“Gender Similarities and Differences” is a 2014 paper written by Janet Shibley Hyde. In it, Hyde consults meta-analyses to investigate whether the genders are more similar than they are different. Instead of discussing her findings, however, let’s consider her choice of title. There’s a chance it might strike the reader as unusual; we are so used to hearing about gender differences in our culture that hearing about gender similarities before differences (or indeed hearing the phrase ‘gender similarities’ at all), might pique one’s curiosity. Does the relative salience that Hyde has given to gender similarities over differences, in virtue of mentioning them first, have any significant consequences for how her readers are likely to think about gender? In this paper, I suggest that it likely does. More specifically, I aim to show that there is a distinctively subtle and yet impactful way in which salience patterns get introduced into our language through the order in which we communicate information.

I begin by introducing the concept of order-based patterns of salience in §2. This concept refers to the ordering of syntactic content, so that certain parts of it are made to stand out more than others. In §3 I suggest an instrumental way of evaluating order-based patterns of salience. Drawing on psychological research into the phenomenon of order effects, I suggest that certain order-based salience patterns are liable to activate epistemically and/or ethically problematic effects – such as false, inaccurate and/or harmful associations and beliefs – and as such can be criticised on the basis of their upshots. In §4, I offer different constitutive ways of evaluating these salience patterns. Certain order-based salience patterns might themselves constitute a subtle type of epistemic or ethical flaw. Finally, in §5 I consider the wider implications that this discussion has for how we ought to communicate.

There are a variety of mechanisms we can employ in language to make certain contents salient. When content is ‘salient’ it is more noticeable – more attention-grabbing – to the audience (see Watzl 2017; and Camp 2017, for a fuller discussion of attention and salience in the mind).1

Simply mentioning something can function to make it salient. For instance, Maxime Lepoutre (2021) focuses on cases in which ‘counterspeech,’ namely, speech that acts to counter falsehoods like those contained in conspiracy theories, can inadvertently make the falsehoods they critique salient. In particular, to debunk a conspiracy theory, counterspeech usually mentions that conspiracy theory, and in so doing, inadvertently makes that conspiracy theory salient, bringing attention to it. This can be problematic for various reasons. One is that by making the conspiracy theory more salient, it becomes more familiar to us, and, according to psychological research, familiar things are more likely to be interpreted as true. If the theory is false, then this is a problem.2

There are several other mechanisms that introduce salience patterns into language. One is to employ phrases that function explicitly to draw attention to some content. For instance, I might say ‘Sean Bean was in Lord of the Rings but remember: he was also in GoldenEye.’ By saying ‘remember,’ I flag to the audience that they should direct their attentional resources to the content coming after that word. Another mechanism involves repeating some content more than others; here, I might discuss Bean’s various film appearances, but keep repeating GoldenEye. Another mechanism still is tone of voice; I might communicate the information about Bean’s performance in GoldenEye using a louder volume, than when I discuss the other films in which he has appeared.

In contrast to these mechanisms, I want to focus on one distinctively subtle way of making certain contents salient in language; the order in which one communicates one’s utterance can make certain contents stand out as more salient than others. Some philosophers have considered syntactic ordering in relation to salience. Susanna Siegel, for instance, talks about how a news story is “more salient, if it is on the front page of a print or digital newspaper than if it is several clicks or pages away” (2022: 239).3 It is this order-based mechanism that I wish to home in on and analyse in this paper.

Consider a psychologist like Hyde, making decisions about how to present their findings about gender comparisons in relation to various psychological traits. They have findings demonstrating differences between females and males; perhaps there are statistically significant differences regarding throwing velocity, and sexual orientation. They also have findings demonstrating striking similarity; perhaps females and males are overwhelmingly similar when it comes to negotiation style, and extraversion. Once the content of what they want to communicate in their article is settled, however, there remains a further decision: with which facts should they begin their paper? They mock up two articles, the only difference between them being whether differences or similarities are mentioned at the top of the page.

