2.1. Transition From Semantic to Conceptual

Concepts are constituents of thoughts. They form thoughts compositionally, just like words form sentences compositionally. However, concepts are not the same as words. Many words express the same concept. For example, “lumi” and “snow” express the same concept SNOW. Moreover, the acquisition of concepts may not require language. A baby can acquire the concept DOG even though the baby cannot talk yet. As Herman Cappelen puts it, concepts are representational devices. Relatedly, concepts are often individuated on the basis of their semantic content:

For example, APPLE and ORANGE are two different concepts because the former is about apples and the latter is about oranges. This idea leads to conceptual discontinuity. Conceptual engineering aims to change the semantic content of concepts and, given semantic individuation, this inevitably leads to conceptual discontinuity. Semantic individuation therefore adheres to transition:

The aim of this article is to reject semantic individuation and to resist transition. The changes in the semantic content of a concept do not entail a change in the identity of a concept. Originalist individuation is offered instead.

Even though Cappelen adheres to semantic individuation, he discusses the connection between semantic individuation and transition in an illuminating way. He points out that Haslanger’s ameliorative project has changed the meaning of “woman” and then he asks whether this should be somehow accounted for in the terminology:

The terminological specification is called for because the change in meaning yields conceptual discontinuity:

The above shows a clear transition from a semantic change of “woman” (as a linguistic item) to a switch in the identity of the concept WOMAN. As constituents of thoughts, pre-amelioration WOMAN and post-amelioration WOMAN* yield different thoughts just because they are distinct concepts.

The passage also reveals a connection between conceptual discontinuity and another challenge known as “change of subject.” It might be phrased in the following way: If conceptual engineering is a meaning revisionary project, then the post-engineered meaning no longer answers the initial question posed at the beginning of the inquiry but, rather, it changes the subject (see Strawson, Reference Strawson and Schilpp1963, p. 506). However, there is a difference. The change of subject can be the result of semantic revision independently of conceptual change. As the passage reveals, it can happen merely by changing the meaning of words. However, whether that leads to conceptual discontinuity depends on the adopted theory concerning concept individuation.

2.2. Originalist Strategy

The general strategy to deal with discontinuity is to find something that stays constant even if there are semantic changes. Cappelen adheres to semantic individuation. As a consequence, he argues that while there can be conceptual changes because there are semantic changes, the broader topic is the constant factor. He develops this through the notion of samesaying (discussed more below). Functionalists like Haslanger, Prinzing, and Amie Thomasson argue that the constant factor is not just the function of a concept but because concepts are individuated by their functions, the concept itself is the constant factor. Even if there are semantic changes, the concept stays the same as long as the function stays the same (elaborated in Section 4). However, it is argued that despite the promising starting point, functionalism is not that different from semantic strategy. Yet, the strategy developed here owes a lot to functionalism. Namely, I agree with functionalism that we need a non-semantic individuation. This provides a very natural source for continuity, i.e., the concept itself.

It is true that we cannot engineer concepts directly. Rather, concepts are engineered through language. The concept WOMAN is engineered by changing the meaning of “woman.” The idea in the passage is that when Haslanger changes the meaning of “woman,” this operation also changes the semantic content of WOMAN. At the same time, there is a switch in the identity of the concept WOMAN, according to semantic individuation. This leads to different thoughts between the ameliorators and the non-ameliorators. The task set in this article is to block that. Specifically, the aim is to provide a view that blocks the transition from semantic changes to conceptual changes. It is argued that originalism provides continuity when other ways of individuation cannot. The originalist proposal is based on two things. First, according to originalism, concepts are not linguistic constructions. Sainsbury and Tye view them as abstract objects, like numbers, with concrete instantiations. Just as sets of three instantiate the number three, uses of concepts in thoughts instantiate concepts. Unlike numbers, however, concepts are not eternal. They are born. They can change over time, fall out of fashion, and even die. In this sense, they are different from Fregean, eternal concepts. According to Sainsbury and Tye, some abstract objects are eternal, some are non-eternal. Marriage is an example of a non-eternal abstract object. Marriages are abstract objects but they can come into existence and they can end. Similarly, concepts come into existence and they can cease to exist, according to Sainsbury and Tye (Reference Sainsbury and Tye2012, pp. 63–66). Importantly, concepts have representational properties. The baby has acquired DOG, because the baby has a representation of a dog. Still, while it is true that concepts are representational devices, as Cappelen puts it, concepts themselves are not representations, according to originalism. The baby’s representation of a dog is not a concept. The concept DOG is a vehicle of that representation (Sainsbury & Tye, Reference Sainsbury and Tye2011, p. 101, Reference Sainsbury and Tye2012, p. 40). Originalism holds that concepts are vehicles of semantic content (Sainsbury & Tye, Reference Sainsbury and Tye2012, pp. 72–79). This view is sometimes called “vehicularism”:

Second, vehicularism enables the central thesis of originalism:

It is argued that together the above two theses allow one to challenge transition.Footnote ¹ Concerning the semantic content of concepts, there has been a guiding thought that the so-called Frege data, that is, the differences between doxastic and epistemic reports containing MORNING STAR and EVENING STAR, have to be accommodated semantically. Even Saul Kripke pointed out that Frege data shows that the purely Millian picture cannot be the whole story (Kripke, Reference Kripke2001, p. 20). Sainsbury and Tye disagree. They are Millians and, as a consequence, their semantics is very austere. They think that Frege data can be met conceptually, not semantically. Even though MORNING STAR and EVENING STAR are semantically indistinguishable, the concepts are distinguishable because they have different origins and this explains the doxastic and epistemic differences between the concepts (Sainsbury & Tye, Reference Sainsbury and Tye2012, p. 45). As a consequence, their semantics is very austere in a Millian sense. The content of an atomic concept like KRIPKE is the referent of the concept. The semantic content of predicative concepts is also the referent, that is, the property to which the concept refers (Sainsbury & Tye, Reference Sainsbury and Tye2012, p. 45). It also needs to be pointed out that nothing said here hinges on the exact nature of the proposed semantics. The originalist framework introduced here hinges on the distinction between concepts and their content as suggested by vehicularism and on originalist individuation which has very little to do with semantics.Footnote ²

