1 Introduction

Philosophers traditionally think that matters of ontology are properly to be reserved to the philosophical discussion. Philosophy must decide what kind of beings must be accepted in the overall ontological domain. At most, a layman may discuss about these matters just in the Sellarsian (1962) manifest image of the world, the one ordinary people mobilize when involved in their ordinary transactions. But it is the scientific image of the world that must tell us what there properly is out there, the image that philosophical reflection continuous with science contributes to articulate.

Yet that a layman’s reactions on this matter cannot contribute to the progress of the philosophical discussion in ontology is not to be taken for granted. This is especially the case if, in the philosophical debate concerning whether entities of a certain kind – in particular, fictional characters: ficta, for short – are to be allowed in the overall ontological domain, ontological positions turn out to be, partially at least, based on philosophers’ intuitions, as Kripke’s semantical argument against descriptivism on proper names has notoriously been charged to do (Machery et al., 2004). For, as such, these intuitions are certainly disputable, for example because they are not shared enough. In particular, it is hard to dismiss such layman’s reactions if they come out of some experimental findings which, although they could certainly be interpreted differently, are best interpreted as enabling a certain ontological interpretation. This is what we aim at showing in this paper, by presenting some data that taken together suggest, in their best interpretation, that on the one hand, in the overall domain of beings, people admit ficta over above ordinary entities such as concrete individuals, and yet, on the other hand, that such an admittance does not allow for fictional characters and concrete individuals to properly interact, for they belong to separate ontological subdomains.

2 The Ontological Debate about Ficta and Its Theoretical Assessment

In ontology, over and above other similar and possibly more famous debates (e.g. on numbers, other minds, or universals) there is a traditional debate as to whether the imaginary protagonists of novels, poems and movies, i.e., fictional characters, must be allowed or not in the overall ontological domain. Notoriously, their metaphysical differences notwithstanding, Meinong-inspired philosophers – Neo-Meinongians following Meinong (1960); Modal Meinongianists à la Berto (2013) and Priest (2016)– Modal Realists à la Lewis (1978), or even Artefactualists following Ingarden (1973), as well as other scholars defending further metaphysical accounts on ficta, have ordinarily been considered as constituting the group of ontological realists on ficta, allowing them in the overall domain of beings. Instead, ontological antirealists in general about ficta, ideally starting from Russell (1905), deny that we should admit such entities.

Various arguments have been provided by both parties in favor of their positions. The controversy originally starts with a Meinong-like argument seemingly refuted by Russell. On the one hand Meinong, in a suitable reconstruction, puts forward the following argument (1960: 83). First premise: in order to merely say truly that a non-existent object, as a fictum is supposed to be, does not exist, one must designate it. Second premise: in order to so designate that object, that object must already be admitted within the overall ontological domain. Conclusion: in order to merely say truly that such an object does not exist, such an object must already be admitted within the overall ontological domain. On the other hand, one of the seeming consequences of Russell’s (1905) allegedly antirealist theory of descriptions is that such a theory allegedly allows him to reject the aforementioned first premise of Meinong’s argument,Footnote 1 by claiming that, once singular terms for non-existent objects are analyzed in terms of that theory, one may truly say that one such object does not exist without committing oneself to one’s designating it (cf. e.g. Berto, 2013: 6, 24).

In the proceedings of the controversy, the most popular arguments – not necessarily the most convincing onesFootnote 2 – have remained the semantically-based ones. Realists claim that it is impossible to account for the real truth-conditions, hence for the real truth-values as well, of a series of fiction-involving sentences (namely, sentences that have either directly or indirectly to do with a linguistic practice of fiction) without being ontologically committed to ficta. Anti-realists deny this claim, by purporting to show that one can provide a truth-conditional account of such sentences that does not appeal to such entities. In particular, this is true of fiction-involving negative existentials, whether general or singular. According to realists, sentences respectively of the form “There is no such thing as a F (general negative existentials) and “a does not exist” (singular negative existentials) respectively quantify over a non-existent fictum and make reference to what is purportedly designated by “a”, i.e., a genuine singular term allegedly for a certain fictum, in order to truly say that either that a fictum satisfying the condition expressed by the predicate “F” does not exist or straightforwardly that such a fictum does not exist. Antirealists deny this, by arguing that one can suitably read those true sentences as making no such quantification or reference, for example by meaning à la Russell that genuine singular term as a disguised definite yet denotationless description.

In a quite recent episode of this controversy, Anthony Everett (2013) attacks artefactualists à la Thomasson (1999) by charging them with the accusation that, if one sees things properly, they do not even account for the datum of the truth of fiction-involving negative existentials. Indeed, he says, since artefactualists are ontologically committed to ficta, they are forced to deny, utterly non-intuitively, that such negative existentials are true. For if one is ontologically committed to ficta, Everett explains, then one is obliged to assert that there is a certain thing such as a F (a given fictum), or equivalently, that a (that very fictum) exists (2013: 148).