These two articles invoke different order-based salience patterns. An order-based salience pattern refers to the ordering of syntactic content in cases where that order does not, pretheoretically, appear to be of consequence informationally-speaking. Two utterances that differ only in the order-based salience pattern applied to them seem (at least pretheroetically) informationally equivalent, such that they have the same truth value. For instance, ‘the scarf is orange and blue’ or ‘the scarf is blue and orange’ seem to be two ways of communicating identical information, namely, the colour of this scarf. I use the word ‘seem’, as I am not interested in strict informational equivalence, but rather its pretheoretical appearance. Perhaps changing the order in which one mentions the scarf’s colours does alter the truth value of the utterance in some subtle sense.4 Indeed, I will argue in §3 that such a change in order can alter the information communicated by an utterance when considered through a pragmatic lens, where one considers how a speaker’s wider communicative context shapes the utterance’s meaning. What is interesting about order-based salience patterns is that they change an utterance in a way that seems inconsequential—something on which I elaborate in §5. The two versions of this psychologist’s article described above, I suggest, fit this bill: the alternative versions seem informationally equivalent.

Despite this, one article makes differences more salient, while the other makes similarities more salient. In other words, order-based salience patterns structure contents by giving some relative salience over others. Salience, in this case, is comparative. But which contents does each article make more salient? To be able to answer this, we need to consult empirical research into how audiences typically process the orders in which utterances are communicated.5 Salience, remember, is about content standing out to an audience.

This research – conducted primarily on English-speaking audiences – indicates that audiences often find more salient content that is presented first, in a phenomenon known as the ‘primacy effect’ (Haugtvedt & Wegner 1994).6 There are some general psychological reasons for this, which I return to throughout this paper. In part, this is down to a mundane fact about mental processing as it occurs in time; content discussed first, in part due to the simple chronological fact that it is mentally processed first, regularly becomes ‘anchored’ in our minds, in the sense that it becomes more heavily relied upon than content presented afterwards. This is known as the ‘anchoring bias’ (Tversky & Kahneman 1974: 1128–1130). This preference for what is presented first is reinforced by the nature of our attention spans; these are such that we generally focus better at the beginning of hearing some information, with our attention waning as time goes on (thus being less likely to notice and retain information communicated later).7 Relatedly, as speakers, we are likely (often unconsciously) to present information that we find most important and relevant to the case at hand first (Kesebir 2017; Formanowicz & Hansen 2022). Evidence also suggests that, as audiences, we tend to infer that content presented first is more important and relevant—something that drives us to focus more acutely on that content (Kesebir 2017).

While general psychological factors, then, contribute to the tendency to find content communicated first most salient, these ‘primacy effects’ are especially likely in certain contexts. In particular, they occur most reliably for topics with which the audience is familiar, or finds interesting or controversial (Haugtvedt & Wegner 1994). These might include topics that are discussed regularly in the news, that come up in everyday conversations, and so on.

There are exceptions, however. In some cases, the content presented last is attended to better by the audience. This latter scenario, known as a ‘recency effect’, generally happens when the content under discussion is unfamiliar, or is of little interest to the audience; here, the content discussed last is typically that which is salient to the audience (Haugtvedt & Wegner 1994). If I am discussing the various attributes of the Helix Aspersa Maxima snail, then, presuming that my audience knows little about snails, and/or shows little interest in this species (no shade intended to snail lovers), the information that I discuss last is liable to stand out to that audience.

As will become clearer in what follows, I am interested in order-based salience patterns as they are applied to familiar topics, which audiences often find interesting and/or controversial—in particular, I am focussing on how we communicate information about gender (something around which we structure a lot of everyday life, a subject that is routinely discussed in the news and media, and which often generates interest and controversy). Therefore, I will assume that discussing content relating to this subject first is the typical way in which one makes it salient.