Before the actual investigation, one further point is worth noting. Within conceptual engineering, Manuel Gustavo Isaac distinguishes between a semantic and a psychological view of concepts. The semantic view conceives of concepts as semantic entities, as compositional constituents of thought. By contrast, the psychological view treats concepts as cognitive entities: structured bodies of information constructed and activated to support higher cognitive processes (Isaac, Reference Isaac2023, p. 2152). Although vehicularism shares a similarity with the psychological view insofar as it treats concepts as containers of information, the present approach remains aligned with the semantic view. The discontinuity challenge is framed in semantic terms: Discontinuity is tied to semantic change. Moreover, vehicularism itself is a semantic project, aimed at resolving semantic issues such as Frege’s puzzle (as seen above). Isaac also distinguishes between two positions in conceptual engineering: those that target concepts themselves and those that target linguistic meaning (Isaac, Reference Isaac2021, p. 290). Cappelen argues that conceptual engineering concerns linguistic meaning rather than concepts (Cappelen, Reference Cappelen2018, p. 61), a view endorsed by Mark Pinder, who treats talk of “concepts” as a façon de parler (Pinder, Reference Pinder2021, p. 143). I align with the first camp. On my view, concepts are distinct from meaning: They are vehicles of meaning, comprising both semantic and non-semantic information. One function of concepts is to enable the accumulation of information about their referents (as seen in Section 6.1). Accordingly, talk of “concepts” is not merely a façon de parler. While engineering targets the content of concepts, concepts themselves are clusters of information irreducible to linguistic meaning.

3.1. Belleri and Continuity in Inquiry

Even though Cappelen provides a very compelling illustration of how semantic individuation leads to conceptual discontinuity, he does not challenge semantic individuation. Rather, he aims to provide continuity by other means. Namely, his solution to discontinuity is the notion of topic. According to him, topics are coarser than semantic properties or conceptual identity. As a consequence, the more coarse-grained topic stays the same even if there are semantic and conceptual changes (see Cappelen, Reference Cappelen2018, pp. 107–121). His solution relies heavily on the notion of samesaying. According to him, two people can be samesayers even if there are semantic differences. Hence, even if there are semantic changes, samesaying ensures topical continuity (Cappelen, Reference Cappelen2018, p. 113).

Belleri criticizes the samesaying strategy. She points out that in many cases, in which there are semantic differences, the intuition is that the speakers are not samesayers. She invites us to consider the definition of torture by the United Nations: “any act inflicting severe suffering, physical or mental, in order to obtain information or to punish.” The UN definition differs significantly from the definition of the U.S. Department of Justice: “any act inflicting pain rising to the level of death, organ failure, or the permanent impairment of a significant body function” (Belleri, Reference Belleri2025a, p. 2946). The definition by the Department of Justice narrows the extension of torture significantly. Belleri also invites us to consider a situation in which someone was indeed tortured so that the victim suffered from an organ failure. In this case, the documents from the UN and from the Department of Justice both state that the prisoner was tortured and in this sense the reports say the same. However, what if the victim was severely tortured, but still miraculously made it without an organ failure or other permanent damages or near-death experience? Then the reports would contradict each other and they would not say the same. One could say that there are two senses of samesaying. In a broad sense, the UN and the Department of Justice are saying the same. However, Belleri aptly points out that often discussions on violations of human rights require greater precision. Belleri concludes: “The upshot of this is that samesaying judgements are unstable” as samesaying can be challenged from the narrow perspective (Belleri, Reference Belleri2025a, p. 2947).

Even though Belleri challenges samesaying, she does not challenge semantic individuation. Her own response to discontinuity is based on a rather intuitive thought: It cannot be that the slightest sharpening of a concept will inevitably lead to conceptual discontinuity. Cappelen also voices this idea (see Cappelen, Reference Cappelen2018, p. 105). In general, it is implausible that the tiniest semantic tweaks turn the concept into something else. Consequently, Belleri argues that semantic changes should not exceed certain limits, as too excessive changes produce discontinuity (Belleri, Reference Belleri2025a, p. 2944). Belleri’s overall strategy is that if semantic changes are controlled and small enough, the methodology secures continuity in inquiry. The problem in discontinuity is that concept C “has undergone excessive semantic change” (Belleri, Reference Belleri2025a, p. 2944). She argues that the inquiry into question Q must be structured so that, at the end, we obtain an answer to Q. This can be achieved with moderate semantic changes (Belleri, Reference Belleri2025a, pp. 2956–2958).

3.2. Inquiry and Concept Identity

Belleri aims to contain semantic changes so that they can be predicted and managed within a particular inquiry in a way that we are still answering the original worry. However, in my view, this is not enough to secure the identity of a concept in the long run and I think we should care about that. Conceptual discontinuity is about the identity of a concept, both within an inquiry and in the long run. David Lewis once pointed out that the lesson from Chisholm’s Paradox is that if you allow small changes in identity, you have to allow big changes. A chain of little differences adds up to big differences (Lewis, Reference Lewis1986, p. 243). According to Roderick Chisholm, it is plausible that Adam, who was 930 years old in our world, W₁, could have been 931 years old in W₂ while still retaining his identity. If it is plausible that, in W₂, Adam is 931 years old, then, surely, it is plausible that in W₃ he is 932 years old. To complicate things, Chisholm then introduces Noah, who was 950 years young in our W₁, and says that Noah could have been the age of 949 in W₂ while still retaining his identity. Little by little, we go on to change all properties of Adam and Noah and we end up with Adam who has all the properties of Noah in W _n and Noah who has all the properties of Adam in the same W _n (Chisholm, Reference Chisholm1967, pp. 1–3). There are only small symmetric changes between two worlds at a time, but there is a massive difference between Adam of W₁ and Adam of W _n who now has all the properties of Noah.