But is the artefactualist really forced to deny something that indeed is a pretheoretical datum all parties in the debate should agree upon; namely, that ficta do not exist? As one may imagine from what we said one paragraph above, the answer to this question is not utterly positive as one might expect in the wake of Everett. In actual fact, for the artefactualist matters are more complex. For her, such negative existentials are indeed true, but only if the scope of the existential quantifier in the general negative existential, or the extension of the first-order predicate of existence in the singular negative existential, is circumstantially restricted to the subset of spatiotemporal beings, i.e., concrete individuals like you and me. Otherwise, if no such restriction holds, i.e., if the quantifier ranges over, or the extension of the first-order existence predicate is the same as, the overall ontological domain of beings, then for the artefactualist such sentences are false. Put alternatively, those sentences work just as sentences like “there is no beer” or “beer does not exist”. If the scope of the existential quantifier or the extension of the first-order existence predicate is restricted to my beerless fridge, the latter sentences are true. Otherwise, if no such restriction holds, such sentences are false; the whole world is full of beer, so to speak (Predelli, 2002; Thomasson, 1999; Voltolini, 2006).

If this is the case, then for artefactualists, and possibly for realists of all sorts, negative existentials have no truth-value per se, but only once they are assessed circumstantially: in certain circumstances (the restricted ones), they are true, in other circumstances (the non-restricted ones) they are false. As Amie Thomasson has summed up: “handling nonexistence claims involving fictional names is tricky: in some cases […], they are clearly true; in other cases […], they may be false” (2010: 112–3).Footnote 3 Of course, for them the opposite holds as regards fiction-involving positive existentials, whether quantified or singular, i.e., sentences respectively of the form “there is such a thing as F” and “a exists”. When they are taken restrictedly, the sentences are false; yet when they are taken non-restrictedly, they are true. For the artefactualist, this non-restricted assessment of such positive existentials is relevant to show that we are ontologically committed to ficta.

Who is right in this controversy mobilizing fiction-involving existentials? Well, one serious possibility is to see that the first contextual factor that is determinant for solving the controversy is co-text, where by “co-text” one means some linguistic material that is added to a sentence in order to favor its overall contextual interpretation. For example, if one adds “While using its lawn mower” to the sentence “Charles III cuts the grass”, it becomes clear that the sentence must be contextually interpreted as saying that His Majesty is mowing the grass and not tearing it in little pieces. In the case at stake, the different linguistic material that is added to existential sentences involving the very same proper names allegedly for ficta is what transforms such sentences in comparative existentials, i.e., sentences in which, as regards existence, a comparison is drawn about different items. For once one asks a layman how to properly assess comparative existentials, one can show that the assessment of existential sentences does not support the alleged intuition of an ontological decommitment about ficta, as the antirealist originally claims. Instead, as we will see in the next Sections, if properly interpreted, it suggests a more nuanced positive ontological interpretation according to which there are ficta in the overall ontological domain, yet taken as belonging to a subdomain failing to include concrete individuals.

In a suitable experiment, we will indeed check how the experiment’s participants evaluate certain kinds of comparative positive existentials. In particular, we will see how such participants evaluate comparative positive existentials involving, on the one hand, proper names allegedly for ficta and for concrete individuals respectively, and, on the other hand, the very same proper names allegedly for ficta and proper names indisputably empty respectively; the latter are proper names about which nobody would ever claim that they refer to something. Some examples of the latter names have been given in the debate. One kind of examples involves misunderstanding; e.g. the case of “Max”, as a case of an expression misheard as a name, in misunderstanding someone else’s utterance of “Go to the max!” (Kroon, 2003: 156), or the case of “Moloch”, a word that is actually a noun (synonymous of “king”), yet it is misunderstood as if it were a name of a deity (Kripke, 2013: 70). Another kind of examples involves making-up: e.g. Kripke’s (2013: 81) “Snazzo”, a name invented by Kripke himself and that nobody has used not even in the attempt at meaning a fictum.

Our hypothesis is that, if we asked the participants of our experiment whether they are in agreement with, or in other and perhaps more technical words, what is the truth-value of a comparative positive existential like:

  1. (1)

    Emma Bovary exists, as well as Yoko Ono;

where “Emma Bovary” is the name allegedly for Flaubert’s unfortunate fictional heroine while “Yoko Ono” stands for the renowned concrete Japanese artist that was John Lennon’s favorite partner, such participants would tendentially react by saying that (1) is false. Yet, our hypothesis continues, this would not be the participants’ reaction if we asked them to evaluate the following comparative positive existentials:

  1. (2)

    Penelope Cruz exists, as well as Yoko Ono;

  2. (3)

    Angela Merkel exists, as well as Yoko Ono;

  3. (4)

    Donald Trump exists, as well as Yoko Ono,

which all involve genuine singular terms for concrete individuals, for they would quite likely be evaluated by those people as true. However, our hypothesis proceeds, the same people would provide a reaction analogous to the reaction they provide as regards (1) if they were asked about the truth-value of:

  1. (5a)

    Emma Bovary exists, as well as Moloch,

  2. (5b)

    Emma Bovary exists, as well as Snazzo,

where “Moloch” and “Snazzo” are the two aforementioned indisputably empty proper names that nobody would ever dream of taking as names for ficta, by disagreeing with both (5a) and (5b), hence taking them to be false. Unlike the cases of:

  1. (6)

    Emma Bovary exists, as well as Anna Karenina;

  2. (7)

    Emma Bovary exists, as well as Desdemona;

  3. (8)

    Emma Bovary exists, as well as Leopold Bloom,

where all the names involved purportedly refer to fictional characters, which would be evaluated by those participants as true, by agreeing with them.

