In 1948, Claude Shannon introduced his version of a concept that was core to Norbert Wiener's cybernetics, namely, information theory. Shannon's formalisms include a physical framework, namely a general communication system having six unique elements. Under this framework, Shannon information theory offers two particularly useful statistics, channel capacity and information transmitted. Remarkably, hundreds of neuroscience laboratories subsequently reported such numbers. But how (and why) did neuroscientists adapt a communications-engineering framework? Surprisingly, the literature offers no clear answers. To therefore first answer (...) "how", 115 authoritative peer-reviewed papers, proceedings, books and book chapters were scrutinized for neuroscientists' characterizations of the elements of Shannon's general communication system. Evidently, many neuroscientists attempted no identification of the system's elements. Others identified only a few of Shannon's system's elements. Indeed, the available neuroscience interpretations show a stunning incoherence, both within and across studies. The interpretational gamut implies hundreds, perhaps thousands, of different possible neuronal versions of Shannon's general communication system. The obvious lack of a definitive, credible interpretation makes neuroscience calculations of channel capacity and information transmitted meaningless. To now answer why Shannon's system was ever adapted for neuroscience, three common features of the neuroscience literature were examined: ignorance of the role of the observer, the presumption of "decoding" of neuronal voltage-spike trains, and the pursuit of ingrained analogies such as information, computation, and machine. Each of these factors facilitated a plethora of interpretations of Shannon's system elements. Finally, let us not ignore the impact of these "informational misadventures" on society at large. It is the same impact as scientific fraud. (shrink)

Purpose – The purpose of this paper is to ask whether a first-order-cybernetics concept, Shannon’s Information Theory, actually allows a far-reaching mathematics of perception allegedly derived from it, Norwich et al.’s “Entropy Theory of Perception”. Design/methodology/approach – All of The Entropy Theory, 35 years of publications, was scrutinized for its characterization of what underlies Shannon Information Theory: Shannon’s “general communication system”. There, “events” are passed by a “source” to a “transmitter”, thence through a “noisy channel” to a “receiver”, that passes (...) “outcomes” (received events) to a “destination”. Findings – In the entropy theory, “events” were sometimes interactions with the stimulus, but could be microscopic stimulus conditions. “Outcomes” often went unnamed; sometimes, the stimulus, or the interaction with it, or the resulting sensation, were “outcomes”. A “source” was often implied to be a “transmitter”, which frequently was a primary afferent neuron; elsewhere, the stimulus was the “transmitter” and perhaps also the “source”. “Channel” was rarely named; once, it was the whole eye; once, the incident photons; elsewhere, the primary or secondary afferent. “Receiver” was usually the sensory receptor, but could be an afferent. “Destination” went unmentioned. In sum, the entropy theory’s idea of Shannon’s “general communication system” was entirely ambiguous. Research limitations/implications – The ambiguities indicate that, contrary to claim, the entropy theory cannot be an “information theoretical description of the process of perception”. Originality/value – Scrutiny of the entropy theory’s use of information theory was overdue and reveals incompatibilities that force a reconsideration of information theory’s possible role in perception models. A second-order-cybernetics approach is suggested. (shrink)

Purpose – In the last half-century, individual sensory neurons have been bestowed with characteristics of the whole human being, such as behavior and its oft-presumed precursor, consciousness. This anthropomorphization is pervasive in the literature. It is also absurd, given what we know about neurons, and it needs to be abolished. This study aims to first understand how it happened, and hence why it persists. Design/methodology/approach – The peer-reviewed sensory-neurophysiology literature extends to hundreds (perhaps thousands) of papers. Here, more than 90 (...) mainstream papers were scrutinized. Findings – Anthropomorphization arose because single neurons were cast as “observers” who “identify”, “categorize”, “recognize”, “distinguish” or “discriminate” the stimuli, using math-based algorithms that reduce (“decode”) the stimulus-evoked spike trains to the particular stimuli inferred to elicit them. Without “decoding”, there is supposedly no perception. However, “decoding” is both unnecessary and unconfirmed. The neuronal “observer” in fact consists of the laboratory staff and the greater society that supports them. In anthropomorphization, the neuron becomes the collective. Research limitations/implications – Anthropomorphization underlies the widespread application to neurons Information Theory and Signal Detection Theory, making both approaches incorrect. Practical implications – A great deal of time, money and effort has been wasted on anthropomorphic Reductionist approaches to understanding perception and consciousness. Those resources should be diverted into more-fruitful approaches. Originality/value – A long-overdue scrutiny of sensory-neuroscience literature reveals that anthropomorphization, a form of Reductionism that involves the presumption of single-neuron consciousness, has run amok in neuroscience. Consciousness is more likely to be an emergent property of the brain. (shrink)

