1. Introdution

In this paper, we argue that ordinary judgments about core normative topics purport to attribute stable, objective properties and relations. Normative concepts such as the concept of being morally right resemble paradigmatically representational concepts of ordinary objects, artifacts, or natural kinds in purporting to pick out stable features of the world. Of course, the presumption of successfully picking something out in thought may be mistaken: it may turn out that there is no phlogiston or moral rightness. Our point is that such discoveries overturn our initial implicit understanding of the relevant concepts.

Our strategy is to first identify which aspects of conceptual competence ground representational purport and then show that these features are shared by normative concepts. In Sections 2 and 3, we offer a general characterization of concepts as cognitively basic ways of keeping track of a topic in thought and talk. We explain how this distinctive cognitive role is tied to both cognitive significance phenomena and to subpersonal cognitive mechanisms for managing information. In Section 4, we focus on paradigmatically representational concepts such as concepts of objects and natural kinds. We identify three broad features of our conceptual practices that ground the representational purport of these concepts. In Section 5, we argue that normative concepts exhibit these same features.

2. Keeping track of a topic

There's an important distinction in the philosophy of mind between thinking about the same topic and thinking about that topic as the same. To get a feel for the distinction, consider two cases. Suppose you’ve encountered Bernard J. Ortcutt on separate occasions and formed thoughts about that very individual – that he's a spy and that he's a pillar of the community – without being in a position to recognize that it's the same person. You might decide to report Ortcutt's suspicious activity to a responsible person like Ortcutt. In this case, you’re thinking about the same man but your thoughts don’t represent him as the same. In contrast, consider the expanding body of beliefs you’ve accumulated about Donald J. Trump. In this case, you have no difficulty recognizing sameness of topic among your thoughts. Reading news about Trump's latest tweet, for instance, might prompt you to remember how Trump behaved as a candidate and to conclude that Trump is living up to your fears. In this train of reasoning, you’re thinking of the same man as the same. This intuitive contrast between thinking of the same topic again and thinking of the same topic as the same is marked in commonsense psychology with talk of concepts: you have two different concepts of Ortcutt, but just one concept of Trump.

Given its role in grounding our intuitions about concepts, it's important to be clear about the epistemic distinction between the two cases. We’d like to highlight four epistemic signatures of thinking of a topic as the same (Schroeter Reference Schroeter2012). First, it seems obvious that the topic of your thoughts all concern the same person: there's no need for reflection or reasoning before you engage in inferences that trade on the presumption that there's just one topic in question. Indeed, the question of sameness simply doesn’t arise as long as you’re engaged in ordinary object-level ruminations on the topic: your focus throughout is simply on Trump himself. Second, sameness of topic seems rationally indubitable: it would be crazy to doubt whether Trump is Trump. As long as your attention remains focused outwards on the world rather than on your own representational states, the question of sameness seems closed. Third, the sameness of topic seems epistemically basic: no object-level evidence about Trump could help justify your assumption that Trump is Trump. Fourth, the appearance of obvious and indubitable sameness disappears when you engage in meta-level reflection about your thoughts. To formulate an open identity question about the Trump case, you just need to step back from your object-level perspective and ask a meta-cognitive question: e.g. whether this thought represents the same individual as those other thoughts do, or whether this man (the man represented by this occurrent belief) is the same person as that man (the one represented by those beliefs).

Notice that this distinctive epistemic signature also occurs in communication. If a friend asks you whether you think Trump will be impeached, you’ll understand her question as pertaining to Trump – the man who is the topic of many of your own attitudes. As long as you retain your ordinary object-level perspective focusing on the world, it will seem obvious and rationally indubitable that Trump (the man she asked about) is Trump (the man who tweeted, seemed unelectable, etc.).

Similar observations hold for your thinking about kinds, properties, relations, normative statuses, etc. Consider a children's game of tag, where the status of being ‘it’ (i.e. being the person whom all the other players try to avoid) is passed from one player to another via being caught and tagged by whoever is currently ‘it’. When you succeed in tagging someone new as ‘it’ in your game, you see that person as acquiring the very same ‘it’ status that you previously had. And when she complains that she wasn’t really tagged so that you’re still ‘it’, you immediately understand her complaint as pertaining to the very same status that you associate with that term. Indeed, it seems obvious and rationally incontrovertible that the same topic is in question. This is why the belief she's expressing – that you’re ‘it’ – seems to directly logically contradict your own belief that you’re not ‘it’.

We’ve argued elsewhere that thinking about a topic as the same is explained by subpersonal cognitive mechanisms (Schroeter Reference Schroeter2012; Schroeter and Schroeter Reference Schroeter and Schroeter2014). We must have some cognitive mechanisms for bundling together information-carrying states that allow us to consciously focus on features of our environment as topics of thought, to recognize these features again, to amass new information about them and weed out mistaken information, to keep different information bundles segregated from each other, and to retrieve information from the same bundle when engaged in reasoning about a topic. These various cognitive abilities amount to an information-management system that binds and segregates information-bearing states about features of our environment. The person-level epistemic signature of thinking of a topic as the same is best seen as a manifestation of the underlying dispositions that structure our perspective as thinkers.

