Fabrizio Calzavarini

In the present chapter, I shall argue that neuropsychological data provide some evidence that inferential and referential competences are functionally isolable, i.e., that they can operate independently of one another. In extreme cases, one competence can be completely damaged without this resulting in an impairment of the other, and vice versa (double dissociation). One side of the dissociation, i.e., selective damage to referential abilities, is well attested in the relevant neuropsychological literature. It is common, for instance, in visual agnosia, an impairment of the capacity to recognise visually presented objects which is crucially unrelated to deficit in elementary vision, language, memory, and general cognitive decline. Critically, the patients’ performances in inferential tasks (e.g., definition naming, word-word matching) are usually perfectly preserved. The second side of the dissociation, i.e., the selective impairment of inferential competence, is rarer but also attested. Deficits in inferential performances can be found in a variety of syndromes, such as semantic dementia, semantic aphasia, and Alzheimer’s Disease. None of the syndromes presented above is selective for inferential competence. Nevertheless, as we shall see, selective damage to inferential competence has been observed several times, and the number of cases has grown considerably in recent years.

To vindicate empirically the distinction between inferential competence and referential competence we must show it to be consistent with anatomical data on brain structure in both normal and pathological populations. In this chapter, I shall do this by discussing some direct electrocorticography stimulation studies investigating the neural basis of inferential (i.e., definition) and referential (i.e., visual) naming in pre-surgical epilepsy patients. As I shall argue, however, very few neuroimaging studies are immediately pertinent to this issue. Indirect insights on the neural basis of the two competences come from some studies investigating the distinction between verbal vs non-verbal stimuli in visual and auditory modalities. Direct insights come from a few functional Magnetic Resonance Imaging [fMRI] studies analysing neural activation during visual naming vs definition naming. Of special importance in this context, as we will see, is an fMRI experiment performed by Marconi and colleagues. The authors contrasted two referential tasks involving pictures as stimuli (i.e., visual naming and word-picture matching) with two inferential tasks involving words as stimuli (i.e., definition naming and word-word matching). Taken together, the results of the studies to be discussed in this chapter suggest that, in the normal population, inferential and referential abilities are mediated by partially dissociated cortical networks Some limitations of these studies must, however, be acknowledged.

So far, I have been concerned with the distinction between inferential and referential competence understood as a conceptual distinction, but this distinction can naturally suggest an intriguing empirical hypothesis: inferential and referential semantic abilities are actually underpinned by distinct cognitive systems. In order to assess empirically such dual theory of semantic competence we must map it onto the linguistic tasks that are routinely used in neuropsychology of language. The primary purpose of this chapter is to introduce a distinction between inferential and referential semantic tasks. A secondary aim of the chapter is to begin to discuss the structure of the inferential and referential semantic systems in the human brain. We will see that certain intuitive considerations together with some neuroscientific data suggest that the inferential and the referential semantic systems are not ‘monolithic’ entities: they consist in various, functionally interrelated sub-systems. A final aim of the chapter is to discuss some issues arising in the field of computational modelling of inferential and referential competence.

In the previous chapters, I have analysed some empirical evidence in favour of the proposed distinction between inferential and referential competence and I have outlined a neurocognitive model supporting this distinction at the level of the human brain. In this conclusive chapter, I shall discuss an important potential complication for this general picture, namely the alleged involvement of referential (visual) cortex during inferential tasks, especially when concrete, highly imageable stimuli (words, sentences) are involved in the tasks. According to a popular view, such involvement might reflect (visual) imagery/simulation processes that are essential to language. This would have critical implications for the theory assessed in this book: the cognitive autonomy of inferential competence and referential competence would have to be significantly weakened. In the present chapter, contrary to this popular view, I shall argue that the involvement of referential visual (and perceptual) cortex in inferential tasks should not be overstated.

In this chapter, I shall argue that meaning postulates cannot account for a significant aspect of human lexical competence, namely the ability to apply words to objects and circumstances in the world (referential competence). I shall also argue that the desired account cannot be provided within the theoretical framework of truth-conditional semantics. This problem is highlighted by Kripke’s and Putnam’s arguments against the idea that reference and truth-conditions are cognitively determined. If these arguments are successful, then truth-conditional semantics requires an objective or metaphysical notion of reference, a notion which cannot easily be implemented by a psychological theory of human semantic competence. As we will see, in philosophical semantics, there are two main possible reactions to this difficulty: the first reaction is to force the notion of objective reference within the limits of a psychological theory of competence; the second reaction is to argue that providing a theory of objective reference and providing a theory of referential competence are two radically different endeavours. I shall opt for the second solution, and, for the remainder of this work, I shall be exclusively concerned with referential (vs inferential) competence, as distinct from objective or ‘metaphysical’ reference.

