This article focuses on the problem of representational content. Accounting for representational content is the central issue in contemporary naturalism: it is the major remaining task facing a naturalistic conception of the world. Representational content is also the central barrier to contemporary cognitive science and artificial intelligence: it is not possible to understand representation in animals nor to construct machines with genuine representation given current (lack of) understanding of what representation is. An elaborated critique (...) is offered to current approaches to representation, arguing that the basic underlying approach is, at root, logically incoherent, and, thus, that standard approaches are doomed to failure. An alternative model of representation - interactivism - is presented that avoids or solves the problems facing standard approaches. Interactivism is framed by a version of functionalism, and a naturalization of that functionalism completes an outline of a naturalization of representation and representational content. (shrink)

This paper sets out a view about the explanatory role of representational content and advocates one approach to naturalising content – to giving a naturalistic account of what makes an entity a representation and in virtue of what it has the content it does. It argues for pluralism about the metaphysics of content and suggests that a good strategy is to ask the content question with respect to a variety of predictively successful information processing (...) models in experimental psychology and cognitive neuroscience; and hence that data from psychology and cognitive neuroscience should play a greater role in theorising about the nature of content. Finally, the contours of the view are illustrated by drawing out and defending a surprising consequence: that individuation of vehicles of content is partly externalist. (shrink)

Carey has argued that there is a system of core numerical cognition – the analog magnitude system – in which cardinal numbers are explicitly represented in iconic format. While the existence of this system is beyond doubt, this paper aims to show that its representations cannot have the combination of features attributed to them by Carey. According to the argument from abstractness, the representation of the cardinal number of a collection of individuals as such requires the representation of individuals as (...) such, and this in turn requires non-iconic format, from which it is concluded that the explicit representation of the cardinal number of some individuals requires non-iconic representational format. In support of the first premise, an account is given of what approximate cardinal numbers might be, and in support of the second, a direct argument is articulated and defended. Finally, in response to an objection, a second argument for the central thesis is provided. While the discussion is couched in the terms of Carey’s work, the considerations it adduces are perfectly general, and the conclusion should therefore be taken into consideration by all those aiming to characterize the AM system. (shrink)

This book defends a novel view of mental representation—of how, as thinkers, we represent the world as being. The book serves as a response to two problems in the philosophy of mind. One is the problem of first-personal, or egocentric, belief: how can we have truly first personal beliefs—beliefs in which we think about ourselves as ourselves—given that beliefs are supposed to be attitudes towards propositions and that propositions are supposed to have their truth values independent of a perspective? The (...) other problem is how we can think about nonexistents given the widespread view that thought essentially involves a relation between a thinker and whatever is being thought about. The standard responses to this puzzle are either to deny that thought is essentially relational or to insist that it is possible to stand in relations to nonexistents. This book offers an error theory to the problem. The responses from this book arise from the same commitment: a commitment to treating talk of propositions—as the things towards which our beliefs are attitudes—as talk of entities that actually exist and that play a constitutive and explanatory role in the activity of thought. (shrink)

In virtue of what does a representational state have the content it does? Several philosophers have recently proposed that a representational state gets its content from its biological function. After explaining the sense of biological function used in these views, I criticise the proposal. I argue that biological function only determines representational content up to extensional equivalence. I maintain that this holds even if biological function is defined in terms of an intensional notion like (...) Sober's "selection for". (shrink)

I consider the question of whether success-linked theories of content – theories like those of Ramsey (1927), Millikan (1984) and Blackburn (2005) which take there to be a definitional link between representational content and behavioral success – are consistent with the plausible claim that we can use content-attributions to explain behavioral success. Peter Godfrey-Smith (1996) argues that success-linked theories of content are too closely linked to success to be able to explain it. Against this, I (...) present a plausible account of how content-attributions make available good explanations of behavioral success, and argue that if we want our content-attributions to be able to do this explanatory work, then we actually need to embrace a success-linked theory of content. (shrink)

In the pair of articles of which this is the first, I shall present a set of problems and philosophical proposals that have in recent years been associated with the term “relativism”. All these problems and proposals concern the question of how we should represent thought and speech about certain topics. The main issue here is whether we should model such mental states or linguistic acts as involving representational contents that are absolutely correct or incorrect, or whether, alternatively, their (...) correctness should be thought of as varying with some (more or less surprising) factor. In this, first, article, I shall discuss the general issue of relativism about representational content. I shall claim that there are legitimate ways of attributing contents that are absolute truth-bearers, and there are also equally legitimate ways of attributing relativistic representational contents. In the companion piece “Relativism 2: Semantic Content”, I look in more detail at the more specific question whether semantic contents (i.e. the contents assigned to linguistic utterances in the semantics of natural language) should be construed in an absolutist or a relativist way. (shrink)

