Presupposition: accepted information or embraced beliefs? The role of informative function and trigger type in separating two levels of accommodation

1 Introduction

Information can be linguistically presented in different ways, and these variations have been observed to influence people’s critical scrutiny of incoming information, i.e., their “epistemic vigilance” (Sperber et al. 2010). For example, a specific piece of content can be framed as a presupposition, signalling that it does not require extensive scrutiny, or as an assertion, prompting a thorough evaluation of such information (Ducrot 1972; Givón 1982; Lombardi Vallauri 2019; Macagno 2023; Pietrandrea 2020; Reboul 2017; Sbisà 2007, 2023]). This paper explores the persuasive influence of presupposition, taking into account for the first time the roles of informative function (i.e., topic-focus articulation) and trigger type. Initially, we define the notion of presupposition and describe how the informative function can impact its triggering. Subsequently, we review the empirical data exploring presupposition’s persuasive effects in concealing false information and in memory creation. Following this, we describe an experimental investigation we conducted to examine participants’ ability to recognise false information immediately and after one week, depending on the manner in which the false content was originally linguistically presented. Finally, we discuss the implications of our findings for presupposition theories.

2 Methods

2.2 Materials

2.2.1 Video

The video is Hot water (duration: 02’25’’), from A Day in the Life of a Cat Owner found in the Simon’s Cat series. The video is on YouTube at the following link: https://www.youtube.com/watch?app=desktop&v=fcjR93Jin4Y&amp%3Bt=291s.

The video has no dialogue and shows a cat entering the bathroom while its owner is having a shower. After unrolling the toilet paper, the cat drinks from the toilet. With its paw, it drops a watch into the toilet and flushes it. Next, the cat eats some toothbrushes, but, from its expression, it is clear it does not like them. Then, the cat makes the toothbrushes fall out of the bowl where they were, to use it as a cat bowl. While all this is happening, the owner continues having his shower without noticing anything. Afterwards, the cat turns on the sink tap to drink some water, and, consequently, some steam comes out of the shower. The cat realises that turning on the tap changes the water temperature in the shower, making it hot. So, it turns on the tap again, and its owner falls from the shower (as he reasonably attempts to avoid the hot water). In the last scene, the cat points to the bowl while the owner is on the ground. A selection of stills from the main events of the video is provided below (Figure 1).

Figure 1:

Selection of captures of the main events in the video.

2.2.2 Audio stimuli

Stimuli were composed of pre-recorded Italian sentences. All the sentences were uttered by a female native speaker of Italian with high competence in the relationship between prosody and topic-focus structure. Audio stimuli were preferred over possible written ones because prosody in audio allows for a more secure transmission of the topic/focus distribution.

Week 1

On Week 1, target stimuli consisted of 6 triplets of sentences. For each triplet, one sentence conveyed a misrepresentation of the video through a topical presupposition (condition “topical presupposition”), one sentence conveyed the same misrepresentation through a focal presupposition (condition “focal presupposition”), and one sentence conveyed the same misrepresentation through an assertion (condition “assertion”) (see Table 1).

Table 1:

Examples of Targets and First Block Fillers. The presuppositional or assertive encoding is typed in bold. Capital letters indicate the main stress/focus. English translation is provided in brackets.