If our experiment will yield such results, it may be interpreted as suggesting two things at one and the same time. First, given the falsity of both sentences like (1) and sentences like (5a-5b), that participants are willing to accept both concrete individuals and ficta as belonging to the overall ontological domain. Yet second, given the specific falsity of sentences like (1), that participants also take concrete individuals and ficta as inhabiting separate subdomains of that domain, which prevents such items from having a proper interaction.

This possible conclusion, especially the second one, would also be in tune with already available results in cognitive sciences. Previous findings in developmental psychology have shown that, unlike young children, adults and mature children recognize that concrete individuals and ficta cannot meet, as they live in distinct realms (Stolnick & Bloom, 2006; Weisberg & Bloom, 2009). Skolnick and Bloom (2006) showed that children are able to distinguish between different fictional worlds and reality, suggesting that by 4 to 6 years of age, children already have knowledge about different ficta and can correctly use this knowledge. Weisberg and Bloom (2009) moreover found that even 3- to 4-year-olds children both believe that ficta that exist in one pretend game do not necessarily exist into another such game and can psychologically separate pretend game worlds when given spatial or temporal cues to their separation are provided.

Different studies have demonstrated that the fantasy/reality distinction develops with age (Martarelli & Mast, 2013; Martarelli et al., 2015): 7–8-year-old children showed a fundamental categorical distinction, comparable to that of adults, whereas 3–4-year-old treated the real world just as one of many worlds. Tested on a fantasy/reality distinction task, i.e., to judge whether particular entities were real or fantastic, Martarelli and Mast (2013) reported that children aged 3 to 8 years old, but not adults, show a tendency to err by judging fantastic entities as real (i.e., response bias toward reality). The authors suggested that the process of classifying items into real versus fantastic categories develops at least until children are 7 to 8 years old. Their analyses also revealed a developmental trend in children’s sensitivity to the fantasy/reality distinction: 3–4-year-old had poorer performance than all other groups, and 5–6 and 7–8-year-old children performed worse than adults. Finally, Martarelli et al. (2015) showed that, after statistically controlling for age, non-verbal intelligence, and language skills, the theory of mind abilities, i.e., the ability to attribute mental state to others, still significantly contributed to the prediction of fantasy understanding.

3 Hypothesis

On the basis of what we said in the previous Section, the hypothesis that we want to test is whether the truth-values that are ascribed to comparative positive existentials are context-, or better cotext-sensitive; notably, whether they depend on the kind of linguistic material that is added to the original existential sentence in order to generate a comparative positive existential. In particular, if the comparative positive existential involves both genuine singular terms for concrete individuals and genuine singular terms purportedly for abstract objects of a particular kind, i.e., fictional characters – as in (1) – the participants will tend to judge that sentence as false, by disagreeing with the sentence. Yet, if the comparative positive existential involves both genuine singular terms purportedly for fictional characters and genuine singular terms that definitely stand for nothing whatsoever – as in (5a-5b) – the participants will tend to provide the same truth-evaluation, by again disagreeing with the sentence. In order to see that this is the case, sentences like (5a-5b) will be flanked by comparative positive existential sentences involving merely genuine singular terms purportedly for fictional characters – sentences like (6), (7), and (8). For in such a case, the truth-evaluation will likely be different: the last sentences will be judged as true, by agreeing with them.

To our knowledge, no experimental studies have previously tested this hypothesis.

4 Experiment

4.1 Method

4.1.1 Participants

107 participants took part to the study [Mean = 23.76 years; SD = 6.18; 94 female]. Such participants were all native Italian speakers. The experiment was administered online. Informed consent was obtained from every participant.

