3.1. Discrepancies between Austin and STS

First, note that while Austin’s project is primarily concerned with singular utterances by individuals, Callon and MacKenzie have switched to talk of entire discourses of entire research fields, conceivably involving any number of utterances by any number of individuals, as well as nonverbal acts. For Callon, this is part of the definition cited in the preceding text, and MacKenzie (Reference MacKenzie2006, 19) makes a similar view explicit. Second, if, as stated, Callon’s goal was indeed to pick up on Austin, it seems peculiar to offer such a specific definition of science being performative only if a discourse changes the things “that it describes,” especially considering that one of Austin’s major moves was to collapse his distinction, arguing that all constatives are kinds of performatives, too. Would we not therefore need to consider all assertions by scientists as performatives? For this reason, the claim by Callon just two pages later that “all science is performative” (Callon Reference Callon, MacKenzie, Muniesa and Siu2007, 318; emphasis added) suddenly seems to be much more in agreement with Austin’s own collapse, but at the cost of standing in direct contradiction with the more specific definition Callon offered just pages before. Undoubtedly, it is not the case that all science contributes to the “creation of reality” in the unique way that economics and BSM are argued to be doing here; astronomy, to take an example, has yet to materialize a new star or planet by way of disseminating a theory about it. Classifying Pluto a planet would be a performative under Austin, for example a kind of assertion, but it would not affect anything about Pluto.

Offering clarification, MacKenzie (Reference MacKenzie2006, 17) proposed what is now a widely cited typology of performativity of economics. He distinguishes between generic performativity, where “an aspect of economics (a theory, model, concept, procedure, data set, etc.) is used by participants in economic processes, regulators, etc.”; effective performativity, where “the practical use of an aspect of economics has an effect on economic processes”; and barnesian performativity, where “practical use of an aspect of economics makes economic processes more like their depiction by economics.”Footnote ⁸ The latter name is a nod to sociologist Barry Barnes (Reference Barnes1983), and the self-fulfilling BSM example is argued to be a case in point.

MacKenzie’s typology is an improvement. In particular, thinking of kinds of performativity within science as subsets of each other clears a path to resolve the contradiction within Callon’s setup: While all scientific practices may be considered generically performative as in having some effect on the world, only a subset of those practices has the specific effects required to shape or bring into being its referents in the way the BSM is argued to be doing. While addressing one concern, however, MacKenzie’s typology introduces a series of more.

First, whereas Austin’s performatives are about utterances issued by a speaker and how their meanings are mediated by conventional environments, MacKenzie defines the performativity of economics in terms of use of theories. Use and utterance of propositional content, however, are different. For instance, an assertion of some theory issued by an economist should still count as performing an assertion in Austin’s sense, irrespective of whether the asserted theory is subsequently used by anyone or not. An illocutionary act of assertion might perfectly be felicitous, on the basis that it is being uttered in the appropriate conventional context such that interlocutors recognize it as an assertion, but it might still fail at its intended perlocutionary effect, namely, scientific uptake and use; people might not believe an assertion, deem it irrelevant, and so on. Austin’s insight that the performance of an utterance and its downstream consequences can come apart is entirely lost in MacKenzie’s typology, centered solely around use.

Second, there is tension between the categories on offer. Note that the label of effective performativity might suggest that generic performativity would have no effects whatsoever. Yet both uptake and use of an economic model could be considered ordinary perlocutionary downstream effects in Austin’s full sense. So what exactly is effective performativity supposed to be effective about? Looking at the definition, it tells us that the use of economics should have “an effect on economic processes,” but this raises even more questions. In particular, how could “participants in economic processes,” as detailed in the definition of generic performativity, switch to some other model without thereby also changing the very processes they are participating in? By virtue of “participants” being literal constituents of the process in question, a change in their behavior would, by construction, necessitate at least some change to that process, too, and MacKenzie offers no guidance as to where to draw the line.

Lastly, there is the question of how other disciplines that are not economics fit into the picture, especially given Callon’s claim that all science is performative. This applies even to the most encompassing category, generic performativity, defined by MacKenzie in terms of “participants” and “use.” Clearly, this does not extend to all science; you may be forgiven for having difficulties in conceiving of participants in, say, remote astronomical processes. Only speaking of economics, MacKenzie once more offers no guidance as toward how to apply his typology elsewhere.

