Hypothetical Identities and Chimerical Reductions

Tom Seppalainen, Portland State University

Commentator 1: David Harker, East Tennessee State University

Commentator 2: Elizabeth Schier, Macquarie Centre for Cognitive Science, Macquarie University & Berlin School of Mind and Brain

5 responses to “Hypothetical Identities and Chimerical Reductions”

  1. Harker’s and Schier’s deep and extensive critical comments provided me an opportunity to rethink the substance of my paper and its argumentative strategy even if I see no reason to change my mind about its main premises and conclusions. But I did learn that I must make explicit many ideas about scientific epistemology and especially models of testing that I originally left tacit; I also learned once again how easily functional models and posits – psychophysical models as they used to be called – are turned into neural and even “physical” theories by reductive color naturalists. So thanks for that and on to the criticisms and the issues they raise.

    Let me start from Churchland’s argument and the role of novel predictions in it. Harker and Schier agree that I’ve misunderstood Churchland’s goal. According to them, Churchland is out to challenge “dualists” and “anti-reductionists” by refuting the claim that physical theories are incapable of explaining and making predictions about the qualitative aspects of subjective experience, qualia. My purported mistake is that I am focusing on the truth of the predictions (and let me grant for argument’s sake that this is identical to the truth of the explanation) whereas Churchland is only after predictions, regardless of their truth, and explanation, regardless of its truth. That’s all it purportedly takes to show the capabilities of science for this philosophical debate.

    Detaching truth – in the empirical or epistemological sense of the term not in any deep metaphysical correspondence sense since the latter is just the issue at stake in a naturalist scientific ontology and it is addressed indirectly through the former sense(s) – from some other “cognitive” and maybe even epistemological goals and achievements of science, is for many philosophers of science and all scientists highly suspect. Empiricists (including the present author), have always been suspicious of infusing science with other epistemological values in addition to truth. Yet in light of Churchland’s past, the premise of Harker’s and Schier’s claim about my misunderstanding may be less implausible than it initially appears. The earlier, eliminativist o was critical of empiricism and on grounds of almost Quinean-wide testing holism. He saw it requiring and justifying extra or super-empirical criteria for hypothesis or “theory” assessment which, notoriously, lead him to celebrate unrestricted theory-ladenness of evidence for purposes of philosophical conclusions. Yet, in the work under discussion, Churchland is a reductionist not an eliminativist. More importantly, he advocates a very traditional if not doctrinaire empiricist assumption, the explanation-prediction symmetry thesis, the “logical” identity of prediction and explanation. The examples Churchland gives of predictions from SM are at the same time illustrations of its explanatory power. This demonstrates the symmetry thesis which, in turn, entails the lack of extra-empirical criteria in theory-assessment. Explanatory matters are in fact evidentiary, in substance. This fits empiricism: truth is the central concept of both explanation and prediction or, more specifically, central to both the D-N model of explanation and H-D model based theories of inductive confirmation.

    But despite this tension within Churchlandia, let me take Schier’s and Harker’s position for granted to further explore the agendas of epistemological and ontological naturalists, what appears in name our collective philosophical camp. I will refer to theoretical explanations that satisfy criteria which fail to include truth “explainy” and use usual cognates (the term is not my own; I learned it from Clark Glymour). The parallel term for theories that satisfy criteria of prediction “generation” yet regardless of their truth is “predicty” (this term might be my own but I doubt it). I also use its cognates. Of course these are fictional not real words and likely so because no epistemologist has ever really used or needed them in theory-assessment. Yet, they must be “concepts,” at least for my critics and, according to them, Churchland. I will pretend they are both. Since my critics emphasize explainy and predicty theories for philosophical naturalism whereas I do not, it is to be expected that we disagree quite a bit on the nature and methodology of naturalism. It is also to be expected that the criteria for explainy and predicty cognitive constructs must be in some deep sense conceptual if not philosophical as opposed to empirical and scientific. Such criteria, when related to concepts and conceptual analysis, as Schier at least does, deserve to be dubbed “Kantian.” By definition, such criteria need not be constrained by empirical processes and procedures, scientific practice, undertaken to track and assess truth but, instead, constrain this practice. So the debate at hand is between Kantian and Empiricist Naturalism. According to the former and Schier, Harker, and their Churchland, naturalists can and should explore the explainyness and predictyness of theories; according to the latter including the present author, this is not enough or possible. But prior to developing my critical replies to the Kantians, let me map these lofty “isms” to the argumentative context we started with initially.

