This commentary is divided into two parts. The first considers a possible role for Gray's predictor plus comparator mechanism in human episodic recognition memory. It draws on the computational specifications of recognition outlined in Humphreys et al. to demonstrate how the logically necessary components of recognition tasks might be mapped onto the mechanism. The second part demonstrates how the mechanism outlined by Gray might be implicated in a form of imitative learning suitable for the acquisition of complex tasks.

Gray mistakenly thinks I have rejected the sort of theoretical enterprise he is undertaking, because, according to him, I think that "more data" is all that is needed to resolve all the issues. Not at all. My stalking horse was the bizarre (often pathetic) claim that no amount of empirical, "third-person point-of-view" science (data plus theory) could ever reduce the residue of mystery about consciousness to zero. This "New Mysterianism" (Flanagan, 1991) is one that he should want to combat as (...) vigorously as I have done. (shrink)

Robust theories concerning the connection between consciousness and brain function should derive not only from empirical evidence but also from a well grounded inind-body ontology. In the case of the comparator hypothesis, Gray develops his ideas relying extensively on empirical evidence, but he bounces irresolutely among logically incompatible metaphysical theses which, in turn, leads him to excessively skeptical conclusions concerning the naturalization of consciousness.

Segmentalized consciousness in schizophrenia reflects a loss of the normal Gestalt organization and contextualization of perception. Grays model explains such segmentalization in terms of septohippocampal dysfunction, which is consistent with known neuropsychological impairment in schizophrenia. However, other considerations suggest that everyday perception and its failure in schizophrenia also involve prefrontal executive mechanisms, which are only minimally elaborated by Gray.

69 Thompson-Schill, S.L. _et al. _(1997) Role of left inferior prefrontal cortex 59 Buckner, R.L. _et al. _(1996) Functional anatomic studies of memory in retrieval of semantic knowledge: a re-evaluation _Proc. Natl. Acad._ retrieval for auditory words and pictures _J. Neurosci. _16, 6219–6235 _Sci. U. S. A. _94, 14792–14797 60 Buckner, R.L. _et al. _(1995) Functional anatomical studies of explicit and 70 Baddeley, A. (1992) Working memory: the interface between memory implicit memory retrieval tasks _J. Neurosci. _15, 12–29 and cognition (...) _J. Cogn. Neurosci. _4, 281–288 61 Bäckman, L. _et al. _(1997) Brain activation in young and older adults 71 Petrides, M. (1994) Frontal lobes and behavior _Curr. Opin. Neurobiol._ during implicit and explicit retrieval _J. Cogn. Neurosci. _9, 378–391. (shrink)

In this paper, we argue that the comparator model is not a satisfactory model of sense of agency (SoA). We present a theoretical argument and experimental studies. We show (1) most studies of SoA neglect a distinction between SoA associated with movements (narrow SoA) and SoA associated with environmental events (broad SoA); (2) the comparator model emerges from experimental studies of sensory consequences narrowly associated with movements; (3) narrow SoA can be explained by a comparator model, but a motor signal (...) model is simpler and explain narrow SoA equally well; and (4) standard experimental paradigms study only broad SoA. Finally, we present results from two experiments, where we have failed to induce illusory narrow SoA in healthy participants. We believe our experimental approaches should have led to illusory SoA, if the comparator model of SoA was correct. The results challenge proponents of the comparator model of narrow SoA. (shrink)

Both children and adults predict the content of upcoming language, suggesting that prediction is useful for learning as well as processing. We present an alternative model which can explain prediction behaviour as a by-product of language learning. We suggest that a consideration of language acquisition places important constraints on Pickering & Garrod's (P&G's) theory.