4.1.2 Stimuli and Procedure

We created forty-five written sentences in Italian (randomly presented). Each sentence consisted in a direct existence comparison of the form “Term1 exists, as well as Term2” (or the other way around: remember that in a conjunction the order of the conjuncts is logically irrelevant), where one of the terms was always a proper name purportedly referring to a fictional character. As for such characters, 5 items included literary fictional characters – e.g. Dr. Frankenstein, Ron Weasley – 5 items included fictional characters taken from comic books – e.g. Mickey Mouse, Spiderman – while 5 fictional characters were from classic tales – e.g. Rapunzel, Snow White. While the 15 fictional characters purportedly referred to by one the terms were kept constant across items, the proper name constituting the other term was manipulated in order to generate three experimental conditions: in 15 of the cases, we had a proper name purportedly referring to a fictum – e.g. “Alladin”, “Merlin” – 15 proper names referred to concrete individuals—realia, to give them a single name – e.g. “Elon Musk”, “Cameron Diaz” – while 15 proper names were utterly non referring; namely, names such as “Ollodin” or “Cerlin” that fail to refer to anything whatsoever –see Discussion. In this way, we created positive existential comparisons in three conditions: 15 ficta vs. ficta (FF), 15 ficta vs. realia (FR) and 15 ficta vs. non-items, where by talking of “non-items” we want to stress that the corresponding names do not refer at all (FnI). This is an example of each positive existential comparison type:

  1. (FF)

    Dr. Frankenstein exists, as well as Faust;

  2. (FR)

    Dr. Frankenstein exists, as well as Bill Gates;

  3. (FnI)

    Dr. Frankenstein exists, as well as Dr. Fruttestaine.

As anticipated before, the order of presentation of the two names involved in the comparison was randomized for each item.

The experiment consisted in a rating task, where the participants were required to express how they agree/disagree with the comparative positive existentials on a 1–5 points Likert scale (i.e., strongly agree, fairly agree, neither agree nor disagree, fairly disagree, strongly disagree) – plus a “don’t answer” option. Accordingly, scores around 1 show a minimum agreement to the comparative positive existentials, scores around 5 show a maximum agreement and scores around 3 represent a midpoint agreement.

One of the main problems in a judgment task is the knowledge problem, namely, the participants’ degree of familiarity with the stimuli proposed. In addition to the “don’t answer” option, to control this problem, at the experiment’s end we provided a list of all the ficta and realia stimuli used in the main task, which did not include the indisputably empty names failing altogether to refer to entities of either kind. We asked the participants to tick the stimuli that they did not know. We removed from the data set the participants’ inconsistent responses; namely, all items including characters ticked by the participants as “unknown” in the final list, as well as all items with “don’t answer” responses.

4.2 Results

The Fig. 1 shows the mean rating scores of the three experimental conditions, namely FF, FR and FnI. The target items received a mean rating score of 4.00(1.29) in condition FF, 1.73(1.15) in condition FR and 2.02(1.37) in condition FnI (Fig. 1). A non-parametric Kruskall-Wallis test was conducted to examine the differences on the participants’ judgement of the target sentences according to the experimental condition. This analysis revealed a significant effect of condition (χ2(2) = 1543; p < 0.001). Pairwise comparisons between conditions were conducted using the Dwass-Steel-Critchlow-Fligner (DSCF) method. This statistic revealed significant differences between all conditions: the target items received a significantly higher rating score in condition FF as compared to both conditions FR (W = -50.70; p < 0.001) and FnI (W = -42.67; p < 0.001); the difference between conditions FR and FnI was also significant (W = 7.56; p < 0.001).

Fig. 1
figure 1

Mean rating scores of target items in conditions FF, FR and FnI

5 Discussion

The results support the psychological plausibility of our hypothesis. First of all, the participants agree more with comparative existential sentences like “Dr. Frankenstein exists, as well as Faust; namely, ficta vs. ficta condition (FF; mean rating score of 4.00) than with comparative existential sentences like “Dr. Frankenstein exists, as well as Bill Gates”; namely, ficta vs. realia condition (FR; score 1.73) or “Dr. Frankenstein exists, as well as Dr. Fruttestaine”; namely, ficta vs. non-items (FnI; score 2.02), with a degree of agreement as FR < FnI < FF. Moreover, both FR (ficta vs. realia) and FnI (ficta vs. non-items) conditions reported rating score below the midpoint (score 3). Thus, when asked to compare a fictional character with concrete individuals as regards existence, the participants tend to say that they behave differently: concrete individuals exist, the character does not. But they tend to say the same yet by having the opposite kind of reaction concerning ficta when they are requested to compare a fictional character with non-items: the participants tend to say that they behave differently as well, but because unlike the latter, the fictional character exists. There remains a significant difference between the FR and the FnI conditions, namely the fact that, although in both cases there is a negative evaluation as regards the comparison of the elements involved – ficta vs. realia, ficta vs. non-items – with respect to their existence, this evaluation is more stressed in the FR than in the FnI condition. This possibly depends on the fact that fictional characters are perceived as standing so to speak in an intermediate condition between concrete individuals and non-items.

As a first reaction, one may put in question whether the data we have found are relevant for semantics – what is the real truth-value of sentences involving existential statements – but not for ontology, i.e., for the issue of whether items of a certain kind should be allowed in the general ontological domain.