    The argumentative context features dualists of various forms – gappists, mysterians, and the ilk. I will call them “anti-reductionists” since, in his paper, Churchland is a reductionist instead of an eliminativist and since dualists of various type and strength commonly deny explanation and other epistemological achievements of science typically only in a reductionist form en route to sweeping conclusions about the mind. Schier and Harker explain my mistaken understanding by underscoring that Churchland is only out to refute the claim that reductive physical theories are incapable of explaining and making predictions about qualia. And for this conclusion he needs only predictions not true or verified predictions. In other words, Churchland’s argument requires only explainy and predicty theories not true theories.

    First, a logical point: if all Churchland wants to establish is a refutation of one of the premises of anti-reductionists, the truth of physicalism (or reductionism, or materialism or…) is left in the twilight zone. Refuting one premise of an argument does not establish the opposite of its conclusion. The point applies, mutatis mutandis, to color reductionism, the identity of color qualia with some neural or “physical” posits. It too is left in the twilight zone. This is an unusually modest goal. I will call it “the modest argument.” Contrary to my critics, I read also many less modest claims about reduction, the truth of the hypothetical identities, even in this paper of Churchland’s. I will refer to this as the “bold argument.” So, second, in my critics, we have a modest argument and a premise about the importance of explainy and predicty epistemological values for naturalism (and for science) to underwrite a categorical distinction between the modest and the bold argument. These premise my misunderstanding of Churchland, for my critics. I also assume this criticism goes beyond Churchland exegesis, that Schier and Harker are actually critical of my empiricist naturalism.

    How to proceed argumentatively? It would be fairly straightforward to reply to my critics (and perhaps even Churchland even though I am unconvinced about his Kantianism) if they gave some reasons for endorsing the values of explainy and predicty theories and especially from Churchland’s paper and/or corpus. They do not. To reply, I will wade through the details of the premises given by Harker and Schier to protect the modest argument from my refutation of the bold argument. The outcome will be the same though: I will show why neither science nor naturalism should be saddled with pseudoscientific criteria, explainy and predicty theories. Executing this strategy will though take more than a few words and I apologize in advance for that. So on with the task.

    Harker thinks it is a “simple confusion” on my part to mistake predictions with their truth, verified predictions. I hope this charge is not premised on an idea that the distinction between the two, predictions and verified predictions, is a simple dichotomy. If it were, a dualism would be established within science between its theoretical or conceptual aspects, the conceptual norms of predicty and explainy science, and its experiential, experimental and material aspects, the truth-based criteria for theory-assessment. Such a dualism would be disastrous both for science and ontological naturalism. If these two “aspects” of science are not intimately connected (in our philosophical reconstructions) we must succumb to idealism or its modern neo-Kantian equivalents, paradigm-specific worlds and the ilk. Perhaps my Kantian critics are not swayed by such ramifications. But in that case, I can make a high prior-probability yet novel prediction that anti-reductionists would be overjoyed and celebrate by developing arguments with even more outlandish criteria of explainyness than the extant ones featuring criteria such as “a priori deducibility,” “intelligibility,” “logical supervenience” and the rest – criteria that Churchland himself has never grown tired of criticizing – while adding some novel imaginary ones concerned with the newly found criterion of predictyness, ones that I predict would put the endless zombie-phenomena to shame.