Purpose – The purpose of this paper is to examine the popular “information transmitted” interpretation of absolute judgments, and to provide an alternative interpretation if one is needed. Design/methodology/approach – The psychologists Garner and Hake and their successors used Shannon’s Information Theory to quantify information transmitted in absolute judgments of sensory stimuli. Here, information theory is briefly reviewed, followed by a description of the absolute judgment experiment, and its information theory analysis. Empirical channel capacities are scrutinized. A remarkable coincidence, the (...) similarity of maximum information transmitted to human memory capacity, is described. Over 60 representative psychology papers on “information transmitted” are inspected for evidence of memory involvement in absolute judgment. Finally, memory is conceptually integrated into absolute judgment through a novel qualitative model that correctly predicts how judgments change with increase in the number of judged stimuli. Findings – Garner and Hake gave conflicting accounts of how absolute judgments represent information transmission. Further, “channel capacity” is an illusion caused by sampling bias and wishful thinking; information transmitted actually peaks and then declines, the peak coinciding with memory capacity. Absolute judgments themselves have numerous idiosyncracies that are incompatible with a Shannon general communication system but which clearly imply memory dependence. Research limitations/implications – Memory capacity limits the correctness of absolute judgments. Memory capacity is already well measured by other means, making redundant the informational analysis of absolute judgments. Originality/value – This paper presents a long-overdue comprehensive critical review of the established interpretation of absolute judgments in terms of “information transmitted”. An inevitable conclusion is reached: that published measurements of information transmitted actually measure memory capacity. A new, qualitative model is offered for the role of memory in absolute judgments. The model is well supported by recently revealed empirical properties of absolute judgments. (shrink)

Shannon’s information theory has been a popular component of first-order cybernetics. It quantifies information transmitted in terms of the number of times a sent symbol is received as itself, or as another possible symbol. Sent symbols were events and received symbols were outcomes. Garner and Hake reinterpreted Shannon, describing events and outcomes as categories of a stimulus attribute, so as to quantify the information transmitted in the psychologist’s category (or absolute judgment) experiment. There, categories are represented by specific stimuli, and (...) the human subject must assign those stimuli, singly and in random order, to the categories that they represent. Hundreds of computations ensued of information transmitted and its alleged asymptote, the sensory channel capacity. The present paper critically re-examines those estimates. It also reviews estimates of memory capacity from memory experiments. It concludes that absolute judgment is memory-limited and that channel capacities are actually memory capacities. In particular, there are factors that affect absolute judgment that are not explainable within Shannon’s theory, factors such as feedback, practice, motivation, and stimulus range, as well as the anchor effect, sequential dependences, the rise in information transmitted with the increase in number of stimulus dimensions, and the phenomena of masking and stimulus duration dependence. It is recommended that absolute judgments be abandoned, because there are already many direct estimates of memory capacity. (shrink)