Sentence type	Trigger	Week 1			Week 2	Video
		Topical presupposition	Focal presupposition	Assertion	Assertion
Target	Definite description	I pettini non sono piaciuti al GATTO. (The CAT did not like the combs.)	Al gatto non sono piaciuti i PETTINI. (The cat did not like the COMBS.)	C’erano dei PETTINI, e non sono piaciuti al GATTO. (There were COMBS, and the cat didn’t like THEM.)	C’erano dei PETTINI. (There were COMBS.)	No combs were present.
	Temporal clause	Quando il gatto si gratta il musetto, il padrone sta cantando sotto la DOCCIA. (When the cat scratches its nose, the master is singing in the SHOWER.)	Il padrone sta cantando sotto la doccia quando il gatto si gratta il MUSETTO. (The master is singing in the shower when the cat scratches its NOSE.)	Il gatto si gratta il MUSETTO, e il padrone sta cantando sotto la DOCCIA. (The cat scratches its NOSE, and the master is singing in the SHOWER.)	A un certo punto il gatto si gratta il MUSETTO. (At some point the cat scratches its NOSE.)	The cat did not scratch its nose.
First block filler (mimicking the targets)	Definite description	La finestra era sopra il LAVANDINO. (The window was above the SINK.)	Sopra il lavandino c’era la FINESTRA. (Above the sink was the WINDOW.)	C’era una FINESTRA, ed era sopra il LAVANDINO. (There was a WINDOW, and it was above the SINK.)	C’era una FINESTRA. (There was a WINDOW.)	There was a window above the sink.
	Temporal clause	Quando il gatto ha tirato lo sciacquone, il padrone non si è accorto di NULLA. (When the cat flushed the toilet, the master didn’t notice ANYTHING.)	Il padrone non si è accorto di nulla quando il gatto ha tirato lo SCIACQUONE. (The master didn’t notice anything when the cat flushed the TOILET.)	Il gatto ha tirato lo SCIACQUONE e il padrone non si è accorto di NULLA. (The cat flushed the TOILET, and the master didn’t notice ANYTHING)	A un certo punto il gatto ha tirato lo SCIACQUONE. (At some point the cat flushed the TOILET.)	The cat flushed the toilet.

Half of the presuppositions were triggered by definite descriptions and half by temporal clauses. For the definite description presuppositions, the assertive counterparts were obtained by rephrasing them as indefinite phrases contained in presentative constructions (for ex., Il gatto “The cat” vs C’era un gatto “There was a cat”) or in possessive constructions (for ex., Il collare del gatto “the cat’s collar” vs Il gatto ha un collare “the cat has a collar”).^[3] On the other hand, assertive versions of temporal clauses were created by turning them into main clauses (for ex. Quando il gatto si gratta il musetto “When the cat scratches its nose” vs Il gatto si gratta il musetto “The cat scratches its nose”).

A first block of filler stimuli consisted of 8 triplets of sentences that mimicked the target stimuli with the exception that this time no misrepresentation was present (see Table 1). Namely, for each triplet, one sentence conveyed certain content through a topical presupposition, one conveyed the same content through a focal presupposition, and one conveyed the same content through an assertion. As in the target stimuli, half of the presuppositions were triggered by definite descriptions and half by temporal clauses. The assertive counterparts were created following the same criteria as for the target stimuli. A second block of filler stimuli consisted of 4 sentences (2 misrepresenting the video, 2 without any misrepresentation) that did not mimic the target stimuli (see Table 2). Two familiarisation sentences (1 misrepresenting the video, 1 without any misrepresentation) were also created, with the aim to make participants familiarise with the task (see Table 2).

Table 2:

Examples of Fillers from the Second Block and Familiarisation Sentences. English translation is provided in brackets.

Type of sentence	Example	Video
Second block filler	Nel cartone, c’era un gatto che aveva due grossi occhi tondi. (In the cartoon, there was a cat with two big round eyes.)	In the cartoon, there was a cat with two big round eyes.
Familiarisation sentence	C’era un cane. (There was a dog.)	No dog was present.

Week 2

On Week 2, the targets (N = 6 sentences) contained the same misrepresentations as in Week 1, but they were all in the assertion condition, i.e., contained the assertive counterparts of definite description-presuppositions (3 sentences) and temporal clause-presuppositions (3 sentences) (see Table 1). The first-block fillers (N = 8 sentences), too, were all in the assertion condition and encoded the same correct information that in Week 1 was either asserted or presupposed (see Table 1). To make the sentences more natural, some target sentences and the first block fillers slightly differed from the Week 1 assertion condition sentences (see Table 1). For example, Week 2 sentences could start with a un certo punto (eng. at some point). However, what can be called the ‘region of interest’ (that is, the assertive encoding of certain information) was precisely the same in Week 1 and Week 2 assertion condition sentences.

The second-block fillers (N = 4 sentences) and the two familiarisation sentences were identical to Week 1 (see Table 2).