3.2. Saving Austin from Mäki

Another perspective staged against the previously mentioned use of performativity was offered by Uskali Mäki (Reference Mäki, Karakostas and Dieks2013) titled Saving Austin from MacKenzie. Austin’s performatives, Mäki argues, are constitutive of a particular action. Uttering “I bet you X” in the right circumstances is to bet. Compare this with MacKenzie’s case, in which the uptake and use of certain theories causally influenced what happened on specific markets. Unlike the utterance of a bet, however, the dissemination of a theory clearly does not constitute market practices; at best, it causally changes them. In this manner, Mäki argues, Callon and MacKenzie would confuse the “authentic meaning of performativity” (Mäki Reference Mäki, Karakostas and Dieks2013, 447).

Mäki is right to point out that there are confusing aspects to the STS departure from Austin. Ironically, however, neither does Mäki’s exegesis appear particularly faithful to Austin, especially once we recall Austin’s own collapse of the distinction between constatives and performatives. As demonstrated in section 2, the term performative is by design ambiguous between a number of ways in which a speaker can perform actions or “do things with words,” only one of them being illocution. Granting that illocution may indeed have the constitutive character Mäki singles out, he mistakenly equates illocution as the only existing or relevant form of performativity, or way of “doing things with words.” This overlooks Austin’s main theme, which was to abandon the constative/performative distinction as an argumentative ladder, revealing that plain descriptive constatives turn out to be performatives, too, and that factual assertions are also members of some subtype of illocutions. An entirely plausible counter against Mäki’s critique would thus be to say that Callon, MacKenzie, and others have, instead of illocution, mostly been interested in particular perlocutionary effects and sequels of economists’ assertions, namely those that affected behavior among market participants and brought into being new social practices that were made possible by the adoption of a model like BSM. Nothing about Austin’s argument would prohibit employing “performativity” to refer to such perlocutionary effects. This way, although for different reasons, both Mäki’s and MacKenzie’s appropriations of performativity turn out equally faulty for invoking a highly ambiguous phrase that should, where it matters, be further specified.Footnote ⁹

It matters, I take it, a great deal what we (do not) mean by “performativity of science.” Recall the debates hinted at in the introduction—strong and potentially far-reaching claims have been made on the brittle basis of the previously mentioned notions of performativity of science, such as those on scientists’ responsibilities or criteria for model evaluation. Second, the concept of perlocution alone does not, as such, capture the more specific perlocutionary effects that, as some have put it, “bring into being” a theory’s referent, or “shape reality,” or which result in “self-fulfilling prophecies.” In what follows, I defend a positive proposal aimed at placing these debates in philosophy of science on a sounder conceptual footing.

4.1. Reflexive performativity as a perlocutionary sequel

Turn again to Austin’s own collapse of the distinction between performatives and constatives (section 2). If one accepts this move, note that one is already committed to at least one all-encompassing category that contains virtually all possible locutionary acts any given scientist might perform. These locutions may or may not have significant perlocutionary downstream effects and may or may not constitute any illocutionary act such as assertion; but, by virtue of being a locutionary act performed by an actual speaker, its execution already constitutes some change or effect on the world in a very trivial sense: The world would be different had it not been uttered.

Naturally, many of the interesting communicative acts performed by scientists often have nontrivial perlocutionary effects. If the IPCC (Reference IPCC, Skea and Reisinger2022, 8) asserts that “emissions have increased since 2010 across all major sectors globally,” perlocutionary effects may take the form of uptake among policy makers, and perlocutionary sequels may take the form of further downstream reactions, such as enacting certain policies in response to this assertion. Similarly, if a government agency asserts that certain dioxins are carcinogenic, there may be certain behavioral changes in response. If astronomers assert the existence of some celestial body, this may change a variety of social practices, as it has numerous times throughout history.