This paper proposes how various disciplinary theories of cognition may be combined into a unifying, sub-symbolic, computational theory of cognition. The following theories are considered for integration: psychological theories, including the theory of event coding, event segmentation theory, the theory of anticipatory behavioral control, and concept development; artificial intelligence and machine learning theories, including reinforcement learning and generative artificial neural networks; and theories from theoretical and computational neuroscience, including predictive coding and free energy-based inference. In the light of such a (...) potential unification, it is discussed how abstract cognitive, conceptualized knowledge and understanding may be learned from actively gathered sensorimotor experiences. The unification rests on the free energy-based inference principle, which essentially implies that the brain builds a predictive, generative model of its environment. Neural activity-oriented inference causes the continuous adaptation of the currently active predictive encodings. Neural structure-oriented inference causes the longer term adaptation of the developing generative model as a whole. Finally, active inference strives for maintaining internal homeostasis, causing goal-directed motor behavior. To learn abstract, hierarchical encodings, however, it is proposed that free energy-based inference needs to be enhanced with structural priors, which bias cognitive development towards the formation of particular, behaviorally suitable encoding structures. As a result, it is hypothesized how abstract concepts can develop from, and thus how they are structured by and grounded in, sensorimotor experiences. Moreover, it is sketched-out how symbol-like thought can be generated by a temporarily active set of predictive encodings, which constitute a distributed neural attractor in the form of an interactive free-energy minimum. The activated, interactive network attractor essentially characterizes the semantics of a concept or a concept composition, such as an actual or imagined situation in our environment. Temporal successions of attractors then encode unfolding semantics, which may be generated by a behavioral or mental interaction with an actual or imagined situation in our environment. Implications, further predictions, possible verification and falsifications, as well as potential enhancements into a fully spelled-out unified theory of cognition are discussed at the end of the paper. (shrink)

The dynamical systems approach in cognitive science offers a potentially useful perspective on both brain and behavior. Indeed, the importation of formal tools from dynamical systems research has already paid off for our field in many ways. However, like some other theoretical perspectives in cognitive science, the dynamical systems approach comes in both moderate or pragmatic and “fundamentalist” varieties (Jones & Love, 2011). In the latter form, dynamical systems theory can rise to some stirring rhetorical heights. However, as argued here, (...) it also triggers a number of serious and specific reservations. (shrink)

This commentary elaborates on Gray's conclusion that his neurophysiological model of consciousness might explain how consciousness arises from the brain, but does not address how consciousness evolved, affects behaviour or confers survival value. The commentary argues that such limitations apply to all neurophysiological or other third-person perspective models. To approach such questions the first-person nature of consciousness needs to be taken seriously in combination with third-person models of the brain.

On the basis of neuropsychological evidence, it is clear that attention should be given a role in any model of consciousness. What is known about the many instances of dissociation between explicit and implicit knowledge after brain damage suggests that conscious experience might not be linked to a restricted area of the brain. Even if it were true that there is a single brain area devoted to consciousness, the subicular area would seem to be an unlikely possibility.

An active role for conscious processes in the production of behaviour is proposed, involving top level controls in a hierarchy of behavioural control. It is suggested that by inhibiting or sensitizing lower levels in the hierarchy conscious processes can play a role in the organization of ongoing behaviour. Conscious control can be more or less evident, according to prevailing circumstances.

When people manipulate a moving object, such as writing with a pen or driving a car, they experience their actions as intimately related to the object’s motion, that is they perceive control. Here, we tested the hypothesis that observers would feel more control over a moving object if an unrelated task drew attention to a location to which the object subsequently moved. Participants steered an object within a narrow path and discriminated the color of a flash that appeared briefly close (...) to the object. Across two experiments, participants provided higher ratings of perceived control when an object moved over a flash’s location than when an object moved away from a flash’s location. This result suggests that we use the location of spatial attention to determine the perception of control. If an object goes where we are attending, we feel like we made it go there. (shrink)

Gray extrapolates from circuit models of psychopathology to propose neural substrates for the contents of consciousness. I raise three concerns: knowledge of synaptic arrangements may be inadequate to fully support his model; latent inhibition deficits in schizophrenia, a focus of this and related models, are complex and deserve replication; and this conjecture omits discussion of the neuropsychological basis for the contents of the unconscious.

Gray unwisely melds together two distinguishable contributions of consciousness: one to epistemology, the other to evolution. He also renders consciousness needlessly invisible behaviorally.

An intelligent model of the incremental capacity curve of an automotive lithium-ferrophosphate battery is presented. The relative heights of the two major peaks of the IC curve can be acquired from high-current discharges, thus enabling the state of health estimation of the battery while the vehicle is being operated and in certain cases, aging mechanisms can be suggested. Our model has been validated using a large dataset representing different degradation scenarios, obtained from a recently available open-source database.

In Gray's conjecture, mismatches in the subicular comparator and matches have equal prominence in consciousness. In rival cognitive views novelty and difficulty especially elicit more conscious modes of cognition and higher levels of self-regulation. The mismatch between Gray's conjecture and these views is discussed.