Measures on materials

Stimuli complexity was measured through a Gulpease calculation (Piemontese 1996). Results were analysed with a linear model, with Condition (Topical presupposition, Focal presupposition, Assertion) and Trigger (Definite description, Temporal clause) as predictors and interaction included. The analysis was carried out using R software (version 4.2.1) and the package lme4. No significance was found. This indicates that the stimuli are comparable in complexity in the different conditions and the two trigger types.

2.2.3 Answer sheet

Google Forms was used to collect participants’ responses. For each sentence, the three possible answers (vero/true, non so/I don’t know, falso/false) were written and presented as icons in the input field (Figure 2). The sentences were not transcribed on the form; only numbers referring to the order by which the sentences were played were present.

Figure 2:

Icons (vero/true, non so/I don’t know, falso/false).

2.3 Procedure

The experiment was carried out through Google Meet to allow the experimenter to have contact with the participants.^[4]

The experimental procedure was the following: on Week 1, participants watched the video and then listened to 20 sentences about it. For each sentence, they had to select between three options on the screen to indicate whether the sentence was ‘true’, ‘false’, or they ‘didn’t know.’ After the test, participants were asked if they had previously seen the video, and their answers were excluded if they had. A week later (Week 2), participants were once again asked in the same modality to indicate whether 20 sentences were ‘true’, ‘false’, or they ‘didn’t know’ without having the possibility of rewatching the video. Again, the sentences were auditorily presented.

Participants had no time pressure while answering, because, in both test sessions, the following sentence was played when all participants had finished answering the former. Participants were tested from their homes. They were divided into different sessions according to their preference on when to take the test.

2.3.1 Presentation of the stimuli

In Week 1, the target stimuli and the first block of fillers were organised in three lists according to a Latin Square design, so that each participant was presented with each sentence in only one condition. Each list also included the second block of filler stimuli and the two familiarisation sentences. So, each list was composed of 20 sentences: 2 familiarisation sentences, 6 targets (1 definite description topical presupposition, 1 definite description focal presupposition, 1 definite description assertive counterpart, 1 temporal clause topical presupposition, 1 temporal clause focal presupposition, 1 temporal clause assertive counterpart), and 12 fillers. For each list, the order of the sentences was pseudo-randomised, and the familiarisation sentences were always played first. There were two pseudo-randomisations per list.

On Week 2, all the participants were presented with a list of 20 sentences, containing the familiarisation sentences (N = 2), the target sentences (N = 6), the first block of filler sentences (N = 8), and the second block of filler sentences (N = 4). Again, the order of the sentences was pseudo-randomised, with familiarisation sentences always played first, and two different pseudo-randomisations were created (Figure 3).

Figure 3:

The pipeline of the Experiment.

3 Results

Figure 4 reports the mean values of participants’ responses in each condition (presupposition in topic, abbreviated as PT/ presupposition in focus, PF / assertion, AS) and in each trigger type (definite description, DEF / temporal clause, SUB). Please remember that, in Week 2, all the targets encoded the misrepresentation through an assertion. So, PT (topical presupposition), PF (focal presupposition), and AS (assertion) refer back to how the misrepresentations were conveyed in Week 1.

Figure 4:

Accuracy in mistake recognition as a function of condition (PT, PF, AS) and trigger (DEF, SUB) in Weeks 1 and 2.

For Week 1, in the DEF subset, people were significantly less accurate in PT condition than in AS condition (estimate = −0.3253, Std. Error = 0.1121, df = 506.0247, t value = −2.903, p = 0.00386) and in PF condition than in AS condition (estimate = −0.3050, Std. Error = 0.1121, df = 506.0247, t value = −2.722, p = 0.00671). Importantly, we can see that the coefficients for PT and PF are highly similar in value and the standard errors are equal, which reveals that they have similar effect sizes. In other words, taking assertion as the baseline level, the model predicts that a similar negative adjustment in estimated accuracy should be made for both kinds of presupposition, suggesting that topical and focal presupposition, when compared to assertion, have a similar lowering effect on accuracy. Moreover, the model revealed a trend for the comparison AS DEF versus AS SUB, with people being less accurate with AS SUB (estimate = −0.3372, Std. Error = 0.1775, df = 7.4128, t value = −1.900, p = 0.09680). Neither an enhancing nor lowering effect was found to be added by the interaction between condition and trigger (for ConPF:TrSUB, estimate = 0.0463, Std. Error = 0.1585, df = 506.0247, t value = 0.292, p = 0.77030; for ConPT:TrSUB, estimate = 0.1399, Std. Error = 0.1585, df = 506.0247, t value = 0.883, p = 0.37782). This indicates that once we account for the considerable difference in accuracy between the assertive versions of definite descriptions and temporal clauses, the adjustment required to estimate the accuracy linked with temporal clause-triggered presuppositions in comparison to those induced by definite descriptions is minimal. In other words, these data suggest that (a) compared to assertion, topical and focal presuppositions have a similar decreasing effect on accuracy, and (b) definite descriptions and subordinate clauses exhibit differing accuracy rates in both their presuppositive forms and their assertive counterparts.