The metaphor of a mental filing system is useful in clarifying the basic structure of the subpersonal mechanisms that ground these dispositions.Footnote ¹ In your office filing cabinet, particular manila file folders serve to bind together a changing bundle of documents and keep it separate from other similar bundles in the other files. Whenever you reach for a particular file (say, your [John Perry] file) your default assumption is that all the documents in it will pertain to the same topic (e.g. they’re all papers written by Perry). You may, of course, discover that this default assumption is incorrect: there could be a paper by Kaplan that has slipped into your [Perry] file. So to find out whether your default assumption is mistaken, you’ll need to step back and actually examine the different document in the file to see if they match the rest of the documents in the file. But it would defeat the whole purpose of a filing system if you always had to compare the content of individual documents before you took them to pertain to the same topic. The key advantage of a well-maintained filing system over a giant disorganized pile of documents is that you can, in general, know that co-filed documents pertain to the same topic before actually reading them. So long as you’ve been a reasonably careful filer who only puts new stuff into existing files when it's appropriately similar to the stuff already in the file, your disposition to treat everything in the same file as pertaining to the same topic will be warranted and the exercise of that disposition will normally suffice for knowledge of sameness of subject matter. Moreover, organizing your information in files allows you to accumulate and store a whole lot of interesting new information, to access it efficiently, and monitor it for mistakes. And the stability of the files, together with your conscientious updating practices, helps explain the sense in which your filing system involves stable topics despite the constant changes in the actual contents of your files.

Mental files are structurally analogous. The central idea behind the file metaphor is that many of our consciously accessible thoughts are organized in a similar way: they are bound together (or co-indexed) by topic. For any topic you’re capable of focusing on in thought, you have an evolving bundle of information-carrying states – attitudes and cognitive dispositions – that you’re automatically disposed to treat as pertaining to the same topic. The mental file metaphor suggests that you have a system of subpersonal mechanisms that serve to bind a body of informational states together and keeps that body separate from other bodies stored in the system. So a particular mental file is just a binding mechanism, like a manila file folder, that bundles together different information-carrying states in such a way that those states are automatically treated as pertaining to the same topic by your cognitive system.

This file metaphor nicely captures the structure of thinking about a topic as the same. By placing the explanation at the subpersonal level, mental file accounts explain why the appearance of sameness is obvious, indubitable and epistemically basic when we use a single file to think about the world. The model also helps explain why this epistemic signature disappears when we formulate explicitly meta-level questions about whether particular thoughts pick out the same semantic contents. In effect, mental files prejudge the question of sameness of content among mental states, so you need to step back from using the same file if you want to formulate a coherent question of sameness of topic.

Because the mental filing structure captures the epistemic signature of concepts, one might be tempted to simply identify mental files with concepts. But this would be a mistake. There are several problems with this simple proposal. To begin with, we haven’t specified how mental files are individuated diachronically or counterfactually. Since concepts can be shared over time and across counterfactual circumstances, the mental file proposal is at best incomplete. A natural way to develop the file model of concepts would appeal to causal-historical relations linking files over time and between individuals. Tracing the links established by the appearance of obvious and indubitable sameness of topic linking individuals’ mental files at a time with their past files and with those of others would allow us to demarcate diachronic and interpersonal filing traditions. Such traditions reflect the participants’ own implicit epistemic perspective on guaranteed sameness of topic (see Schroeter and Schroeter Reference Schroeter and Schroeter2014).

But there are still structural mismatches between continuous mental filing traditions and concepts. For instance, concepts have their semantic contents essentially, but mental files do not. Mental files are subpersonal cognitive mechanisms, which can acquire different semantic interpretations due to changes in the external circumstances or to changes in the body of attitudes bound together. A single continuous filing tradition with ‘Madagascar’ or ‘water’ could shift its semantic content over time, but such a shift would constitute a shift in the concept (Burge Reference Burge1988; Evans Reference Evans1973; Schroeter and Schroeter Reference Schroeter and Schroeter2014). The relational nature of these filing traditions also raises problems. Sameness of concept seems to be an equivalence relation (reflexive, symmetric and transitive), but the fact that filing traditions can merge or split will generate failures of transitivity (Fine Reference Fine2007; Pinillos Reference Pinillos2011).

Nevertheless, we’ve argued that there is an internal connection between mental files and concepts. Concepts, we suggest, are individuated by the semantic disambiguation of our mental filing traditions. In effect, concepts are representational state types demarcated by semantically fine-tuning our filing traditions. The basic unit for semantic interpretation is the filing tradition as a whole, not token elements of thought considered independently. When an entire tradition has two (or more) equally justified interpretations, ideal semantic interpretation will seek to partition that tradition in such a way that each cell in the partition can be assigned a univocal interpretation. Disambiguated filing traditions capture the direct epistemic relations forged among token elements of thought by a mental filing system while at the same time assigning a univocal semantic content to all tokens within the same partition.Footnote ²

For present purposes, the distinction between mental filing traditions and concepts won’t matter. Our question is how our basic ways of keeping track of a topic can generate representational purport. The possibility of shifts in representational content and worries about defining equivalence relations are orthogonal to this central concern.

