1 Introducing an ‘old’ idea

Many contemporary philosophers endorse some form of Representationalism about perception:


Representationalism: If an agent enjoys a perceptual experience as of P, then an agent enjoys a perceptual representation with P as its intentional content.

According to representationalism, perceptual states, and not just post-perceptual states (e.g., cognitive judgment, belief), are representational. While this generic version of representationalism does take a position on what having an intentional content consists in, nearly all contemporary representationalists endorse a more committal thesis we call ‘Orthodox representationalism’:


Orthodox representationalism: A perceptual representation’s intentional content has constitutive veridicality (truth, accuracy or satisfaction) conditions.Footnote 1

Traditionally, orthodox representationalism identifies perceptual content with propositions (or conceptual content more generally), maintaining that perceptual content is belief- or judgment-like in that it does something akin to asserting or ‘telling’ us something about the world (Cohen, 2015; Glüer, 2009). Other, more ‘relaxed’ orthodox representationalists replace propositional contents with non-propositional and/or non-conceptual contents. Importantly, however, they maintain that representational success is descriptive in nature, i.e., that perceptual representations have constitutive veridicality conditions, and that perceptual representational contents instantiate a relaxed version of predicative or subject-predicate form (though perceptual representations may have non-linguistic, e.g., imagistic, formats).Footnote 2 For example, according to Tyler Burge (2010, 2022), perceptual states employ a non-conceptual predicate-like attributive under which a subject-matter or “singular element” is picked out. Similarly, (Nanay, 2020) claims that “representing something means attributing properties to this thing”.

Contemporary philosophers who reject orthodox representationalism typically endorse some variety of anti-representationalism (Chemero, 2009; Gallagher, 2017; Hutto & Myin, 2013). Accordingly, the contemporary discourse about representation appears to offer two choices: either one endorses orthodox representationalism or one rejects representationalism altogether. But these haven’t always been, and needn’t continue to be, the only options. The widespread popularity of orthodox representationalism among contemporary analytic philosophers contrasts sharply with its widespread rejection amongst earlier philosophers like Descartes, Malebranche, and Helmholtz. Yet these philosophers didn’t reject representationalism. We’ll argue that this ‘old’ idea about the nature of perception points to a middle path: a genuinely practical, and therefore ‘unorthodox’ conception of perceptual representation (Sect. 2).

How helpful can such a conception be to contemporary philosophers of perception? We suggest that there’s good reason for representationalists to dissociate the notion of representation from veridicality conditions and to explore the possibility of robust and philosophically interesting representations even while having practical success conditions instead of veridicality conditions. We then address the objection that the idea we advocate recovering is too old— that contemporary perceptual science vindicates orthodox representationalism. As a reply, we argue that contemporary perceptual science further motivates an attempt to articulate a view of practical perceptual representation (Sect. 3). But what would such a view look like?

We outline a ‘practical access’ account of representation as a promising candidate. This account suggests that representations have a practical nature in general, and that the fundamental standard of success that applies to representation as such is a standard of practical appropriateness. Some kinds of representations continue to have constitutive veridicality conditions on this account. However, their veridicality conditions are grounded in appropriateness conditions. In addition, the sorts of representations that have constitutive veridicality conditions presuppose ‘de agendo’ representations. De agendo representations have constitutive appropriateness conditions rather than veridicality conditions. We suggest, programmatically, that perceptual representations may have de agendo contents (Sect. 4).

Some old ideas are still good ideas. We hope to convince the reader that practical perceptual representations are at least a good enough idea to be explored, and, more generally, that there may be more ways of being a representation than orthodox representationalism is able to acknowledge.

2 Practical perceptual representation in the past

Certain early modern philosophers appear to maintain that perceptual states are representational despite rejecting orthodox representationalism (e.g., Simmons, 2008). Descartes, for instance, did not seem to think that perception is in the business of producing true/false judgments about the world -–that, rather, is the business of the intellect. What, then, is the job of sensory perception? Descartes tells us in the Meditations (Meditation VI) that the deliverances of sensory perception are fundamentally practical: they ‘teach us’ about the needs and goals deriving from our embodied nature, and about possibilities of interaction with the environment and other bodies.