    In my view, even the conceptual distinction between prediction and verification is far from simple. The evidentiary goal of science is testable or falsifiable predictions (and hypotheses). Not any old logical consequence of a theory deserves to be called a “prediction” in empirical science. A prediction is something that can be false. And to make a judgment about falsifiability, we must know something about the empirical domains of study, true real phenomena. In fact we need to know a lot about these matters. This judgment and its bases percolates all the way to the far reaches of theory via models, analogies, metaphors and other “semantic” tools. The idea can be made clear from the semantic point of view. The sphere of theories and their logical consequences, predictions, despite its conceptual nature, gains implicit meaning from sources based in verifiability conditions, experimental truth-conditions. In the case at hand, a history of inquiry into color phenomena has generated content to theories, hypotheses, models and other representations through their links and looser associations to color phenomena. These associations have brought empirical content via truth-conditions to the “equations” or “conceptualizations,” theories and predictions. Had this history been just predicty, without truth, the conceptualizations would still be wanting in cognitive content, meaning. This much must be correct about semantic verificationism. Hopefully these remarks show that the truth of predictions cannot just be “beside the point,” as Harker claims.
    So let me continue with my critics’ premises. Harker provides a segway to the interconnected themes of my paper, verification, falsifiability and novelty in his further articulation of the reason for my alleged “confusion” over prediction and verification. To bolster his distinction between prediction and verification, Harker writes that, “showing that empirical evidence can be brought to bear on some debate is distinct from establishing that the evidence favours a particular theoretical model.” I am in full agreement. But I fail to see how this justifies a dichotomy between prediction and verification so that one would be only a “conceptual” issue not to mention one useful for Churchland who has a view of Harker’s predictyness – “how evidence is brought to bear” – that I criticize for its favouritism. To show how evidence is brought to bear on theories and theoretical debates will feature true predictions somewhere along the way, in history. And showing how empirical evidence is brought to bear on scientific controversy is precisely what undermines predictivity as a criterion independent from truth. The above statement of Harker’s is not about a conceptual distinction between verification and prediction but a methodological dictum. If showing how the world is brought to bear on theory would be the same as endorsing one theory, we would have only unfalsifiable theories. If the two practices were run together so that showing empirical evidence’s bearing would always be tantamount to establishing a specific theory we would be in the realm of pseudoscience. Real predictions are ones that can be judged in truth-value and for that to be possible they must have the potentiality of being false. But predictions from unfalsifiable theories are not testable. Popperians and logical empiricists would conclude that such theories have no information or semantic cognitive content, respectively. I have no idea why Harker thinks this formulation supports his charge about my conceptual confusion. After all, I am emphasizing falsifiability which requires articulating potential falsifiers. To do this, we must leave the confines of concepts and deal with the empirical world and address how our hypotheses “deal” with it, ultimately, for purposes of testing, but, also, for purposes of cognitive content.

    Let me summarize. I believe the major disagreement between Harker and me has vanished. Truth of predictions, the issue of verification, matters only if the predictions are falsifiable, could be other than true. This methodological axiom does not establish a dichotomy between prediction and verification but a definition of what a prediction is and, to boot, one based in the empirical methodologies of the science(s) in question. One does indeed have to know how truth is decided empirically, what empirical truth is as opposed to some mere conceptual possibility, to fulfil the definition of a prediction. Logical consequences of theories are not enough. So let us turn to the real issue of my paper, the one so eloquently raised by Harker’s own statement above. Let us explore how empirical evidence is brought to bear on hypotheses of SM-style color science and further explore whether this is achieved in a manner that remains open with respect to support for rival “theoretical models,” falsifiability.

    Churchland endorses a hypothetical identity that takes the form of an isomorphism between color responses and neural posits. This is not just a scientific neural hypothesis but also a rule about how evidence is brought to bear on this hypothesis and, for Churchland, probably, on other if not all hypotheses in this area of science. The evidentiary model, the model of testing, is itself structured in terms of an isomorphism. From a comparative perspective to scientific epistemology, I find this situation unusual even if not unique. Moreover, it tells us directly how empirical evidence must be brought to bear on theories so it is a way to answer Harker’s methodological question about how “empirical evidence can be brought to bear on some debate.” It is brought to bear through isomorphisms. But does this rule for using and creating evidence satisfy Harker’s full dictum that calls us to distinguish the evidentiary model from the verification of some one theoretical model? Let us loosen up by asking a few rhetorical questions.