Returning to conceptual identity, the problem is semantic individuation coupled with the fact that small semantic changes are allowed in the first place. This combination does not work well with identity, which is transitive. Even if the constraints on semantic changes are sufficient to secure continuity in inquiry, the constraints are not enough to secure the identity of a concept (if we operate with semantic individuation). Even if we have small, controlled semantic changes in concept C₁ so that an inquiry I₁ still answers the initial question Q₁, it could be that when we reach I₇, the semantic changes are already excessive concerning the identity of the concept C₇. C₁ and C₇ cannot be the same concept, according to semantic individuation. The problem with the inquiry-based strategy is that, while inquiry can be well constrained, it is very difficult to constrain a string of inquiries. Just as Adam little by little turns into Noah, little by little moderate and controlled changes within an inquiry turn into immoderate changes in a string of inquiries. Specifically, there is a threat that inquiries become incommensurable. If inquiry I₁ aims to inquire questions concerning marriage between same-sex couples and I₇ aims to inquire whether marriage can involve more than two people, the results of these inquiries may not be added together because the inquiries involve different concepts of marriage. To overcome this would mean that in addition to constraints on inquiry I₁, we also had constraints on conceptual engineering as a whole, which would be detrimental for the future of engineering. To illustrate, what if conceptual engineering sets a limit on MARRIAGE so that heteronormative MARRIAGE is the absolute limit and the extension just cannot be stretched further? Conceptual engineering is distinguished from conceptual analysis because conceptual engineering is not so much interested in analyzing our current concepts as it is in exploring the concepts we ought to have (see Prinzing, Reference Prinzing2018, p. 854). But the threat of discontinuity prohibits us from exploring the conceptual possibilities too much.

As an upshot, Belleri’s view cannot be used to contain conceptual discontinuity. My diagnosis is that the aim of conceptual engineering, the effort to change the semantic content of a concept, is incompatible with the idea that concepts are individuated by their semantic content. Through transition, this inevitably leads to conceptual discontinuity. Hence, conceptual discontinuity cannot be addressed without addressing transition directly.

4.1. Functions and Discontinuity

Functionalism provides a non-semantic individuation of concepts. The functional view holds that semantic changes need not lead to conceptual changes. This blocks transition.

Functional individuation has gained popularity recently. Thomasson suggests that functions might provide a better way of individuating concepts than “precise intensions and extensions” (Thomasson, Reference Thomasson, Burgess, Cappelen and Plunkett2020, p. 443). Mona Simion and Christoph Kelp introduce what might be called an “innovative functional view.” Their view emphasizes the innovation of new concepts instead of repairing old ones. According to them, the function of a concept has a central role in innovation (Simion & Kelp, Reference Simion and Kelp2020). (Simion and Kelp’s view is revisited in Section 6.2 as I explain their positive influence on my originalist engineering.) Still, the most prominent functionalist is Haslanger. She introduces the idea that functions can provide conceptual continuity even if there is semantic discontinuity (Haslanger, Reference Haslanger2000, Reference Haslanger, Burgess, Cappelen and Plunkett2020). However, in the following, I will discuss Prinzing’s functionalism in detail. Interestingly, Prinzing and I share almost all of the underlying assumptions concerning discontinuity. Yet, we arrive at different conclusions concerning MARRIAGE. This pinpoints my disagreement with functional individuation.

4.2. Prinzing’s Functional View

According to Prinzing, concepts are functional kinds, “cognitive tools that we use when thinking about and interacting with the world.” Because they are tools, he argues, they have a purpose: Concepts are for something (Prinzing, Reference Prinzing2018, pp. 858–868, especially p. 858).

This view relates to Prinzing’s two basic claims. First, he adheres to non-semantic individuation. Prinzing is very explicit about the nature of the problem. He says that his main concern in discontinuity is “a question about the degree to which a given concept can change” (Prinzing, Reference Prinzing2018, p. 855). For Prinzing, the question is about the identity of concepts and then continuity in subject matter is preserved through conceptual identity (Prinzing, Reference Prinzing2018, pp. 857–861). Since Prinzing thinks that concepts are functional kinds, they are also individuated by their function, according to him. Prinzing explicitly denies semantic individuation: “Some philosophers have thought […] that conceptual identity requires extensional equivalence. I deny this” (Prinzing, Reference Prinzing2018, p. 871).Footnote ³ Instead of semantic individuation, Prinzing argues that concepts are individuated by their functions: “the identity criterion for concepts is functional equivalence” (Prinzing, Reference Prinzing2018, p. 867). This is how there can be conceptual continuity despite semantic differences. Second, Prinzing thinks that conceptual engineering needs an essentialist criterion for the identity of a concept. Prinzing sees the function of the concept as an essential feature of a concept: If the function changes, then a concept is no longer the same (Prinzing, Reference Prinzing2018, p. 862). These claims align well with my own thoughts on concepts and discontinuity. I also adhere to non-semantic individuation, albeit originalist individuation. I think concepts are individuated by their origin. Prinzing has also convinced me regarding essentialism. However, instead of seeing function as an essential feature of a concept, I think that essentialism relates to the origin of a concept (as explained in Section 5.3).

