On Resemblance



In his book Sprachen der Kunst Nelson Goodman develops an account of representation. This relation between objects on Goodman’s view, has different modes. The most basic such modes are denotation and exemplification. In the early stages of his argument, he rules out resemblance as a possible criterion for representation. He does this almost en passant in his work, but his critique of resemblance accounts of representation, more specifically of pictorial representation, has sparked great discussion in its wake. Ruling out resemblance as a necessary or sufficient condition for representation poses a problem for traditional accounts of resemblance going back to Plato. This article will introduce Christina Abell’s account of a resemblance theory of depiction as a solution to the problem posed by Goodman. In a first section I am going to briefly present Goodman’s critique of resemblance as either a necessary or sufficient condition for depiction. In section II, two responses to the problem will be discussed. Section III evaluates how successful the responses are with a focus on Abell’s account of a resemblance theory of depiction.

I. Goodman’s critique of resemblance as a criterion for representation

The main problem Goodman sees with representation as a relation between objects based on resemblance is that resemblance is symmetrical in nature. If we assume a naive view of resemblance accounts of depiction the following holds: A represents B if and only if A clearly resembles B. This, according to Goodman, cannot be sufficient for a distinct relation of representation for two reasons. First, two cars in an assembly line are close to perfect copies of another and therefore clearly resemble another, but neither of them refers to or represents the other. Second, all things resemble themselves, resemblance is reflexive and symmetrical. Representations usually are not (Goodman, 15).

If we try to eliminate the problems of the naïve view, by substituting one of the objects with a painting we get the following formula: If A is a picture or painting and clearly resembles B, then A represents B. Again, he identifies two problems with the view. First, if object A, in his example a painting of the Duke of Wellington, resembles object B, the person who was the Duke of Wellington, then it holds that B resembles A as well because of the symmetrical nature of resemblance. Evidently this poses some difficulty because the painting of the Duke resembles all other paintings in an exhibition far more than the Duke himself. It shares with them its texture, shape, size etc. Second, to constrain the entities that fall under B to non-pictorial entities does not solve the problem, because pictures can represent other pictures (Goodman, 16-17). From these considerations, Goodman concludes that resemblance can be neither a necessary nor sufficient condition for representation. The core feature of representation according to him is denotation, a mode of representation that relies on convention.

II. The responses

There are different ways in which this challenge to resemblance accounts of representation can be met. It is important to note that while Goodman discusses various kinds of representation in art, the opponents of his critique discussed here focus exclusively on pictorial representation. In addition, Goodman claims that pictures do not represent actual objects, but propositional content. A picture of a specific horse does not depict the horse, but the fact that x is that specific horse. His opponents however do not elaborate on this point.

II. a) Craig Files

One approach that challenges Goodman’s critique is that of Craig Files. Files proposes two things, first, that the symmetric nature of resemblance does not force us to abandon resemblance as a key component of a theory of pictorial representation; second, that Goodman’s critique fails because he conflates (1) the  conditions that are required for an object to represent and (2) that which confers representational content (Files, 398-399). The basic idea is that there are two questions to be answered about representation. A theory of the nature of representation must tell us by virtue of what a representation-bearer is about something; which property makes it a representation-bearer. A theory of content then must tell us by virtue of what a representation-bearer has content or which properties it picks out for representing (Files, 399-400). These are different questions that allow for independent answers. Files' approach is to claim that Goodman only discusses whether resemblance confers representational content, but conflates this with providing an answer to (1). A theory of the nature of representation, however, is broader and need not rely on resemblance even if a theory of content does. Files now claims that the resemblance theory answers (2) and that resemblance can confer pictorial content. He incorporates resemblance as part of a theory of content into a broader theory of the nature of representation.