Gray's comparator model fails to provide an adequate explanation of consciousness for two reasons. First, it is based on a narrow definition of consciousness that excludes basic phenomenology and active functions of consciousness. Second, match/mismatch decisions can be made without producing an experience of consciousness. The model thus violates the sufficiency criterion.

In line with Allen Newell's challenge to develop complete cognitive architectures, and motivated by a recent proposal for a unifying subsymbolic computational theory of cognition, we introduce the cognitive control architecture SEMLINCS. SEMLINCS models the development of an embodied cognitive agent that learns discrete production rule-like structures from its own, autonomously gathered, continuous sensorimotor experiences. Moreover, the agent uses the developing knowledge to plan and control environmental interactions in a versatile, goal-directed, and self-motivated manner. Thus, in contrast to several well-known (...) symbolic cognitive architectures, SEMLINCS is not provided with production rules and the involved symbols, but it learns them. In this paper, the actual implementation of SEMLINCS causes learning and self-motivated, autonomous behavioral control of the game figure Mario in a clone of the computer game Super Mario Bros. Our evaluations highlight the successful development of behavioral versatility as well as the learning of suitable production rules and the involved symbols from sensorimotor experiences. Moreover, knowledge- and motivation-dependent individualizations of the agents’ behavioral tendencies are shown. Finally, interaction sequences can be planned on the sensorimotor-grounded production rule level. Current limitations directly point toward the need for several further enhancements, which may be integrated into SEMLINCS in the near future. Overall, SEMLINCS may be viewed as an architecture that allows the functional and computational modeling of embodied cognitive development, whereby the current main focus lies on the development of production rules from sensorimotor experiences. (shrink)

The communicative aspects of the contents of consciousness are analyzed in the framework of a neural network model of animal communication. We discuss some issues raised by Gray, such as the control of the contents of consciousness, the adaptive value of consciousness, conscious and unconscious behaviors, and the nature of a model's consciousness.

This paper introduces a special issue of Cognitive Science initiated on the 25th anniversary of the publication of Parallel Distributed Processing (PDP), a two-volume work that introduced the use of neural network models as vehicles for understanding cognition. The collection surveys the core commitments of the PDP framework, the key issues the framework has addressed, and the debates the framework has spawned, and presents viewpoints on the current status of these issues. The articles focus on both historical roots and contemporary (...) developments in learning, optimality theory, perception, memory, language, conceptual knowledge, cognitive control, and consciousness. Here we consider the approach more generally, reviewing the original motivations, the resulting framework, and the central tenets of the underlying theory. We then evaluate the impact of PDP both on the field at large and within specific subdomains of cognitive science and consider the current role of PDP models within the broader landscape of contemporary theoretical frameworks in cognitive science. Looking to the future, we consider the implications for cognitive science of the recent success of machine learning systems called “deep networks”—systems that build on key ideas presented in the PDP volumes. (shrink)

In this commentary, I point out some weaknesses in Gray's target article and, in the light of that discussion, I attempt to delineate the kinds of problem a cognitive neuroscience of consciousness faces on its way to a scientific understanding of subjective experience.

Gray, like other recent authors, seeks a scientific approach to consciousness, but fails to provide a biologically convincing description, partly because he implicitly bases his model on a computationalist foundation that embeds the contents of thought in irreducible symbolic representations. When patterns of neural activity instantiating conscious thought are shorn of homuncular observers, it appears most likely that these patterns and the circuitry that compares them with memories and plans should be found distributed over large regions of neocortex.

If sensations were behaviorally conceived, as they should be, as complex functional patterns of interaction between overt behavior and the environment, there would be no point in searching for them as instantaneous psychic elements within the brain or as internal products of the brain.

Our target article proposed that language production and comprehension are interwoven, with speakers making predictions of their own utterances and comprehenders making predictions of other people's utterances at different linguistic levels. Here, we respond to comments about such issues as cognitive architecture and its neural basis, learning and development, monitoring, the nature of forward models, communicative intentions, and dialogue.