Purpose – A key cybernetics concept, information transmitted in a system, was quantified by Shannon. It quickly gained prominence, inspiring a version by Harvard psychologists Garner and Hake for “absolute identification” experiments. There, human subjects “categorize” sensory stimuli, affording “information transmitted” in perception. The Garner-Hake formulation has been in continuous use for 62 years, exerting enormous influence. But some experienced theorists and reviewers have criticized it as uninformative. They could not explain why, and were ignored. Here, the “why” is answered. (...) The paper aims to discuss these issues. Design/methodology/approach – A key Shannon data-organizing tool is the confusion matrix. Its columns and rows are, respectively, labeled by “symbol sent” (event) and “symbol received” (outcome), such that matrix entries represent how often outcomes actually corresponded to events. Garner and Hake made their own version of the matrix, which deserves scrutiny, and is minutely examined here. Findings – The Garner-Hake confusion-matrix columns represent “stimulus categories”, ranges of some physical stimulus attribute (usually intensity), and its rows represent “response categories” of the subject’s identification of the attribute. The matrix entries thus show how often an identification empirically corresponds to an intensity, such that “outcomes” and “events” differ in kind (unlike Shannon’s). Obtaining a true “information transmitted” therefore requires stimulus categorizations to be converted to hypothetical evoking stimuli, achievable (in principle) by relating categorization to sensation to intensity. But those relations are actually unknown, perhaps unknowable. Originality/value – The author achieves an important understanding: why “absolute identification” experiments do not illuminate sensory processes. (shrink)

Purpose – For half a century, neuroscientists have used Shannon Information Theory to calculate “information transmitted,” a hypothetical measure of how well neurons “discriminate” amongst stimuli. Neuroscientists’ computations, however, fail to meet even the technical requirements for credibility. Ultimately, the reasons must be conceptual. That conclusion is confirmed here, with crucial implications for neuroscience. The paper aims to discuss these issues. Design/methodology/approach – Shannon Information Theory depends upon a physical model, Shannon’s “general communication system.” Neuroscientists’ interpretation of that model is (...) scrutinized here. Findings – In Shannon’s system, a recipient receives a message composed of symbols. The symbols received, the symbols sent, and their hypothetical occurrence probabilities altogether allow calculation of “information transmitted.” Significantly, Shannon’s system’s “reception” (decoding) side physically mirrors its “transmission” (encoding) side. However, neurons lack the “reception” side; neuroscientists nonetheless insisted that decoding must happen. They turned to Homunculus, an internal humanoid who infers stimuli from neuronal firing. However, Homunculus must contain a Homunculus, and so on ad infinitum – unless it is super-human. But any need for Homunculi, as in “theories of consciousness,” is obviated if consciousness proves to be “emergent.” Research limitations/implications – Neuroscientists’ “information transmitted” indicates, at best, how well neuroscientists themselves can use neuronal firing to discriminate amongst the stimuli given to the research animal. Originality/value – A long-overdue examination unmasks a hidden element in neuroscientists’ use of Shannon Information Theory, namely, Homunculus. Almost 50 years’ worth of computations are recognized as irrelevant, mandating fresh approaches to understanding “discriminability.”. (shrink)

Introduction & Objectives: Norwich’s Entropy Theory of Perception (1975 [1] -present) stands alone. It explains many firing-rate behaviors and psychophysical laws from bare theory. To do so, it demands a unique sort of interaction between receptor and brain, one that Norwich never substantiated. Can it now be confirmed, given the accumulation of empirical sensory neuroscience? Background: Norwich conjoined sensation and a mathematical model of communication, Shannon’s Information Theory, as follows: “In the entropic view of sensation, magnitude of sensation is regarded (...) as a measure of the entropy or uncertainty of the stimulus signal” [2]. “To be uncertain about the outcome of an event, one must first be aware of a set of alternative outcomes” [3]. “The entropy-establishing process begins with the generation of a [internal] sensory signal by the stimulus generator. This is followed by receipt of the [external] stimulus by the sensory receptor, transmission of action potentials by the sensory neurons, and finally recapture of the [response to the internal] signal by the generator” [4]. The latter “recapture” differentiates external from internal stimuli. The hypothetical “stimulus generators” are internal emitters, that generate photons in vision, audible sounds in audition (to Norwich, the spontaneous otoacoustic emissions [SOAEs]), “temperatures in excess of local skin temperature” in skin temperature sensation [4], etc. Method (1): Several decades of empirical sensory physiology literature was scrutinized for internal “stimulus generators”. Results (1): Spontaneous photopigment isomerization (“dark light”) does not involve visible light. SOAEs are electromechanical basilar-membrane artefacts that rarely produce audible tones. The skin’s temperature sensors do not raise skin temperature, etc. Method (2): The putative action of the brain-and-sensory-receptor loop was carefully reexamined. Results (2): The sensory receptor allegedly “perceives”, experiences “awareness”, possesses “memory”, and has a “mind”. But those traits describe the whole human. The receptor, thus anthropomorphized, must therefore contain its own perceptual loop, containing a receptor, containing a perceptual loop, etc. Summary & Conclusions: The Entropy Theory demands sensory awareness of alternatives, through an imagined brain-and-sensory-receptor loop containing internal “stimulus generators”. But (1) no internal “stimulus generators” seem to exist and (2) the loop would be the outermost of an infinite nesting of identical loops. (shrink)