In a similar vein, Malebranche held that “sense perception…is not concerned with truth, but only with the self-preservation of the organism” (Davies, 1924). Indeed, he writes: “We shall make it clear, then, that we should rely on the testimony of sight not in order to judge concerning the truth of things in themselves but only to discover the relation they have to the preservation of body” (The Search for Truth, Book I, Chapter 6). On the one hand, in speaking of the “testimony of sight,” Malebranche seems to echo orthodox representationalists claiming that perceptions are ‘tellings’ of some sort. On the other hand, like Descartes, Malebranche resists construing these tellings on the model of assertion or true/false judgment. On what we take to be the Malebranchian view, perceptual states are not the kind of things that we can evaluate for accuracy or inaccuracy (Simmons, 2009).

The view that transpires from these brief glimpses into Cartesian and Malebranchian thought is difficult to place in the contemporary dialectic. While some orthodox representationalists contest that perception is propositional or conceptual, they still consider it, in its general form, analogous to intellectual judgment. Perceptual representations either consist in descriptions or attribute properties to environmental particulars and those attributions are supposed to be accurate or veridical with respect to those particulars. Accordingly, even those orthodox representationalists who opt for non-propositional or otherwise non-conceptual perceptual content are still committed to such content consisting in something very much like a theoretical judgment. In other words, they are still committed to the view that perception, to use Kathleen Akins’s phrase, is engaged in an “ontological project” of categorizing environmental particulars and their properties (Akins, 1996, p. 396). Whereas, following Simmons (2009), we’ve suggested that the early modern philosophers we mentioned take perception instead to be engaged in something closer to what Akins calls a “sensory motor project” of positioning a perceiver to respond appropriately to particulars in her environment.

Moreover, while a number of philosophers have offered accounts of action-oriented or ‘practical’ perceptual representations, they remain orthodox. This is because ‘practical’ representations, on these views, are either (i) just standard veridicality-conditional representations that are underwritten by embodied skills, or (ii) imperatival rather than indicative, or (iii) ‘practical’ in the sense of attributing practical properties or ‘affordances’ like edibility and climbability, or some combination thereof.Footnote 3 On such views, practical perceptual representational content remains a matter of veridicality conditions.Footnote 4 Meanwhile, a number of anti-representationalists appeal to the action-oriented nature of perception as a reason to drop the idea of perceptual representations altogether. Thus, a view that wants to hold onto perceptual representations while rejecting veridicality conditions, consistent with the Cartesian and Malebranchian insights mentioned above, fails to fit neatly into either orthodox representationalism or anti-representationalism.

But one needn’t even go as far back as the early moderns to find a view that doesn’t quite fit the contemporary dialectic: some nineteenth and early twentieth century accounts of perception in philosophy and psychology are similarly difficult to fit into current anti-representationalist and orthodox representationalist categories. To give just one example, many contemporary interpreters place the nineteenth century philosopher-physicist-physiologist Hermann von Helmholtz firmly in the orthodox representationalism camp, even suggesting that he might be the most important precursor of David Marr’s computational approach to perception (Marr, 1982). However, Isaac (Isaac, 2019, 2020) argues that Helmholtz was rather skeptical of the idea that perception’s goal is to provide true descriptions of the world by means of a direct correspondence between subjective sensations and their external causes. This interpretation seems to be truer to the text, given Helmholtz’s own claim that “there cannot be any other truth of our ideas except of a practical truth” (Helmholtz, 1950). So, while Helmholtz was no doubt some kind of representationalist, he seems to have endorsed a kind of ‘practical representationalist’ view in the vicinity of Descartes and Malebranche.

We think that the fact that these ‘old’ ideas are hard to place within the contemporary philosophical discourse alone provides good motivation for exploring them further. This motivation is strengthened by the fact that philosophers who stress the genuinely practical (i.e. action-guiding) nature of perception are often inclined towards anti-representationalism—a view that does not lack its own problems. If veridicality conditions really are constitutive of representations as such, anti-representationalists reason, and if perception’s telos is practical success as opposed to truth or veridicality, how can perception be constitutively representational? Hence, the familiar conclusion: perception simply cannot be representational.