Despite these similarities, there is still an important difference. MARRIAGE shows this vividly. MARRIAGE seems to have more than one function, and if the function provides the criterion for identity, this is problematic. Prinzing discusses the idea of multifunctional concepts. He thinks that since MARRIAGE is a heavily contested concept, it is a good candidate for a multifunctional concept. The functions of MARRIAGE, in the ordinary use of the word “marriage,” are among other things: (1) keep track of and endorse the legal aspects of marriage; (2) uphold a religious sacrament involving relationships; and (3) keep track of a special kind of exclusive romantic relationship (Prinzing, Reference Prinzing2018, p. 873). Just because MARRIAGE has multiple functions, Prinzing is inclined to think that there is more than one concept at play. Moreover, it does seem that some of the functions are conflicting. MARRIAGE cannot keep track of a sacred relationship between a man and a woman while simultaneously keeping track of legal aspects of MARRIAGE from the point of view of equality. Taking inspiration from Kevin Scharp’s thoughts on conflicting concepts, Prinzing concludes that if the functions of multifunctional concepts cannot be reconciled, then they must be substituted or eliminated “but they cannot be revised because revisions preserve all the functions of the pre-engineering concept” (Prinzing, Reference Prinzing2018, p. 874). This means that the single concept MARRIAGE has to be substituted with heteronormative MARRIAGE and equal MARRIAGE, as persistent disagreement over the function of a concept is evidence that the disagreeing parties are using different concepts, even though it appears that there is only one (Prinzing, Reference Prinzing2018, pp. 874–875). It seems that contestation reflects not only semantic differences, but also functional differences. Hence, functional individuation yields the same result as semantic individuation.Footnote ⁴ Next, I offer an alternative theory of concept individuation. Crucially, originalism allows for conceptual continuity despite semantic or functional differences.

5.1. Vehicularism and the Sameness of Thoughts

To start the exposition of originalism, Prinzing notes the following concern. If semantic revisions are allowed, then one is committed to the view that in the old days it was true that marital rape was impossible. Nevertheless, nowadays it is true that marital rape is possible. He goes on to point out that this is pretty strange and then admits that evaluating the truth of old claims made with old versions of concepts can be somewhat tricky business (Prinzing, Reference Prinzing2018, p. 873). My originalism-inspired response to this worry is that the use of the same concept does not necessarily lead to the same truth value; using the same content in the same way does. The content of MEAT has changed significantly. It has narrowed from all edible stuff to animal flesh. I do think that the thought “Onions are meat” is the same in the 12^th century and nowadays but the truth value has changed. More specifically, originalism makes a sharp distinction between a concept and its content and this is the first step in solving conceptual discontinuity. Sainsbury and Tye point out: “Concepts are constituents of thoughts, vehicles of representation, tools used in thinking. […] They can be combined into larger structures, including thoughts […]” (Sainsbury & Tye, Reference Sainsbury and Tye2012, p. 40). This is a statement of vehicularism.

In one sense, the functionalists are right. Concepts have a function. They store and organize information. Concepts function as vehicles of representational information. They store and organize information about the world. However, if originalism is right, then the problem for functionalism is that all concepts share this function. So it cannot be used to individuate concepts. At the same time, thoughts are structured from constituent concepts. Thoughts can be true or false but the structures of concepts are not true or false by themselves. Content determines whether thoughts are true or false. Thoughts and their semantics are two different things. My own originalist-inspired view is that while the content of a concept plays a part in compositional structure, it does not determine the identity of thoughts. The compositional contribution can be stated in terms of type theory. Just because “Sainsbury” refers to a singular object, it can only feature as an input for a predicate of type <e,t>. In this sense, the content limits the compositional input of a concept. However, the target of a compositional structure is a truth value. That is, the compositional structure is the semantic content that then determines the truth value. Nevertheless, content does not determine identity. The identity of a thought traces back to the identity of its constituents and to their origin. Therefore, “Hesperus is Hesperus” and “Hesperus is Phosphorus” are two different thoughts with the same content. This view accommodates the idea that the thought “Onions are meat” is the same thought in the 12^th century and nowadays. The constituents of the thought are the same concepts but the semantic content of some of the concepts has changed, leading to the different truth values.Footnote ⁵ The originalist view of the sameness of thoughts is in contrast with Cappelen’s view. It is clear that people in the 12^th century and people nowadays cannot be samesayers. Samesaying seems to be a thesis, not only about our thoughts and their constituents, but also about the content of our thoughts. Consider an utterance: “Barack Obama is the current U.S. president.” It was true in 2015 but it is not true, if uttered now. In this case, the content is the same but the world has changed. The same semantic compositional structure yields different truth values in different settings. In our case about onions, the world is the same, but there are differences in the content of the constituents of the thought. Because of these differences, people in the old days and people nowadays assign different truth values to their respective thoughts. The truth value of a thought is determined compositionally by the content of the constituents and, given semantic individuation, the different truth values reflect the fact that the thoughts are composed of different concepts. Hence, the thoughts cannot be the same. With the distinction between concepts and their content, originalism can resist this conclusion. I believe this originalist view gives the engineers what they want, as conceptual engineering is not just about changing the content of concepts, but also about influencing thoughts and beliefs composed of concepts. It can be argued that MARRIAGE is engineered so that the thought “Legally speaking, there are no obstacles for Jack and John to get married” is true, even though in the old days the same thought might have been false.

5.2. Originalist Individuation: Blocking Transition

The distinction between concepts and their content is the first step. The second is the actual individuation and this is what helps to avoid transition. Originalist individuation blocks the transition from semantic changes to changes in the identity of concepts. According to originalism, semantic content has very little to do with the identity of concepts: “concepts are to be individuated by their historical origins, as opposed to their semantic […] properties” (Sainsbury & Tye, Reference Sainsbury and Tye2012, p. 40). There are plenty of examples in which the semantic content has undergone major changes but, plausibly, the concept is still the same. As already mentioned, Sainsbury and Tye report that the extension of MEAT has gradually narrowed from anything edible, which was its content in the 12^th century, to flesh. Everyday concepts like MANSION and SALAD have also gone through extensional changes. Not too long ago, probably less than 100 years ago, a two-storey, five-bedroom house might have qualified as a mansion. But nowadays it does not qualify as a mansion due to the general upsizing of houses. As Cian Dorr and John Hawthorne point out, in the old days SALAD referred to dishes in which green leaves were the main ingredient but nowadays the use of SALAD is much more liberal. It also applies to dishes that have very little to do with green leaves (Dorr & Hawthorne, Reference Dorr and Hawthorne2014, pp. 284–285; see also Prinzing, Reference Prinzing2018, p. 860; Simion & Kelp, Reference Simion and Kelp2020, pp. 996–997). Crucially, even though the extension of MANSION has narrowed and the extension of SALAD has expanded, it does seem that neither MANSION nor SALAD has changed to another concept. More likely, their content has changed, not the concepts themselves.