A second model with the same structure was used to analyse Week 2 results.^[6] In the DEF subset, a trend was found in the comparison between PT and AS, with people being less accurate in the presence of the presupposition (estimate = −0.17534, Std. Error = 0.10410, df = 421.01885, t value = −1.684, p = 0.0928). In contrast, no significance nor trend was found comparing PF and AS (estimate = −0.02191, Std. Error = 0.10410, df = 421.01885, t value = −0.210, p = 0.8334). No statistical significance was also found comparing PT and PF (estimate = 0.1534, Std. Error = 0.104, df = 421, t value = 1.474, p = 0.3045), although the average accuracy rates with PT and PF are quite different (M_PT = −0.31, M_PF = −0.16). One possibility is that the variability in the data, as indicated by the error bars in the graph, may have obscured the distinction between the two conditions. Increasing the sample size is expected to reduce this variability, potentially yielding clearer and more conclusive results. No significance was also found in the comparison between AS DEF and AS SUB (estimate = −0.01380, Std. Error = 0.17017, df = 7.09339, t value = −0.081, p = 0.9376). Thus, contrary to Week 1, providing accurate true-or-false judgments for assertions on simple entities does not seem easier than for assertions on complex events in Week 2. As for the interplay between condition and trigger, the coefficient for the interaction CONPT:TRSUB is small in value (estimate = −0.06448, Std. Error = 0.14721, df = 421.01885, t value = −0.438, p = 0.6616). Conversely, the coefficient for the interaction CONPF:TRSUB is notably high in absolute value if compared with the other ones provided by the model (estimate = −0.19646, Std. Error = 0.14721, df = 421.01885, t value = −1.334, p = 0.1828). Thus, to estimate accuracy when transitioning from PF in the DEF subset to PF in the SUB subset, one has to account for the difference in accuracy between the assertive counterparts of definite descriptions and temporal clauses, which is minimal, and then apply a substantial adjustment with a negative value. This means that there is a strengthening of the misleading effect of focal presupposition in the SUB subset compared to the DEF subset. Indeed, if we take SUB as the reference value, and consider the SUB subset, we see that people were significantly less accurate in PT condition than in AS condition (estimate = −0.23982, Std. Error = 0.10410, df = 421.01885, t value = −2.304, p = 0.0217) and also in PF condition than in AS condition (estimate = −0.21836, Std. Error = 0.10410, df = 421.01885, t value = −2.098, p = 0.0365). It is worth noting that the coefficients for temporal clause presuppositions in topic and in focus are highly similar in value and the standard errors are identical, which implies that also the corresponding effect sizes are similar.

We also contrasted the accuracy associated with asserted mistakes in Week 1 with the accuracy in Week 2 corresponding to the items that were asserted mistakes one week earlier (i.e., AS in Week 1 and 2). Our analysis revealed a main effect of Week (p < 0.001) and an interaction effect between Week and Trigger (p = 0.03) (where “Trigger” here refers to the assertive versions at stake since no presupposition trigger was present). More specifically, accuracy decreases from Week 1 to Week 2, but only for the assertive versions of definite descriptions (for definite description assertive versions, estimate = 0.3870, Std. Error = 0.107 df = 254, p < 0.001; for temporal clause assertive versions, estimate = 0.0614, Std. Error = 0.107, df = 254, p = 0.57).