Another qualification is necessary. It should be noted that while we can use this general rule of thumb for order-based salience patterns—that, the content discussed first is made more salient (especially regarding content that is familiar and/or interesting to the audience)—this picture can be complicated by the presence of other mechanisms of salience, including those mentioned earlier. Perhaps certain salience mechanisms, like the addition of the phrase ‘but interestingly’ inserted before the second-discussed content usually overrides the order-based salience mechanism that makes the initial content more salient. To keep things simple, I focus on cases where these extra mechanisms of salience are not present.8

So, toying with the order in which one communicates information constitutes one mechanism for introducing patterns of salience into one’s utterance. The order-based salience mechanism is particularly interesting, I suggest, because it shows that one can shift the salience pattern in one’s utterance without needing to add new, or remove existing, syntactic content. Consider again Lepoutre’s discussion of salience, where simply mentioning a conspiracy theory makes it salient (2021). Lepoutre implies that, to avoid making a conspiracy theory salient when debunking it, one would need to avoid mention of (i.e. remove syntactic content regarding) that theory. Of course, this would be difficult to do; to debunk a conspiracy theory, one must mention it! Hence the seeming intractability of the problem that Lepoutre identifies, whereby debunking conspiracy theories seems necessarily to come with giving that conspiracy theory undue salience.

By contrast, order-based salience patterns can shift what is salient simply through reorganising the content communicated—without needing to subtract or add new syntactic content. One can communicate exactly the same words across two utterances, and yet alter what one makes salient just by changing how those words are ordered (in a way that, pretheoretically, seems inconsequential). This feature of order-based salience patterns makes this particular salience mechanism especially subtle and difficult to detect. The consequences of this will be addressed in §5.

Hyde, then, in the title of her paper, “Gender Similarities and Differences,” is adopting a particular order-based salience pattern, which makes gender similarities more salient than differences.

Do the order-based salience patterns we apply to language matter? Here, I expand upon the empirical evidence mentioned in the previous section, to analyse the impact that the primacy effect, generated by certain order-based salience patterns, can have. In particular, I draw on research regarding the ‘order effect.’ This shows how simply changing the order-based salience pattern in some content can lead to significantly different audience responses to that content, including the activation of substantive beliefs and associations in the audience—beliefs and associations that we might find problematic.

So far, I have suggested an instrumental critique of order-based salience patterns; this entails that the problems with a pattern are derivative of the problems with its upshots. Can order-based salience patterns instead be problematic in and of themselves? Here, I suggest that order-based patterns of salience can themselves constitute a subtle form of epistemic or moral flaw. To see this, let us take a short detour to consider a related proposal, made by Elizabeth Anderson.

In her 1995 paper “Knowledge, Human Interests, and Objectivity in Feminist Epistemology,” Anderson claims that a linguistic account of a phenomenon that contains only truths and yet omits other, relevant truths constitutes an epistemically biased account of that phenomenon. To illustrate this idea, Anderson cites a book by the controversial American Black Nationalist Louis Farrakhan. Farrakhan’s 1991 book The Secret Relationship between Blacks and Jews describes the role of Jews in the Atlantic slave system. Anderson lists many claims made by the book that are true. For example, “that Jews had considerable investments in the Dutch West India Company, which played a significant role in the seventeenth century Atlantic slave trade…[and] that a larger percentage of Jews living in the U.S. South owned slaves than did Southern whites as a whole” (Anderson 1995: 38).

The problem, Anderson claims, is not that Farrakhan’s book contains falsehoods; arguably, it does not. It is that it does not put the facts it discusses into a wider context that accurately represents the significance of those facts. Taken by themselves, these facts give the impression that Jewish people played a particularly significant role in the Atlantic slave system, more significant than other ethnic groups. The larger context, however, shows this impression (though not the facts themselves, purportedly) to be false. As Anderson notes, “The share of the Jewish investment in the Dutch West India Company was small, and the Dutch played a significant role in the Atlantic slave trade only in the seventeenth century, when the trade was small … [Further, a] greater proportion of U.S. Southern Jews owned slaves than other Southern whites only because they were concentrated in urban areas, where rates of slave ownership were higher” (Anderson 1995: 38).