It is quite common that the semantic content of concepts changes but it does not necessarily entail that the identity of a concept changes. Subsequently, Sainsbury and Tye propose their non-semantic individuation for concepts. According to them, concepts are individuated by their historical origin, not by their semantic content. The semantic content of a concept can change but the concept remains the same as long as its origin remains the same. To set the current view in its proper contextual setting, I will discuss two aspects of originalism. I will first discuss the essentialist aspect of originalism. In Section 6, I go on to discuss the origin of a concept. This also specifies the role of semantic content within originalism. Finally, I argue that originalism leads to a freedom of choice in conceptual engineering. Semantic or functional changes do not inevitably lead to conceptual discontinuity but, with originalist engineering, the engineer retains the freedom of choice whether to stay revisionary or adopt a substitutional approach.

5.3. Essentialism: Fissions and Fusions

Just as Prinzing recognizes the possibility of multifunctional concepts, Sainsbury and Tye discuss the possibility of concepts with multiple origins. However, my diagnosis is that the existence of multifunctional concepts is highly problematic for functional individuation, whereas the problem of multiple origins is not a devastating problem for originalism. The problem is accommodated with the notions of fission and fusion. Specifically, originalism can accommodate the clash between essentialism and the fact that some concepts appear to have more than one origin.

According to originalism, the origin is the most important feature of the identity of a concept. Specifically, it is essential that a concept has only one origin. In the same vein, it cannot be that two concepts have the same origin. However, there seem to be obvious counterexamples to Sainsbury and Tye’s essentialism concerning individuation. They point out that if Isaac Newton introduced MASS and this introduction later led to the development of RELATIVISTIC MASS and INERTIAL MASS, then it would seem that this is a case of two concepts with a single origin. This is a problem because it suggests that there is more than one concept with the same identity. Sainsbury and Tye handle this with the notion of fission. According to originalism, there are three concepts at play with three originating uses. First, we have the undifferentiated MASS that was introduced by Newton. We also have two new concepts with their own originating uses as a result of a fission.

An example of a single concept with seemingly multiple origins is QUARK. Again, this is a problem because QUARK suggests that a single concept has more than one identity. Sainsbury and Tye refer to an apparently standard story about how QUARK came about. It has been suggested that Murray Gell-Mann and George Zweig introduced the concept QUARK independently of each other. Thereby, it appears that QUARK has two origins. The originalist account of this is again that there are actually three distinct concepts at play: Gell-Mann’s QUARK, Zweig’s ACES, and the fusion of these two concepts, which is distinct from the two. Crucially, the origin of the third concept is the fusion of the two previous ones, both of which ceased to exist at the moment of the fusion (Sainsbury & Tye, Reference Sainsbury and Tye2012, pp. 66–68.) With the notions of fission and fusion, Sainsbury and Tye are able to maintain the thesis that every concept has only one origin and that no concept can have more than one origin. For originalism, there are many features that are contingent concerning the identity of a concept. One such feature is the semantic content but the origin is an essential feature. The concept cannot survive as the same concept if the origin changes.

It is plausible that the constituents of a conceptual fusion can persist as distinct concepts. Consider CAUSAL REFERENCE. According to the well-known story, Hilary Putnam and Kripke developed the idea independently. Originalism would suggest that their concepts were later fused into one. This general CAUSAL REFERENCE then subsumed all of the information of the constituent concepts. However, it’s likely that important differences remain — for example, Putnam might have insisted that his version differs from Kripke’s due to his emphasis on stereotypes. In that case, we might identify three concepts: general CAUSAL REFERENCE (the result of the fusion), Putnam’s CAUSAL REFERENCE, and Kripke’s CAUSAL REFERENCE, which all have their own origins. The general concept arises from the fusion, while the other two persist as distinct because the creator of Putnam’s CAUSAL REFERENCE insisted so. In what follows, we examine the role of the creator more closely.

6.1. Concept Creation and Thinker’s Intentions

At this point, one might doubt that semantic content does not influence individuation. Moreover, discontinuity is not always bad. Consider PHLOGISTON and OXYGEN: Here, discontinuity was a welcome advance. In the case of MEAT, one might think that the concept has changed because its content has changed too much, suggesting that originalism yields the wrong result by preserving continuity where discontinuity is needed.Footnote ⁶ Admittedly, intuitions about MEAT vary, making it hard to decide between continuity and discontinuity. I will argue, however, that originalism does not give the wrong answer but instead allows freedom of choice. Semantic or functional considerations are not irrelevant; they inform the decision to revise a concept or replace it. Within originalism, however, such engineering is a matter of choice rather than a necessity imposed by semantic or functional individuation.

This section explores originalist engineering through three related themes. First, it outlines the originalist framework, arguing that engineers’ intentions are central to conceptual continuity. Second, it examines externalism, showing that conceptual discontinuity is linked to the marketing challenge — how to persuade others to adopt an engineered concept. Finally, it applies these ideas to the philosophy of race, illustrating how originalism handles continuity case by case. Rather than relying on fixed semantic or functional criteria for concept identity, originalism prioritizes engineers’ intentions, informed by semantic and functional considerations. These considerations support continuity not through individuation, but by helping engineers meet the marketing challenge.