    What if a theory of color vision does not employ a hypothetical isomorphism-based identity? What if rival hypotheses feature some other maybe “looser” hypothetico-deductive relationship between theory and evidence? Or, to make matters worse or maybe just diverse, what if the prediction is a probability distribution that we find difficult to capture in a deductive relation at all maybe because of some specific interpretation of probability? The questions are rhetorical. Sensory neuroscience can only be limited to an isomorphism-model of the evidentiary relationship and, thereby, rule out these and other alternative evidentiary models on extra-scientific grounds. On scientific grounds, one addresses this diversity in “styles of” predictions or maybe even predictyness on grounds of truth, verifications. This brings us back to the Kantianism of my foes. In disassociating truth or verification from explanation and prediction, I assume they would answer the above challenge “from diverse empirical practices” by maintaining that isomorphisms are explainy and predicty. But Kantianism is precisely the question at stake. In emphasizing truth or verification instead of truth-independent criteria, explainyness and predictyness, I am addressing the plausibility of Kantian naturalism, which in the case at hand takes the form of an isomorphism assumption.

    I do not know whether scientists endorse isomorphisms on grounds of Kantian extra-empirical “reasons,” their imagined and perceived past success, metaphysical intuitions, pragmatic limitations related to current levels of experimental practice, or something else. There is evidence from the literature that some indeed do take it as a “constraint.” But, at least in principle, their reasons are beside the point. The task of scientific practice is to test hypotheses. It is truth that counts instead of the many imaginable ways that hypotheses can map on to evidence and regardless of whether only some of these ways are predicty or make sense conceptually, “a priori.” At any given point in history, we need independent empirical criteria to judge the merits of hypotheses isomorphism-based ones included. This platitude brings us directly to the debate between Churchland and me, the nature and importance of novel predictions for falsifiability

    But, before that, let me update my response to my critics. Hopefully Harker has now some idea why I am asking for novel predictions that introduce new structural relations. The main reason for demanding novel evidence is that the isomorphism assumption is not just any old hypothesis subjected to test. It is also a specific interpretation of the evidentiary relationship which in this case concerns qualia and neural posits. And for those who believe in the symmetry thesis between prediction and explanation, like Schier and Churchland, it is also a specific articulation of the criterion of an “explainy” hypothesis. Isomorphisms are explainy regardless of truth, verification, in their view. I am asking for reasons on empirical grounds and apart from a priori explainyness since I do not believe in such criteria and surely not without reasons and Schier and Churchland neglect to give any. The crux of the issue is that, if we stick with this model of evidence, what it is to support a hypothetical identity, we must derive novel predictions about structures (that turn out to be true). How do we do achieve that?

    First a few words on how not to achieve it and why that mattered in and for my paper. The evidence I referred to as “old” in my paper does not test the isomorphism assumption. R. De Valois’ and his collaborators’ seminal classification and statistical analysis of the small sample of single cell data from macaque LGN that forms the bedrock of SN was based in psychophysical evidence, Hurvich and Jameson’s functional opponent model. These architects of SM were quite explicit about all this. Their use of evidence was fully theory-laden (or, better, model-laden). This did not stop earlier naturalist color ontologists, Hardin and Clark, from basing their famous reductionism, color subjectivism, on this very interdisciplinary model. My interest in Churchland’s article has that as its backdrop. Churchland promised novel predictions and evidence to support SM and the ontological color reduction, unlike the earlier reductionists. Novel evidence creates severe tests and these are the ones used to fight allegations of theory-ladeness, unfalsifiability, and other methodological evidentiary concerns. And to achieve novel predictions when dealing with a hypothetical identity structured in terms of an isomorphism requires new structural relations (the alternative is to entertain altogether different evidentiary models and, again, judge them through predictions, a topic I did not explore in my paper). Before exploring novelty, I am now at a point to reply directly to Schier’s version of my purportedly mistaken interpretation of Churchland’s goals.
    Schier thinks I have misunderstood the conceptual task Churchland and other naturalists are engaged in. For her, naturalists are the ones open to radical conceptual change behest empirical discoveries and the ones that serve these to the analytic, intuitionist or armchair, “pure” conceptual analysts. I have nothing other than deep admiration for this agenda. The issue is whether this is an accurate description of the case at hand. Does a systematic if not dogmatic allegiance to an isomorphism assumption really result in “empirical discoveries” about the mind-brain relationship? That is the issue at hand and, for me, verifications, empirical truth, must and can only decide. Schier apparently disagrees. But maybe she will change her mind if I point out that truly celebrating the conceptual broadening of ontological horizons, in this case, surely requires entertaining different articulations of the relationship between evidence or testing the isomorphism one through novel predictions. Only this manner do we achieve a naturalistic, science-based assessment of the very psychoneural assumption that Churchland presents as a hypothesis. This requires addressing verifications and those of novel predictions to which I will now turn.