Anderson concludes that ‘The Secret Relationship’ constitutes a biased representation of the Jewish role in the slave trade in virtue of omitting significant facts that are crucial to our understanding of that topic. In this way, Anderson adds to a large literature suggesting that we need to look beyond simply truth when judging the adequacy of our language (cf. Elgin 1996; Richard 2008; Gibbard 1990).21 In so doing, Anderson suggests a constitutive epistemic critique of ‘The Secret Relationship’ that does not rely on merely assessing its truth-conditions. One way in which we can directly assess the epistemic standing of a linguistic account of a subject, in other words, is by assessing whether it includes all pertinent and significant facts. A subtle way in which one’s account of a phenomenon can be biased, then, is by omitting relevant, significant facts about that phenomenon.22

What about order-based salience patterns? Making the ‘wrong’ content relatively more salient than the ‘right’ content might also be constitutive of a subtle bias. Imagine an amended version of Farrakhan’s book, which included all significant truths, but ordered them so that all the negative truths about the Jewish role in the slave trade were discussed first. The complaint here would be that, organising these (significant) facts in a way that makes salient the negative features of the Jewish role in the slave trade counts as a subtle way in which one’s account is biased.23 One subtle way in which a representation can be biased, in other words, consists in treating the ‘wrong’ facts as more salient than the ‘right’ facts. The order-based salience pattern employed by this hypothetical version of Farrakhan’s book might also count as a subtle way of being anti-Semitic. Where these states, of bias and anti-Semitism, are constitutively epistemically and/or ethically flawed, then the order-based salience pattern that counts as a version of these phenomena is also constitutively flawed.

How more specifically are we to understand this constitutive claim?24 One way is non-instrumentally. Imagine the sorts of order-based salience patterns described in this piece never finding an audience, and thus not having problematic effects (presume too they don’t encourage the speaker to develop or embolden their prejudices,25 or some other negative effect relating to the speaker’s later conduct). The patterns might structure syntactic contents in a private journal entry, an academic paper that never sees the light of day, a person’s utterances when speaking to themselves, and so on.26 To take one example, imagine someone who keeps a private diary in which they discuss their thoughts about a range of people. Whenever they mention people who are women, they describe their appearance before they describe their personalities, behaviours, careers, and so on. The opposite is true for when this individual writes about men. I think that these patterns are problematic in themselves, despite their lack of problematic upshots.

One way of justifying this is to suggest that the order-based salience pattern constitutes a failure of a moral duty, such as treating people fairly. This strategy has been employed to defend the constitutive evaluation of things like certain beliefs and utterances that lack negative effects. For instance, Rima Basu (2023: 5) suggests that certain of our beliefs, whether or not they generate negative effects, can wrong others by constituting a failure in our duty to adopt what P. F. Strawson calls the ‘participant stance’ with other people—referring to the need to treat others, in deed and thought, as persons with agency, instead of as mere things. Discussing the broader category of derogatory attitudes, Ralph DiFranco and Andrew Morgan (2023: 2) suggest that these “involve failing in one’s moral duty to be affectively open, that is, in the duty to avoid reacting to others with hostility.” I find these arguments plausible, despite their likely demanding implications. Instead of defending them here, though, I will focus on two other ways of developing a constitutive critique of order-based salience patterns that I believe will be less controversial.

Consider one response to the journal entry above. A critic might suggest that any epistemic or moral problem here would be derivative on a prior mental attitude influencing that order-based salience pattern, meaning that it’s the prior mental attitude where the problems reside; perhaps the salience pattern of talking about women’s appearances before more substantive traits is bad only in virtue of the objectifying outlook of the individual uttering it.27 Critically, I want to suggest that even if we concede the relevance of a prior mental state—insofar as the prior mental attitude is necessary for making a principled judgment between order-based salience patterns that are acceptable versus those that are not—that needn’t mean that the order-based salience pattern itself loses its normative quality.