When dealing with ʻmeaningʼ and related notions, one cannot ignore what for a long time was the dominant paradigm in semantics, namely what I shall refer to as truth-conditional cognitivism. According to this paradigm, truth-conditional formal semantics for natural language, in Montagovian or Davidsonian form, is a theory of semantic competence. In this chapter, I shall argue that truth-conditional formal semantics can account for the structural or compositional aspect of human semantic competence, i.e., the ability to understand complex linguistic expressions on the basis of the knowledge of the meanings of the component words and the way in which these words are combined. Furthermore, I shall argue that adding a set of meaning postulates to a standard Montagovian semantic system makes it possible to also model a significant portion of human lexical competence. This is the aspect that is responsible for performances such as material inference, judgments based on semantic relations, word definition, and so on (inferential competence). In the following chapter, however, I shall argue that meaning postulates are insufficient when it comes to explicating the full content of lexical competence.

The primary aim of the article is to take a fresh look at the debate between amodal and grounded (modality-specific) theories of conceptual representations by formulating what I call the challenge from (global) supramodality. The challenge proceeds from the observation that, in recent years, many neuroscience data suggest that extensive portions of what are traditionally considered modality-specific cortices are in fact supramodal in nature; that is, they can process external information independently from the perceptual modality. According to a strong or global interpretation of these data, the entire brain has a task-specific and sensory-independent organization. Within this extreme view, the traditional distinction between modality-specific and amodal representations loses much of its sense. The article aims to explore the implications of (global) supramodality for the debate on the conceptual format. I shall argue that none of the (few) attempts to deal with this issue is ultimately successful, so the challenge remains unsolved.

In this paper, we want to tackle the Molyneux question thoroughly, by addressing it in terms of both ordinary perception and pictorial perception: if a congenitally blind person recovered sight, could she recognize visually the 3D shapes she already recognized tactilely, both when such shapes are given to her directly and when they are given to her pictorially, i.e., as depicted shapes? We want to claim that empirical evidence suggests that the question can be positively answered in both cases. For in the first case, such evidence shows that perception of 3D shapes is supramodal; namely, it can be equivalently achieved in different sense modalities, notably touch and vision, independently of the sensory input such shapes are accessed. While in the second case, such evidence shows that one can satisfy both in vision and in touch the condition for depicted shapes, which are typically not where the perceiver is, to be grasped by that perceiver in a picture’s subject, i.e., what the picture presents. This condition states that the picture’s vehicle, i.e., the typically 2D physical basis of a picture, is enriched by adding to its properties the 3D grouping properties that allow for a figure/ground segmentation to be performed in that vehicle’s elements.

In The Performance of Reading, Peter Kivy introduces, on a purely phenomenological basis, an interesting and potentially fruitful analogy between the experience of silently reading literary texts and the experience of silently reading musical scores. In Kivy’s view, both mental experiences involve a critical element of auditory mental imagery, consisting in having a performance “in the head” or the mind’s ear. This analogy might have significant implications for the ontological status of literary works, as well as for the theoretical relations between music and language. Nevertheless, Kivy’s hypothesis has never been investigated and discussed in its empirical merits. In the present paper, we shall claim that neuroscience data support, at least in part, Kivy’s phenomenological observations about the relation between reading musical scores and reading texts. Despite being functionally and anatomically dissociated at the cognitive level, the two reading experiences both involve an auditory simulation of the content, which seems to be functionally critical for a deep and rich experience of literary texts and musical scores.

This volume brings together new papers advancing contemporary debates in foundational, conceptual, and methodological issues in cognitive neuroscience. The different perspectives presented in each chapter have previously been discussed between the authors, as the volume builds on the experience of Neural Mechanisms Online – webinar series on the philosophy of neuroscience organized by the editors of this volume. The contributed chapters pertain to five core areas in current philosophy of neuroscience. It surveys the novel forms of explanation developed in cognitive neuroscience, and looks at new concepts, methods and techniques used in the field. The book also highlights the metaphysical challenges raised by recent neuroscience and demonstrates the relation between neuroscience and mechanistic philosophy. Finally, the book dives into the issue of neural computations and representations. Assembling contributions from leading philosophers of neuroscience, this work draws upon the expertise of both established scholars and promising early career researchers.