3. Concepts vs Fregean senses

Our model of keeping track of a topic in thought turns on subpersonal cognitive mechanisms that bind informational states at a time, over time and between individuals. The resulting binding relations explain both our cognitive ability to keep track of information as about the same topic and the resulting person-level epistemic signature of thinking of a topic as the same. This relational approach to concepts is very different from the more familiar Fregean sense model, which appeals to matching patterns of substantive understanding.

Two token words express the same Fregean sense just in case each token is governed by the same pattern of understanding, which suffices to fix the same semantic content for each token (relative to specified facts about the empirical context of use). So if two token uses of ‘water’ are each governed by precisely the same implicit reference-fixing criterion, the thoughts expressed by those tokens are guaranteed to co-refer (assuming that they occur in the same historical environment). A precise match in criteria, moreover, is supposed to explain Fregean cognitive significance phenomena: why it sometimes seems obvious, indubitable, and epistemically basic from the thinker's point of view that there is just one topic in question.

Critics have raised numerous worries about the adequacy of this Fregean model. Do we really have indubitable commitments that are rich enough to uniquely determine the semantic content? Could such criteria really be shared over time and between individuals? Would a perfect match in implicit criteria really be obvious? In short, it's far from clear that the Fregean model can fully explain the epistemic signature of keeping track of a topic as the same (see e.g. Byrne and Pryor Reference Byrne, Pryor, Garcia-Carprintero and Macia2006; Fine Reference Fine2007; Kripke Reference Kripke1980; Putnam Reference Putnam, Kiefer and Munitz1970, Reference Putnam1973; Schroeter Reference Schroeter and Zalta2017; Stalnaker Reference Stalnaker1990).

For the purposes of this paper, we’d like to highlight the core structural difference between the relational model of concepts and the Fregean model of sense. As we’ve seen, Fregean senses are individuated by matching criteria for using an expression (perhaps together with contextual factors). In contrast, a relational model of concepts has no structural commitment to a precise match in topic-fixing criteria. What explains the direct epistemic appearance of sameness is being linked to the same tradition, which is consistent with open-ended variation in substantive understanding over time and between individuals (Schroeter Reference Schroeter2012; Schroeter and Schroeter Reference Schroeter and Schroeter2014). And even when there is a precise match in individuals’ topic-fixing criteria, that match does not suffice for competence with the same concept on the relational model: there must still be a relevant causal-historical link between individuals’ mental files for those individuals to share a concept.

It's also worth stressing a difference in scope between Fregean sense and concepts. The Fregean's focus on linguistic meaning encourages one to think of concepts as rules governing the correct use of particular expressions. But our understanding of concepts as ways of keeping track of a topic in thought is much more restrictive. You may not be thinking about any topic at all when you respect conventional criteria for using expressions. Consider the case of indexicals. Linguistic conventions governing ‘here’ do not correspond to a concept in the sense of keeping track of the same topic as the same. The ability to use ‘here’ is not an ability to focus our attention on the topic of hereness as such and to treat all ‘here’ thoughts as pertaining to that same topic. Hereness is not a topic of ordinary thought. Instead, ‘here’ is used to pick out different topics on different occasions: particular uses direct our attention to the location of the utterance. What we keep track of in thought and talk are the locations themselves – Melbourne, or Vancouver, or Angell Hall – not the relational property of being the location of an arbitrary speaker. But mastering semantic rules governing ‘here’ does not entail a conceptual capacity to keep track of Melbourne or Angell Hall as such. So linguistic competence with the meaning of ‘here’ doesn’t require conceptual competence in the sense we’ve been articulating: the ability to keep track of a particular topic in thought.

4. Mental files and representational purport

In this section, we take a closer look at the distinctive cognitive and epistemic practices associated with concepts of paradigm representational topics like (i) ordinary physical objects and (ii) natural and artifact kinds. Our aim is to isolate the aspects of our conceptual practices that ground their representational purport. This will put us in a position to ask whether our ways of keeping track of core normative topics like being morally right and being all told right exhibit the same representational purport.

Our conceptual capacity to represent ordinary physical objects entails a rich set of abilities that single out persisting physical individuals as such and allows us to think of them as the same across different episodes of thought. We want to bring out three key features of the mental filing system, which we take to ground the representational purport of our concepts of ordinary objects.

Long-term binding and storage: A central function of concepts of individual objects is to allow for learning about a particular individual over time. To fulfill this role, a mental filing system requires mechanisms for stably binding persisting attitudes and cognitive dispositions together over time. A mental filing system allows for the accumulation of a body of information about an object's nature and history that can be stored and later redeployed in perceptual recognition or offline reasoning about the object.