Let us assume that some organisms, humans at least and the other higher animals, have internal states and behavioral states that represent things external to themselves. One of the questions that everyone would like answered about these states is: In virtue of what does such a representational state get the specific content that it has? An answer to this question that’s popular just now is: In virtue of its biological function. I believe there is a deep reason why (...) such an answer can’t work. I shall present that reason in this paper.First, we need to be clear about the notion of biological function that’s being appealed to here. In particular, we need to distinguish two fairly different ways of thinking about function. We could think of something’s function as what it contributes to the present functioning or output of some complex system of which it is a part. Roughly, this is Cummins’ view in (Cummins 1984). (shrink)

In this essay, I specify types of representational content that can be attributed to Kant’s account of representation. The more specific aim is to examine which of these types of content can be regarded as possible without the application of concepts. In order to answer the question, I proceed as follows. First, I show how intuition (in Kant’s sense) can be seen as providing indexical content independently of empirical concepts. Second, I show in what sense the (...) generation of spatial content can be regarded as non-categorial. A key distinction is that a perceptual examination of an object can be understood as thoroughly sensible and particular, whereas a conceptual determination always grasps the object via its generalisable features. Third, I propose that the faculties of sensibility and understanding are not only separable in principle, but that their contributions remain in a certain sense separate in actual cognition as well. This is to say that a conceptual determination of an object does not entail that the object ceases to be non-conceptually available to the perceiver, which further suggests the autonomy of sensibility and its perceptual content. Finally, I raise difficulties in attributing non-conceptual representational content to Kant’s judgment-centered stance on representation and experience, only to emphasise how these difficulties easily lead to a misappreciation of Kant’s fundamental distinction between sensibility and understanding and their unique cognitive contributions. (shrink)

Two debates loom large in current discussions on phenomenal consciousness. One debate concerns the relation between phenomenal character and representational content. Representationalism affirms, whereas “content separatism” denies, that phenomenal character is exhausted by representational content. Another debate concerns the relation between phenomenal consciousness and cognitive access. “Access separatism” affirms, whereas, e.g., the global workspace model denies, that there are phenomenally conscious states that are not cognitively accessed. I will argue that the two separatist views are (...) related. Access separatism supports content separatism by undermining the most prominent sort of arguments in favor of representationalism, namely ones that appeal to the phenomenology of perceptual experiences. (shrink)

How can it be true that one sees a lake when looking at a picture of a lake, since one's gaze is directed upon a flat dry surface covered in paint? An adequate contemporary explanation cannot avoid taking a theoretical stand on some fundamental cognitive science issues concerning the nature of perception, of pictorial content, and of perceptual reference to items that, strictly speaking, have no physical existence. A solution is proposed that invokes a broadly functionalist, naturalistic theory of (...) perception, plus a double content analysis of perceptual interpretation, which permits non-supervenient, culturally autonomous modes of reference to be generated and artistically exploited even in a purely physical world. In addition, a functionalist concept of broad or 'spread' reference replaces the traditional precise intentional concept of reference, which previously made reference to non-existent items theoretically intractable. (shrink)

Tracking representationalism is the theory that phenomenal consciousness is a matter of tracking physical properties in an appropriate way. This theory holds that phenomenal character can be explained in terms of representational content, and it also entails that there is unlikely to be a strong correlation between phenomenal character and neural states. However, the empirical evidence shows that both claims cannot be true. So, tracking representationalism is wrong. Its fault is due to ignoring the internal correlation of experience, (...) the existence of which shows that phenomenal character is shaped by neural states to a large extent, so it cannot be wholly explained by representational content. (shrink)

