The HIT framework accepts a number of assumptions that are widely held as plausible or even well established in the literature on category-specific agnosia. We point out that a number of these elementary conjectures, now almost taken for granted, have received little in the way of convincing empirical support.

Most current theories of category-specific semantic deficits appeal to the role of sensory and functional knowledge types in explaining patients' impairments. We discuss why this binary classification is inadequate, point to a more detailed knowledge type taxonomy, and suggest how it may provide insight into the relationships between category-specific semantic deficits and impairments of specific aspects of knowledge.

The purpose of the commentary is not to contradict HIT but rather to question whether its increase in predictive power outweighs the decrease in parsimony. For the refutable aspects of HIT, a simpler model for naming appears to achieve as much. Both models better fit the facts concerning naming performance than describe category-specificity.

Humphreys and Forde fail to account for the ontology of the “features” that they claim are constitutive of concepts. This failure is common to decompositional theories of conceptual representation. Category-specific deficits can be better explained by a theory that takes inferential relations among atomic concepts to be the key characteristic of conceptual representation and processing.

Humphreys and Forde conceptualize object representations as structural descriptions, without discussing the implications of structural description models. We argue that structural description models entail two major assumptions – a part-structure assumption and an invariance assumption. The invariance assumption is highly problematic because it contradicts a large body of findings which indicate that recognition performance depends on orientation and size. We will delineate relevant findings and outline an alternative conception.

Although Humphreys & Forde's HIT provides a comprehensive account of category-specific deficits, standard models of categorization and identification may also be able to explain many aspects of such deficits. The assumptions of an exemplar-based account of category- specific deficits are presented, and it is argued that exemplar models may be able to explain key findings on impaired object identification and categorization.

Recent investigations and theorising about category-specific deficits have begun to focus upon patients with progressive brain disease such as semantic dementia and Alzheimer's disease. In this commentary we briefly review what insights have been gained from studying patients of this type. We concentrate on four specific issues: the sensory/functional distinction, correlation between features, neuroanatomical considerations, and confounding factors.

Humphreys and Forde (H&F) propose that greater within- category structural similarity makes living things more difficult to name. However, recent studies show that normal subjects find it easier to name living than nonliving things when these are matched across category for potential artefacts. Additionally, at the level of single pixels, visual overlap appears to be greater for nonliving things.

What aspects of the data from studies of acquired category- specific impairments are relevant to theories of knowledge representation? Discussion in the target article focuses on the living/nonliving dichotomy. However, many case studies reveal considerably more complex patterns of impaired and preserved performance that undermine this distinction. We consider this evidence and discuss its implications for theories of knowledge representation.

The HIT model explains the existence of semantic category-specific deficits by assuming that sensory knowledge is crucially important in processing living things, while functional knowledge is crucially important in processing nonliving things – the sensory/functional assumption. Here we argue that the sensory/functional assumption as implemented in HIT is neither theoretically nor empirically grounded and that, in any case, there is neuropsychological evidence which invalidates this assumption, thereby undermining the HIT model as a whole.

To account for category-specific semantic deficits, Humphreys and Forde propose to fractionate semantic memory into multiple sensory and functional knowledge stores. There are reasons to doubt the empirical productivity of this proposal, unless theoretically motivated principles of distinguishing and weighting the different kinds of object knowledge can be spelled out in detail.

The HIT model comes close to a view suggested by Donald Hebb, that cognitive representations are organized as distributed neuron webs, cell assemblies, whose components are mutually connected and whose internal connections provide continuous information exchange among sub-components of the representation. Two questions are asked related to (1) the organization of internal connections of a concept representation and (2) the conditions under which information exchange between components are assumed in the HIT model.

We offer additional points that support a distributed semantic memory: (1) the activation of representations that are modality-specific; (2) patients with inferotemporal lesions fail to activate visual object representations in semantic tasks, although normal subjects do; (3) direct activation of action systems from pictorial information, but not from words; (4) patients who demonstrate superiority with abstract words fail to access perceptual representations.

This commentary questions whether the category properties of an object can be determined independent of the experience of the categorizer. Expertise studies have shown that the judged properties of an object can differ from expert to novice and from expert to expert. The expertise findings indicate that object properties exist not only in the world, but in the mind of the categorizer.

Humphreys & Forde concentrate on the living/nonliving dissociation. However, further dissociations have been reported, including selective loss or preservation in recognizing body parts and numbers. This commentary outlines the relevance of the number category for understanding the organising principles of semantic memory.

Humphreys and Forde argue that semantic memory is divided into separate substores for different kinds of information. However, the neuro-imaging results cited in support of this view are inconsistent and often methodologically and statistically unreliable. Our own data indicate no regional specialisation as a function of semantic category or domain and support instead a distributed unitary account.

The authors' model reduces the literature on conceptual representation to a single node: “encyclopedic knowledge.” The structure of conceptual knowledge is not so trivial. By ignoring the phenomena central to reasoning about living things, the authors base their dismissal of semantic systems on inadequate descriptive ground. A better descriptive account is available in the conceptual development literature. Neuropsychologists could import the insights and tasks from cognitive development to improve their studies.

We summarise and respond to the main points made by the commentators on our target article, which concern: (1) whether structural similarity can play a causal role in normal object identification and in neuropsychological deficits for living things, (2) the nature of our structural knowledge of the world, (3) the relations between sensory and functional knowledge of objects, and the nature of our functional knowledge about living things, (4) whether we need to posit a “core” semantic system, (5) arguments that can be marshalled from evidence on functional imaging, (6) the causal mechanisms by which category differences can emerge in object representations, and (7) the nature of our knowledge about categories other than living and nonliving things. We also highlight points raised in our article that seem to be accepted.

When one takes the evolution of operant behavior as prototype, one sees that the term replication is too tied to the peculiarities of genetic evolution. A more general term is recurrence. The important problem raised by recurrence is not “information” but relationship: deciding when two occurrences belong to the same lineage. That is solved by looking at common environmental effects.

Authors frequently refer to gene-based selection in biological evolution, the reaction of the immune system to antigens, and operant learning as exemplifying selection processes in the same sense of this term. However, as obvious as this claim may seem on the surface, setting out an account of “selection” that is general enough to incorporate all three of these processes without becoming so general as to be vacuous is far from easy. In this target article, we set out such a general account of selection to see how well it accommodates these very different sorts of selection. The three fundamental elements of this account are replication, variation, and environmental interaction. For selection to occur, these three processes must be related in a very specific way. In particular, replication must alternate with environmental interaction so that any changes that occur in replication are passed on differentially because of environmental interaction.

One of the main differences among the three sorts of selection that we investigate concerns the role of organisms. In traditional biological evolution, organisms play a central role with respect to environmental interaction. Although environmental interaction can occur at other levels of the organizational hierarchy, organisms are the primary focus of environmental interaction. In the functioning of the immune system, organisms function as containers. The interactions that result in selection of antibodies during a lifetime are between entities (antibodies and antigens) contained within the organism. Resulting changes in the immune system of one organism are not passed on to later organisms. Nor are changes in operant behavior resulting from behavioral selection passed on to later organisms. But operant behavior is not contained in the organism because most of the interactions that lead to differential replication include parts of the world outside the organism. Changes in the organism's nervous system are the effects of those interactions. The role of genes also varies in these three systems. Biological evolution is gene-based (i.e., genes are the primary replicators). Genes play very different roles in operant behavior and the immune system. However, in all three systems, iteration is central. All three selection processes are also incredibly wasteful and inefficient. They can generate complexity and novelty primarily because they are so wasteful and inefficient.