Consider DiFranco and Morgan’s (2023) argument regarding slurs that are uttered privately and thereby generate no negative consequences. The slurring utterance will, they concede, likely be influenced by a derogatory attitude had by the slurrer. So, they ask, why not find the moral fault in that attitude instead of the slur? Their response is this: “if a particular derogatory attitude is morally defective, then it is wrong to express that attitude as well”. Here the utterance of a slur is considered an expressive manifestation of that attitude, endorsing and manifesting that attitude in a way that inherits its badness. Consider a person who (privately) says ‘I hate gay people,’ in a way that expresses (endorses) their homophobic attitude. Compare this with an individual who, at the end of a therapy session aimed at changing their objectionable attitudes, says the same thing. DiFranco and Morgan (2023: 7) suggest that this latter instance is simply “a neutral and well-informed evidential self-report on her mental life.” I take it that we have different moral evaluations of these utterances. We can make sense of those different evaluations by acknowledging that some utterances simply report an attitude of hatred, while others action-express them. Utterances that action-express hatred, for instance, endorse and manifest that hatred in a way that inherits the moral badness of that attitude.28

In a recent Ethics blogpost, Zachary Irving (2023) makes a related claim, this time about attentional patterns.29 Irving gives an example of someone who has prejudicial biases, like sexist, objectifying beliefs. In one scenario, this influences their attentional patterns so that they find women’s bodies more salient than their conversational contributions. In another scenario in which they still have those biases, the individual successfully exercises attentional control to suppress the usual attentional patterns above. The former is worse than the latter, Irving suggests—an intuition we can make sense of by adopting what he calls the ‘process view’ where the guiding state and the resulting salience pattern in attention are morally bad. In other words, instead of insisting that the mental state guiding the attentional patterns is the locus of evaluation, one might say it’s the whole process of the mental state guiding a salience-based structure in attention. We might extend this idea to order-based salience patterns in language; the locus of evaluation can be the whole process, whereby the mental state guides the order-based salience pattern applied to the individual’s utterance. This gives us a constitutive evaluation of order-based salience patterns in language, albeit restricted to cases where those patterns are influenced (in the right way) by a prior defective attitude.

There are other ways of developing a constitutive evaluation of these order-based salience patterns. A rich vein in the philosophy of language has used Austinian speech act theory to argue that certain utterances, beyond causing problematic effects, can themselves constitute actions (e.g. Langton 1993; Maitra & McGowan 2012). Saying “I pronounce you married” in a marriage ceremony does not just cause two people to become married; it constitutes the act of marrying. Saying ‘Run – there’s a fire!’, said by a person in a burning building, constitutes the act of warning. Sometimes, the act constituted by the utterance is harmful, or otherwise problematic. Rae Langton (1993: 302–303) gives the example of Apartheid law. She says that the utterance ‘Black people are not permitted to vote’ uttered in certain contexts, constitutes subordination; it constitutes the acts of (unfairly) ranking black people as inferior, (unjustly) depriving them of rights and powers, and legitimating discrimination against them. Where those acts are harmful, that makes the utterance, insofar as it is those acts, constitutively harmful. This constitutive route, unlike those described above, is not non-instrumental; it does not attempt entirely to screen off the effects of the utterances. Indeed, for the speech act to be successful, certain upshots are required, including hearer uptake. What it does do, however, is draw a norm-driven, as opposed to causal, connection between the utterance and its effects.30

If we extended this idea to order-based salience patterns, this would mean restricting the subset of order-based salience patterns considered constitutively problematic to those with problematic effects—specifically, those that do actually contribute to relevant problematic social activities (thus excluding the private utterances described above). Still, it provides a way of finding those order-based salience patterns constitutively problematic.

We have at least three separate ways of arriving at a constitutive critique for assessing order-based salience patterns, then. These involve locating the problem: solely in the salience pattern itself; in the combination of the salience pattern and a guiding prior attitude; and/or in the combination of the salience pattern and its effects. Different readers might be persuaded by different routes to a constitutive evaluation.

With these routes sketched, let’s return to the particularly common linguistic structuring of mentioning men before women. I considered an instrumental critique of this in §3, referring to the gender stereotypes this is liable to activate. The suggestion here is that, in addition to these potential problematic upshots, this common order-based salience pattern might itself count as a subtle form of androcentrism in one or all the ways described in this section. Perhaps, for instance, it constitutes the (speech) act of unfairly placing men at the centre of things.