Purpose – This study aims to examine the observer’s role in “infant psychophysics”. Infant psychophysics was developed because the diagnosis of perceptual deficits should be done as early in a patient’s life as possible, to provide efficacious treatment and thereby reduce potential long-term costs. Infants, however, cannot report their perceptions. Hence, the intensity of a stimulus at which the infant can detect it, the “threshold”, must be inferred from the infant’s behavior, as judged by observers (watchers). But whose abilities are (...) actually being inferred? The answer affects all behavior-based conclusions about infants’ perceptions, including the well-proselytized notion that auditory stimulus-detection thresholds improve rapidly during infancy. Design/methodology/approach – In total, 55 years of infant psychophysics is scrutinized, starting with seminal studies in infant vision, followed by the studies that they inspired in infant hearing. Findings – The inferred stimulus-detection thresholds are those of the infant-plus-watcher and, more broadly, the entire laboratory. The thresholds are therefore tenuous, because infants’ actions may differ with stimulus intensity; expressiveness may differ between infants; different watchers may judge infants differently; etc. Particularly, the watcher’s ability to “read” the infant may improve with the infant’s age, confounding any interpretation of perceptual maturation. Further, the infant’s gaze duration, an assumed cue to stimulus detection, may lengthen or shorten nonlinearly with infant age. Research limitations/implications – Infant psychophysics investigators have neglected the role of the observer, resulting in an accumulation of data that requires substantial re-interpretation. Altogether, infant psychophysics has proven far too resilient for its own good. Originality/value – Infant psychophysics is examined for the first time through second-order cybernetics. The approach reveals serious unresolved issues. (shrink)

A ‘Black Box’ cannot be opened to reveal its mechanism. Rather, its operations are inferred through input from (and output to) an ‘observer’. All of us are observers, who attempt to understand the Black Boxes that are Minds. The Black Box and its observer constitute a system, differing from either component alone: a ‘greater’ Black Box to any further-external-observer. To Glanville (1982), the further-external-observer probes the greater-Black-Box by interacting directly with its core Black Box, ignoring that Box’s immediate observer. In (...) later accounts, however, Glanville’s greater-Black-Box inexplicably becomes unitary. Why the discrepancy? To resolve it, we start with von Foerster’s archetype ‘machines’, that are of two kinds: ‘Trivial’ (predictable) and ‘Non-Trivial’ (non-predictable). Early-on, Glanville treated the core Black Box and its observer as Trivial Machines, that gradually ‘whiten’ (reveal) each other though input and output, becoming ‘white boxes’. Later, however, Black Box and observer became Non-Trivial Machines, never fully ‘whitenable’. But Non-Trivial Machines can be concatenated from Trivial Machines, and are the only true Black Boxes; any greater-Black-Box (Non-Trivial Machine) may (within its core Black Box) involve white boxes (that are Trivial Machines). White boxes, therefore, could be the ultimate source of the greatest Black Box of all: the Mind. (shrink)