All three of those latter examples, however, lack the more specific perlocutionary effects which Callon, MacKenzie, and much of the related literature in philosophy of science have in mind; those effects that render a model shaping “the reality that it describes” (Callon Reference Callon, MacKenzie, Muniesa and Siu2007, 316), or “creating the phenomena it describes” (MacKenzie and Millo Reference MacKenzie and Millo2003, 108). Recall MacKenzie’s argument that traders at options markets priced options radically differently after BSM was disseminated and incorporated into trading procedures, in a way that changed actual market behavior to be more in accordance with BSM. These cases involve much more specific perlocutionary effects that do not bear on just any practices, but specifically on the referents or target systems that the theory, model, and so forth, is intended to address. Against this background, in a first step, it appears apt to explicate these differences as follows.Footnote ¹⁰ Asserting (or otherwise disseminating) a claim, prediction, concept, theory, model, or similar concept is

1. (i) Generically performative if it has any impact on the world at all, and it is

2. (ii) Reflexively performative if it has not just any impact on the world, but specifically on the referent or target system it is designed to depict, such that the changes caused by this impact bear on the claim’s veracity (such as, but not limited to, predictive accuracy, model-target-fit, and referential stability).

This requires more corroboration, but first, some notes on terminology. First, MacKenzie’s label of “generic” is retained; making the implied all-encompassing category explicit is useful, if just to distinguish it from more specific subsets. Nevertheless, its definition is swapped to be even more unapologetically generic, both to align with Austin’s collapse and to encompass scientific disciplines beyond economics. Second, both categories are quite broad, lumping together claims, predictions, concepts, models, and so on. This is as it should be; all those are things scientists conceivably communicate, and there is, as of now, no reason to focus on any particular one. Third, as argued previously, MacKenzie’s previous terminology of “effective” does not succeed in cashing out a salient or clear subcategory of performativity of science. I replace it here with reflexivity, for two positive reasons. For one, reflexivity is already somewhat common among philosophers of science to refer to the feedback effects of scientific practices that are at issue here (Buck Reference Buck1963; Guala Reference Guala2016b; Northcott Reference Northcott2022). Also, the term does a better job at eliciting the self-referential character exhibited by the cases in the literature (more on this shortly). Fourth, reflexive performativity subsumes MacKenzie’s “Barnesian” and “counterperformativity,” as well as other related types, such as self-fulfilling and self-defeating prophecies, reactive kinds, and others. The focus of this article will not be on delineating those; it will be on consolidating the preceding distinction with speech act theory.

4.2. Consolidating STS and Austin

As detailed earlier, STS scholars such as MacKenzie and Callon have not been particularly interested in analyses of individuals’ utterances. Much more broadly, they have been employing performativity to refer to a variety of ways in which scientific theories are mobilized to affect and bring into being all sorts of social practices, including practices that said theories have been meant merely to represent. If notions of performativity of science, like the previously mentioned, rely solely on individuals’ acts of assertion, however, this may appear insufficient for analyzing the kinds of processes and institutions required to enable such mobilization. Perhaps, then, performativity-talk is not suited to the purposes of STS and philosophy of science scholars after all.

This would be too quick a dismissal. Even while respecting speech act theory’s focus on individuals’ utterances, there is sufficient conceptual overlap for both disciplines to draw productively on each other’s resources. Note that I have taken some liberty in defining performativity not only in terms of assertions but also in terms of other ways of disseminating a claim, model, or theory, and so on. Dissemination and assertion, however, are distinct. In Austin’s and Searle’s frameworks, assertions involve a commitment of the speaker to the truth of the uttered proposition; an utterance only counts as an assertion if it is evident to hearers that the speaker takes the uttered to be true. Disseminations, in contrast, may occur whenever a claim or model is being replicated or circulated without any particular truth commitment on the part of the speaker. This may happen, say, in an undergraduate course where various standard models of economics are being taught, but no particular claim is being made as to their empirical adequacy. Other conceivable ways of model dissemination may involve, for instance, committing code to a software repository, copying pictures of physical models, or selling computed model outputs (as happened with the BSM model and Fischer-Black’s price sheets that were heavily used in trading pits). These are precisely the kinds of acts that historically have interested STS scholars. Yet, by virtue of not necessarily involving a literal speaker, let alone a speaker’s truth commitment, this broader class of disseminative acts has no obvious place in Searle’s or Austin’s frameworks.Footnote ¹¹ They are neither assertions nor easily classifiable as commissives, expressives, declarations, or the like.

