene by a singleton that only contains this scene [see (46a)]. Since the information content of the that-clause in (38) contains all past situations that exactly verify the sentence A spider was hanging from the ceiling [see (46b)], it includes the propositional information content of a spider in (34a). To capture the pastness of the relevant situations (see the past tense in ‘was hanging $\ldots$’), I restrict the set in (46b) by demanding that s is located at an earlier point in time than @ (formally: ‘$t_{s} \prec t_{@}$’).

Arguably, the replacement of sets of (total) worlds by sets of exact truthmakers [= of (partial) situations] effects that we can no longer capture semantic inclusion through set-theoretic inclusion (contra Sect. 2.2). To answer this challenge, I replace set-theoretic inclusion with a relation, $\sqsubseteq$, of pointwise informational inclusion in (46). I assume that a set of situations, $\varSigma$, informationally includes another set $\varSigma '$, i.e. $\varSigma \sqsubseteq \varSigma '$, iff $\forall s' \in \varSigma ': (\exists s \in \varSigma : s' \le s)$, where $\le$ is Moltmann’s (2020, pp. 167–168) partial ordering on situations (see Liefke and Werning, 2018, p. 659 for a description of this ordering). On this ordering, a situation s includes a situation $s'$, i.e. $s' \le s$, if s contains all information that is contained in $s'$. This condition requires that the location $l_{s}$ and time $t_{s}$ of the world-part about which s contains information includes the location $l_{s'}$ and time $t_{s'}$ of the world-part about which $s'$ contains information (s.t. $l_{s}$ maintains or expands the perimeters of $l_{s'}$ and $t_{s}$ starts before or simultaneously with $t_{s'}$ and ends after or simultaneously with $t_{s'}$).

In Fig. 3, $\le$ holds e.g. between $\sigma _{0}^{ii}$ (in which a spider is hanging from the ceiling and Samuel is eating a fish) and its spatio(-temporal) part $\sigma _{0}^{i}$ (in which a spider is hanging from the ceiling, and nothing else), s.t. $\sigma _{0}^{i} \le \sigma _{0}^{ii}$. Since Moltmann’s ordering is not a proper ordering (unlike <), this ordering can even hold between a situation and itself (s.t. $\sigma _{0}^{i} \le \sigma _{0}^{i}$). The propositional content of the direct object, a spider, from (34a) is given in (46a).

Fig. 3

The propositional representation of Ron’s remembered scene in (34) (white) vs. of Ron’s remembered fact in (38) (dashed)

Assuming that $\sigma = \sigma _{0}^{iii}$, the propositional information contents from (46a) and (46b) are identified with $\{\sigma _{0}^{iii}\}$ (= $\llbracket$a spider$\rrbracket ^{c}$) respectively with $\{\sigma _{0}^{i}, \sigma _{0}^{iii}, \sigma _{2}^{ii}\}$ (= $\llbracket$that a spider is hanging from the ceiling$\rrbracket ^{c}$). In Fig. 3, the members of these sets are marked in white (for $\llbracket$a spider$\rrbracket ^{c}$), respectively are circled by a dashed line (for $\llbracket$that a spider $\ldots \rrbracket ^{c}$). Since each member of $\{\sigma _{0}^{iii}\}$ is an informational part of some member (here: $\sigma _{0}^{iii}$) of $\{\sigma _{0}^{i}, \sigma _{0}^{iii}, \sigma _{2}^{ii}\}$ (see the above definition of pointwise informational inclusion), the propositional content of the that-clause in (38) is semantically included in the content of the direct object DP in (34a) [see (46)].

5 Preserving the Merits of Propositionalism

With the propositional representation of the objects of experiential attitudes in place, I am now ready to show how my account preserves the merits of attitudinal propositionalism. I start with the modelling of natural language entailment (see Sect. 2.2).

5.1 Entailment

I have already observed that, on my proposed account, the direct object DPs in experiential imagination reports have the same information content as the conjunction of all sentences that are exactly true in the scene denoted by these DPs [see (43), (45)]. This identity straightforwardly accounts for the entailment in (47), whose second report, i.e. (47b), only uses a small fragment (viz. a unicorn was cantering) of the informationally much richer sentence (42) in its complement [see (48)].

A similar account to the one above can be provided for intuitive entailments between ‘direct object’ and ‘that-clause’ perception and memory reports [e.g. (34a), (38); see (49)]:

The only difference between my account of (47) and my account of (49) lies in the spatial and worldly restriction on the members of the sets of situations that serve as the information content of the DPs in (47a) and (49a): since the direct object a spider in (49a) refers to a specific scene, its information content is represented by a singleton set—unlike the information content of a unicorn in (47a).