Unlike perceptual tracking of objects, concepts bind information derived from different sensory modalities, testimony, and inference, which may be formatted in different ways. Your mental file for that particular ambulance, for instance, may bind visual indices tracking that object in your visual field, auditory perceptions of the noise of the siren, an awareness of your own emotional response to the object, a rich bundle of beliefs and inferential dispositions that constitute your implicit conception of what kind of thing an ambulance is (e.g. it's a spatio-temporally continuous physical object, it has the characteristic physical and social functions of ambulances), and standing object-specific attitudes about this particular ambulance (e.g. its being from your local hospital, it's the one driven by your friend Al). All of this information can be thought of as propositional attitudes (e.g. imaginings, beliefs, questions, intentions, even perceptions that the object is instantiated) or cognitive dispositions (e.g. recognitional, inferential, or motivational dispositions) bound to the very same mental file. Mental files integrate these different states and make them available to conscious reasoning as pertaining to the same topic. This integration of persisting attitudes and dispositions allows you to recognize the incoming information as pertaining to the same objects – an ability that underwrites both induction and learning about particular individuals as such.

Epistemic oversight: Our conceptual thinking is subject to reflective epistemic oversight to ensure coherence and stability of topic. As we learn more about an object, we acquire more ways of reidentifying it. The more you learn about Donald J. Trump (his changing appearance over time, his business and political associations, his family history, etc.) the more new ways you have of deciding when that individual has been involved in some event you learn about. A growing bundle of attitudes and dispositions means that you’ll be disposed to add new information on the basis of many different kinds of evidence. With each new addition to the bundle, however, there is greater potential for both internal inconsistencies and variation in the external sources of information. Managing such a flexible information-bundling system thus requires epistemic vigilance.

In non-reflective creatures, incompatibilities among beliefs might be resolved through a simple battle among dispositions, with stronger credences within the creature's overall understanding winning out over weaker ones. As reflective thinkers, however, we seek to integrate different elements in our conception of a topic and are actively on the lookout for internal inconsistencies. Along the way, we reflectively refine our understanding in ways that allow us to home in on the features we judge most relevant to our background interest in the topic. For instance, we might discover that the cause of our ‘Trump’ beliefs for the past forty years is not an individual human being but a robot controlled from Moscow. Or we might discover that the whole ‘Trump’ story was a giant media hoax. In that case, what is Trump? The young tycoon? a robot? a fictional character? nothing? When we try to answer these questions, we are implicitly revising our earlier conception of the topic. The correct verdict is not determined by simply letting our strongest prior credences win out: instead we justify one verdict over others by reflecting on our own interests and priorities to determine which feature of the world has been most relevant to our practice in trying to keep track of something as Trump (see Schroeter and Schroeter Reference Schroeter, Schroeter and Daly2015).

Reinterpretation: A mental filing system generates the appearance of a stable topic over time despite revisions in one's bound attitudes and dispositions. Consider an occurrent memory of Trump descending his golden escalator to announce his candidacy. The memory presents itself as a direct window on your past judgment from which it derives: it seems obvious and indubitable that there's just a single topic over time. At the same time, your [Trump] file binds this memory to a bundle of contemporary attitudes and dispositions, many of which you did not have back then. Even if you’ve changed your conception of Trump, you ‘read back’ your current conception as the correct way of identifying the topic of your earlier thoughts. In effect, revisions in your understanding of the topic are treated as a reinterpretation of the content of your past thoughts. If you now think that Trump is a robot, then you’ll take your memory to be about that robot, and you’ll take your prior beliefs that Trump was a human being to be false beliefs about that robot. Thus, your mental filing system can generate the appearance of stability of topic even through radical changes in substantive understanding of the topic.

It's important to see that ‘reading back’ your current conception of a topic as giving the content of your past thoughts isn’t an arbitrary imposition of your current perspective onto your past self. On the contrary, it reflects core epistemic commitments of your past self. After all, you never took your conception of the topic to be infallible. If you had, you would not have been disposed to refine your conception of the topic in the light of empirical inquiry and reflection. In revising your conception of Trump, you saw yourself as correcting your prior conception. And reading this correction back into your understanding of your past thoughts vindicates your general epistemic ambition to keep track of a genuinely interesting and important feature of your environment. Far from changing the topic, then, reinterpretation seeks to identify the topic you were aiming at all along. If your reflective epistemic oversight has been successful – if it has identified criteria that you would have continued to endorse after ideal, empirically informed reflection – ‘reading back’ your reinterpretation is actually the most charitable interpretation of the semantic content of your past thoughts (Schroeter Reference Schroeter2008; Schroeter and Schroeter Reference Schroeter and Schroeter2014).