    My critics were not persuaded by my idea that SM needs support from novel structural relations. Partly the reason may be that I did not review extant use or definitions of novelty. I agree fully with Harker’s suggestion that I should place the argument in a comparative epistemological context. This is the only possible methodology of philosophical naturalism at least for empiricists like myself. If we are to understand the methodology of psychology and especially for purposes of ontological conclusions we need to place it in a comparative scientific context. Yet I am not convinced that Harker’s specific comparative epistemology, the nature and use of novel predictions in physics and chemistry, is the best one for neuropsychology. These might in fact constitute a point of contrast given the broad unifying theories present in these domains. Biology seems more promising with model organisms and, thus, analogies, carrying the weight of unification. Suffice to say that my discussion of novelty was too brief. The situation is also affected by, one, Churchland’s indifference to conceptualizations of novel predictions present in the literature and, two, that his use covers psychological, heuristic, and theoretical senses of novelty. For this reply, I will start with the theoretical sense of novelty according to which novelty requires a comparative context of alternative hypotheses judged in terms of their potential relative to some prediction. With that in mind, let me turn to my critics.

    Schier accepts the theoretical sense of novelty. Yet, once again I am “confused” when asking for novel predictions with novel structural relations. She accepts my point that Churchland’s predictions are not novel but mere applications. Yet she underscores that “they are novel in that the predictions would not have been made without this theory.” I cannot find an argument for her claim. Churchland has shown nothing about the predictive prowess of any other theory. Churchland does not even discuss alternative theories. The theoretical sense of novelty requires that only the theory in question can make the prediction in question or, in probabilistic-to-incremental models of the method, that the likelihood of the evidence in light of alternative hypotheses is low whereas it is high for the hypothesis in question. To make matters worse, Churchland has not even shown that a “physical” theory is necessary for his predictions; he has not ruled out psychophysical models without any “physical” posits as sufficient for his predictions. He should have shown this to take Schier’s escape route. Let me elaborate on the last point.
    At this point in my original argument I was trying to be as charitable as possible to Churchland. I was asking only for novel structural predictions. I was thereby in a sense willingly limiting the relevant alternative hypotheses to those that assume an isomorphism relationship between theory and evidence. This is a significant limitation. There is nothing sacrosanct about this “hypothesis” about the mind-brain or evidence connection. Were we assessing theoretical novelty in an empiricist fashion, we would open up the competition for other kinds of hypotheses and evidence rules. But the base-line methodological requirement for Churchland’s hypothetical isomorphism identity is the derivation of novel predictions about structures. It seems Schier has granted that no such predictions are forthcoming from SM as currently practiced and/or represented and used by Churchland.

    Schier retorts, once again, to the “real intent” of Churchland and locates novel predictions exclusively within the confines of the “modest argument.” For her, the alternative hypotheses (or, maybe better, hypothesis) come outside of science altogether. Her argumentative maneuver transports novel predictions into a pseudo-epistemological context. We are asked to compare scientific hypotheses with linguistic practices outside of science and, to boot, in terms of predictyness including novelty. I doubt any one of us three-to-four is an expert on the ordinary language(s) of color. And more importantly, I doubt any one of us knows what such languages even try to predict qua colors. Ordinary language of colors does not capture knowledge of the type that should be called “theoretical.” It does not have much intra-color-linguistic structure or use for methodological criteria associated with theories except for robust presence of correlations between colors and other “properties,” most of these cultural qua objects. Its purposes are pragmatic and, most importantly for the case at hand, linguistic intuitions have nothing to do with aperture colors and other highly constrained stimulus conditions that are required for the science of color. So how could it make predictions about such stimuli or “objects?” The color laboratory is not the context where ordinary color terms and communication with them emerged or proceeds. We did not coin or teach color terms by making novices peer at un-localizable colors through peep holes. Ordinary color language is about colors in their ordinary environmental setting, object colors, colors of scenes and so on. And finally, on a more programmatic level, Schier’s style of argumentation reminds me of eliminativism not reductionism which is the case at hand. If Churchland thinks there is a “folk-theory” of color comparable to and in competition with a scientific one, he should sketch it. I for one cannot see how the two forms of knowledge could ever compete since their domains do not overlap enough.