For this he leans on Peirce’s theory of representation. Peirce proposes that representation is a triadic relation. It holds between the representation-bearer (e.g. a painting), the representational object (e.g. the Duke of Wellington) and the agent (e.g. a visitor in the museum) through the thought produced by the representation-bearer. The interpretant, the thought that is evoked in the agent by the representation-bearer, links the representational object to the agent by interacting with other mental states and issuing behaviour that conforms or relates to the representational object (Files, 402). Here, we have a theory of the nature of representation that does not rely on resemblance. Now Files needs to answer (2). This is where resemblance comes into play to explain the relation between representation-bearer and representational content. The resemblance in the case of pictorial representation holds between representation-bearer and representational object. They share many intrinsic properties. We can easily recognize depicted objects because of this. The representation-bearer produces a visual experience (the interpretant) in us that resembles the representational object (Files, 403). In this way, the symmetric nature of resemblance is accounted for. This holds only for pictorial representations where there are intrinsic shared properties. In language this does not hold because of the arbitrary relations that constitute it. Here, conventionalism in Goodman’s fashion applies, but it is not necessary for depiction.

Recall that Goodman’s point was that resemblance is ubiquitous and occurs without representation and that therefore resemblance cannot account for the difference between the relation of two cars in an assembly line and the relation between a painting of the Duke and the Duke himself. This, however, is the task of (1) and not of (2). Here, Goodman conflates the two questions. Resemblance only has to account for (2). It follows that resemblance pertains to the question of content, while the triadic nature of representation explains the aboutness of representation-bearers in respect to the conventions about certain objects being representation-bearers (Files, 408).

II. b) Catharina Abell

A further defender of a resemblance view is Catharine Abell. She provides the richest resemblance account on offer, drawing on the notion of experienced resemblance. Abell starts off with a basic distinction. While both figurative and abstract paintings represent, all and only figurative representations depict. It is this latter relation she discusses in this article, and she uses the term 'picture' exclusively to refer to it. As an example, she takes a picture that may symbolize Christ by depicting a lamb. Here, it is the lamb that is depicted, not Christ (Abell, 183).

In a similar fashion to Files, Abell splits the topic of depiction into two questions, one regarding the nature of representation and the other regarding the content of representation.

(a) Metaphysical: what is it for one thing to depict another?

(b) Epistemological: how do we work out that one thing depicts another?

Her goal is to answer both questions, and she supposes that “depiction is governed by aspects of resemblance that are intended by their makers or obtain between picture and object, or are related indirectly to intentions through stylistic conventions” (Abell, 185).

First, she identifies three objections to resemblance accounts that are easily dealt with and have been tackled by Files. First, the symmetrical nature of resemblance is not a problem if resemblance is merely necessary but not sufficient for depiction. Files’ argument works analogously. Second, the ubiquity of resemblance central to Goodman’s car assembly example is dealt with by constraining resemblance to particular aspects. Third, the difficulty of fictional objects facing resemblance views - how can an object resemble a fictional thing - can be met by either a counterfactual resemblance relation or the notion of experienced resemblance (Abell, 188). Abell opts for the counterfactual resemblance relation to deal with the third problem. One objection is not easily met. Dominic Lopes, an accomplished author on the topic of representation, claims that a resemblance account of depiction (RAOD) must meet two conditions, which are mutually exclusive. Abell develops her account of a resemblance view in response to his objection.

Diversity constraint: 
RAOD must accommodate a whole range of pictorial styles and types (realistic and cubist paintings etc.) → there is no single common respect in which they resemble their objects (e.g. shape, colour etc.)

Independence constraint:
RAOD must appeal only to resemblances that interpreters can access without prior knowledge of depictive content. (This applies to accounts that explain both the metaphysical and the epistemological question.)

Lopes’ conclusion is the following: if all types of depiction resemble their objects in the same way, the independence constraint is met, but if that is not the case, then we cannot know how a picture depicts before we know what it depicts, and the independence constraint is not met (Abell, 187-188). Abell now explains why several other resemblance accounts cannot solve this problem. I will not relay this discussion in full but focus on her treatment of the problem. I will solely discuss this in relation to the article previously considered. It seems that Files can accommodate the first constraint, since he does not restrict the types of resemblance properties, but he struggles to meet the second constraint because of the role the agent plays in his view. Abell's solution might, however, prove viable for Files’ account as well.