Return too to Hyde. Perhaps the title that Hyde avoided, which made gender differences more salient than similarities, would itself constitute a subtle form of gender essentialism. It might, for example, be a way of expressing a problematic essentialist outlook.

As has been suggested in work on speech act theory, offering a way of arguing that certain speech implicates epistemic or ethical problems in a way that does not rely on establishing causal relationships can be very helpful. For instance, writing about the debate concerning pornographic speech and its purported effects on women’s subordination, McGowan recommends further exploring pornography’s constitutive harms (in the third sense considered above), saying “As is well known, it is notoriously difficult to establish the truth of…complex [causal] claims” (2005: 28). Indeed, it takes a significant amount of empirical work to establish the sorts of findings discussed in §3.1. Further, while I believe this would be to underestimate the robustness of the tendencies identified in that section, the importance of context means that it is always possible for the critic to say, ‘maybe features of this context mean that this order-based salience pattern doesn’t activate the sorts of problematic inferences you are concerned about.’ Indeed, it is implausible to think we can empirically test the effects of an order-based salience pattern across every conceivable communicative context (across different types of speaker, either pro- and preceded by different utterances, delivered with different tones of voice, etc.). If we can instead suggest that a given order-based salience pattern constitutes an epistemic or moral bias, then one can condemn the utterance in question without needing to prove those causal claims.

What is the fallout of this discussion? Consider first how these instrumental and constitutive critiques of order-based salience patterns might help to shed light on various existing complaints. To expand on an example mentioned earlier, it is common to hear frustrations raised about how routinely news articles, novels, obituaries, and so on, describe women by their appearance before their personalities and careers—an order-based salience pattern that is much less often applied to men (Buchanan & Crucchiola 2018). Relatedly, many from marginalised ethnic backgrounds complain about utterances that give undue salience to their demographic properties; for instance, Zoe Kravitz, referring to a famous quote by Jean-Michel Basquiat, in which he claims that he is not a black artist, but rather an artist, says this: “Happy to be black. Just don’t need to say it in front of everything” (Kravitz, in Willis 2017).31 The tools in this paper provide ways of criticising these order-based salience patterns on instrumental and constitutive grounds. We might find that the former activates objectificatory beliefs about women, for instance, while the latter constitutes a way of expressing disrespectful, othering attitudes.

What, practically speaking, are we to do in light of these findings? Note that salience is pervasive. All language must be structured along dimensions of salience. We cannot communicate without making some contents more salient than others, not least because we must inevitably say one sentence before another. If, as I have suggested, these order-based salience patterns have implications regarding whether we activate a false, inaccurate, and/or harmful belief, or whether they constitute some epistemically or ethically problematic state, then we should reflect on which patterns of salience we use, and which are being communicated to us. Do we tend to say ‘men and women’ when discussing gender in everyday conversations, in teaching scenarios, in the papers that we write? Do we tend to hear gender differences mentioned before similarities in the studies that we read?

Suggesting that we reflect on the order-based salience patterns that we use and hear is a modest proposal, but it has significance. We tend not to reflect on the salience patterns that we and others employ in language, in part because we are prone to treat them as insignificant. There is a sense in which philosophers, for instance, are aware of the importance of ordering their arguments so that they build nicely, and ultimately are more persuasive to the reader. But the more radical effects of order-based salience patterns discussed here—where a change in order can mean the difference between communicating something false or biased, and not—are rarely acknowledged.