Indeed, Burge (2010, 2022) has argued that views that try to ground mental representations in merely biological functions yield only “deflationary”, as opposed to “robust”, representations. That’s because non-trivial veridicality conditions do not simply fall out of biological functions. Since ‘biosemantical’ theories (Millikan, 1984, 2004) are merely describing states that function to covary with other things because their doing so often enough fulfills a biological function, Burge claims the explanatory work they give to veridicality conditions is “redundant” (2010, p. 9). In such cases, veridicality conditions merely redescribe biologically functional covariation of some kind; they don’t imply a concern with objectivity or getting the world right. Although representational functions presuppose, are accompanied by, and often play biological functions, Burge claims that they are not themselves biological functions. Biological functions have a practical telos, so they yield only practical success conditions, whereas representations have a representational telos: they must function to represent veridically (75, 301). If Burge is right, then if perception has only a practical telos, there’s no genuinely constitutive sense in which we can say perception is representational. But prima facie, Descartes, Malebranche, and Helmholtz would disagree. If sense can be made of representations that are representations in a robust, philosophically and psychologically interesting sense, while lacking constitutive veridicality conditions, this would presumably be a happy result for representationalists.

On the other hand, Burge would argue that Descartes, Malebranche, and Helmholtz simply got perception’s function wrong. That view of perception is very old and not based in science, or, to the extent that it was based in science, it was extremely early and now extremely outdated science. For Burge (2010, 2022) has argued that contemporary perceptual science vindicates orthodox representationalism: it shows that perception is concerned with objectivity and veridicality.

We take seriously the worry that the old idea we seek to resuscitate is too outdated to be of use to contemporary philosophy. So, in the next section, we will argue that contemporary scientific accounts of perception do not provide strong support for orthodox representationalism and are in fact compatible with the old idea that perceptual representation takes a practical form.

3 Perceptual science and robust representation

It seems clear that, at least in some cases, perceptual science employs representational concepts merely instrumentally (Burge, 2005, 2010; Egan, 2010, 2018, 2020; Hutto & Myin, 2017; Orlandi, 2014; Ramsey, 2007). Scientific realism about perceptual representations, as opposed to instrumentalism or fictionalism, requires more than noting that representations are often postulated as part of explanations in perceptual science. What needs to be shown is that some of the talk about perceptual representations (a) picks out things that must be understood as having properties that are distinctive of representations, and (b) is ineliminable from successful scientific explanations. In other words, a realist about perceptual representations needs to show that if we get rid of representations, i.e., posits that have certain features that are constitutive of representations, we suffer a genuine explanatory loss, not just in terms of elegance or ease of expression or comprehension, but in terms of explanatory power, parsimony, coherence, rigor, or some other plausibly truth-tracking scientific virtue.

As we have seen, Burge and other orthodox representationalists claim that veridicality conditions are constitutive of representations. What Burge calls “deflationary” representations are information carrying states that play biological functional roles but have veridicality conditions only trivially. To play their explanatory role, it is essential only that they carry information and contribute to appropriate behavior. Saying that such states have veridicality conditions adds nothing to the explanatory work they do. In this way, they have veridicality conditions only trivially, and not constitutively. In contrast, what Burge calls “robust” representations have constitutive, hence non-trivial, veridicality conditions. Burge lays down the following conditions for robust representation:

  1. (i)

    They are constitutively involved in “primitive agency” (2010, 2022)Footnote 5: They occur in creatures that are capable of purposeful action, that is, behavior that is directed at the world and that’s supposed to serve a creature’s needs, rather than behavior that’s merely responsive or reflexive, even if it also serves an organism’s biological ends. (2010, p. 342)

  2. (ii)

    They are constitutively perspectival: Representations constitute a creature’s perspectives on or modes of presentation of the things their representations are about (2010, p. 38, 342, 531).Footnote 6 Perspectives are connected with the sensory capacities of primitive agents on the one hand and capacities objectivity, i.e., mind-independent world-directedness, on the other.

  3. (iii)

    They are constitutively “individual level” (2010, part III): Because they are immediately connected with whole organism action and because they constitute perspectives, they are primarily (though not necessarily exclusively) attributable to a whole organism, i.e., an individual, rather than an organism’s sub-systems or parts.

  4. (iv)

    They have distinctively representational functions and hence conditions of success: While representations are of course biological phenomena, they are also distinctively psychological phenomena. Their telos (function) has features not found in (other types of) biological functions. Hence representations mark the domain of the distinctively psychological. Their function is broadly epistemic in that it is a matter of whole agents, i.e., individual subjects (subjectivity), grasping and in some sense understanding objects (objectivity). Representational success consists in correctly grasping or understanding a mind-independent subject-matter. To be successful, representations must “fit” (be correct with respect to) their subject-matter.