The three key features of our internal mental filing system can be extended to the interpersonal level via stable lexical meanings. Names in a public language can trigger the immediate appearance of sameness of topic that's structurally analogous to stability of topic over time. When you hear someone else use a name in a sentence like, ‘Trump is a moron’, you automatically hear their claim as pertaining to the same topic as the one you associate with that name. This epistemic signature of obvious and indubitable sameness is grounded by stable subpersonal mechanisms for linguistic interpretation that recruit the mental file you associate with ‘Trump’ to interpret the content of others’ claims. The result is an interpersonal mental filing system, in which appropriately linked individuals have a coordinated mental filing tradition with the three key features we highlighted above:

Long-term binding and storage: Coordinating their internal mental filing systems allows members of a community to vastly expand the body of bound information available to each individual. Others’ beliefs are taken as pertaining to the same topic, allowing for direct logical agreement or disagreement and direct acceptance of testimony. Given a baseline of epistemic trust and cooperation within the community, this direct semantic coordination affords a vast improvement over isolated individuals’ abilities to learn about particular topics as such.Footnote ³

Epistemic oversight: Although we have a prima facie disposition to trust others’ testimony, we need epistemic vigilance to deal with conflicts among individuals’ commitments. Just as we take our own pool of standing attitudes into account in this process, we need to take the communal pool of attitudes into account in refining our understanding of the content of our own mental files. Given the greater variability among bound attitudes within a linguistic community, the need for oversight is particularly pressing for shared filing traditions.

Reinterpretation: Any revisions in one's own understanding of the topic are presumed to apply to all bound states. In effect, you ‘read back’ your own current understanding of the topic as providing the correct interpretation of what all of us have been thinking about all along. In the intra-personal case, the correct interpretation should be justifiable from your past perspective. Similarly, in the interpersonal case, the correct interpretation should be mutually justifiable from the perspective of all those to whom it is applied (Schroeter and Schroeter Reference Schroeter and Schroeter2014, Reference Schroeter, Schroeter and Daly2015).

These three features – long-term binding, epistemic oversight, and reinterpretation – capture our ways, as reflective thinkers, of managing our subpersonal system for accumulating empirical information as pertaining to particular objects. Let's now focus on the question of representational purport. How exactly do these three features ground the representational purport of a mental file?

Let's start with long-term binding. It's not obvious that mental files – whose function is to present thoughts as pertaining to the same topic – must always involve long-term accumulation of information about the nature of that topic. Recall our earlier example of ‘being it’: in this case, one might argue that there is an internal binding device, like a mental file, that simply serves to coordinate action schemas across individuals at a time, so that we’re disposed to do and expect certain things depending on who is ‘it’. Such a coordinating practice doesn’t require any ability to accumulate a stable body of information about the topic – i.e. assumptions about the nature of itness itself or a body of contingent facts about that status. In that case, there is little pressure to interpret mastery of ‘being it’ as an ability to represent the abstract status as such, as opposed to interpreting it as a mere expressive device for practical interpersonal coordination.

Another contrast case would be perceptual tracking of objects. Zenon Pylyshyn has argued that we have ‘visual indices’ that bind incoming streams of visual information into discrete units, which we track across the visual field (Pylyshyn Reference Pylyshyn1989, Reference Pylyshyn2007). It would, of course, be appropriate for an external interpreter to describe these visual indices as ‘tracking’, ‘locking onto’, or ‘representing’ the objects which cause the incoming stream of information they bind together. But ascribing representational purport to visual indices seems unwarranted. We take representational purport to be a person-level phenomenon that must be grounded in the internal structure of the thinker's conscious perspective. Visual indices are merely subpersonal mechanisms automatically triggered by fleeting proximal stimuli: they do not persist beyond a single episode of perceptual tracking, and so do not afford long-term storage for bound information. From the thinker's conscious perspective, they are merely attention-grabbing aspects of their visual experience.Footnote ⁴

In contrast, a mental file associated with a name like ‘Trump’ serves to accumulate and store a rich and stable body of empirical information as pertaining to a single topic. In effect, the thinker is mining the world for more information about a putatively stable topic. The growing body of information is used in outward-directed engagement with features of the environment: re-identifying the feature as the same on different occasions, redeploying accumulated information to that case, and storing information derived from these new cases. Thus, long-term information storage via a mental filing system can set up an informational feed-back loop for accumulating new information about specific features of the environment. From the thinker's conscious perspective, this accumulation of information from the world is presented as all pertaining to a stable topic. This combination of stable topic and outward directed epistemic activity in gleaning new information provides rich internal grounds for attributing representational purport. From your internal perspective, you are actively using your [Trump] file to isolate and learn about a stable feature of the world.

This representational purport is strengthened by epistemic oversight of one's mental filing system. Two aspects of oversight are crucial here: (i) coherence and (ii) epistemic ambition.

Maintaining coherence within a file is consistent with the lack of representational purport. In a mental filing practice with ‘being it’, for instance, we have a coherence requirement that only one person can be judged ‘it’ at a time. So if two players claim that different individuals are ‘it’, they will feel pressure to resolve the dispute by relying on the implicitly grasped rules of the game. For instance, a conflict over who's ‘it’ might generate an argument over who was really tagged. But it doesn’t follow their thoughts purport to represent the status of itness itself. As we noted, competent users of the term need not have any ability to focus on the status as such in thought or accumulate information about it. Seeking to resolve contradictions does not alter this lack of representational purport.