On a second reading, however, speech act theory at large should, in principle, have no particular problem accounting for diverse acts of dissemination, for two independent reasons. For one, Austin already counted various forms of communication, such as written language or hand signs, as capable of carrying out speech acts. So, even when we consider Austin’s acts of assertion, we already have to picture a diverse range of possible communicative practices, extending far beyond utterances.Footnote ¹² Against that background, then, “dissemination” is helpfully ambiguous between the many ways in which claims, models, predictions, and theories may proliferate through social networks, extending to the replication of all kinds of cultural artifacts beyond written or uttered language, and not necessarily involving a speaker’s truth commitment.Footnote ¹³ Although this broader class of disseminative acts has no explicit place in either Austin’s or Searle’s frameworks, appreciating disseminatives as a distinct type of speech act should not introduce too much tension, at the very least because asserting a claim, model, and so forth, may just count as one among many possible ways of disseminating a claim, model, and so on.

Second, even when one does not want to count all forms of disseminations as types of speech acts, that is if one denies the existence of disseminatives, acknowledging the existence of diverse ways of downstream replication of someone’s locutionary content should still not disrupt speech act theory’s grand picture: To pick up on a prior point, a replication or use of someone’s claim may perfectly be regarded as a perlocutionary effect or sequel of said assertion.

In this manner, talk of reflexive performativity in science can, in principle, cater to both camps. Defined in terms of assertions and other ways of disseminating scientific outputs, it retains Austin’s focus on a speaker’s (or disseminator’s) act, while also accommodating MacKenzie’s emphasis on the diverse ways of mobilizing the products of science, and their diverse consequences. As such, the distinction between generic and reflexive performativity requires hardly a concession from either side: It neither confines reflexive performativity of science to individuals’ utterances, nor does it require a particular stance on whether disseminations are a distinct type of speech act. At minimum, it requires that there can be perlocutionary effects that bear on a claim’s, model’s, and so forth referent or target system.Footnote ¹⁴

4.3. From illocutions to referent malleability: An orthogonal typology

With reflexive performativity, we have now introduced a term that is willfully ambiguous between asserting and other ways of disseminating scientific content. As a consequence, however, its relationship with other well-trodden distinctions in philosophy of language remains vague. Foremost, this applies to attempts at classifying types of illocutions, like John Searle’s famous typology. In this section, I will argue that the proposed framework is best seen as orthogonal to Searle’s typology of illocutions and that reflexivity as a feature of speech acts can cut across all types of illocutions. To that end, I will introduce the notion of degrees of referent malleability and a third subset of the typology, constitutive performativity.

Building on Austin, Searle proposed a range of criteria by which various types of illocutions differ. The most important criterion is the purpose or “illocutionary point” (Searle Reference Searle1979, 2–4). The point of assertions, for instance, is to fit the world as it is; assertions are assessable along dimensions of truth. In contrast, the point of commissives is to change a social fact about the world; betting, for instance, commits one to a future action, a commitment that has not existed prior to one’s utterance. Similarly, declarations are intended to bring about a correspondence between the uttered proposition and a social fact; if someone successfully nominates you a candidate, then you are a candidate. Searle calls this element of the illocutionary point “direction of fit,” which can go in the direction of word-to-world, as in the case of assertions, or world-to-word, as in the case of commissives, or in both directions, as in the case of declarations.

Assertions that are reflexively performative sit rather awkwardly in this typology. While they might begin as plain assertions at the time of speaking, with their illocutionary point being to match the world, reflexive performativity is conditioned by the idea that there can be perlocutionary downstream effects that bear on the word-to-world fit, too. Consider a stylized example of a self-fulfilling bank run (Merton Reference Merton1948). An economist might assert that a bank is unstable, attempting to fit their utterance to the world. The perlocutionary effect of causing distress among the bank’s clients, however, makes this an assertion that does not simply match some state of affairs, but that functions as an intervention on the assertion’s referent; it causes the bank to be less stable. Similarly, in a different scenario, an economist might try to do the opposite and avert a bank run by asserting that, against their best knowledge, the bank is stable and that clients should not worry. Notably, the economist’s attempt to avert harm would work if clients perceived the economist’s utterance as a trustworthy factual assertion, the true purpose of which, however, is not just to match the world, or to lie, but to bring about its world-to-word match. Suddenly, then, reflexively performative assertions come to share important features with other types of illocutions like commissives or declarations, blurring their illocutionary status.