    In my mind, the most charitable interpretation of Schier’s criticism and her argument is that, with this new comparative context, one that pits science against non-science, Schier has changed the topic. We are not debating over the theoretical sense of novelty because we are not comparing theories; we are debating over a psychological notion according to which sheer “surprise” is the essence of novelty (Mandik does just that in one of his papers that uses this piece of Churchland’s). But I am not party to that debate. With respect to psychological reactions, I have little to say. I can attest to the fact that some are “surprised” by Churchland’s novel colors. I witnessed such reactions at Linnfield College a few years back when Churchland demod his experiments in a darkened library atrium. Personally, I was more surprised to see Churchland at that little known college campus than his color show. The rest of the audience would have probably been equally surprised upon seeing mere monochromatic stimuli – if it was their “first.” But this (social?) psychology hardly matters for the issue at stake.

    Schier’s allegations about “confusions” on my part should not distract the readers from the real issue. Novel predictions are used in “scientific realism” debates and for a reason – evidentiary power. Churchland proposes a realistic interpretation of a scientific hypothesis, a psychoneural identity. The likelihood of its truth is increased through verified novel predictions; it may even be advanced from a regular “mere” hypothesis to an “absolutely” confirmed one through extended incremental novel confirmations. The issue (in addition to the other problems I showed in many paper) for the real argument about reduction is whether Churchland-style predictions can be made without his favorite “physical” theory. I have not seen an argument to the effect that they cannot in Churchland’s paper or received one from Schier. In my view, Churchland’s predictions can be made without any significant physical or, better, neural, or, even more accurately, non-physical neurocomputational hypothesis. It seems to me from Mollon and Webster and some other publications that they can be predicted from traditional psychophysical models. I suspect they could be predicted by just experimenting with color stimuli, for fun, without any hypotheses in mind. But that kind of fun does require some lab equipment not just “folksy colored objects.”

    Let me end the discussion of the topic of novel predictions with a few remarks to Harker’s concerns. Harker expresses a difficulty in understanding why “predictions of novel colour experiences, given exposure to unusual stimuli, fail as genuinely novel predictions.” I hope I have said enough about theoretical novelty to explain why. Novelty here is understood relative to alternative hypotheses. That determines novelty. “Unusual stimuli” should be unpacked the same way. Let alternative hypotheses deal with these and other, usual stimuli, in the psychophysical sense(s) of the term(s).. Both “novel” and “unusual” are inherently relative notions. From the standpoint that Schier approaches the matter, “ordinary knowledge and exposure,” every stimulus used in color science is novel and unusual. Normal humans just do not view “thin slices of electromagnetic radiation” in labs through apertures. Similarly, the colors we see as normal observers are complex mixtures. Without some scientific context, “novel” and “unusual” are free-floating notions.

    Harker also wonders whether my problems with the “kind” of novelty presented by Churchland relate to my claim about unfalsifiability. If there is a kind for Churchland, they do and in a straightforward manner. The reason why I am calling for the prediction of novel structural relations is not only to satisfy the evidentiary strictures for a neuropsychology with an isomorphism based evidence-rule. I also want to ensure that the interpretation of structural relations from experimental data satisfies falsifiability. I sort of built that into novelty in my paper, rightly or wrongly. The point is that the predictions must be testable and in a manner that does not allow for after the fact ad hoc modifications. Scientists, even more than the rest of us, are creative in seeing patterns in data including structural relations. SM and its history is a case in point. To satisfy falsifiability, the kinds of relations expected to be discovered in the data should be specified in advance. I see no other way to achieve this then by making predictions about novel relations in phenomenology or behaviour on the basis of neural hypotheses (or vice versa) and specifying the novel relations expected in as specific a form as possible. Such a testing protocol might have even received a nod of approval from Sir Karl. None of this happened when the foundations of SM were created and hence, from a methodological perspective, that evidence consists at best of accommodations. The data is “theory-laden” and by the very “theory” in question.