In 1947, Hardy, Wolff, and Goodell achieved a psychophysics milestone: they built a putative sensation-growth scale, for skin pain, from pain-difference limens. Limens were found using the “dolorimeter”, a device first made by Hardy & co. to evoke pain for pain-threshold measurements. Scant years later, though, H.K. Beecher (MD) discredited the pain scale – according to Paterson (2019), citing the historian Tousignant. Yet Hardy & co. receive approval in the literature. Intrigued, we scrutinized their methods, then Beecher’s critiques, and Tousignant’s (...) history of threshold dolorimetry. Beecher decried dolorimetry as irrelevant, favoring clinical trials of pain relief. But he failed to discredit dolorimetry. (shrink)

An ongoing mystery in sensory science is how sensation magnitude F(I), such as loudness, increases with increasing stimulus intensity I. No credible, direct experimental measures exist. Nonetheless, F(I) can be inferred algebraically. Differences in sensation have empirical (but non-quantifiable) minimum sizes called just-noticeable sensation differences, ∆F, which correspond to empirically-measurable just-noticeable intensity differences, ∆I. The ∆Is presumably cumulate from an empirical stimulus-detection threshold I_th up to the intensity of interest, I. Likewise, corresponding ∆Fs cumulate from the sensation at the stimulus-detection (...) threshold, F(I_th ), up to F(I). Regarding the ∆Is, however, it is unlikely that all of them will be known experimentally; the procedures are too lengthy. The customary approach, then, is to find ∆I at a few widely-spaced intensities, and then use those ∆Is to interpolate all ∆Is using some smooth continuous function. The most popular of those functions is Weber’s Law, ∆I⁄I=K. But that is often not even a credible approximation to the data. However, there are other equations for ∆I⁄I. Any such equation for ∆I⁄I can be combined with any equation for ∆F, through calculus, to altogether obtain F(I). Here, two assumptions for ∆F are considered: ∆F=B (Fechner’s Law) and (∆F⁄F)=g (Ekman’s Law). The respective integrals involve lower bounds I_th and F(I_th ). This stands in broad contrast to the literature, which heavily favors non-bounded integrals. We, hence, obtain 24 new, alternative equations for sensation magnitude F(I) (12 equations for (∆I⁄I) × 2 equations for ∆F). (shrink)

This paper concerns the Black Box. It is not the engineer’s black box that can be opened to reveal its mechanism, but rather one whose operations are inferred through input from (and output to) a companion observer. We are observers ourselves, and we attempt to understand minds through interactions with their host organisms. To this end, Ranulph Glanville followed W. Ross Ashby in elaborating the Black Box. The Black Box and its observer together form a system having different properties than (...) either component alone, making it a greater Black Box to any further-external observer. How far into this greater box can a further-external observer probe? The answer is crucial to understanding Black Boxes, and so an answer is offered here. It employs von Foerster’s machines, abstract entities having mechano-electrical bases, just like putative Black Boxes. Von Foerster follows Turing, Ashby, E. F. Moore, and G. H. Mealy in recognizing archetype machines that he calls trivial (predictable) and non-trivial (non-predictable). It is argued here that non-trivial machines are the only true Black Boxes. But non-trivial machines can be concatenated from trivial machines. Hence, the utter core of any greater Black Box (a non-trivial machine) may involve two (or more) White Boxes (trivial machines). This is how an unpredictable thing emerges from predictable parts. Interactions of White Boxes—of trivial machines—may be the ultimate source of the mind. Keywords: . (shrink)