But when we have a mental file with long term information storage, like that associated with ‘Trump’, epistemic oversight aims for a much richer type of coherence in substantive understanding. Given that your ‘Trump’ thoughts purport to represent a single topic, there is rational pressure to ensure that you have a coherent overall understanding of that topic: e.g. what kind of thing Trump is, his history, habits, and other characteristics. You’ll typically have some implicit ‘guiding conception’ of what kind of thing Trump is, which constrains your understanding of coherent possibilities. If you take Trump to be a physical object, for instance, you assume that he's located in space and time, causally unified at a time and over time, that he traces out a continuous spatiotemporal path, that the relations between his parts are stable, etc. (Dickie Reference Dickie and Jeshion2010, 227). Given this guiding conception, it is incoherent to suppose that Trump was in two places at one time. A guiding conception structures your overall understanding of a topic. As a consequence, it helps focus the general outward-directed representational purport of your mental file on a particular aspect of external reality: a persisting physical object. From your own perspective, your ‘Trump’ thoughts purport to represent an object, a particular human being, rather than a human kind in general or Trumpish behaviour.

The second aspect of reflective oversight is epistemic ambition. What we mean by this is that you don’t take your current guiding conception to be a fail-safe guide to the nature of the topic you’re thinking about. As a rational epistemic agent, you have an interest in keeping track of genuinely interesting and relevant features of the world, as opposed to mere ontological shadows of your current guiding conception. For instance, you could discover that Trump is not in fact an ordinary empirical object – perhaps Trump is just a fictional character played by multiple actors. Although this type of radical revision in guiding conceptions is rare, it is in principle justifiable on the basis of your prior epistemic ambition and the actual empirical history of your filing tradition. This radical corrigibility of your guiding conception reflects your implicit commitment to representing objective features of the world rather than projections of your current categorizing criteria.Footnote ⁵

This commitment to objective representation is also manifest in the third feature we highlighted, Reinterpretation. When you ‘read back’ your revised understanding of Trump as giving the content of your past thoughts (and those of others), you’re implicitly treating earlier divergent conceptions as mistaken. That is, you take your current understanding to be a correct characterization of the very same topic you and others were thinking about all along.

By itself, a disposition to treat your past self (and others in your filing tradition) as mistaken about a topic does not entail a commitment to stable representational purport. It could just be a manifestation of ‘bad manners’ – foisting your current standards onto others without regard for what's relevant from their point of view. In fact, some theorists suggest our linguistic practice with ‘tasty’ is governed by a convention mandating this type of bad manners: individuals rely on their own current standards of taste to claim that anyone with different standards have false beliefs about what's tasty (Kölbel Reference Kölbel2002; MacFarlane Reference MacFarlane2014). The proposed ‘bad manners’ convention for ‘tasty’ would coordinate expectations about how to express gustatory preferences, without committing participants to a property with stable instantiation conditions picked out by all uses of the expression. The practice is thus like the one we imagined for ‘being it’: there's interpersonal coordination without representational purport.

The hallmark of objective representational purport here is not reinterpretation per se, but the epistemic norms that govern reinterpretation. The substantive understanding that you ‘read back’ must be justifiable to other perspectives within your tradition. Holding your reinterpretations to mutually justifiable epistemic standards indicates that you’re treating your filing tradition as representing stable, intersubjectively relevant features of the world. This standard of mutual justifiability is met in our practice with proper names. To be warranted, a conclusion that Trump is in fact a fictional character played by multiple actors can’t simply rest on an idiosyncractic taste for conspiracy theories rejected by others. Both the empirical facts and shared background interests that have sustained your filing tradition are relevant to determining whether such conclusions are warranted. In general, our judgments about the falsity of our prior conception are backed by a commitment to finding a mutually justifiable interpretation of the semantic content of ‘Trump’ judgments.

In sum, we have suggested that the representational purport of our mental filing system for ordinary objects is richly grounded in the thinker's own perspective. Together, the three key features of this system – (1) long-term binding, (2) epistemic oversight, and (3) reinterpretation – allow the thinker to gather information about individual objects, whose essential nature she may not fully understand, and to correct mistakes about their nature and history. Long-term binding allows for outward-directed accumulation of information about a stable topic over time. Epistemic oversight and reinterpretation seek to refine our current understanding of the topic while holding ourselves and others to mutually justifiable standards of correctness. This process of self-regulation is organized around an implicit ‘guiding conception’, which places substantive coherence requirements on the subject's understanding of the topic. Epistemic ambition and mutual justifiability aim to refine the thinker's current understanding to arrive at a coherent understanding of objective features that are relevant to shared background interests. These more holistic epistemic standards can justify revisions to one's current guiding conception. Overall, the core features (1) – (3) ground the objective representational purport of our mental filing system for ordinary objects: they manifest our internal efforts to keep track of objective features of the environment that are relevant to our projects and purposes.

Singular thought is the parade case for motivating and explaining how our practices in managing a mental filing system can ground objective representational purport. But we find similar practices with other topics that we keep track of and learn about. Natural kinds like being a horse and artifact kinds like being a moped are clear cases in point. The way we think about these kinds requires a filing system that respects the same three key features we identified as grounding representational purport in the case of our conceptual representation of objects – long-term binding, epistemic oversight, and reinterpretation of a shared mental filing tradition.