Yet this doubt would be scarcely motivated. For as we saw in Section 1, the ontological debate about ficta precisely moves from the very assessment of the truth-value people ascribe to positive existential sentences purportedly about fictional characters. Antirealists ground their negative ontological conviction partially at least on the intuition that such sentences are taken to be false; realists are charged to ground their positive ontological conviction on the contrary judgement that such sentences must be taken to be true only when the relevant existence predicate is taken as non-restricted. Of course, as we will repeatedly stress throughout the paper, such intuitions scarcely are ontologically decisive. Yet, if such philosophers already take their semantic intuitions as relevant for ontological matters, the more should be the case when, instead of intuitions, one can rely on shared assessment data concerning comparative existentials. This sharing legitimizes us to say that what the existential comparisons suggest is that ficta must be included in the ontological domain as along with realia, while there are no such things as non-items, for the terms purportedly referring to them are indisputably empty.

Yet at this point, one may still retort that our overall methodology is ungrounded. What ordinary people, untrained in philosophical matters, think as regards the existence of items of different kinds is one thing, what one must conclude in ontology as regards the admission of such items in the overall domain of beings is quite another thing. One cannot derive ontological results from the layman’s ontological opinions, so to say.

Granted, we agree on that there is no deductive derivation from the layman’s ontological opinions to ontological results. Yet we want not only to stress that the data we have found are very interesting in themselves and moreover that, as data concerning comparative existential sentences that were never considered before in the literature, they surely have to be considered in the philosophical debate (Castañeda, 1980). We also want to argue that, in an inductive manner, our ontological interpretation of such data is merely their best explanation. This is what we will try to prove in the remainder of the discussion, by discarding any other conceivable interpretation of the data as less plausible than ours.

To begin with, a skeptic may precisely rejoinder that the data do not support the ontological conclusions we draw from them. For since our comparative positive existentials are conjunctions, the fact that they are assessed to be false may simply depend on the fact that their shared first conjunct, e.g.:

  1. (9)

    Emma Bovary exists

both in (1) and in (5a-5b) is ordinarily taken to be false by all parties in the debate.

Yet this rejoinder would not be correct. Antirealists of all sorts, from old descriptivists à la Russell up to followers of a pretense-theoretic account saying that a sentence like (9) involves in their real, not fictional, truth-conditions some form of pretense or a betrayal of it (e.g. Everett, 2013; Walton, 1990), may say that such a sentence is false. Yet, as we said in Section 2, realist artefactualists claim that a positive existential like (9) is false when taken restrictedly, with the existence predicate as ranging only on concrete individuals, but is true when taken non-restrictedly, as ranging on all items whatsoever. This is what our data corroborate. For, when taken as a conjunct of the conjunctions (6–8), (9) turns out to be assessed as true just as those conjunctions themselves, as it must be the case if the existence predicate occurs there non-restrictedly, while when occurring in the conjunction (1), it turns out to be assessed as false just as the conjunction (1) itself, unlike the other conjunct of (1), as it must be the case if the existence predicate occurs there restrictedly. Hence, the most plausible explanation as to why the conjunctions (5a-5b) are assessed as false is that, as occurring there, (9) is assessed as true while the other conjuncts respectively containing the indisputably empty proper names “Moloch” and “Snazzo” are assessed as false, for there the existence predicate occurs non-restrictedly.

Moreover, one may retort that, as far as ficta are concerned, one must draw a distinction between internal uses of sentences, concerning what happens within a story, and external uses of sentences, concerning what happens outside a story. So, one may deal with a sentence like (9) just as one does with a sentence like “Emma Bovary is a woman”, which is true within Flaubert’s story but not outside that story, since outside that story, if Emma is ever something, it is a fictional character, not a woman. So, one may say that (9) stresses that Emma exists within the story, but not out of it. In this vein, a Quine-like antirealist philosopher who is prompted to deny that, from an ontological point of view, there is a difference between ficta and non-items – neither of them belongs to the overall domain of beings, she claims – may first equate the internal use of a sentence allegedly involving ficta with a corresponding “in the fiction”-sentence, so as to have, in the case of an existential sentence internally used, a corresponding sentence of the form “in the fiction, a exists”, where “a” is a proper name purportedly referring to a fictum. Then, she may say that the data show that, whenever it is a matter of attribution of existence to a fictional character, the only thing that is true is that in fiction, that character exists, in its there being a full-fledged entity. So for example, there is absolutely no such thing as Emma Bovary; the only thing that is true is that in the Flaubert story, i.e., Madame Bovary, Emma Bovary exists (in fact, Flaubert does not write in Madame Bovary that Emma is a trick of light or something like that, for he instead writes that she is a woman standing in a quite complicated sentimental situation).