Speculatively, it is precisely this blurred family resemblance between reflexive assertions and other illocutions that made performativity talk initially attractive to STS scholars like Callon and MacKenzie. However, without clarifying its roots, an overreliance on mere family resemblance has arguably led to confusion among their readership (see section 3). Conceiving of the relationship between (a) reflexivity and (b) Searle’s typology as orthogonal, I argue, resolves much of it. Let me expand.

What Searle has in mind with his direction-of-fit scheme is a certain convention-enabled social malleability; performing a speech act like a commissive brings about a new or different social fact, and this is an affordance the speaker only has because of the specific conventional environment in which the speaker happens to be. The conventional environment matters, as, for instance, only if your interlocutors understand what a bet is can you meaningfully issue one by way of an utterance. Contrast this with the bank run case. While certain conventions need to be in place for the economist’s utterance to be recognized as an assertion, the fact that the economist’s assertion has the transformative powers that it has is not mediated by conventions in the same manner. Conceivably, the bank’s clients would have recognized the economist’s utterance as an assertion, believed it as a fact of the matter, and then they would have made their own decision based on the information they received. Crucially, this is not the same as if there had been a hypothetical convention in place that would have the bank be unstable if an economist declared it to be; akin, perhaps, to how a public servant could declare a bank as insolvent by way of an administrative act. In addition to having the illocutionary affordance of asserting something about the bank, in principle, this endows the economist with a sort of perlocutionary affordance to intervene on the state of the bank by way of an utterance. In contrast to Searle’s convention-enabled social malleability, an assertion that is reflexively performative hence points to a certain nonconventional malleability of the world, mediated by perlocutionary sequels.

Besides being nonconventional, such perlocutionary malleability also need not occur only with social referents. A growing literature on what one could call the perlocutionary effects of classifications focuses on the way in which biological systems, like species or ecosystems, are affected by human ways of conceptualizing them (Cooper Reference Cooper2004; Khalidi Reference Khalidi2010; Fagerberg Reference Fagerberg2022; Witteveen Reference Witteveenn.d.). Khalidi (Reference Khalidi2010, 346), for example, argued that human concepts of tame and non-tame wolves played an important part in the selective environment of wolves, affecting their evolutionary trajectory and eventually giving rise to an entirely different species: the domestic dog. Similarly, Fagerberg (Reference Fagerberg2022) argued that classifications of viruses affect their evolutionary trajectories, making virus kinds essentially dependent on human classifications.

Note, however, that both Searle’s convention-enabled-social-malleability and the other, neither-necessarily-social-nor-necessarily-convention-enabled malleability, share a common feature, which I will refer to as referent malleability. In both instances, what happens is that whatever the locution in question is about or can be said to refer to is also being affected by the performance of that speech act, one way or the other. In the case of declarations, such as “I resign!,” it brings about a particular social arrangement that had not existed prior, yet to which the locution simultaneously refers. In the case of an assertion like the economist’s “The bank is unstable!,” the bank’s status is being referred to, but it is also intervened on by the assertion’s dissemination and by way of a longer and indirect chain of perlocutionary sequels. In both cases, however, the performance of the utterance ultimately bears on its referent’s properties. In this manner, both exhibit a certain degree of referent malleability, understood as the degree to which the referent of a locution is being affected by its performance in a way that bears on the locution’s veracity, which does not require assuming any particular mechanism by which this bearing has to take place.Footnote ¹⁵