We have stable bodies of standing attitudes and cognitive dispositions about topics like horses or mopeds, which we use to identify new instances of the kind and to learn more about that kind. So we must have mental files that bind and store these persisting informational states as pertaining to the same topic. Moreover, this mental filing system must afford long-term binding and storage if it's to underwrite learning about the kind.

As a consequence of this open-ended learning, we need to exercise epistemic oversight to ensure the coherence and relevance of the substantive understanding in the mental file. A crucial distinction between our ways of keeping track of objects and our ways of keeping track of kinds lies in our guiding conception of the topic. Unlike our conception of individual objects, our guiding conception of kinds does not require spatiotemporal continuity among instantiations of the kind. Files that purport to keep track of kinds are governed by the presumption of coherent instantiation conditions. One important aspect of coherence is formal consistency: we take kinds to capture a dimension of empirical similarity shared by all instances that ensures that our file does not cross-classify cases as both in and out of the kind.

A more substantive type of coherence, however, involves a guiding conception of the specific kind in question. For instance, there's an important distinction between our guiding conception of natural as opposed to artifact kinds. Consider your [horse] file: the bundle of attitudes and dispositions grouped together involves a presumption that horses form something like a biological species: e.g. you implicitly assume that horses don’t just share similar surface appearances but share similar biological characteristics such as characteristic modes of nutrition and reproduction, that there is some ‘hidden essence’ that explains these similarities and is passed on through reproduction, that kind membership supports induction to unobserved horses, that being a horse excludes being a member of other biological species (e.g. no horse is a donkey), and that being a horse figures in a hierarchical taxonomy of biological kinds (e.g. horses are grouped together with donkeys and zebras as all belonging to a higher-order biological kind (equus), there are distinct subspecies or breeds of horse) (cf. Bird and Tobin Reference Bird, Tobin and Zalta2017).Footnote ⁶ In contrast, your mental file for mopeds is governed by a presumption that they are created by humans to function in a certain way in order to serve certain human purposes. You implicitly assume that mopeds are small two-wheeled vehicles designed for short city transport, that they have a characteristic profile that distinguishes them from related vehicles (such as cars, motorcycles and bicycles), that kinds of vehicle aren’t exhaustive or mutually exclusive (there can be an intermediate case between mopeds and motorcycle, or hybrid moped/jetskis), that there's no intrinsic ‘hidden essence’ to mopeds – their classification depends on what they can do and what they were intended to do. These initial guiding conceptions of natural and artifacts kinds, however, can be rationally refined or rejected in the light of empirical discoveries. We could discover that the things we’ve been classifying as horses are really artifacts controlled by Martians, and the things we’ve been classifying as mopeds are really the Martians themselves. In that case, we would have good reason to revise our guiding conceptions of our two filing practices.

When we do revise our initial conceptions in the light of further reflection or empirical discoveries, we ‘read back’ our revised conception as giving the real content of the thoughts in our mental filing tradition. As in the case of our thoughts about ordinary objects, this sort of reinterpretation is epistemically warranted insofar as it vindicates our prior epistemic ambition: all along, we took ourselves to be keeping track of features that are genuinely important and relevant to justifying our shared filing tradition – not just projections of our current conceptions.

Together, these core features ground the objective representational purport of our mental filing system for natural and artifact kinds: our filing practices of accumulating and managing information reflect an implicit effort to keep track of stable features of the world that are relevant to our shared background interests.

5. Representational purport in the normative case

We keep track of normative properties like being morally right in thought in much the same ways that we keep track of paradigm representational topics like objects, natural kinds and artifact kinds. Over time we acquire and store a wide variety of attitudes and dispositions as pertaining to the topic of moral rightness: commitments about particular empirical cases, rules of thumb about what sort of features make an action morally right, procedures for adjudicating hard cases, assumptions about when one's own moral judgments may be unreliable, criteria for deciding which moral testimony is most trustworthy, theoretical assumptions about the metaphysical character of moral rightness, and so on. The public language term ‘morally right’ helps to bind this evolving bundle of attitudes and dispositions with those of others in our community, presenting them as pertaining to the same topic. Thus our shared mental filing tradition grounds long-term accumulation, storage and binding of information as about the property of morally rightness. As in the case of paradigm representational concepts, moreover, we exercise reflective epistemic oversight to maintain the coherence and relevance of the bound information, and we rely on our revised understanding to reinterpret the semantic content of all attitudes in our filing tradition. So the concept expressed by ‘morally right’ seems to fit the pattern we’ve identified as grounding representational purport in uncontroversial cases.