Especially in the appendix to the opening part of his Ethics, Spinoza discusses teleology in a manner that has earned him the status of a staunch critic of final causes. Much of the recent lively discussion concerning this complex and difficult issue has revolved around the writings of Jonathan Bennett who maintains that Spinoza does, in fact, reject all teleology. Especially important has been the argument claiming that because of his basic ontology, Spinoza cannot but reject thoughtful teleology, that is, (...) teleology involved in the actions of conscious cognitive beings who have thoughts of future states of affairs. For Spinoza, a particular idea is a modification of the thinking substance the object of which is a certain modification of the extended substance, and Bennett’s central argumentative move is to claim that there is no room in Spinoza’s system for a key ingredient in thoughtful teleology, the tenet that representative content of ideas is causally efficacious. In what follows, I begin by presenting Bennett’s argument. As his position has received much criticism, I then take up the ways in which it has been discussed and found wanting. I think that Bennett’s position really is something that should not be endorsed; however, and despite the lively discussion, it also seems to me that there is more to be said about what is at stake here. Thus, I aim at offering an analysis of the nature of Bennett’s argument and the ensuing discussion with the aim of discerning the philosophical source from which Bennett’s interpretation draws its force. (shrink)

Teleosemantics attempts to explain the content of mental representations through an appeal to functions, and typically attributes function to selection history. The narrowest cases focus on only evolutionary fitness benefit through natural selection, while broader theories have come to accept multiple levels of selection, including those over the course of a lifetime such as neural selection. The precise way to define function has given rise to many debates over the content of hypothetical mental representations. In this paper, I (...) argue that defining function through the lens of Nikolas Tinbergen’s levels of analysis provides a valuable framework in organizing and analyzing teleological theories of function, and can advance the debate into a more organized discourse. In particular, after defining and defending them on theoretical grounds, I go through three classic teleosemantic debates about content ascription and function and attempt to organize them in a Tinbergian framework in order to demonstrate their general applicability, followed by two more examples from humans and primates in order to give insight on the role of each level of analysis in cases where we have more direct internal access to the content of such representations. I finally suggest that a pluralist viewpoint of the content of representations in complex organisms may be justified. I argue that in light of Tinbergen’s levels of analysis framework, teleosemantic theories of mental content can be more clearly discussed and advanced. (shrink)

Steve Stich used to be an eliminativist. As far as I can tell, he renounced eliminativism about the time that he moved from the west to the east pole.1 Stich was right to reject eliminativism, though I am not convinced that he rejected it for the right reasons. Stich 1983 contains a comprehensive attack on representational content, a central feature of both folk psychology and the Representational Theory of Mind, the leading philosophical construal of scientific psychology. Stich’s (...) current position on the role of content in psychological explanation is not entirely clear. One of my aims in this chapter is simply to invite Stich to clarify his views on representational content; the question that forms the title of this chapter is therefore addressed directly to Stich. I begin by sketching his original anti-content argument. I then trace some later developments in his thinking about content. I argue that content does play an important role in scientific psychology, for precisely the reasons that Stich identified in his original argument against content. I conclude with some general remarks on eliminativism. (shrink)

Questions concerning the nature of representation and what representations are about have been a staple of Western philosophy since Aristotle. Recently, these same questions have begun to concern neuroscientists, who have developed new techniques and theories for understanding how the locus of neurobiological representation, the brain, operates. My dissertation draws on philosophy and neuroscience to develop a novel theory of representational content.

Normative standards are often applied to emotions. Are there normative standards that apply to emotions in virtue solely of facts about their nature? I will argue that the answer is no. The psychological, behavioural, and neurological evidence suggests that emotions are representational brain states with various kinds of biological functions. Facts about biological functions are not (and do not by themselves entail) normative facts. Hence, there are no nor- mative standards that apply to emotions just in virtue of their (...) having various kinds of biolog- ical functions. Moreover, the peculiar features of emotions make the view that representational content is essentially normative very implausible. Hence, the representational properties of emotions cannot be seen as entailing normative standards. The conclusion is that there are no normative standards that apply to emotions solely in virtue of their nature. Ó 2005 Elsevier Ltd. All rights reserved. (shrink)