For Burge, of course, (iv) is a matter of functioning to represent veridically. Robust representations are directed at mind-independent subject-matters and they are supposed to match them; they “fit” their objects (i.e. are correct with respect to them) iff they are true, accurate, or (in the case of descriptive reference) satisfied of, to, or by their objects. Robust representations are in this way essentially connected with objectivity or the function of objectifying, i.e., getting at (and getting ‘right’) mind-independent objects —more on this in the next section.

Before we move on, however, we want to highlight the fact that Burge considers primitive agency a necessary condition for robust representation. Because Burge spills a fair amount of ink denying that sensory registrations that guide action are sufficient for robust representation, it’s easy to miss the point about the importance of robust representations’ action-guiding role. Nonetheless, by Burge’s own light, action-guidance is essential for robust representation. Where trivial veridicality conditions can get a hold on just about anything (Burge’s example (2010, p. 292) is an automatic water pump) non-trivial veridicality conditions only get a hold on the states of agents—the sorts of things that can have perspectives. Plausibly, this is because agency is a necessary ingredient for subjectivity. An agent has needs and acts to satisfy those needs: causal impacts from objects in the world get connected with movements appropriate (for the organism, given its needs) to their causes. It is in virtue of this that things in the world can mean something to, or have significance for, agents. In this way, agency is a condition for there being a subject for which things in the world can be objects.

While necessary, Burge denies that primitive agency and the action-guiding role of sensory states are sufficient for robust representation. Indeed, on his view sensory states that merely guide action are paradigmatic deflationary representations. For while primitive agency gets subjectivity into the picture, Burge thinks that it isn’t enough to get objectivity into the picture. And until objectivity enters the scene, you don’t really have aboutness. So, you don’t really have representation. Burge thinks the perceptual constancy mechanisms are what takes a creature from merely sensing and acting to properly perceiving, i.e., representing an objective subject-matter.

In what follows, we’ll argue that perceptual constancy mechanisms, as described by perceptual science, do not provide strong evidence of sensory systems having objectifying functions in the way Burge requires. For as much as Burge wants to de-intellectualize objective representation, the kind of objectivity he requires remains quite demanding: the constancy mechanisms must be understood as functioning to correctly characterize (descriptively categorize in terms of a type of attribute) things in the world in terms of objective attributes they in fact instantiate. In other words, the constancy mechanisms will have to establish that the sensory systems that exhibit them are indeed engaged in a kind of “ontological project” rather than just a “sensory-motor project” (Akins, 1996). We take issue with this—as noted, in the next section we argue that constancy mechanisms most likely are not up to the task. We also take issue with the idea that primitive agency gets subjectivity into the picture without bringing objectivity with it. You don’t need ontology for objectivity—for directedness upon the objective world. Subjectivity and objectivity come in species or degrees, and they come together: a creature’s capacity to act in ways that are sensitive to the nature of the things they’re directed at suffices for a minimal kind of objectivity.

Accordingly, we’ll argue that the action-guiding role of sensory states should be considered a sufficient condition for robust representation. At the same time, however, we agree with Burge that this role isn’t sufficient for non-trivial veridicality conditions. So, if we’re right that perceptual science does not provide strong evidence that constancy mechanisms have objectifying functions, this empirical avenue for defending orthodox representationalism fails. However, we’ll argue that the situation is sunnier for representationalism of the generic kind with which we began this essay, for perceptual representations might be de agendo representations (Springle & Humphreys, 2021). De agendo representations have constitutive appropriateness conditions rather than veridicality conditions; they fit the environmental items and situations they pick-out not by matching or accurately describing or depicting them, but rather by being appropriate to them. In other words, when it comes to de agendo representations, representational success is a matter of practical fittingness.

3.1 Perceptual constancies and objectifying functions

Burge’s claim that perceptual systems have objectifying functions centers on perceptual constancies. According to Burge, the constancies constitute “central instances of perceptual objectification” (2010, p. 397). They are capacities for “separating out” information about how a creature is being affected from information about the world. In other words, the constancies are capacities for getting the objective world right. They determine, via “internalized constraints,” i.e., the perceptual system’s assumptions about the environment, what physical categories (attributes) to posit upon the receipt of various patterns of proximal simulation. They thereby track objective, mind-independent properties of the distal causes of proximal stimuli. With the constancies, Burge thinks that perceptual science has discovered and successfully described capacities to track objective properties of external particulars and, consequently, that science itself directly supports the claim that perceptual systems have objectifying functions.