We take it to be uncontroversial that our concept of moral rightness conforms to this general pattern. And the fact that we accumulate information as about this topic over time provides support for the objective representational purport of our filing practice. But there may be room to worry about the reflective standards of coherence governing normative concepts. As we saw in the case of ‘being it’ and ‘tasty’, purely formal coherence and reinterpretation norms don’t suffice to ground representational purport. The keys to objective representational purport at the reflective level, we suggested, are: (i) the substantive guiding conception of the topic and (ii) the epistemic ambition to track important and relevant features of the world. These two features shape our practices of reflective epistemic oversight and reinterpretation of other attitudes in our filing tradition. Without the right kind of representational discipline at the level of reflective epistemic practices, your filing practices may fall short of objective representational purport. For instance, your concept of moral rightness might just a case of ‘bad manners’ – using your current evolving moral standards to evaluate others irrespective of their epistemic perspectives.

However, our reflective epistemic practices do not support this interpretation of normative topics like moral rightness. Our guiding conception of moral rightness is representational in much the same ways as our guiding conceptions of objects, natural kinds or artifact kinds. A guiding conception functions as a mental model of the topic that (i) guides one's inductive practices of learning and generalizing from one case to the next and (ii) determines one's substantive standards for coherence and incoherence for the topic. It's clear that we have a rough implicit guiding conception of moral rightness. We take moral rightness to be a property of actions, agents and circumstances that supervenes on non-normative descriptive properties. Roughly, morally rightness is supposed to distinguish actions that respect the practical interests and concerns of affected parties, while weighing everyone's interests equally. Like our guiding conceptions of natural or artifact kinds, our guiding conception of normative properties seeks to identify a stable principle of categorization, which determines a coherent intension across all possible worlds. The intension should not cross-classify cases and it should not vary from one episode of thought to another. Moreover, moral rightness demarcates a genuine pattern of empirical similarity: we are able to recognize the instantiation of moral rightness on the basis of ordinary empirical facts, and this empirical recognition capacity can be projected to new cases.Footnote ⁷

In addition, our reflective moral epistemology evinces the epistemic ambition to identify a genuinely important and relevant property – rather than a mere projection of our current patterns of moral understanding. We take our current conception of moral rightness to be corrigible in the light of further inquiry and reflection – not just with respect to particular cases but also with respect to general principles and even aspects of our initial guiding conception. One important impetus for such corrections is an effort to reconcile different aspects of our moral understanding through narrow reflective equilibrium: for instance, we may realize that we have been classifying similar cases differently or that our moral rules of thumb conflict. Another impetus can come from new empirical information about our psychology or our circumstances: for instance, we may come to appreciate the harms others experience from our ‘jokes’ or we might become convinced of the existence of a moral law-giving deity. A third impetus for correcting our guiding conception comes from reflecting on facts about our shared filing tradition. For instance, learning that others in our filing tradition assume moral rightness takes into account the interests of animals as well as humans may lead us to adjust our own conception accordingly. Or more radically, learning of the role judgments of moral rightness have played in reinforcing social hierarchies over time might justify the conclusion that moral rightness marks conventional social norms for reinforcing elite social status. These examples are intended to illustrate the ways that we take our guiding conception of moral rightness to be corrigible via holistic reflection on our filing tradition and the empirical circumstances in which it developed. This openness to correction is an indication that we take moral rightness, like paradigm representational topics, to be an objectively relevant feature of the world, and not merely a projection of our current criteria.

We’d like to emphasize that the notion of representational purport at stake here is a robust one. We’re not simply claiming that it's appropriate for participants in the filing tradition to speak as if there were a stable property picked out by that tradition. We’re claiming that participants are committed to there actually being a single stable property picked out. In thinking about moral rightness, our reflective epistemic oversight seeks a single intension that can be justified from all different perspectives within one's filing tradition and determines the intension of the tradition as a whole. In short, we are actively looking for a single, univocal intension in refining our understanding of moral rightness through wide reflective equilibrium. This stable representational purport clashes with contextualist, truth-relativist, inferentialist, and pragmatist interpretations of our concept of moral rightness. Perhaps these non-representational semantic interpretations can be supported on other grounds – but they fail to capture the internal epistemic perspective of conceptually competent thinkers.

A second clarification concerns the properties normative concepts purport to represent. The mere fact that a concept has representational purport does not entail that the feature represented enjoys a privileged metaphysical status. Consider your concept of being a moped: the represented kind doesn’t play any deep role in scientific explanation and it is not metaphysically fundamental. Nonetheless, being a moped is an objective feature of the world: it's not a mere projection of your current criterion for recognizing mopeds. Instead, your conception is corrigible in the light of further information about what's important and relevant to your shared filing tradition. Mutatis mutandis for your concept of morally right.

We’d like to end this section with a caveat. Although we think that many normative concepts have objective representational purport (e.g. our concepts of morally right, all-told right and practical reason), we do not claim this is true of all normative concepts. In particular, many of the concepts expressed by ‘ought’ seem to lack such representational purport. Epistemic modals are a case in point: as many theorists have pointed out, there is no pressure to find a single context-invariant intension for claims about what one ought to believe. The representational purport of normative concepts must be established on a case-by-case basis.Footnote ⁸

Disclosure statement

No potential conflict of interest was reported by the authors.