Because it comes in degrees, the concept of referent malleability enables mapping kinds of performatives along its axis. Recall the typology as it has been set up so far: Rephrased in terms of referent malleability, it would follow that reflexive performativity exhibits at least some referent malleability, whereas generic performativity exhibits none. Still, if one accepts generic performativity as the all-encompassing category that Austin collapsed it into, we should expect all Searle’s illocution types to surface somewhere in that framework, too. Put in terms of referent malleability, those speech acts that are entirely self-referential constitute a limiting case and could be said to feature the highest possible degrees of or complete referent malleability. For example, the locutionary referent of a declaration like “I resign!” is the performed act of resigning; uttering it makes it true because in uttering it felicitously one resigns. Contingent on the perlocutionary effects achieved, all Searle’s illocutions can, in principle, pop up across the full range of referent malleability. Importantly, that also applies to assertions, although self-referential assertions are admittedly peculiar, such as the liar’s paradox: “This statement is false.” In a nod to Mäki’s critique (see section 3.2), one might call locutions that exhibit this kind of complete referent malleability constitutive performativity—although note that, unlike in Mäki’s critique, this is decidedly not about the constitution of a particular type of illocutionary speech act (saying “this sentence is false” may still constitute a kind of assertion). What matters for referent malleability is that the act of uttering constitutes the utterance’s own locutionary referent. Although assertions perhaps have their paradigmatic place in generic performativity, and other illocutions have their paradigmatic place in constitutive performativity, between these extremes, reflexive performativity occupies an in-between class of partial referent malleability that, in principle, all illocutionary speech acts can inhabit, explicitly including acts of assertion. See the Figure 1 for a summary of the subsets.

Figure 1.

Classifying kinds of performativity along degrees of referent malleability.

Admittedly, as a stand-alone category, constitutive performativity is presumably not of much interest to philosophers of science. What matters, however, is to have a framework for conceiving of the different ways in which scholars have meant science to be performative—one that has a comprehensible relationship with its conceptual origins, and which enables other pressing debates to be had. The relationship on offer, then, is this: Searle’s types of illocutionary acts are mainly classified by the speaker’s intent, involving the illocutionary point. In contrast, the proposed typology cuts across a speech act’s perlocutionary effects or sequels, which may occur irrespective of the speaker’s intent. While the two can come together, they need not, rendering both typologies orthogonal.Footnote ¹⁶

4.4. What’s a referent?

The proposed definition of reflexive performativity asks three main questions. What aspect of the world is being changed due to the dissemination of scientific content? What is the referent of said content? And are they the same? Crucially, this makes the question of whether something is, in fact, an instance of reflexive, generic, or constitutive performativity conditional on what one takes the referent or target system to be.

This runs counter to a tendency in the literature to declare phenomena as reflexive irrespective of particular communicative acts, let alone their locutionary referents. Robert Northcott (Reference Northcott2022, 43), for instance, classifies “schizophrenia” and “domestic dogs” as reflexive kinds, but not “invasive species” or, curiously, “electric toothbrushes.” There is, however, nothing about invasive species that would, in principle, make them immune to the kind of coevolution that gave rise to domestic dogs or viruses (as argued by Khalidi and Fagerberg, see section 4.3), and there is nothing about toothbrushes that would make them immune to a kind of cultural evolution that occurs in response to perlocutionary effects of findings on, say, dental health. Most things change over time, one way or the other, but reflexivity is (or should be) about whether they change as a result of the dissemination of particular claims, models, theories, and so forth that refer to it.

The fact that the proposed typology demands a specification of the referent is, I contend, a merit. Among other things, it highlights that the question of what type of performativity one is dealing with can pose challenges in its own right, empirically and conceptually. Consider again the example from sociology of economics: The BSM model for option pricing has surely worked as a useful action-guiding piece of algebra for many traders, as evidenced by its widespread adoption. Yet, from this, it does not follow that it had a discernible real-life market as its referent to start. In short, what the BSM model does is, given personal risk preferences and estimates of volatility, to tell one how to value options under certain rationality assumptions. In other words, it gives traders tools to act consistently with a specified set of beliefs and assumptions. As a result, while the dissemination of this action-guiding algebra may have affected many trades and real-life markets, its relevant epistemic properties remain unchanged; given certain assumptions on rationality, it stands true that a certain price would be rational, whether you follow that advice or not. It could be argued then that the BSM model has not been reflexively performative after all, at most generically performative, precisely because it never was supposed to have a real-life referent, and at most serves as an abstract and counterfactual conditional.Footnote ¹⁷ Interestingly, in their original paper, Black and Scholes (Reference Black and Scholes1973) appear to jump in and out of this view. At times, they adopt the trader’s eye view as a guide to maximize profit (conditional view). At other times, they test back whether their model is consistent with actual market behavior, that is whether it is empirically adequate as a representation of actual markets (representational view).