First of all, let us reply that it would be very ad hoc as regards existential comparisons concerning ficta to interpret the existential conjunct purportedly regarding a fictum as internally, not externally, used. For of course we might have asked the experiment’s participants to also evaluate comparative existential sentences involving, over and above concrete individuals, numbers or universals, where no internal use of the relevant existential sentence is ever at stake.Footnote 4 But moreover and more relevantly, the above antirealist interpretation of the data may at most support the fact that, as we said at the end of Section 1, comparative existential sentences merely purportedly involving ficta, such as (6)-(7)-(8) before, are evaluated as true. For this evaluation is quite compatible with rereading them as:

  1. (6F)

    In fiction, Emma Bovary exists, as well as Anna Karenina;

  2. (7F)

    In fiction, Emma Bovary exists, as well as Desdemona;

  3. (8F)

    In fiction, Emma Bovary exists, as well as Leopold Bloom.Footnote 5

Yet by the same token, (1) cannot be reinterpreted as:

  1. (1F)

    In fiction, Emma Bovary exists, as well as Yoko Ono.

For unlike (1), sentences of the (1F)-kind can be assessed by ordinary people as true, not as false. Suppose that “Yoko Ono” were replaced by “Napoleon Bonaparte”. Since its concrete designatum, i.e., the French emperor, appears in Tolstoj’s War and Peace, we would quite likely have that:

  1. (1FN)

    In fiction, Emma Bovary exists, as well as Napoleon Bonaparte.

would be evaluated as true, not as false. The same would actually be the case with “Yoko Ono”, if we took the sit-com Mad About You as a fiction involving the renowned artist.Footnote 6

At this point, from the opposite side of the ontological realists, an ontologically luxuriant Meinongian-Platonic fan of the idea of ways of being, that is, of the idea that different items have a different mode for them to be, might say that the data show that the participants rank items according to their different modes of existence, which may – but need not – be matched by different degrees for items to be (McDaniel, 2017). It is false both that Emma Bovary exists like Yoko Ono and that she exists like Moloch and Snazzo. For on the one hand, she does not have existence in the same mode as Yoko Ono, since, unlike the artist, she is not a concrete individual, but has a different way of being, i.e., it is a fictional character, a kind of abstract object. Yet on the other hand, she still has a higher mode of existence in terms of its abstractedness than the two non-items Moloch and Snazzo. More in general, according to this interpretation, in the overall hierarchically organized chain of beings, on the one hand (actually, pace Plato) concrete individuals have a way of being higher than that of abstract objects, as often fictional characters are metaphysically taken to be. Yet on the other hand, qua abstract objects fictional characters have a way of being still higher than mere non-items like Moloch and Snazzo.

From the point of view of philosophy of language, this interpretation of the data is a contextualist one (e.g. Recanati, 2003). For it holds that the relevant predicate of existence changes its meaning – from meaning concretely existing to meaning abstractly existing – in the different comparisons mobilized by the respective sentences. More precisely, (1) and (5a)-(5b) are respectively taken to mean the same as:

  1. (1C)

    Emma Bovary exists in the same concrete way as Yoko Ono;

  2. (5aC)

    Emma Bovary exists in the same abstract way as Moloch;

  3. (5bC)

    Emma Bovary exists in the same abstract way as Snazzo.

Yet not only this contextualist interpretation overintellectualizes the participants’ reactions, by ascribing them a sort of implicit metaphysics of beings that they may well not possess, qua ordinary people that have no particular interest in metaphysics. Folks may believe in ghosts and spirits just as they may believe in Emma Bovary and Anna Karenina, but they plausibly have no proper metaphysical idea of what kind of entities they respectively are. But also, it attributes to non-items a sort of existence, the lowest one in the supposed chain of beings, where this attribution is not supported by the data. It is a philosophical conception to hold that, whenever a predication occurs in a sentence, there is something of which the property that is so predicated, a property of existence in this case, applies, by simultaneously postulating in our case three ways of being that respectively amount to three different species of the existence property as a common genus: i.e., existence as a concrete object, existence as an abstract object, and existence as a non-item, whatever this means. Meinong (1960) for one, was tempted by this idea (in order however to ultimately reject it), by ascribing to non-items quasi-being (Mulligan, 2019, Taieb, 2020). But it is hard to ascribe to ordinary people untrained in philosophy that philosophical conception.

Fortunately enough, however, there is a less demanding way of interpreting the data, which we espouse. This way begins by remarking that one and the same general property of existence is predicated in the relevant comparison, but continues by noting that such a predication is differently evaluated with respect to different domains. Hence on the one hand, with respect to a restricted context viz. subdomain of concrete individuals, the one a sentence like (1) mobilizes, (1) is taken to be false. For in that context, existence only holds of such individuals, not of fictional characters, since the domain in question is just the restricted domain of concrete individuals. Yet on the other hand, with respect to the unrestricted context sentences like (5a)-(5b) mobilize, the context appealing to the overall domain of beings, (5a)-(5b) are also taken to be false. For existence also holds of fictional characters, insofar as they figure in that overall domain, but not of anything outside that domain, for there is no such thing. In actual fact, that anything is not a thing at all, since as regards non-items the only things that there are, properly speaking, are just indisputably empty names like “Moloch” and “Snazzo”.