Overall, the results for Week 2 point to different effects of Condition depending on the presupposition trigger type of Week 1. Specifically, having been exposed (one week before) to topical or focal presuppositions has different decreasing effects on accuracy (one week later) when the presupposition was triggered by a definite description or a temporal clause. When the presupposition was triggered by a definite description, a significant accuracy lowering effect on Week 2 responses appears to be present only if the presupposition was topical in Week 1, while the accuracy lowering effect is observed for presuppositions of both topical and focal subordinate clauses.

4 Discussion

This study investigated the impact of presupposition on epistemic vigilance, considering both the trigger type and the topical or focal status of presupposition. Specifically, using two different trigger types (definite descriptions and temporal subordinate clauses), we tested people’s accuracy in misrepresentation recognition between different informational conditions (i.e., topical presupposition, focal presupposition, and assertion) twice: first, immediately (Week 1) and then after one week (Week 2).

Overall, these results confirm that informational conditions such as presupposing and asserting impact epistemic vigilance both in immediate mistake recognition rates and in memory creation. Four main patterns of results emerged. First, when immediate judgments are solicited (Week 1), presupposition proves more effective than assertion in diverting attention from a misrepresentation, irrespective of its topical or focal status. Second, always in immediate judgements (Week 1), truth assessments are more difficult on temporal clauses and their assertive counterparts than on definite descriptions and their corresponding assertions. Third, this difference seems to diminish when memory representations are concerned (Week 2). Fourth, information structure (i.e., topic-focus articulation) has different impacts on the creation of memorial representations in Week 2 depending on the presupposition trigger: with definite descriptions, only topical presuppositions are more misleading than assertions, while with subordinate clauses both topical and focal presuppositions are more misleading than assertions. We elaborate on these points in turn below.

1. Topic-focus status does not affect immediate information acceptance

In Week 1, people noticed misrepresentations to a significantly greater extent when they were asserted rather than presupposed. In contrast, the topical versus focal status of presupposition was not found to affect misrepresentation recognition rates. This aligns with theoretical literature acknowledging the distracting influence of presupposition in general, and with previous empirical findings indicating that wrong information is less easily noticeable when presupposed (Ducrot 1972; Erickson and Mattson 1981; Givón 1982; Loftus 1975; Lombardi Vallauri 2019; Macagno 2023; Pietrandrea 2020; Reboul 2017; Sbisà 2007, 2023]; Schwarz 2015; see Section 1 above). However, it adds to the existing literature that the impact of informative status (i.e., topical vs focal) on the working of presuppositions (Dahl 1974; Lombardi Vallauri 2009; Strawson 1964) may not always be decisive. All in all, our data seem to indicate that in the immediate, the turning of the presupposed information into a “secondary main point” (Abrusán 2016) does not significantly raise people’s vigilance towards that piece of information. In other words, people tend to accommodate presupposed information quite independently from the information structure of the utterance.

It is also to underline that the misleading effect of presupposition was found even if the participants had the possibility of answering with “I don’t know”, indicating the robustness of the misrepresentation-concealing effect of presupposition.

1. When immediate judgments are solicited, complex events are more difficult to judge as compared to bare entities

Results indicate that in Week 1, definite descriptions and subordinate clauses display differing accuracy rates in both their presuppositive and assertive forms, with errors in subordinate clauses and their assertive counterparts proving more challenging to detect. A possible explanation for this pattern lies in the differing levels of complexity inherent in the information these linguistic structures encode. Evaluating subordinate clauses or their assertive versions requires making judgments about complex events, whereas assessing definite descriptions or their assertive versions involves judgments about the existence of entities in context. It is likely that making judgments about complex events is in general more difficult than making judgments of existence. Indeed, a difference in accuracy was also found for the assertive counterparts of presuppositions, i.e., for sentences where no presupposition was present.