For the typology, this is not much of a problem, as in principle, it should be possible for any particular model to be involved in both generic and reflexive performativity over the course of its lifetime, contingent on perlocutionary effects and on its particular referents. MacKenzie, for instance, has argued that, after its initial publication, it took years of mobilization and institutional changes for actual option prices to follow the model’s implications. Only as adoption among traders grew did it also become more adequate at representing actual markets. In this way, different uses of the same model can refer to different referents and trigger different perlocutionary effects, and thus also obtain different performative status.

4.5. The typology in action

The above is not merely an academic exercise. Specifying the supposed referent and being clear about what kind of performativity is at play is especially relevant for current debates on the ethics and politics of “performative” science. Consider the following.

Not distinguishing between generic and reflexive performativity, van Basshuysen et al. (Reference van Basshuysen, White, Khosrowi and Frisch2021) compare advice that epidemiologists give to the public during a pandemic to advice that doctors gave to patients. Both regularly project what would happen given some action, like if a patient changed their exercise regime, or if the public adhered to a social distancing policy. Reasonably, both the doctor and the epidemiologist may have difficulties predicting what a patient or the public will do upon receiving their advice. In short, they may be sure about “if p, then q,” but they need not be sure whether “p” upon giving their advice; perlocutionary effects can be hard to predict. This is then where van Basshuysen et al. (Reference van Basshuysen, White, Khosrowi and Frisch2021, 121) develop a strong claim: “Just as we should not think that a doctor is unqualified because…she cannot…forecast how much longer, exactly, a patient would live under a changed exercise regime,…it is unhelpful to assess the utility of epidemiological models based on their predictive abilities alone.” Instead of valuing only their accuracy, the argument goes, we could appraise the ability of an epidemiological model to change people’s behavior in a way that prevents harm.

This argument has been met with resistance (Winsberg and Harvard Reference Winsberg and Harvard2022; see Ortmann Reference Ortmann2025 for a discussion). Here, employing the preceding notion of referent malleability, I will just point out that the comparison to doctors appears misleading precisely because it conflates reflexive with generic performativity. Under normal circumstances, when doctors give advice, one expects them to utter true conditionals that, crucially, remain true (or false) irrespective of what the patient does. In the terminology of this article, the doctor’s example strikes as a case of generic performativity because their conditional speech act does not affect its locutionary referent in a way that changes its veracity—there is no referent malleability. The mechanisms that link exercise to health remain as they are after the doctor’s speech act, and they remain as they are irrespective of what the patient does upon hearing the advice. Against this background, taking performativity as a legitimate purpose, and taking it to be distinct from epistemic purposes, could be understood as suggesting that the doctor should outright lie to their patient: Falsely exaggerating negative effects of their lack of exercise might spur them into a healthier lifestyle. This, of course, seems unacceptable.Footnote ¹⁸

Conditionals that are reflexively performative are structurally different. In particular, what would need to happen for a conditional to be reflexively performative is that its dissemination bears on the circumstances that made the conditional true or false in the first place. That said, although they overlook this crucial difference in their simpler examples, van Basshuysen et al. (Reference van Basshuysen, White, Khosrowi and Frisch2021) do suggest a feedback loop that might have made this happen in the epidemiologist’s case: Policy compliance is a key parameter in pandemic models, but compliance is also dependent on public perceptions of the actual danger, which in turn are informed by disseminated model projections. This plausibly makes policy compliance a moving target, estimations of which also have the potential to influence it. This contrasts with the doctors’ example, where the referent in question consists of, say, biological mechanisms that link exercise to life expectancy, which remain unaffected by the doctor’s conditional assertion. While van Basshuysen et al. document cases of generic performativity and suggest a plausible case of reflexive performativity, they base a strong normative claim on the latter, but without much evidence for it. Partly, I suspect, this is because showing a change in the truth of the conditionals amounts to a much more demanding empirical exercise, and partly because they do not distinguish between generic and reflexive performativity.