To explore the role of semantic content in conceptual continuity, it is best to start from the beginning and examine the role intentions have in concept creation and in the subsequent deference use. According to originalism, semantic content has a pivotal role in concept creation. Sainsbury and Tye first give, as a crude illustration, a view of a double-duty of origin. The referent of a concept is simply fixed at the origin. From there on, it is preserved by the same mechanism that preserves the identity of the concept. This double-duty view subscribes to semantic individuation. The mechanism for preserving the semantic content also preserves the identity of a concept. Needless to say, Sainsbury and Tye deny semantic individuation. So they challenge the latter part of the double-duty, the part about the content preservation. As it is well established now, originalism argues that the referent of a concept can shift while the identity of the concept stays the same but Sainsbury and Tye also emphasize intentions, not only in concept creation, but also in the deference use of concepts. It is fairly obvious that, in concept creation, the intentions of the thinker do have a role. Concepts are created intentionally. This is analogous to baptism. Things are named intentionally or, at the very least, intentions often have a role in naming things (see Sainsbury, Reference Sainsbury2005, pp. 106–113). A baptism of an object starts the practice of using that word for that object. Sainsbury and Tye argue that there is a parallel story concerning concepts. The baptism also introduces a new concept and, once a concept has been created, intentions play an important role in the deference use of that concept.

As Sainsbury and Tye point out, most public concepts have the content they have before we as thinkers acquire them.Footnote ⁷ We generally aim to use them as they are used in the community of thinkers. As Max Kölbel points out, the possession and employment of many concepts requires, not only that a thinker has the ability to form thoughts and beliefs, but also that the thinker stands in a social relation of deference (Kölbel, Reference Kölbel2023, p. 29). Sainsbury and Tye call this kind of use “non-originating use.” Non-originating use has two constraints: (1) it is based on deference; and (2) it involves accumulation of information. These constraints apply primarily to the semantic content of concepts. A concept user just introduced to the use of concept C aims to preserve the content that concept C has in the thoughts of old users. Old users also aim to preserve the content of a concept in a way that accommodates their own previous use and the use of others. Nevertheless, constraint (2) can also involve new information about semantic content. In fact, the accumulation of information is very important for vehicularism.Footnote ⁸ One might say that that is what the “files” enable. Not only are they containers of (semantic) information, but they also provide a platform for acquiring new information about the referent. (The two constraints are revisited in Section 6.3 in relation to racial antirealism.)

It is very tempting to say that if the intentions to defer to existing use or the intentions to accumulate information are not satisfied, “some kind of significant failure has occurred” (Sainsbury & Tye, Reference Sainsbury and Tye2012, p. 70). While acknowledging the role of intentions in deference, originalism still opposes the conditional. MANSION and SALAD are counterexamples to the conditional. It is very unlikely that a significant mistake has happened in these cases even though the intentions to defer are not satisfied. The point can be illustrated with the case of MADAGASCAR. Sainsbury and Tye give two versions of the story. In both stories the facts are the same but the intentions differ. First, we have a deference story:

The second story is about concept creation:

In the first story, the map-makers aimed to defer to Polo’s use but the intentions to do so were not satisfied. In the second story, the map-makers knowingly diverted from Polo’s use of the term and perhaps unwittingly came to create a new concept. The creation is pinpointed to the utterance: “let’s just name the island Madagascar.” The MADAGASCAR they created had a new origin and the content was fixed at the origin, as Sainsbury and Tye insist. It is more likely that the first story is a more accurate depiction of events and this might be why philosophers of language often speak of a mistake in the case of Madagascar, perhaps because the switch in the referent is so dramatic. If pressed, philosophers might say that it was a mistake because of the failure to satisfy the intentions. In contrast, I think this thought is erroneous. Granted that the switch is more dramatic than in MANSION or SALAD but that is all. It is more dramatic than a smooth and gradual shift but still no significant failure occurred. The same result of the stories reinforces this thought.

This turns out to be good news for conceptual engineering. Namely, conceptual engineers are in the driver’s seat. It is up to them to defer to the existing use, even if there is a significant divergence from the standard deference practice. It is also up to them to come up with a new concept if they so wish. There might be various reasons that engineers want to substitute the concept. The decision might be influenced by semantic or functional factors, but within the originalist framework, these become pragmatic or, perhaps more accurately, instrumental considerations, as the goal is to achieve amelioration through engineered concepts.

6.2. Originalist Engineering and the Marketing Challenge

Originalism brings good news for conceptual engineers. It liberates conceptual engineering from the grip of forced conceptual changes, giving the control to engineers. In the following, I discuss liberated engineering in relation to two seemingly different issues — the distinction between revisionary and substitutional engineering and the issue of externalism. In the context of originalist engineering, these issues are related. According to originalist engineering, conceptual discontinuity is no longer forced by semantics or by functions. If the semantics of a concept changes enough, then the concept changes to another, according to semantic individuation. According to functionalism, if a function changes, then the identity of the concept also changes. These changes happen inevitably and regardless of the engineer. With originalism, the engineer remains in control. The distinction between revisionary and substitutional engineering remains a choice. Moreover, the choice is related to externalism. This section discusses how externalism is related to the question how the engineer can persuade other people to adopt the engineered concept. In my view, this is at the heart of semantic externalism.