Conversely, in Week 2, judging the truth or falseness of a sentence based on our memory of a complex event does not seem to be more difficult per se than judging the truth or falseness of a sentence based on what we recall about an entity. Also, the accuracy for the assertive versions of the stimuli decreases from Week 1 to Week 2, but only for the assertive versions of definite descriptions. Tentatively, both patterns could be put down to the fact that our cognitive system compresses representations when stocking data (Christiansen and Chater 2016). This compression likely reduces the distinction between recalling information about entities and events, contributing to the reduced differences between definite descriptions and their assertive counterparts, on one hand, and subordinate clauses and their assertive counterparts, on the other. Regarding the decreased accuracy for assertive versions of definite descriptions, asserted information in Week 1 likely benefited from its explicitness and semantic simplicity, making it easier to process and evaluate at the time of exposure. In contrast, presupposed information, being embedded within the structure of the statement and taken for granted, required greater inferential effort, likely making it less readily critically evaluable in Week 1. By Week 2, the task shifted to recalling and evaluating previously encoded information. Given the inherently imprecise nature of memory retrieval, distinctions between the explicitness of asserted information and the implicitness of presupposed information may have blurred. This “levelling effect” could have diminished the sharp distinction between the two types of information over time.

1. The belief-shaping power of presupposition appears affected by the interplay between trigger type and information status

In line with Loftus (1975), Week 2 results indicate that presuppositions have a greater effectiveness than assertions in the forming of wrong memorial representations. However, this time, both the focal versus topical status of presupposition and the trigger type play an important role. The fact that the two presupposition triggers are associated with different response patterns in Week 2 indicates that the heterogeneity between triggers also reverberates on the presupposition’s capacity to affect memory formation. Specifically, presuppositions presented in Week 1 with a temporal clause, no matter if topical or focal, have (as compared to assertions) a significantly stronger misleading effect on people’s answers in Week 2. Thus, with temporal clauses, the presupposition versus assertion opposition seems to be the decisive condition that determines the formation of long-term memories. Conversely, definite description focal presupposition appears to leave traces in memory more similar to its assertive counterpart than to topical presupposition after one week. It is worth noting once more that this effect of presupposition on memory formation was found even if the participants had the option of answering with “I don’t know”, highlighting the robustness of the phenomenon.

We would like to explore a possible, though speculative, interpretation of the described pattern. Summarising, at the moment of their first presentation, presuppositions triggered by both definite descriptions and temporal clauses align together in being significantly more misleading than assertions, irrespective from their being topical or focal presuppositions. But one week later, a split has occurred in the mnemonic traces left by the same presuppositions: temporal clause presuppositions prove still more misleading than assertions both if they had been presented one week before in the topical or in the focal status, but definite description presuppositions keep their full misleading effect only if they had been presented as topical presuppositions, while focal ones rather align with the lower misleading effect of assertions. This split may reveal that presupposition accommodation is not a single, uniform process, but rather a dual mechanism, consisting in (1) a provisional pragmatic acceptance for the purposes of the ongoing communicative interaction, and (2) a cognitive acceptance for the purposes of one’s belief updating.

Actually, these two possibly different processes have already been described separately as two possibly different causes leading to accommodation, i.e. to the acceptance of a presupposed content even if it is questionable. In our introduction above, we have recalled some literature^[7] suggesting that framing a specific piece of content as a presupposition often signals that it does not require extensive scrutiny, possibly causing a cognitive bias, namely, lesser attention on its being questionable or false. On the other hand, it has been suggested several times in the literature that there is another reason why accommodation happens, driven more by pragmatic cooperation than by a cognitive bias.^[8] As Ducrot (1972]: 91) puts it, presuppositions “set the price” for the conversation to continue. This is further clarified by Sbisà (2007), who says that refusing to accommodate a presupposition means delegitimising and, thus, marginalising from the conversation, the speaker who produced it. Consequently, addressees tend to accept presuppositions to maintain the communicative relationship. Macagno (2015) further observes that challenging a presupposition also comes with the price of being recognised as naive and unprepared. Finally, commenting on a possible joint effect of pragmatic cooperativity and cognitive bias for the persuasive impact of presuppositions, Lombardi Vallauri (2024) builds on the fact that, for the reasons just recalled, objecting to a presupposition is undesirable, and the behaviour we are by far most used to is to accommodate presuppositions, i.e., to treat them as if their content is true. When presuppositions appear, we prepare for accepting them without evaluating them in depth. Of course we could still cognitively evaluate and reject the content of a presupposition without telling the interlocutor. However, since we avoid to challenge presuppositions explicitly, questioning them interiorly is typically of too little use for the effort to be worth it. As a consequence, language users mostly do not challenge presuppositions at all, neither interiorly nor explicitly.