This situation does not specifically concern sentences involving the “existence”- predicate. Even if we had sentences like “Boris Johnson is dancing, as well as Rudolph Nurejev” and “Boris Johnson is dancing, as well as that turtle over there”, both sentences would likely be evaluated as false insofar as the extension of the predicate “to dance” involved in such sentences is respectively restricted to elegant dancers in the first case and non-restricted in the second case.

Still from the point of view of philosophy of language, our interpretation is a relativistic interpretation, in which one and the same sentence with the very same semantic content is evaluated with respect to different domains. Consider a sentence like “It rains”. A relativist may say that the sentence with its fixed semantic content is true in a domain restricted to Iceland, but false in a domain restricted to Qatar (Recanati, 2007). Likewise, our interpretation holds that the relevant existence predicate always means the same thing, i.e., the very same property of existence, yet the sentences in which that property is predicated are evaluated with respect to different contexts in the sense of different domains of items whose extension thereby differs: notably, the restricted domain of concrete individuals and the unrestricted overall domain of beings (Voltolini, 2012, 2018). Now first, as Meinong himself (1960) ended up by understanding, merely figuring in the overall domain does not require for something to have any way of being whatsoever, in the technical sense of the expression. As we saw in Section 2, for Meinong a thing is given, so that it belongs to the overall domain, hence one has it at one’s disposal if one thinks of it, period. Additionally, for him some of such things have Sein and some other do not, where having Sein is a non-universal first-order property, but this is irrelevant for ontology. Second, pace McDaniel (2017), quantifier restriction does not entail ways of being either. Granted, as regards sentences like (1), the existential domain is restricted to the subset of items sharing the property of being a concrete individual. Yet, in order to provide such a restriction, there is no need for the relevant property shared by the items in the relevant subset to be a way of being. As our original example with “there is no beer” shows, restricting the domain to items lying in a fridge does not mean that lying in a fridge is a way of being. For the items sharing that property may belong to different categorical kinds – for example, sadly enough, the fridge in question may contain both animate and inanimate concrete individuals, and less sadly, within inanimate concrete individuals, both vegetables and minerals, etc.

Now, not only this relativistic way of interpreting the data is less ontologically and metaphysically demanding than the contextualist way, but it has a larger explanatory power. For it may also be appealed to in other close situations that do not involve existence. Consider a couple of sentences inspired by Charles Travis’ (1997) famous Pia’s example. Suppose that Pia paints in green some of the russet leaves of her plants, while painting in another color, say violet, some of her russet pears. If one chromatically compares such leaves with such pears, one obtains a sentence that is likely false:

  1. (10)

    Pia’s leaves are as green as her pears.

Yet suppose now that Pia’s leaves are chromatically compared with some vegetables that grow in her garden without any depictive intervention, say, Pia’s naturally violet aubergines. One thus obtains another sentence that is likely false just as (10):

  1. (11)

    Pia’s leaves are as green as her aubergines.

An easy interpretation of this situation is that, while (10) tends to be evaluated as false with respect to a context viz. a restricted domain that contains only artefactually painted objects, some of which – the leaves – are artefactually green while some others – the pears – are artefactually violet, (11) tends to be evaluated as false with respect to a context viz. an unrestricted domain that contains both artefactually painted green objects – the same leaves – and naturally colored violet objects – the aubergines.

A consequence of this interpretation is that the anaphora implicitly occurring in sentence like (5a) and (5b) must be properly reinterpreted. For, insofar as names such as “Moloch” and “Snazzo” do not refer to anything whatsoever, not even a non-existent fictum, the conjuncts of such sentences implicitly involving predications of the same existence predicate respectively have an ordinary and a metalinguistic meaning, as if (5a) said that Emma Bovary exists while “Moloch” refers and (5b) said that Emma Bovary exists while “Snazzo” refers. But this use-mention shift with anaphora is not particularly problematic. For it also holds in completely different sentences, such as e.g. Partee’s (1973: 412) “As soon as he asked ‘Where is Jane?’, she arrived”.

Yet once room is found for interpreting the data partially metalinguistically, the antirealist may rise up again and put forward a wholly metalinguistic interpretation of the data, by rereading the relevant phrase of the kind “NN exists” occurring in the comparative positive existential as meaning something like “‘N.N.’ refers”.

Granted, this radical metalinguistic interpretation accounts for the evaluation of both (1) and (5) as false that the participants to our experiment provide. For when they are respectively reread as saying:

  1. (1M)

    “Emma Bovary” refers, as well as “Yoko Ono”;

  2. (5aM)

    “Emma Bovary” refers, as well as “Moloch”;

  3. (5bM)

    “Emma Bovary” refers, as well as “Snazzo”,

all such sentences will be judged to be false, insofar as in (1M) the first conjunct is false for the antirealist – for her, “Emma Bovary” is in the same predicament as “Moloch” and “Snazzo”, that is, it fails to refer to something tout court – while both in (5aM) and in (5bM) both conjuncts are false – for the aforementioned reason.