지각 경험의 내용이 전적으로 개념적 내용으로만 구성되지 않음을 주장하는 비개념주의의 이론 중 하나는 동물과 같은 비인간적 존재들의 지각 경험을 근거로 지각적 표상 내용의 비개념적 측면이 있음을 강조하는 것이다. 비개념주의자 피콕과 헐리에 따르면, 동물과 같은 존재들은 우리 인간과 마찬가지로 지각 경험을하는 것이 가능하고, 우리와 동일한 표상적 내용을 갖는다. 하지만 이들은 우리가 갖는 개념 혹은 개념적 능력을 소유하고 있지않기 때문에 이들의 지각적 표상의 내용은 비개념적 내용에 해당된다. 만일 이들의 주장이 옳다면, 비인간적 존재들의 지각 경험과 그러한 경험을 구성하는 내용은 특정한 방식으로 증명될 수 (...) 있고, 지각 경험의 표상적 내용은 지각의 주체가 소유한 개념 혹은개념적 능력으로 부터 독립되어 구성된다는 것 또한 증명될 수 있다. 본 논문은 이러한 접근 방식이 갖는 두 가지 문제를 지적한다. 첫째, 비인간적 존재들의 내적 경험을 ‘지각의 가능조건(an enabling condition of perception)’만으로 판단 할 수 없다는 것이고 둘째는, 지각 주체와 대상사이의 ‘지향적 인과관계(intentional causation)’의 성립을 명확히 밝힐 수 없다는 점이다. 이로부터 본 논문은 지각 경험의 내용의 개념성과 비개념성을판단하는 데 있어 동물의 지각적 표상의 가능성을 전제로 논의하는 것은 ‘마치 그럴듯한 귀속(as-if attribution)’에 기반한 논증임을 주장할 것이다. (shrink)

This chapter contains sections titled: I II III IV V References.

In Knowledge and the Flow of Information, Fred Dretske explains representational content by appealing to natural indication: a mental representation has its content in virtue of being a reliable natural indicator of a particular type of state of the world. His account fails for several reasons, not the least of which is that it cannot account for misrepresentation. Recognizing this, Dretske adds a twist in his more recent work on representational content (sketched in 'Misrepresentation' and (...) elaborated in Explaining Behavior): a mental representation acquires its semantic content when the fact that it naturally indicates some type of state of the world acquires explanatory relevance. This shift in emphasis from natural indication to function (and, so, to behaviour) is intended to address the disjunction problem. Whether or not it succeeds (I argue that it does not), new problems emerge as a direct result of the shift to emphasis on behaviour. First, although Dretske appeals to appropriate response behaviours for particular types of representations, it is not at all clear that such behaviours exist. Second, Dretske's theory cannot account for representations that never figure in the behaviour of the organism of which they are a part. In response to these problems, I suggest two possible routes to recovery: one, an appeal to implicit beliefs (or dispositions); and, two, an appeal to compositionality. Although it seems clear that Dretske must enlist one of these options if he is to formulate a successful causal naturalist theory of representational content, given his theoretical commitments, neither option is readily available to him. (shrink)

In giving an account of the content of perceptual experience, several authors, including Fred Dretske, Gareth Evans, Christopher Peacocke, and Michael Tye, have employed the notion of nonconceptual representational content.[1].

Michael Tye argues for two crucial theses: (1) that experiences of pain have representational content (essentially); (2) that the representational content can be specified in terms of something like damage in parts of the body. (Different types of pain are connected with different types of damage.) I reject both of these theses. In my view experiences of pain carry nonconceptual content, but do not represent essentially. Rather they are apt to represent when the subject attends (...) to them. The experiences carry nonconceptual content not only about tissue damage, but about many other qualities as well, including dispositional qualities. (shrink)

This paper aims at answering the question “When is informational content explicitly represented in a cognitive system?”. I first distinguish the explicitness this question is about from other kinds of explicitness that are currently investigated in philosophy of mind, and situate the components of the question within the various conceptual frameworks that are used to study mental representations. I then present and criticize, on conceptual and empirical grounds, two basic ways of answering the question, the first one coming from (...) the classical computational theory of mind, the latter one issued from a procedural conception of informational contents. I then argue for a new answer to the initial question, an answer that retains some valuable insights of the criticized theories, but which underlines the importance and the interpretational source of the distinctiveness proper to explicitly represented contents. (shrink)

Despite the fact that reproduction of experiments by peers has traditionally been regarded as of the utmost importance in enabling the intersubjectivity of scientific practice, reproductions may yield discordant results and deciding which result should be favored may not be an easy task. According to Harry Collins, experimental disagreement is resolved by the action of social, political and economic factors, but not by means of epistemic and scientific, or so-called internal reasons. His motivation for such a claim is the presence (...) of an infinite regress at the core of experimental activity that, according to him, cannot be stopped with scientific resources alone: the experimenters’ regress. The goal of this... (shrink)