Currently, production and comprehension are regarded as quite distinct in accounts of language processing. In rejecting this dichotomy, we instead assert that producing and understanding are interwoven, and that this interweaving is what enables people to predict themselves and each other. We start by noting that production and comprehension are forms of action and action perception. We then consider the evidence for interweaving in action, action perception, and joint action, and explain such evidence in terms of prediction. Specifically, we assume (...) that actors construct forward models of their actions before they execute those actions, and that perceivers of others' actions covertly imitate those actions, then construct forward models of those actions. We use these accounts of action, action perception, and joint action to develop accounts of production, comprehension, and interactive language. Importantly, they incorporate well-defined levels of linguistic representation. We show how speakers and comprehenders use covert imitation and forward modeling to make predictions at these levels of representation, how they interweave production and comprehension processes, and how they use these predictions to monitor the upcoming utterances. We show how these accounts explain a range of behavioral and neuroscientific data on language processing and discuss some of the implications of our proposal. (shrink)

Humans and other animals are able not only to coordinate their actions with their current sensorimotor state, but also to imagine, plan and act in view of the future, and to realize distal goals. In this paper we discuss whether or not their future-oriented conducts imply (future-oriented) representations. We illustrate the role played by anticipatory mechanisms in natural and artificial agents, and we propose a notion of representation that is grounded in the agent’s predictive capabilities. Therefore, we argue that the (...) ability that characterizes and defines a true cognitive mind, as opposed to a merely adaptive system, is that of building representations of the non-existent, of what is not currently (yet) true or perceivable, of what is desired. A real mental activity begins when the organism is able to endogenously (i.e. not as the consequence of current perceptual stimuli) produce an internal representation of the world in order to select and guide its conduct goal-directed: the mind serves to coordinate with the future. (shrink)

Experimental studies of speech production involving compensations for auditory and somatosensory perturbations and adaptation after training suggest that both types of sensory information are considered to plan and monitor speech production. Interestingly, individual sensory preferences have been observed in this context: subjects who compensate less for somatosensory perturbations compensate more for auditory perturbations, and \textit{vice versa}. We propose to integrate this sensory preference phenomenon in a model of speech motor planning using a probabilistic model in which speech units are characterized (...) both in auditory and somatosensory terms. Sensory preference is implemented in the model according to two approaches. In the first approach, which is often used in motor control models accounting for sensory integration, sensory preference is attributed to the relative precision (i.e., inverse of the variance) of the sensory characterization of the speech motor goals associated with phonological units (which are phonemes in the context of this paper). In the second, ``more original’’ variant, sensory preference is implemented by modulating the sensitivity of the comparison between the predicted sensory consequences of motor commands and the sensory characterizations of the phonemes. We present simulation results using these two variants, in the context of the adaptation to an auditory perturbation, implemented in a 2-dimensional biomechanical model of the tongue. Simulation results show that both variants lead to qualitatively similar results. Distinguishing them experimentally would require precise analyses of partial compensation patterns. However, the second proposed variant implements sensory preference without changing the sensory characterizations of the phonemes. This dissociates sensory preference and sensory characterizations of the phonemes, and makes the account of sensory preference more flexible. Indeed, in the second variant the sensory characterizations of the phonemes can remain stable, when sensory preference varies as a response to cognitive or attentional control. This opens new perspectives for capturing speech production variability associated with aging, disorders and speaking conditions. (shrink)

Gray hypothesizes that the contents of consciousness correspond to the outputs of a subicular (hippocampal/temporal lobe) comparator that compares the current state of the organism's perceptual world with a predicted state. I argue that Gray has identified a key contributing system to conscious awareness, but that his model is inadequate for explaining how conscious contents are generated in the brain. An alternative model is offered.

Gray's neuropsychological model of consciousness uses a hierarchical feedback loop framework that has been extensively discussed by many others in psychology. This commentary therefore urges Gray to integrate with, or at least acknowledge previous models. It also points out flaws in his feedback model and suggests directions for further theoretical work.

In this article we present symmetric diffusion networks, a family of networks that instantiate the principles of continuous, stochastic, adaptive and interactive propagation of information. Using methods of Markovion diffusion theory, we formalize the activation dynamics of these networks and then show that they can be trained to reproduce entire multivariate probability distributions on their outputs using the contrastive Hebbion learning rule (CHL). We show that CHL performs gradient descent on an error function that captures differences between desired and obtained (...) continuous multivariate probability distributions. This allows the learning algorithm to go beyond expected values of output units and to approximate complete probability distributions on continuous multivariate activation spaces. We argue that learning continuous distributions is an important task underlying a variety of real‐life situations that were beyond the scope of previous connectionist networks. Deterministic networks, like back propagation, cannot learn this task because they are limited to learning average values of independent output units. Previous stochastic connectionist networks could learn probability distributions but they were limited to discrete variables. Simulations show that symmetric diffusion networks can be trained with the CHL rule to approximate discrete and continuous probability distributions of various types. (shrink)