Yet once again, the antirealist is unable to account for how the participants evaluate comparative existential sentences involving just names purportedly for ficta such as (6), (7), and (8). For as we said, such participants tend to evaluate such sentences as true. Yet when metalinguistically reread as:

  1. (6M)

    “Emma Bovary” refers, as well as “Anna Karenina”;

  2. (7M)

    “Emma Bovary” refers, as well as “Desdemona”;

  3. (8M)

    “Emma Bovary” refers, as well as “Leopold Bloom”,

for the antirealist they should turn out to be false, because of the alleged falsity of all their respective conjuncts given the alleged failure (for her) of all those names’ reference.

To be sure, a moderate antirealist may acknowledge that there is a difference in failure of reference between names like “Emma Bovary” on the one hand and names like “Moloch” and “Snazzo” on the other hand. While “Emma Bovary” is used in order to actually unsatisfactorily refer to something, neither “Moloch” nor “Snazzo” are so used and simply do not refer at all. This acknowledgment may allow her to suitably reinterpret metalinguistically our comparative positive existentials (1) and (5a-5b):

  1. (1MM)

    “Emma Bovary” merely purportedly refers, as well as “Yoko Ono”;

  2. (5aMM)

    “Emma Bovary” merely purportedly refers, as well as “Moloch”;

  3. (5bMM)

    “Emma Bovary” merely purportedly refers, as well as “Snazzo”.

First, this moderate metalinguistic account may explain why our participants judge both (1) and (5a)-(5b) to be false. For this account, on the one hand, (1) is judged to be false, for unlike “Emma Bovary”, which merely purportedly refers to something, “Yoko Ono” fully refers to something. Yet on the other hand, (5a)-(5b) are both judged by the experiment’s participants to be false, for unlike “Emma Bovary”, which again merely purportedly refers to something, “Moloch” and “Snazzo” utterly fail to refer to something. Second, this account may also explain why the experiment’s participants take (6–8) as true, when reread as:

  1. (6MM)

    “Emma Bovary” merely purportedly refers, as well as “Anna Karenina”;

  2. (7MM)

    “Emma Bovary” merely purportedly refers, as well as “Desdemona”;

  3. (8MM)

    “Emma Bovary” merely purportedly refers, as well as “Leopold Bloom,”Footnote 7

Yet once things are put this way, in her interpretation the moderate antirealist appeals to modes of referring: some names fully refer, others merely purportedly refer, still others utterly fail to refer. But if, as we saw before, appealing to modes of being sounded to overintellectualize matters, the same also holds of appealing to modes of referring. Granted, one may say that a philosophical interpretation of the data may disregard overintellectualizations, if it is philosophically convincing. But, as we already know from Kripke (2013), metalinguistic interpretations of existential sentences hardly provide the right modal content to such sentences. Moderate such interpretations do not escape this problem. For example, there is a possible world where sentences (2–4) are true, yet their moderate metalinguistic reinterpretations such as:

  1. (2MM)

    “Penelope Cruz” fully refers, as well as “Yoko Ono”;

  2. (3MM)

    “Angela Merkel” fully refers, as well as “Yoko Ono”;

  3. (4MM)

    “Donald Trump” fully refers, as well as “Yoko Ono”,

are false.

In the end, therefore, for us the most plausible interpretation of our data is the aforementioned relativistic one. According to it, by means of their truth-value assessments, the participants of our experiment show that they believe that concrete individuals, ficta, and non-items are in a different predicament as regards existence: on the one hand, ficta and concrete individuals are ranked together as belonging to the overall domain of beings; on the other hand, they are distinguished as belonging to different subdomains. For on the one hand, according to this interpretation, the ascribed falsity both to (1) and to (5a-b) suggests that both concrete individuals and ficta are taken by the participants together while they ask to be differentiated from non-items that do not exist at all, for the only things that there are in the overall domain of beings on their concern are the referential failures of the relevant singular terms involved (e.g., “Moloch” and “Snazzo”). Yet on the other hand, for this interpretation the specific ascribed falsity of (1) also suggests such participants recognize that concrete individuals and ficta exist separately, viz. in different subdomains (actually, those of abstract objects and of concrete individuals respectively). This interpretation does not overintellectualize the situation, for the practice of enlarging and restricting domains is a usual practice people adopt in ordinary lives (remember the “beer” example we appealed to at the beginning of the paper). Seen this way, as we said in Section 2, our data square with other findings that have already been collected in the literature: unlike young children, mature children, as well as adults, recognize that concrete beings and ficta cannot meet, for they live in somehow separate realms which prevents them from having causal interactions (while sorts of non-causal interactions, such as being moved by ficta, are possible) (Stolnick & Bloom, 2006; Weisberg & Bloom, 2009). It would be nice to replicate our findings with existential comparisons involving other items whose ontological status is controversial, e.g. Platonic entities such as universals like The Bold and the Beautiful and numbers such the Four or its square root. Our guess is precisely that they would be separated from concrete individuals and yet also kept ontologically together when it is a matter of telling all them from non-items again. But this is a matter for another work.