Indeed, order effects—and the wider research into ‘framing effects’32—receive so much press precisely because our responses surprise us. In part, this is because not only are these patterns of salience themselves rarely consciously noticed by the audience, but any inferences that those patterns solicit us to make are also rarely consciously registered (Sher & McKenzie 2006; Holbrook et al. 2000). This has real significance. Research demonstrates that where inferences are activated under our conscious awareness, such as those to gender stereotypes as discussed in §3, the beliefs and associations that we have inferred are especially likely to go on to influence our thought and behaviour (Banaji et al. 1993). This is because, unaware of how a salience pattern is soliciting us to think and act, we do not attempt to block any problematic inferences that we might be making. The study participants in Kesebir’s and Bruckmüller and colleagues’ studies did not simply call to mind gender stereotypes. They then thought in ways congruent with those stereotypes, going on to attribute more masculine stereotypes to men, and more feminine stereotypes to women (as opposed to rejecting the stereotypes that had been called to mind). The under the radar nature that order-based salience patterns (and comparative question framings of the sort Bruckmüller et al. examined) have, then, means that any problems that they cause and constitute can more easily continue unchecked.33

This arguably gives order-based salience patterns an insidious power. The very thing that makes the impact of order-based salience patterns surprising is also what makes them powerful. This concern about the insidious strength of order-based salience patterns might motivate us to go further than the modest suggestion above. Beyond simply reflecting on which order-based salience patterns we employ and encounter, we might have reason to employ salience patterns that actively counter problematic ones. Perhaps we have a directed epistemic duty to our audience, such that we owe truth- or accuracy-conducive utterances to our audiences. This might mean that we owe to our audience order-based salience patterns that avoid the likely activation of false beliefs, and so on, as well as patterns that do not constitute a type of epistemic flaw. Perhaps more intuitively we have a duty to avoid wronging, disrespecting, or harming others. This might give us a reason to avoid order-based salience patterns that are liable to trigger wrongful, disrespectful or harmful beliefs (and so on) as well as patterns that constitute a moral wrong. Hyde has, of course, opted to reverse the standard order-based salience pattern, which makes gender differences more salient than similarities. We might see her as fulfilling epistemic and/or ethical duties in doing so, by disrupting the likely activation of cognitively accessibly false and harmful beliefs that the order ‘gender differences and similarities’ is liable to generate.

So, ought we adopt order-based salience patterns that counter problematic conventional ones? There are many things to discuss regarding this proposal. One thing to reiterate is that context matters. In some contexts, talking about gender differences before similarities (perhaps even talking only about gender differences), for instance, might not cause or constitute problems. Consider, for instance, the need to correct for androcentric biases in medicine, by emphasising how a medicine tested on male subjects is in fact much less effective on average in female bodies (Criado-Perez 2019); emphasising sex differences here seems epistemically and ethically important. In other contexts, even if we do think that this salience pattern is liable to activate or constitute gender essentialism, any concerns we might have in this regard may be outweighed by other (moral, political, epistemic, practical) considerations. A second point of contention for this latter proposal relates to epistemic paternalism. The proposal at hand recommends using patterns of salience to shape our audience’s inferential patterns for (what we deem is) the better, without consulting them on the issue. One would need to show that this is acceptable (see Jackson 2022: 134, 144, for a persuasive defence of epistemic paternalism in relation to order effects). These issues deserve further attention, which I do not have the space for here. They do, however, demonstrate the relevance of this discussion to the lively debates about nudge theory (Niker 2018; Noggle 2017)—a theory using insights from behavioural economics to suggest that simple, low-cost interventions can influence people to think and act better (Thaler & Sunstein 2008).

Suppose that an article communicates only truths about a subject, such as gender psychology. Further, suppose (rather idealistically) that it communicates all relevant truths about that subject. Might there nevertheless be grounds for criticising that article for giving relative salience to certain truths over others? In other words, while we cannot criticise the article for being untrue, or perhaps for being misleading in virtue of omitting certain truths, can we nevertheless criticise it for its order-based salience patterns? This paper has suggested that we can. We might evaluate an order-based salience pattern on instrumental grounds, for its liability to activate an epistemically and/or ethically problematic belief, ideology, and so on. Or we might evaluate it on constitutive grounds, finding the salience pattern itself constitutive of some epistemic and/or ethical flaw. In light of this, I suggest that we pay more attention to the order-based salience patterns that shape all communication, given their potential epistemic and ethical significance.