Bayesian approaches to human cognition have been extensively advocated in the last decades, but sharp objections have been raised too within cognitive science. In this paper, we outline a diagnosis of what has gone wrong with the prevalent strand of Bayesian cognitive science (here labelled pure Bayesian cognitive science), relying on selected illustrations from the psychology of reasoning and tools from the philosophy of science. Bayesians’ reliance on so-called method of rational analysis is a key point of our discussion. We tentatively conclude on a constructive note, though: an appropriately modified variant of Bayesian cognitive science can still be coherently pursued, as some scholars have noted.

There are some complex experiences, such as the experiences that allow us to understand linguistic expressions and pictures respectively, which seem to be very similar. For they are stratified experiences in which, on top of grasping certain low-level properties, one also grasps some high-level semantic-like properties. Yet first of all, those similarities notwithstanding, a phenomenologically-based reflection shows that such experiences are different. For a meaning experience has a high-level fold, in which one grasps the relevant expression’s meaning, which is not perceptual, but is only based on a low-level perceptual fold that merely grasps that expression in its acoustically relevant properties. While a pictorial experience, a seeing-in experience, has two folds, the configurational and the recognitional fold, in which one respectively grasps the physical basis of a picture, its vehicle, and what the picture presents, its subject, that are both perceptual, insofar as they are intimately connected. For unlike a meaning experience, in a seeing-in experience one can perceptually read off the picture’s subject from the picture’s vehicle. Moreover, this phenomenological difference is neurologically implemented. For not only the cerebral areas that respectively implement such experiences are different, at least as far as the access to those experiences’ respective high-level content is concerned. As is shown by the fact that one can selectively be impaired in the area respectively implementing the meaning vs. the seeing-in experience without losing one’s pictorial vs. semantic competence respectively. But also, unlike meaning experiences, the area implementing the seeing-in experiential folds is perceptual as a whole. For not only a picture’s subject can be accessed earlier than an expression’s meaning, but also the neural underpinnings of such folds are located in the perceptual areas of the brain.

Semantic perceptualism is the thesis that meaning experiences are forms of perceptual experiences. According to its defenders, this view is motivated not only by philosophical considerations, but also by empirical evidence. In the present article, I shall provide the first comprehensive and critical review of the empirical evidence in support of semantic perceptualism, including a detailed analysis of the relevant neuroanatomical data. The conclusions of my analysis are largely pessimistic. I believe that the relevant behavioral, cognitive, and patient data are suggestive but hardly conclusive. Moreover, neuroanatomical data speak strongly against semantic perceptualism.

According to the semantic view of computation, computations cannot be individuated without invoking semantic properties. A traditional argument for the semantic view is what we shall refer to as the argument from the cognitive science practice. In its general form, this argument rests on the idea that, since cognitive scientists describe computations (in explanations and theories) in semantic terms, computations are individuated semantically. Although commonly invoked in the computational literature, the argument from the cognitive science practice has never been discussed in detail. In this paper, we shall provide a critical reconstruction of this argument and an extensive analysis of its prospects, taking into account some ways of defending it that have never been explored so far. We shall argue that explanatory considerations support at best a weak version of the argument from the cognitive science practice, according to which semantic properties concur with formal syntactic properties in individuating computations in cognitive science, but not a strong version, according to which computation individuation in cognitive science is semantic as opposed to formal syntactic.

Reverse inference is a crucial inferential strategy used in cognitive neuroscience to derive conclusions about the engagement of cognitive processes from patterns of brain activation. While widely employed in experimental studies, it is now viewed with increasing scepticism within the neuroscience community. One problem with reverse inference is that it is logically invalid, being an instance of abduction in Peirce’s sense. In this paper, we offer the first systematic analysis of reverse inference as a form of abductive reasoning and highlight some relevant implications for the current debate. We start by formalising an important distinction that has been entirely neglected in the literature, namely the distinction between weak and strong reverse inference. Then, we rely on case studies from recent neuroscientific research to systematically discuss the role and limits of both strong and weak reverse inference; in particular, we offer the first exploration of weak reverse inference as a discovery strategy within cognitive neuroscience.