However, as Green (2023) points out, there is currently no general consensus about what perceptual constancy is. If the constancy mechanisms are objectifying in Burge’s sense, then, presumably, when constancy mechanisms appear to be performing the function that’s responsible for their selection, they in fact faithfully represent properties distal objects and their properties. At the very least, it should be apparent that they are trying to do this. But this isn’t apparent.

Springle (2019) and Hatfield (2009) argue that this does not hold true for size or shape constancy. Studies (in particular, Granrud, 2004) suggest that when perceivers represent the objective size and shape of things they perceive, this is partly a function of judgement, and Buccella (2021), Buccella and Chemero (2022) argues that an alternative overall conception of what constancy is, i.e. one that privileges the tracking of global relational invariants over the extraction of objective properties, is at least equally if not more compatible with available empirical evidence. Moreover, as Springle (2019) notes, cognitively ‘simpler’ creatures still exhibit behaviors suggesting the presence of constancy mechanisms. However, as they appear in simpler creatures, constancy mechanisms don’t look to be the objectifying capacities Burge talks about, even though he claims otherwise. Consider frogs. As Hutto and Myin (2013, p. 119) note, Burge “definitively places frogs (as well as fish and octopi) on the list of perceivers…because they can perceive basic constancies”. Frogs and other anurans (toads) also exhibit the capacity to localize—i.e., determine the direction and distance—of a “distal source of stimuli without serial sampling” (Burge, 2010, p. 427), where Burge considers the capacity to localize the most reliable indicator that any individual is an objectifying perceiver. The question is whether frog perception, by virtue of involving constancy mechanisms and localizing, is thereby objective in Burge’s sense. State-of-the-art science concerning the brain and behavior of anurans suggest that the answer is ‘No.’ According to Neander (2006) the working assumptions of these sciences do not include a commitment to anuran perceptual representation with constitutive veridicality conditions in any obvious or robust way. Instead, they treat anuran perceptual representation as a fundamentally practical matter of tracking appropriate or inappropriate targets for consumption. Chirimuuta (2015) draws a similar conclusion about the nature of the frog’s ‘bug detector’, noting that:

... It would be inappropriate to impose some abstract standard of correctness on the frog’s visual states, such as the precision with which they relate the fly’s shape and contour, or even the ability to deliver internal states that reliably correlate with the presence of specific external objects. (110)

Thus, as Hutto and Myin (2013) also argue, treating constancy mechanisms not as objectifying capacities but as capacities for robust responsiveness—a matter of enabling appropriate action rather than veridical representation–appears to be in line with actual scientific practice. In contrast, Burge’s interpretation appears unparsimonious and overly normatively demanding (Adams & Maher, 2017).Footnote 7

3.2 Categories

Burge might argue that the constancies constitute objectifying capacities by appealing to the categories he understands to be involved in perceptual representation. According to Burge, perceptual systems categorize distal stimuli based on objective attributes such as object shape, color, and size. The constancies are the capacities thanks to which perceptual systems categorize distal stimuli in this way. So, Burge might conclude, the constancies themselves must be after those very same objective attributes. This, in turn, could justify sticking to a traditional understanding of the constancies insofar as one buys into the rest of Burge’s picture.

However, the sorting and classifying done by constancy mechanisms—provided they do it at all—is entirely sub-individual, and the outputs of such sorting and classifying are not available to the individual (for instance: we see unified objects, not collections of unbounded features, even though different features are processed by different mechanisms in the brain. We cannot ‘undo’ feature binding at the individual level, even if we wanted to). Burge assumes that individual-level perceptions involve the very categories involved in the sub-individual processing. Not only is this not obviously a good assumption (Buccella, 2022); it’s not even a good assumption by his own lights. For Burge denies that we can infer robust psychological states from scientific descriptions of sub-individual processes, as these often deploy only deflationary psychological notions. Just as Burge does not accept that perception involves inference in any robust way just because perceptual processing is often described (modeled) in such terms, so we should not accept that perceptual representation in any robust way involves the categories in terms of which the sub-individual processing that leads up to it is often modeled or described.

Now, Burge might respond that the constancies and the categories he attributes to them do show-up at the individual-level. This much seems evident. But the question remains: in what form do they show up at the individual-level? The categories (e.g., objective color, shape), we’ve suggested, are supplied by conceptualization of perceptual phenomenology (e.g., space looking contracted at a distance) which does not itself contain such properties but may instead prompt acts of judgement that deploy such concepts.