To solve the diversity constraint, Abell introduces a concept suggested by Robert Hopkins. With regard to pictorial content, Hopkins introduces two levels. The first level is the content of seeing-in. This is determined by experienced resemblance in occlusion shape. It is what I see in the picture. The second level is that of actual pictorial content. It is determined by the content of seeing in and the intentions of the picture’s maker. As an example, Abell uses the drawing of a stick-figure. Here the content of seeing-in is a humanoid shape with a huge head and disproportionately thin body. The pictorial content is what is actually depicted: a human of indistinct shape (Abell, 190). The important conclusions from this are two. First, intentions constrain the contribution that experienced resemblance makes to pictorial content. Second, intentions determine which respects of experienced resemblance contribute to pictorial content. This solves the diversity constraint because now a resemblance view can simply state that different respects of resemblances govern different instances of depiction. They can be constrained to intrinsic, objective features like texture or to objective, non-intrinsic features like occlusion shape or aperture colour or to response-dependent features (Abell, 196). The picture maker’s intentions are key to determining which respects govern a particular instance of depiction. The artist may take colour or shape to be the centre of a piece of art. This constrains the choice of relevant resemblance features.

Abell now faces a second hurdle. The viewer’s knowledge about a picture maker’s possible intentions—and therefore the depictive content implied in her solution to the first problem—indicate that her approach falls short of meeting the second constraint. To solve the independence constraint, Abell turns to Grice’s theory of non-natural meaning. To accommodate the second constraint, Abell must explain how we can guess at the picture maker’s intentions without knowing what the depictive content is. Grice offers exactly this. The idea is simple: Grice states that we can identify a person’s communicative intentions from their communicative behaviour and its products. The ability to do the same regarding pieces of art stems from and is part of this communicative practice and ability (Abell, 198-201). This allows Abell to meet the independence constraint as well as the diversity constraint and solve the problem posed by Lopes.

III. Can resemblance account for depiction?

Both Files and Abell formulate promising answers to the problems posed by Goodman. I will examine them in turn in this last section.

Files correctly identifies the conflation of two crucial—and despite being necessarily connected in some ways—distinct questions about depiction. This allows him to answer both questions individually, using resemblance only to account for a theory of content but not for a theory of the nature of representation. Resemblance, on his view, remains a necessary feature of depiction but is not sufficient for it. Files introduces experienced resemblance to link the representation-bearer to the representational object via the agent. He further proposes that in the case of depiction there are intrinsic resemblances between representation-bearer and representational object. The notion of experienced resemblance in his view does not substitute for the actual resemblance relation regarding depiction. While this argument is convincing, it is possible to account for a broader range of representation if we assume mere experienced resemblance.

Abell, like Files, identifies two central questions for any account of representation. In      this way, she implicitly gives the same argument as Files (a conflation argument) against Goodman. Her theory is superior to Files since she accommodates both constraints introduced by Lopes. It is easy to imagine how Files could account for the diversity constraint, but it is not evident that he has a ready solution for the independence constraint. The position sketched in his article could accommodate the solution Abell offers, but it is not clear whether Files would endorse this. There are several  strengths to Abell’s account. First, like Files, she answers not one but both questions relevant for a theory of representation. Second, the inclusion of the picture maker’s intention as part of pictorial content (following Hopkins) allows her to cleanly answer the diversity constraint. Third, and most importantly she grounds her answer to the independence constraint in Grice’s theory of communication. In this way, she manages to use a basic explanation of an essential feature of human life, communication, to explain depiction relations. This move allows her to explain and justify the strong intuition that resemblance plays at least some part in figural paintings. One difficulty I see with Abell’s approach is her solution to the problem of representation of fictional entities. Contrary to Files, she opts for a counterfactual solution, which I think is not much more than sleight of hand. Excluding experience as a relevant factor, she avoids having to account for the psychological salience of the resemblance relation instead addressing the problem of the nature of actual resemblance to fictional entities only with a counterfactual story. Despite this reservation, her account succeeds in presenting a viable argument for a resemblance theory of depiction. 


Abell, Catharine (2009). «Canny Resemblance», Philosophical Review, 118(2): 183–223. doi:10.1215/00318108-2008-041

Files, Craig (1996). «Goodman’s Rejection of Resemblance», British Journal of Aesthetics, 36: 398–412.

Goodman, Nelson (2019). Sprachen der Kunst: Entwurf einer Symboltheorie. 9. Aufl. Frankfurt a.M.: Suhrkamp.

Hopkins, Robert. (1994). «Resemblance and Misrepresentation», Mind, 103(412), 421–438. doi:10.1093/mind/103.412.421 

Lopes, Dominic (1996). Understanding Pictures. Oxford: Clarendon.