Now, if these two different biases both distinguish presuppositions from assertions, and if they result in two different processing mechanisms constituting what is usually just called “accommodation”, it is no wonder that under specific circumstances they can occasionally split. This may be what we observed in our data. In the case of definite descriptions, their being focused may not prevent the provisional pragmatic acceptance of the presupposed content for the purposes of the ongoing exchange, but it may selectively be able to prevent their referent from becoming permanently ingrained in the addressee’s set of beliefs, resulting in the observed difference between topical and focal definite description presupposition in Week 2. In contrast, when temporal clauses are concerned, focus status seems incapable of selectively hindering the belief-shaping effect of the presupposition: both with topical and focal presuppositions, the two accommodation processes run in parallel. A possible cause may be that, in the case of subordinate clauses, the increased complexity of the mental representation (as compared to – merely nominal – definite descriptions) could make the separate management of the two mental processes (provisional acceptance within the ongoing exchange vs. acceptance as a permanent belief) more difficult. Thus, the differences among the triggers appear to be related to the semantic complexity of the constructs they produce. In other words, reacting to presupposed information on two distinct tracks might be easier when dealing with simpler information, like in the case of definite descriptions, but it might become more difficult when more complex mental representations are concerned, such as in the case of temporal clauses. In other words, a lower cognitive load may allow for the diverging managements of two separate mental processes, while a higher cognitive load should lead to merging them into a single, more ‘opaque’ process. The latter would lead to the same accommodating behaviour for both the provisional pragmatic purposes of the exchange and the permanent cognitive purposes of belief formation, resulting in no difference between topical and focal temporal clause presupposition in Week 2.

If these hypotheses are true, they may represent an important and quite new issue for the theoretical inquiry on presupposition. Our data suggest that what is usually called “accommodation” not only has (at least) two different causes, but it may also encompass two corresponding separate mental processes. Furthermore, how these processes are engaged may depend on the concurrent cognitive load, i.e. on the processing effort associated with computing the content of the presupposition. Specifically, the two processes may be more easily separated when the cognitive load is lower, like in the case of simple-entity-encoding definite descriptions. Conversely, when the cognitive load increases, such as in the case of complex-event-encoding temporal clauses, distinguishing provisional, pragmatic acceptance from definitive cognitive refusal could become too demanding.

5 Conclusions and future lines of research

Overall, our results add further evidence to the claim that presupposition is linked to lower levels of epistemic vigilance as compared to assertion, in line with previous theoretical and experimental research (Ducrot 1972; Erickson and Mattson 1981; Givón 1982; Loftus 1975; Lombardi Vallauri 2019; Macagno 2023; Pietrandrea 2020; Reboul 2017; Sbisà 2007, 2023]; Schwarz 2015). This is also reflected on the shaping of memories. However, this study enriches the literature by experimentally comparing different information statuses (topic or focus) of presuppositions and contrasting different triggers (definite descriptions and temporal clauses). While the turning of the presupposed information into a (secondary) main point when it is focused does not seem to affect mistake recognition rates straightaway, it has an impact on the long-term shaping of memorial representations, but only in the presence of a certain type of trigger (i.e., definite descriptions).

These findings suggest that accommodating presuppositions can encompass two different processes: provisionally accepting information for the sake of the conversational interchange, and forming one’s beliefs. The data we collected suggest that these processes apply differently depending on the information structure of the utterance and the type of trigger involved. More specifically, we put forward the hypothesis that one of the two processes can be more easily inhibited separately from the other when the concurrent cognitive load is relatively low, while when the cognitive load increases, they tend to be managed jointly.

This hypothesis obviously needs further experimental investigation, specifically designed to verify it under varying cognitive load conditions. Valuable insights may also come from studying presupposition accommodation in development or late adulthood. These two populations are recognised to have lower cognitive abilities (e.g., in terms of working memory capacity) than young adults (Cappell et al. 2010; Reynolds et al. 2022). Hence, examining their (possibly lower) capacity to process the two mentioned accommodation components separately could provide data on the role played by cognitive load.