And by Burge’s own description, the constancies show up practically, i.e., in the guidance of action (2010, p. 409). The constancies contribute to perception by enabling subjects to respond to categories of things appropriately. But it does not follow from this that perception must therefore represent those very categories as such. So, the way the constancies show up at the individual-level need not include the ‘categories’ referred to in descriptions of the sub-individual processing involved in preparing perceiver responses. We conclude, then, that perceptual science does not support Burge’s contention that perceptual states are objective in a way that entails that they possess non-trivial veridicality conditions. In the next section, we argue that it might nonetheless offer such support for a different kind of representationalism.

4 Practical perceptual representation: a promising program

On the ‘old’ (early modern) view, sense perception provides a fundamentally practical grasp of the world which isn’t well characterized in terms of truth-conditional content but that nonetheless seems to be representational. In this section, we’ll sketch what we call a ‘practical access’ analysis of representation in terms of which contemporary philosophers might conceptualize perceptual content as fundamentally practical. We will also briefly explain why we think that explanations in perceptual science must posit practical representations.

The practical access analysis of representation comes from Springle’s larger “problem-solving framework,” an action-first metaphysical and conceptual framework for understanding intentional phenomena (Springle, 2021; Springle & Humphreys, 2021). The problem-solving framework is action-first in that it treats intentional actions as prior to representations in the order of understanding analysis, meaning that an analysis of representation must mention intentional action or whatever analysis is given for intentional actions. Accordingly, the practical access analysis analyzes representations in terms of intentional actions.Footnote 8

Springle (2021; Springle & Humphreys, 2021) develops and defends an “inverted” theory of intentional phenomena.Footnote 9 According to Springle, representational intentionality or “aboutness” just is or derives from the intentional directedness of actions, rather than, as many philosophers assume, the other way round. Accordingly, the intentional directedness of actions is not a product of being caused and guided by prior intentional mental states. Instead, intentional actions (broadly understood to include mental actions and the actions of non-human animals but excluding the actions of sub-systems) are processes agents produce that function to “solve practical problems” by satisfying a creature’s needs for flourishing. For instance, the action of eating functions to satisfy a creature’s need for nourishment.

According to Springle, the fundamental way creatures represent their environments is in terms of the ways in which they are ready to intentionally respond. For instance, a creature that knows how to do only one thing with an apple, namely, eat it, practically represents a particular apple when it’s the target of a potentiated intentional act of eating it. Other creatures, like humans, have a much richer practical understanding of the apple; a human practically represents the apple in terms of all the actions directed at it that she’s positioned to produce, including acts of judging that it’s an apple, that it’s red, etc. The salient representations will correspond to the actions that serve whichever of the agent’s needs for flourishing are most pressing at the time. Such representations are successful when the potentiated actions that constitute them are appropriate; they misrepresent when the actions are inappropriate. An appropriate action is one that, if it were successfully performed, would non-accidentally solve the practical problem (satisfy the need) it functions to solve.

A full description of Springle’s account of intentional action exceeds what it is possible and necessary to cover in this essay. We can abstract from such details because what’s crucial for our purposes is simply that Springle offers a framework in terms of which it is possible to be a representationalist without being an orthodox representationalist. So, we turn now to the key features of Springle’s account of representation.

First, a representation is something that potentiates an action that is itself causally directed at and supposed to be appropriate with respect to that target. When an agent’s action is potentiated in this sense, we can say that the agent is in a position to respond to the target at which the action is directed (this does not entail that the action will be performed). Springle refers to being so positioned as having “Practical Access”.

Practical Access: An agent A has practical access to a target T iff A is in a position to respond to T where A is so positioned iff A is primed to produce one or more actions that are directed at T. Finally, A is primed to produce such actions iff her capacities for producing such actions have been activated so as to potentiate instances of such actions.

Representation: R is a representation iff R provides (or is supposed to provide) an agent with practical access.

In other words, a representation is something that is supposed to position an agent to produce an appropriate action with respect to some target. They succeed if they ‘practically fit’ the situation (or target) at which they’re directed, i.e., if they’re appropriate. So, on this analysis, the normative standard essential to representation is appropriateness rather than veridicality. That said, as we’ll see below, veridicality conditions may be analyzed as a special variety of appropriateness conditions.