A major question in sensory science is how a sensation of magnitude F (such as loudness) depends upon a sensory stimulus of physical intensity I (such as a sound-pressure-wave of root-mean-square sound-pressure-level). An empirical just-noticeable sensation difference (∆F)_j at F_j specifies a just-noticeable intensity difference (∆I)_j at I_j. Classically, intensity differences accumulate from a stimulus-detection threshold I_th up to a desired intensity I. The corresponding sensation differences likewise accumulate up to F(I) from F(I_th ), the non-zero sensation (as suggested by (...) hearing studies) at I_th. Consequently, sensation growth F(I) can be obtained through classic Fechnerian integration, in which some empirically-based relation for the Weber Fraction, ∆I⁄I, is combined with either Fechner’s Law ∆F=B or Ekman’s Law (∆F⁄F)=g. The number of steps in I is equated to the number of steps in F; an infinite series ensues, whose higher-order terms are traditionally ignored (Fechnerian integration). But also, remarkably, so are the integration bounds I_th and F(I_th ). Here, we depart from orthodoxy by including those bounds. Bounded Fechnerian integration is first used to derive hypothetical sensation-growth equations for which the differential ∆F(I)=F(I+∆I)-F(I) does indeed return either Fechner’s Law or Ekman’s Law respectively. One relation emerges: linear growth of sensation F with intensity I. Subsequently, 24 sensation-growth equations F(I) that the author had derived using bounded Fechnerian integration (12 equations for the Weber Fraction (∆I⁄I), each combined with either Fechner’s Law or with Ekman’s Law) are scrutinized for whether their differentials F(I+∆I)-F(I) return the respective Fechner’s Law or Ekman’s Law, particularly in the previously-unexamined limits (∆I⁄I)≪1 and (∆I⁄I)→0. Classic claims made by Luce and Edwards (1958) are then examined, viz., that three popular forms of the Weber Fraction, when combined with Fechner’s Law, produce sensation-magnitude equations that subsequently return the selfsame Fechner’s Law. When sensation-growth equations are derived here using bounded Fechnerian integration, Luce and Edwards (1958) prove to be wrong. (shrink)

Purpose – Neuroscientists act as proxies for implied anthropomorphic signal- processing beings within the brain, Homunculi. The latter examine the arriving neuronal spike-trains to infer internal and external states. But a Homunculus needs a brain of its own, to coordinate its capabilities – a brain that necessarily contains a Homunculus and so on indefinitely. Such infinity is impossible – and in well-cited papers, Attneave and later Dennett claim to eliminate it. How do their approaches differ and do they (in fact) (...) obviate the Homunculi? Design/methodology/approach – The Attneave and Dennett approaches are carefully scrutinized. To Attneave, Homunculi are effectively “decision-making” neurons that control behaviors. Attneave presumes that Homunculi, when successively nested, become successively “stupider”, limiting their numbers by diminishing their responsibilities. Dennett likewise postulates neuronal Homunculi that become “stupider” – but brain-wards, where greater sophistication might have been expected. Findings – Attneave’s argument is Reductionist and it simply assumes-away the Homuncular infinity. Dennett’s scheme, which evidently derives from Attneave’s, ultimately involves the same mistakes. Attneave and Dennett fail, because they attempt to reduce intentionality to non-intentionality. Research limitations/implications – Homunculus has been successively recognized over the centuries by philosophers, psychologists and (some) neuroscientists as a crucial conundrum of cognitive science. It still is. Practical implications – Cognitive-science researchers need to recognize that Reductionist explanations of cognition may actually devolve to Homunculi, rather than eliminating them. Originality/value – Two notable Reductionist arguments against the infinity of Homunculi are proven wrong. In their place, a non-Reductionist treatment of the mind, “Emergence”, is discussed as a means of rendering Homunculi irrelevant. (shrink)