Semantic externalism states that the meaning of words and the content of concepts depend on external factors, such as the external world, environmental influences, chains of reference, and the linguistic community. Externalism, rooted in Putnam’s famous phrase that “meaning ain’t in the head,” emphasizes that meaning is not solely confined to individual cognition but is tied to communal linguistic understanding (Putnam, Reference Putnam1979, pp. 215–271, especially p. 227; see also Haukioja, Reference Haukioja and Raatikainen2023). Despite the challenge imposed by the communal nature of meaning, conceptual engineering aims to reshape the semantic content of concepts. As an originalist, I think there are two separate issues at play. The first one concerns how I can change the semantic content of APPLE. The second question is how I sell that change to the rest of the linguistic community. In principal, the first part is not that difficult. I can change the content of my concept APPLE just like that. Sainsbury and Tye call this “thinkers’-reference” (Sainsbury & Tye, Reference Sainsbury and Tye2012, pp. 71–72). While changing the thinkers’-reference is easy in principal, there are practical constraints on changing the content. If the goal is scientific precision, it brings one set of constraints; if the goal is societal amelioration, it brings another set of constraints. In any case, the engineering of the content already involves the normative dimension of conceptual engineering. It requires analyzing the current concept MARRIAGE and asking whether it meets the goal of amelioration. If not, the engineer must consider what version of MARRIAGE would better serve society today. Nevertheless, if this is not difficult enough, the second part is even more difficult. Given the linguistic tendency to defer to the other thinker’s use and to one’s own previous use, it is difficult to persuade fellow thinkers to switch to engineered content especially if the change is dramatic. Pinder offers a similar view. He uses the distinction between speaker-meaning and semantic-meaning.Footnote ⁹ According to him, success in shifting speaker-meaning is fairly easily accomplished. To convince others to follow the engineered speaker-meaning is the difficult part. Pinder also argues that the success in speaker-meaning oriented engineering is no small feat: “[I]n most cases, it gives conceptual engineers what they want. Carefully articulated speaker-meanings can convey new information and new ideas” (Pinder, Reference Pinder2021, p. 156). If the engineered speaker-meaning is well constrained with respect to the set goals, it can yield new ideas about how to carve up reality. While I think that this is a helpful idea, I also think that conceptual engineering cannot function as an ameliorative project unless there is a wider deference to the engineered meaning. This leads to what might be called a “marketing challenge”:

In the face of the marketing challenge, conceptual discontinuity can be an asset. I argue that the choice between staying revisionary or going substitutional is a strategic one, aimed at persuading others to adopt the engineered content, and originalist engineering offers key tools to evaluate these strategies. While revised concepts are often easier to market, framing a concept as entirely new can sometimes be more effective. I owe this line of thought to Simion and Kelp. They emphasize innovation in conceptual engineering. For them, conceptual engineers are designers: They repurpose old concepts or create new ones with new functions, which are then “launched on a competitive market of concepts” (Simion & Kelp, Reference Simion and Kelp2020, p. 989). In such a market, every detail matters. I’m comfortable with this marketing perspective, as the ultimate aim of conceptual engineering is to create more egalitarian, less discriminatory concepts.Footnote ¹⁰ There are many variables that affect conceptual engineering. As we will see in the next section, conceptual engineering of RACE involves many issues, most importantly social justice and equality. It also raises an immediate normative question: How can the concept RACE help achieve these goals? These form the main components of conceptual engineering. The marketing challenge introduces an additional set of parameters, and here, originalism can be beneficial.

While originalism always adheres to the originalist individuation of concepts, the originalist concept creation can utilize the insights of other ways of individuating. Namely, it can take advantage of the motivations behind semantic and functional individuation. An originalist can, for example, use Belleri’s idea of semantically constrained inquiry as a guideline in conceptual engineering. Likewise, functional sameness or differences can also provide valuable insight into continuity considerations. The freedom that comes with originalism is tied to the fact that originalist engineers can use both semantic and functional considerations in their assessments of whether to stay revisionary or go substitutional — and, to repeat, this choice is helpful concerning the marketing challenge. The next brief case study aims to illustrate the advantages. Namely, it turns out that originalism yields revisionary racial antirealism without the problems that Haslanger’s revisionary social constructionism faces.

6.3. Originalism and (Provisionally) Revisionary Racial Antirealism

In this section, the ideas presented above are put to use. Namely, the discussion of racial antirealism shows that while semantic and functional individuation face significant pressure to go substitutional, originalist engineering still provides the choice between revisionist and substitutional racial antirealism.

The debate between social constructionism and racial antirealism can be framed as one between revisionary and substitutional conceptual engineering. Social constructionists seek to revise RACE from a biological to a social concept, while antirealists advocate eliminating RACE entirely, replacing it with RACIALIZATION. For antirealists, the biological concept RACE is empty and rooted in the justification of oppression — so why preserve or rehabilitate it? Instead, they argue we should focus on the real phenomenon: racialization. At the same time, social constructionists like Haslanger agree that RACE lacks biological grounding, but argue that, redefined socially, it retains explanatory value.

Given the marketing challenge, it seems that social constructionism has the upper hand. Even though there are significant changes to the content of RACE, the function is still the same. The function of RACE is to provide a description of the ideology that carves up social reality into a hierarchical structure. The concept RACE serves the purpose and “[t]he need to identify and explain persistent inequalities […] between people of different ‘colors’” (Haslanger, Reference Haslanger2000, p. 36). According to Haslanger, the function of explaining the hierarchical structure of social reality is the same despite the changes in content. Hence, the revisionary nature of the proposal provides a solution to the marketing challenge. In the context of philosophy of race, the marketing challenge is similar but not identical to the mismatch worry. Picking up the term from Ron Mallon, Joshua Glasgow raises the worry against contemporary non-essentialist biological realism about race (Glasgow, Reference Glasgow, Glasgow, Haslanger, Jeffers and Spencer2019, pp. 118–120; Mallon, Reference Mallon2006, pp. 529–539). His idea is that contemporary biological realism does not match the folk idea of race. The actual genetic population branches, to which the racial classification is supposed to refer, do not match with the ordinary racial classification. Glasgow expands the mismatch argument to social constructionism. There is a mismatch between the ordinary concept RACE and RACE as a social concept, as folks generally think that race is biological, not social (see Díaz-León, Reference Díaz-León, Marques and Wikforss2020, p. 190). It is important to distinguish the marketing issue from the mismatch worry. The mismatch concerns only the content of RACE but the marketing challenge accepts differences in content. The mismatch is picked up again below, but now it should be made clear that Haslanger offers her a social constructivist view to tackle the marketing issue by staying revisionary. According to the view, the identity of the concept RACE is the same because the function of a concept is the same.