Recent discussions in the philosophy of psychology have examined the use and legitimacy of such notions as “representation”, “content”, “computation”, and “inference” within a scientific psychology. While the resulting assessments have varied widely, ranging from outright rejection of some or all of these notions to full vindication of their use, there has been notable agreement on the considerations deemed relevant for making an assessment. The answer to the question of whether the notion of, say, representational content may (...) be admitted into a scientific psychology has often been made to hinge upon whether the notion can be squared with our “ordinary” or “folk” style of psychological explanation, with its alleged commitment to the idiom of beliefs and desires. In this article, I proceed the other way around, starting with experimental psychology itself and asking whether such notions as the above play a legitimate role within a particular area of contemporary theory, the psychology of perception. My conclusion will be that especially the first three have a legitimate place in theories of perception, albeit one that differs from that ascribed to them in a significant portion of the philosophical literature. Along the way I develop a notion of noncognitive functional analysis, which separates the notion of contentful perceptual processing and perceptual representation from cognitive or conceptual content. (This paper has been reprinted, with stylistic revisions, in Perception and Cognition, Essays in the Philosophy of Psychology, Oxford, Clarendon Press, 2009, pp. 50-87.). (shrink)

The on-going debate over the ‘admissible contents of perceptual experience’ concerns the range of properties that human beings are directly acquainted with in perceptual experience. Regarding vision, it is relatively uncontroversial that the following properties can figure in the contents of visual experience: colour, shape, illumination, spatial relations, motion, and texture. The controversy begins when we ask whether any properties besides these figure in visual experience. We argue that ‘ensemble properties’ should be added to the list of visually admissible properties. (...) Ensemble properties are features that belong to a set of perceptible objects as a whole as opposed to the individuals that constitute that set. They include such features as the mean size of an array of shapes or the average emotional expression of an array of faces. Recent work in vision science has yielded compelling evidence that the visual system routinely encodes such properties. We argue that epistemological considerations provide strong reasons to think that these properties figure in visual experience. Judgements about ensemble properties are immediately warranted by our perceptual experience, and the only plausible way that a perceptual experience could confer this warrant is if it confers awareness of ensemble properties. (shrink)

Nicholas Shea offers Varitel Semantics as a naturalistic account of mental content. I argue that the account secures determinate content only by appeal to pragmatic considerations, and so it fails to respect naturalism. But that is fine, because representational content is not, strictly speaking, necessary for explanation in cognitive science. Even in Shea’s own account, content serves only a variety of heuristic functions.

In this paper, we introduce a novel difficulty for teleosemantics, viz., its inability to account for what we call unexploited content—content a representation has, but which the system that harbors it is currently unable to exploit. In section two, we give a characterization of teleosemantics. Since our critique does not depend on any special details that distinguish the variations in the literature, the characterization is broad, brief and abstract. In section three, we explain what we mean by unexploited (...) content, and argue that any theory of content adequate to ground representationalist theories in cognitive science must allow for it.1 In section four, we show that teleosemantic theories of the sort we identify in section two cannot accommodate unexploited content, and are therefore unacceptable if intended as attempts to ground representationalist cognitive science. Finally, in section five, we speculate that the existence and importance of unexploited content has likely been obscured by a failure to distinguish representation from indication, and by a tendency to think of representation as reference. (shrink)

A promising strategy for defending the role that representation plays in explanations of cognition frames the concept in terms of internal models or map‐like mechanisms. “Structural representation” offers an account of representation that is grounded in well‐specified, empirical criteria. However, anti‐representationalists continue to press the issue of how to account for the paradigmatic semantic properties of representation at the subpersonal level. In this paper, I offer an account of how the proponent of structural representation should think about content. There (...) are really two problems of content any account of representation must overcome: the “hard problem of content” and the “content determination problem.” I argue that the structural representation account has the resources to provide a naturalistically respectable notion of content that overcomes both problems. (shrink)

A number of authors have recently cited phenomenal effects of covert attention as a source of objection to representationalism. These authors maintain that covert attention brings about changes to phenomenology that cannot be explained by changes in representational content. This paper deals with two related issues that are central to this debate: (1) how attention interacts with representational content, and (2) how variations in the precision or determinacy of representational content should be incorporated into (...) representationalist accounts of perceptual phenomenology. I address these issues in the course of responding to several recent anti-representationalist arguments that have been advance by Ned Block. I argue that attention can increase the precision of range content and that this provides a basis for defusing both the concerns raised by Block and more general concerns related to covert attention. I conclude by situating my response with respect to analogous proposals that have been developed around the determination relation. (shrink)