In Cybernetics (1961 Edition), Professor Norbert Wiener noted that “The role of information and the technique of measuring and transmitting information constitute a whole discipline for the engineer, for the neuroscientist, for the psychologist, and for the sociologist”. Sociology aside, the neuroscientists and the psychologists inferred “information transmitted” using the discrete summations from Shannon Information Theory. The present author has since scrutinized the psychologists’ approach in depth, and found it wrong. The neuroscientists’ approach is highly related, but remains unexamined. Neuroscientists (...) quantified “the ability of [physiological sensory] receptors (or other signal-processing elements) to transmit information about stimulus parameters”. Such parameters could vary along a single continuum (e.g., intensity), or along multiple dimensions that altogether provide a Gestalt – such as a face. Here, unprecedented scrutiny is given to how 23 neuroscience papers computed “information transmitted” in terms of stimulus parameters and the evoked neuronal spikes. The computations relied upon Shannon’s “confusion matrix”, which quantifies the fidelity of a “general communication system”. Shannon’s matrix is square, with the same labels for columns and for rows. Nonetheless, neuroscientists labelled the columns by “stimulus category” and the rows by “spike-count category”. The resulting “information transmitted” is spurious, unless the evoked spike-counts are worked backwards to infer the hypothetical evoking stimuli. The latter task is probabilistic and, regardless, requires that the confusion matrix be square. Was it? For these 23 significant papers, the answer is No. (shrink)

Information flow in a system is a core cybernetics concept. It has been used frequently in Sensory Psychology since 1951. There, Shannon Information Theory was used to calculate "information transmitted" in "absolute identification" experiments involving human subjects. Originally, in Shannon's "system", any symbol received ("outcome") is among the symbols sent ("events"). Not all symbols are received as transmitted, hence an indirect noise measure is calculated, "information transmitted", which requires knowing the confusion matrix, its columns labeled by "event" and its rows (...) labeled by "outcome". Each matrix entry is dependent upon the frequency with which a particular outcome corresponds to a particular event. However, for the sensory psychologist, stimulus intensities are "events"; the experimenter partitions the intensity continuum into ranges called "stimulus categories" and "response categories", such that each confusion-matrix entry represents the frequency with which a stimulus from a stimulus category falls within a particular response category. Of course, a stimulus evokes a sensation, and the subject's immediate memory of it is compared to the memories of sensations learned during practice, to make a categorization. Categorizing thus introduces "false noise", which is only removed if categorizations can be converted back to their hypothetical evoking stimuli. But sensations and categorizations are both statistically distributed, and the stimulus that corresponds to a given mean categorization cannot be known from only the latter; the relation of intensity to mean sensation, and of mean sensation to mean categorization, are needed. Neither, however, are presently knowable. This is a quandary, which arose because sensory psychologists ignored an ubiquitous component of Shannon's "system", the uninvolved observer, who calculates "information transmitted". Human sensory systems, however, are within de facto observers, making "false noise" inevitable. (shrink)

This paper reveals errors within Norwich et al.’s Entropy Theory of Perception, errors that have broad implications for our understanding of perception. What Norwich and coauthors dubbed their “informational theory of neural coding” is based on cybernetics, that is, control and communication in man and machine. The Entropy Theory uses information theory to interpret human performance in absolute judgments. There, the continuum of the intensity of a sensory stimulus is cut into categories and the subject is shown exemplar stimuli of (...) each category. The subject must then identify individual exemplars by category. The identifications are recorded in the Garner-Hake version of the Shannon “confusion matrix”. The matrix yields “H”, the entropy (degree of uncertainty) about what stimulus was presented. Hypothetically, uncertainty drops as a stimulus lengthens, i.e. a plot of H vs. stimulus duration should fall monotonically. Such “adaptation” is known for both sensation and firing rate. Hence, because “the physiological adaptation curve has the same general shape as the psychophysical adaptation curve”, Norwich et al. assumed that both have the same time course; sensation and firing rate were thus both declared proportional to H. However, a closer look reveals insurmountable contradictions. First, the peripheral neuron hypothetically cannot fire in response to a stimulus of a given intensity until after somehow computing H from its responses to stimuli of various intensities. Thus no sensation occurs until firing rate adapts, i.e. attains its spontaneous rate. But hypothetically, once adaptation is complete, certainty is reached and perception ends. Altogether, then, perception cannot occur until perception is over. Secondly, sensations, firing rates, and H’s are empirically not synchronous, contrary to assumption. In sum, the core concept of the cybernetics-based Entropy Theory of Perception, that is, that uncertainty reduction is the basis for perception, is irrational. (shrink)