One might question the consistency of Haslanger’s functionalism, as it appears multiple functions of RACE are in play. A key idea in racial antirealism is that RACE has historically served as a tool of oppression and discrimination (Glasgow, Reference Glasgow, Glasgow, Haslanger, Jeffers and Spencer2019, pp. 111–118), a view Haslanger shares. Yet she also sees RACE as explanatorily useful: It helps us understand social structures precisely because it organizes social structures hierarchically. Jared Riggs highlights this tension, noting that while RACE may have originally functioned to legitimize pseudo-scientific ideologies, it now serves different purposes, such as identifying and resisting racial injustice (Riggs, Reference Riggs2021, p. 11573). For Riggs, this shift illustrates a broader problem: the “messiness” of functions. Functions, he argues, are being asked to do too much (Riggs, Reference Riggs2021, pp. 11576–11579). This point is revisited below but now it suffices to say that, in my opinion, one of the symptoms of the messiness is that it cannot provide an alternative to semantic individuation, as suggested in Section 4.2. Ultimately, Haslanger’s functionalism seeks to resolve the meaning of RACE through functional continuity, but the dual roles, both oppressive and ameliorative, render the account unstable.

Concerning racial antirealism, the good news is that it does not suffer from the above-described messiness. Although antirealism recognizes the two aspects of racial discussions, the workload is divided between two concepts: RACE and RACIALIZATION. The original purpose of RACE is indeed the legitimation or normalization of discrimination but the explanatory work is done by RACIALIZATION.Footnote ¹¹ As social constructionists point out, RACIALIZATION indeed is a social notion but while social constructionists believe the social RACE can be built on RACIALIZATION, antirealists disagree. The two concepts should be kept separate. To remind, non-originating use of a concept requires two conditions: an intention to defer to others’ usage, and a continued flow of information. Both are absent in the antirealist use of RACIALIZATION. The original function of RACE was to legitimize discrimination under pseudo-scientific terms. Antirealists argue that this concept shouldn’t be made more respectable by assigning it further explanatory value via RACIALIZATION. Instead, RACIALIZATION alone should do the explanatory work. Thus, information should be divided: Leave the legitimation of oppression with RACE and channel explanatory value through RACIALIZATION. This also explains the antirealists’ reluctance to defer to RACE. As pointed out, the flow of information is important in the distinction between originating and non-originating uses. Consider the Oxygen-Phlogiston case. Antoine Lavoisier introduced a new concept because he was reluctant to build his theory of combustion on top of the idea that combustion releases phlogiston. His theory was not an add-on to the phlogiston theory, but a competing one. In contrast, consider WHALE. It is plausible that WHALE was not substituted but revised even though it was discovered that whales are not fish. The information that whales are mammals was added to the set of existing beliefs (they live in oceans, are large, etc.). The antirealists want to distinguish RACIALIZATION from RACE by cutting the flow of information, just as Lavoisier did. This move also makes antirealism substitutional. At the same time, the substitutional nature of antirealism makes it more exposed to the marketing challenge. As noted above, the substitutional approach can sometimes be a solution to the marketing worry but not in this case. The following passage from a defender of antirealism, Adam Hochman, highlights the problem deriving from the substitutional nature. Hochman cites an objection to antirealism. The objection goes that although RACIALIZATION is supposed to replace RACE, RACE is still widely used. This discredits the usefulness of RACIALIZATION as “[t]he implication seems to be that if the racialization concept were really useful, we could do away with the term ‘race’” (Hochman, Reference Hochman2019, p. 1257). His answer to this objection is the following:

The passage suggests that there is a genuine need for RACE, unlike for PHLOGISTON. We need to talk about race and we need to teach about race and yet antirealism tries to eliminate and substitute it with RACIALIZATION. Rather than providing an answer, the passage highlights the marketing challenge in a vivid manner. Despite its somewhat light name, the marketing challenge poses a serious problem for the credibility of racial antirealism.

In response to the marketing challenge, it’s worth exploring what revisionary racial antirealism might look like under originalist engineering. As Hochman notes, RACE may eventually lose its discriminatory connotations — but antirealism shouldn’t depend on that future. Using originalist resources, I propose maintaining two concepts: RACE and RACIALIZATION. For now, antirealists can use RACE as long as it incorporates the idea of racialization and acknowledges the biological aspects of the folk concept. This results in a concept containing incompatible elements — which is precisely a point racial antirealism highlights. As Glasgow’s mismatch worry suggests, RACE already includes conflicting components. The engineered concept RACE retains its oppressive origins, but this is a provisional measure. The long-term aim is to shift public understanding toward RACIALIZATION, a concept that explains hierarchical social structures without internal inconsistency or historical injustice. Still, at this stage, antirealists must continue engaging in discourse about RACE and educating others. The shift toward widespread deference to RACIALIZATION will likely be gradual.

In sum, one might ask what exactly originalism brings to the table. My answer is twofold. First, it shifts the focus of conceptual engineering from discontinuity to the marketing challenge. Semantic and functional considerations remain important, not because of semantic or functional individuation, but because semantic and functional considerations guide engineers with the marketing challenge. As a consequence, originalist engineering gives the freedom of choice over revisionary and substitutional engineering in the face of the marketing challenge. It has been suggested that a revisionary strategy is a rather safe approach in the face of the marketing challenge but there are also numerous cases, like PHLOGISTON and OXYGEN, in which a substitutional approach has been successful.

Second, originalism broadens the tools of conceptual engineering. In the face of the marketing challenge, originalism provides tools to assess semantic changes as well as functional changes. Neither conceptual engineering based on semantic individuation nor engineering based on functions can deliver both. Moreover, originalism gives these tools the right weight. Riggs argues that instead of placing too heavy a workload on functions, like concept individuation, functions can have a useful deflationary role. Functions can have a context-sensitive role within an inquiry but there is no need to provide a systematic theory of functions, like functional individuation does. Likewise, Belleri’s thoughts on moderate semantic changes can provide good insight when an engineer assesses the strategies between revisionary and substitutional engineering.