Informational theories of semantic content have been recently gaining prominence in the debate on the notion of mental representation. In this paper we examine new-wave informational theories which have a special focus on cognitive science. In particular, we argue that these theories face four important difficulties: they do not fully solve the problem of error, fall prey to the wrong distality attribution problem, have serious difficulties accounting for ambiguous and redundant representations and fail to deliver a metasemantic theory of (...) representation. Furthermore, we argue that these difficulties derive from their exclusive reliance on the notion of information, so we suggest that pure informational accounts should be complemented with functional approaches. (shrink)

The current essay introduces the guidance theory of representation, according to which the content and intentionality of representations can be accounted for in terms of the way they provide guidance for action. The guidance theory offers a way of fixing representational content that gives the causal and evolutionary history of the subject only an indirect role, and an account of representational error, based on failure of action, that does not rely on any such notions as proper (...) functions, ideal conditions, or normal circumstances. Moreover, because the notion of error is defined in terms of failure of action, the guidance theory meets the “meta-epistemological requirement” that representational error should be potentially detectable by the representing system itself. In this essay, we offer a brief account of the biological origins of representation, a formal characterization of the guidance theory, some examples of its use, and show how the guidance theory handles some traditional problem cases for representation: the representation of fictional and abstract entities. Being both representational and actiongrounded, the guidance theory may provide some common ground between embodied and cognitivist approaches to the study of the mind. (shrink)

A central idea in Ruth Millikan’s biosemantics is that a representation’s content is restricted to conditions required for the normal success of actions that it has as its function to guide. This paper raises and responds to a problem for this idea. The problem is that the success requirement seems to block us from saying that epistemic modal judgments represent our epistemic circumstances. For the normal success of actions guided by these judgments seems to depend on what is actually (...) the case, not on whether or to what extent various possibilities were supported by our evidence. In response, I argue, first, that actions guided by epistemic modal judgments have as their function to implement strategies for handling epistemic circumstances, second, that the successful performance of this function requires that aspects of these circumstances obtain, and, third, that biosemantics can thus understand epistemic modal judgments as representing these aspects. The recognition of such strategic contents introduces complications; I further argue that these are benign. (shrink)

Recent discussions in the philosophy of psychology have examined the use and legitimacy of such notions as ‘representation’, ‘content’, ‘computation’, and ‘inference’ within a scientific psychology. While the resulting assessments have varied widely, ranging from outright rejection of some or all of these notions to full vindication of their use, there has been notable agreement on the considerations deemed relevant for making an assessment. The answer to the question of whether the notion of, say, representational content may (...) be admitted into a scientific psychology has often been made to hinge upon whether the notion can be squared with our ‘ordinary’ or ‘folk’ style of psychological explanation, with its alleged commitment to the idiom of beliefs and desires. The emphasis on ‘folk psychology’ in arguments concerning the status of various concepts within scientific psychology is unfortunate, because it runs afoul of the ideal that the philosophy of psychology, by analogy with the philosophy of biology and the philosophy of physics, should have at its core the discussion of ongoing psychological theories. For while it may be admitted that the status of appeals to ‘folk psychology’ in order to justify a role for the belief-desire idiom within psychology proper is a matter that is of interest in its own right, it is not at all clear that mainstream experimental psychology has adopted either the language or the conceptions of ‘belief-desire’ psychology. (shrink)

A straightforward way of thinking about perception is in terms of perceptual representation. Perception is the construction of perceptual representations that represent the world correctly or incorrectly. This way of thinking about perception has been questioned recently by those who deny that there are perceptual representations. This article examines some reasons for and against the concept of perceptual representation and explores some potential ways of resolving this debate. Then it analyzes what perceptual representations may be: if they attribute properties to (...) entities, what are these attributed properties, and what are the entities they are attributed to. (shrink)

If connectionism is to be an adequate theory of mind, we must have a theory of representation for neural networks that allows for individual differences in weighting and architecture while preserving sameness, or at least similarity, of content. In this paper we propose a procedure for measuring sameness of content of neural representations. We argue that the correct way to compare neural representations is through analysis of the distances between neural activations, and we present a method for doing (...) so. We then use the technique to demonstrate empirically that different artificial neural networks trained by backpropagation on the same categorization task, even with different representational encodings of the input patterns and different numbers of hidden units, reach states in which representations at the hidden units are similar. We discuss how this work provides a rebuttal to Fodor and Lepore's critique of Paul Churchland's state space semantics. (shrink)