5.2 Intensionality

As one may expect, the propositional information content of experiential attitudes also preserves the intensional behavior of reports of these attitudes [see (12)] (Sect. 2.1). This is so since the representation of the target scene by a set of situations preserves the possibility of interpreting the direct object DP inside the scope of the attitude verb, and of evaluating this DP at a non-@ situation. In particular, since the sets of unicorns and of griffins need not be the same in all situations in which a white unicorn with a peach belly and a rainbow-colored mane and tail is cantering [see (43)], my account ensures the referential opacity of imagine [see (50a)]. The same holds for embedded indefinites’ lack of existential import [see (50c)]: since the existential quantifiers that are introduced by embedded DPs still enter into scope relations with the matrix attitude verb (in (50): imagine), my propositionalist account also allows these DPs to receive a non-specific interpretation [see (50b)].

To avoid spelling out the details²³ of a formal semantics for experiential attitude verbs, I interpret a unicorn in (50) against the specific context from Fig. 1:

The referential opacity of imagine is further due to the fact that different things may be the case in the scenes denoted by a unicorn in (50) and by a griffin in (50a) (see the phrasing ‘is performing certain actions $\ldots$’).

6 Challenges and Prospectives

6.1 A Challenge: Non-singular Scenes

Above, I have defended propositionalism for experiential attitudes by showing that the scenes that serve as the objects of these attitudes allow for a straightforward representation in terms of propositions [= sets of truthmakers/situations]. My defense has been based on the assumption that agents hold experiential attitudes towards (specific or unspecific) singular scenes. However, this assumption does not hold in general, as is shown by (51):

In contrast to (34a), the direct object DP in (51a) is intuitively not equivalent to the result of supplementing it with some contextually salient predicate. This is so since the remembered cat has different properties at Oscar’s different ‘cat seeing’-events, and since Oscar’s remembering the cat may also involve him thinking of the cat when it was not around. The former excludes the DP’s supplementation with a contextually determined predicate [along the lines of (7b)]. The latter excludes the DP’s supplementation with an existence predicate [along the lines of (7c)].

The above notwithstanding, my propositionalist account can still be adapted to capture cases like (51a). This adaptation involves replacing a propositionally represented single scene by the propositionalist representation of a set of scenes (for (51a): the set of Oscar’s [perceptually or counterfactually] experienced scenes that feature said cat). My strategy for the propositionalist representation of this set [in (52)] follows the strategy for the representation of single scenes [in (41); copied below]:

The strategy from (52) differs from the ‘single scene’-strategy in taking as input sets of scenes [see the ‘$\lambda p$’ in (52)], rather than individual scenes. The need to still ‘convert’ these sets into sets of situations is due to the fact that some scenes in the input set are non-specific and are, hence, not situations. The conversion is further required by the wish to capture intuitive entailments between experiential attitude reports with event- and with concrete object-denoting DPs²⁴ [e.g. (53)]:

Following (52), the set of Oscar’s experienced ‘cat’-scenes is represented by the union of the propositional [= singleton] representation of specific scenes (e.g. a particular visual scene of Oscar’s in which the cat was eating tuna out of a can in his garden) and the propositional [= non-singleton] representation of unspecific scenes (e.g. Oscar’s counterfactual buletic scene in which the cat is curling up on his sofa).

It is an open question whether the proposed generalization can be extended to ‘true’ objectual attitudes (e.g. fearing Superman). However, the proposed account is a first, and important, step in the ‘divide and conquer’-defense of attitudinal propositionalism.

6.2 Outlook: Just Propositional Attitudes?

The possibility of providing a propositional characterization of experiential attitude content is an old hat in psychology and cognitive science that goes back (at least) to Tulving (1985). In the attempt to provide a uniquely identifying feature of experiential remembering, Tulving noticed that his earlier what-where-when-criterion for episodic memory (that identified experiential remembering through its information content and the latter’s spatio-temporal ‘anchoring’; see Tulving, 1972) fell short of this purpose. This is so since propositional remembering can be equally detailed (see the complex conjunctive sentence in (42); copied below) and spatio-temporally specific [see the de re-reading, (54), of (38)] as experiential remembering (see also Suddendorf and Busby, 2003; Klein, 2013; Sant’Anna, 2018).

To still distinguish experiential from propositional remembering, later work has identified several special phenomenological features of episodic recall. The latter include an experiencer perspective and a sense of self in episodic memory (see e.g. Klein and Nichols, 2012; Tulving, 2005), a perspective on the experienced scene (Nigro and Neisser, 1983; Rice and Rubin, 2009), the presence of mental imagery (Mahr, 2020; Michaelian, 2016), and a degree of phenomenal transparency that gives rise to experience-likeness (Cheng et al., 2016; Harman, 1990). I expect that a fully-fledged theory of experiential attitudes will need to provide a careful account of these different properties. While recent work in semantics and the philosophy of language has started to, e.g., provide more realistic accounts of visual perspective (beyond centered propositions; see Lücking et al., 2015), it has only taken the first few steps in this direction.