Philosophers tend to identify representations with things that ‘stand-in-for’ other things in the manner of sentences, maps, models and pictures. But according to Springle, the more basic representational relation is “standing-for,” i.e., the aboutness of an action that’s directed at something for some purpose (solving a practical problem). Standing-in-for is a sub-species of standing-for, and both are a matter of practical access. Representations are things that stand-in-for targets because they are (or are supposed to be) potentiations of actions that stand-for those targets. What distinguishes them is the species of practical access they afford.

There are two key varieties of practical access: practical access (which is unmediated) and mediated practical access. An agent A has practical access (unmediated) to a target T iff A is positioned to produce an action that is a direct response to T, where a response is direct when its immediate target is T. An agent has mediated practical access to a target T iff A is positioned to produce an intentional action that is an indirect response to T, where a response is indirect when its immediate target is something that takes the place of T, i.e., something A acts on directly so as to act on T indirectly. Corresponding to these two kinds of practical access are two species of representation:

De agendo Representation: R is a de agendo representation iff R provides unmediated practical access.

De substituto Representation: R is a de substituto representation iff R provides (or is supposed to provide) mediated practical access.

“De agendo” means “about the action”. De agendo representations are constituted by an agent’s potentiated actions. They are thus embodied representations. They stand-for (but not in-for) the intentional actions of which they are potentiations and ipso facto the targets at which those actions are directed. De agendo representations present agents with targets in terms of the potentiated actions that constitute them. In other words, de agendo contents (potentiated actions) are embodied, non-descriptive modes of presentation. When an agent is in a position to respond to a target, she has a de agendo representation of that target; she grasps the target in terms of the potentiated action corresponding to the response she’s positioned to produce.Footnote 10

“De substituto” means “about that which is substituted,” or “about the thing a stand-in stands in for.” De substituto representations are potentiations (or potentiators) of indirect actions: if they have the right features (and if agents employ them appropriately), agents produce actions that are indirectly directed at the targets for which the de substituto representations stand-in. De substituto representations thus provide practical access to targets that agents cannot practically access (act on) directly. Indeed, at least paradigmatically, such representations are themselves products of intentional actions that function to solve practical problems having to do with limitations to (immediate) practical access. In this sense, they have “derived” intentionality. For example, say a botanist needs to learn to identify some species of orchid that only grows in some remote part of the world she hasn’t had a chance to visit. A photograph of an orchid of the relevant species can substitute for seeing an actual flower: by studying the photo, the botanist can learn how to identify the orchid species without ever directly encountering it (A written or spoken description could also do this, though perhaps not as effectively). Of course, the mere existence of a stand-in will not make it the case that one has mediated practical access to what it stands-in-for. De substituto representations provide mediated practical access to the situations they stand in for only when agents have de agendo representations giving them practical access to de substituto representational vehicles; without (unmediated) practical access, one could not intentionally interact with a stand-in (i.e., a de substituto representational vehicle) as such.Footnote 11

De agendo and de substituto representational contents constitute different ways of being intentionally directed at things (in particular, targets). What makes them different is not (at least not primarily) the types of things they intentionally relate one to, but rather the way in which they relate one to them. Take for example an action of eating an apple. Both species of representation can represent such an action, including the apple at which it is directed, but they do so in different ways—by virtue of standing in different kinds of relations to the action (and apple). A de agendo representation is a potentiation of (the process of) an act of eating that’s directed at the apple and that constitutes the agent’s being in a position to eat the apple; the agent de agendo represents the apple in terms of her readiness to eat it. A de substituto representation stands-in-for the action by naming or describing it, and likewise the apple, and thereby potentiates actions (such as theoretical, social, and mental actions) that are indirectly directed at the process or event involving the agent, the apple, and the agent’s eating of the apple. For example, such a de substituto representation may position an agent to answer the question “what is that person doing?” or (if the agent herself ate the apple), “what is it that you’re eating?” or to later plan her afternoon a bit by reasoning, say ten minutes after the event, that she probably won’t be hungry for lunch right away, etc.

The different ways these two species of content are about the things they represent is reflected in their success conditions. Since de agendo contents do not represent by standing- in-for things (actions and situations), but rather by being potentiations of actions that are themselves intentionally (teleologically) directed at targets, they have constitutive appropriateness conditions, but not constitutive veridicality conditions. In contrast, because de substituto contents do represent by standing-in-for things (actions and situations), they do have constitutive veridicality conditions, only understood as an aspect of their appropriateness conditions. This is because whether the indirect action they potentiate is appropriate depends on how they stand-in-for a target and whether they do so “faithfully” or “truly” in the ways required for the indirect action to (non-accidentally) achieve its end.