In philosophy of language, a distinction has been proposed by Diego Marconi between two aspects of lexical competence, i.e. referential and inferential competence. The former accounts for the relation-ship of words to the world, the latter for the relationship of words among themselves. The aim of the pa-per is to offer a critical discussion of the kind of formalisms and computational techniques that can be used in Artificial Intelligence to model the two aspects of lexical competence, and of the main difficulties related to the use of these computational techniques. The first conclusion of our discussion is that the dis-tinction between inferential and referential semantics is instantiated in the literature of Artificial Intelli-gence by the distinction between symbolic and connectionist approaches. The second conclusion of our discussion is that the modelling of lexical competence needs the advent of hybrid models integrating symbolic and connectionist frameworks. Our hypothesis is that Conceptual Spaces, a framework devel-oped by Gärdenfors more than fifteen years ago, can offer a lingua franca that allows to unify and gener-alize many aspects of the representational approaches mentioned above and to integrate “inferential” (=symbolic) and “referential” (=connectionist) computational approaches on common ground.

In philosophical semantics, a distinction has been proposed between inferential and referential lexical semantic competence. The former accounts for the relationship of words to the world, the latter for the relationship of words among themselves. Recent neuroscience research suggests that the distinction might be actually neurally implemented. That is, that inferential and referential abilities might be underpinned by two functionally independent cognitive architectures, with partly different neural realizations. This hypothesis is consistent with brain patient data, supporting the notion of a functional double dissociation between inferential and referential abilities, and with neuroscience data, suggesting that inferential and referential abilities are underpinned by at least partly different regions of the human brain. The principal aim of this article is to provide the first comprehensive and critical review of the empirical evidence in favour of such hypothesis.

This monograph offers a novel, neurocognitive theory concerning words and language. It explores the distinction between inferential and referential semantic competence. The former accounts for the relationship of words among themselves, the latter for the relationship of words to the world. The author discusses this distinction at the level of the human brain on both theoretical and neuroscientific grounds. In addition, this investigation considers the relation between the inf/ref neurocognitive theory and other accounts of semantic cognition proposed in the field of neurosemantics, as well as some potential implications of the theory for clinical neuroscience and the philosophy of semantics. Overall, the book offers an important contribution to the debate about lexical semantic competence. It combines a strong philosophical and linguistic background with a comprehensive and critical analysis of neurosemantic literature. Topics discussed lie at the intersection of philosophical semantics, linguistics, neurolinguistics, cognitive science, artificial intelligence, cognitive neuroscience, and clinical psychology. Due to its interdisciplinary orientation, coverage is rich in introductory remarks and not overly technical, therefore it is accessible to non-experts as well.

Advocates of the pictorial theory of meaning claimed that the meaning of a word is a mental picture and that lexical semantic competence --i.e. knowledge of word meaning -- is closely connected to visual imagery. As a philosophical theory of meaning, the pictorial theory was discredited in the twentieth century. Nevertheless, there is evidence that visual imagery -- the generation of visual mental images -- does play a role in semantic processing and is not just a possible side effect of it. The aim of this article is to critically assess the neuroscientific status of the pictorial theory of meaning as a hypothesis about the neurocognitive underpinnings of lexical semantic competence. I will assess relevant neuroimaging, behavioural, and patient data, as well as data from congenital blindness studies.

When dealing with ‘meaning’ or related notions, one cannot ignore what for a long time was the dominant paradigm in semantics. According to such paradigm, truth-conditional formal semantics for natural language is a theory of semantic competence. In this article, I shall discuss a foundational problem for such semantic program. I shall first be following authors who claim that truth-conditional formal semantics is unable to provide a complete account of lexical competence, and, therefore, it suffers from incompleteness. Moreover, as a few authors observed, a coherent completion of lexical meaning/competence cannot be easily provided in the philosophical framework of truth-conditional semantics. The problem is a consequence of Kripke’s and Putnam’s arguments against the view that semantic values of words are cognitively determined. I shall argue that these arguments undermine the notion that formal semantics can be, at the same time, a theory of truth-conditions and a theory of semantic competence. I shall outline and critically discuss two general attempts to solve this problem, which I shall refer to as the social/ideal route and the concepts route. The conclusions of the article are deliberately open. I shall argue that both these attempts are exposed to conceptual troubles and counterintuitive consequences, and so the lexical problem still remains a deep and unresolved issue for semantics.