Springle’s “practical access” framework provides a way for representationalists to reject the common assumption, to which Burge and orthodox representationalists more generally are explicitly committed, that the essential feature of representations are veridicality conditions. This is because it allows us to conceptualize perceptual representations as de agendo representations which have constitutive appropriateness rather than veridicality conditions. We expect that some reader may have an impulse to object that de agendo representations aren’t really representations. But on what grounds? If it’s that they don’t have constitutive veridicality conditions, the objection just begs the question.

De agendo representations have a legitimate claim to being called representations. Not only are they part of a unified analysis of representation as practical access, but they also have classical intentional properties. First, they have the property of being able to exist without their objects existing, i.e. what Brentano called “intentional inexistence” (Brentano, 1874). They represent the actions of which they are potentiations, and potentiated actions are not yet and may never be actualized (fully or even partially performed). Second, they have constitutive success conditions, namely, appropriateness conditions. And this means that de agendo representations pass a common test for counting as representations (Dretske, 1995; Ramsey, 2007): they can misrepresent. De agendo representations misrepresent when inappropriate, i.e., when the actions of which they are potentiations, and which constitute the way a creature practically grasps a situation (target), are inappropriate. Accordingly, Springle’s framework allows us to distinguish orthodox representationalism from the generic variety of representationalism with which this essay began—what we referred to as the ‘old’ idea; the latter does not necessarily collapse into the former. And this is a good thing, for representationalism is empirically defensible even if orthodox representationalism is not (Springle, 2019).

5 Conclusion

It is uncontroversial that perceptual science is fully committed to perception having a practical, action-guiding function. We have argued that this function does not necessitate representational states with non-trivial veridicality conditions, that there’s empirical reason to think that producing such representations is not the essential function of perceptual systems. All of this speaks against perceptual representations with non-trivial veridicality conditions, as orthodox representationalism requires. But it is entirely compatible with non-trivial appropriateness conditions., i.e., with de agendo perceptual representations. Indeed, the reader might recall (Sect. 3.1) that although Neander (2006, p. 184) notes that while neuroethologists rarely invoke veridicality conditions in describing frog visual capacities, they do invoke standards of appropriateness.

We agree with Burge that robust representation is what distinguishes the subject matter of the psychological sciences from other sciences (Burge, 2010, p. 9). Representations figure in explanations of the individual-level behavior of individual organisms and social groups of organisms. To explain behavior in terms of perceptual representations is to explain behavior in terms of a creature’s sensory perspective—how that creature takes the world to be due to facts about its own needs, sensory system, current and typical environment, individual learning history, etc. Where orthodox representationalism goes wrong is in thinking that veridicality conditions are the essential mark of robust representation, i.e., that if veridicality conditions are eliminable, so are perceptual representations. For the fact that we can eliminate representations with constitutive veridicality conditions from psychological explanations doesn’t entail that we can eliminate representations in toto.

So, while this essay agrees with contemporary criticisms of orthodox representationalism, and especially Burge’s version, it parts ways with most critics, as they have endorsed anti-representationalism full-stop. We think these critics are right to emphasize, with Akins, Descartes, Malebranche, and Helmholtz, that perception’s business is sensory-motor as opposed to ontological. But by our lights, anti-representationalists are no more justified in throwing out perceptual representationalism altogether than orthodox representationalists are justified in attributing (dumbed-down) intellectual powers to perceptual systems. This is because, unlike accounts of ‘practical’ perceptual representations that remain committed to orthodox representationalism, thereby turning the sensory-motor project into a just another type of ontological project, de agendo representations allow us to treat the sensory-motor project as essentially representational without over-intellectualizing perception.

On the proposed view, orthodox representationalism mistakenly treats perceptual representations as de substituto representations when they are actually de agendo representations. Perception positions creatures to directly practically respond to their environments. In the case of rational creatures, some of these practical responses are or involve acts of conceptualizing and ontologizing their environments by producing or affirming de substito representations. But perceptual representations themselves are still de agendo, and can only instruct such responses in rational, self-conscious creatures who possess intellectual, de subsituto representational capacities.

The point of this programmatic discussion of de agendo representations is not to establish that this is the right way to think about practical perceptual representations; only that it’s a promising one. The real point of this essay is to encourage philosophers to liberalize their conception of representation without deflating it, and to explore the possibility that robust perceptual representation might be a purely practical affair.