    I take to have responded to the three big criticisms. Before concluding on an even bigger issue for naturalistic ontology I want to briefly address some of Schier’s comments concerning the details of SM. I must start with a caveat: our disagreement has in part a sociological aspect of what via who to include in the SM paradigm. I do not know how to stipulate this. Paradigms are tricky to individuate as the many criticisms of Kuhn showed back in the 60s. As I said in my paper, there are many data models from different experimental paradigms that assume some sense of opponency. My attempt there at ordering was weak: I made the mathematical sense the common core. This is almost trivial in that mathematical models are always the least committed ones. But I did so because I do not really think there is a theory in this case at all and I was trying to identify something reasonably similar to one: Churchland himself mixes freely “theory” and “model” in his paper. But in the context of the current debate with Schier, I take the presence of a neural color code hypothesis a necessary condition for SM. That is my heuristic in responding here to her many other allegations of “misunderstanding and confusions.” The reason for it is that Schier emphasizes the neuroscientific aspects of SM for its predictive and explanatory prowess. I think she been carried away in her advocacy of Churchland-style color science with “opponent ideas” and sees hypothetical psychophysical posits as neural hypotheses and theories – but on to some of the details.
    Schier lists some facts well known in color science to show that there is fine detail to the structure of color space beyond opponency – which she claims is my exclusive focus. And such detail is important for judging predictive and explanatory power. This is as expected. I certainly did not mean to suggest SM is only about the neural color code. Mixing freely some lingo from (in)famous macro-epistemologists, assuming some “hard core,” a theoretical assumption among other less cognitive things, scientists add on to it other hypotheses, “protective belts,” when carrying on with “puzzle-solving” protocols of “normal science” across experimental circumstances. Scientists fine-tune the fit and extend the fit between a paradigm and evidence. But so what? I did mention testing holism. But my issue was falsifiability and novel predictions, progressive paradigms. But let us review some of Schier’s additions to “phenomenal structure” if only to briefly assess how exclusive SM is in its explainyness and predictyness, SM’s normal scientific methodology, and its progressive nature.

    Schier’s first addition concerns unique hues. But is this an addition? Opponency and unique hues are intimately connected. I cannot envision what shape SM might take as a neural opponent theory and one of phenomenology, as Schier and Churchland emphasize, without opponent unique hues. And less neurally yet still in version(s) of SM fitting my stipulation above, unique hues determine the very axes of the color space (planes), the representation of phenomenology, among other things. Thus, they define the whole framework for further data. After this, Schier lists some psychophysical effects and expresses these in the context of spatial representations, phenomenal color space. She returns to its 3-D nature and mentions saturation “asymmetries.” But three-dimensionality adds next to nothing to her earlier point since we already had a “space” erected for the phenomena. If anything, its mere three-dimensional nature is a weaker “effect” in probative power than the unique-hue determined space. Furthermore, both of these additional color facts have been known since the dawn of experimental color science. And, naturally, “theoretical” interpretations of them have varied. Here’s one for Schier to consider in light of her unwavering confidence in SM’s epistemological powers.

    Helmholtz’s color matching experiments of the 1850’s, carried out prior to the publication of his Handbuch that governed color and many other close-by sciences for decades, showed that mixing three simple spectral primaries to match the rest of the simple spectrum had to deal with “desaturation” and to varying degrees, depending on the choice of the primaries. The Young-Helmholtz three-color theory posited three fundamental sensations into the mind-brain and hypothesized about their “supersaturated” nature to account for facts of color matching. Contrary to SM, this theory equates white with desaturation and construes it as an “additive chromatic compound.” It is a sort of by-product of hues. Moreover, Helmholtz, the last true Renaissance man of science – whose expertise ranged from psychophysics through physiology to thermodynamics – and a die-hard empiricist – whose criticisms of Kantian epistemology are still worth reading – refused to accept the “psychoneural parallelism” or isomorphism axiom of his chief rival, Hering, the father of “opponent theories.” Consequently, Helmholtz cared little about the opponent unique-hue phenomenology even if he engaged in as much testing, prediction as any scientists ever. This “phenomenal space” was not a unique explanandum or uniquely probative effect for Helmholtz because his testing model was different from SM. Yet Helmholtz had his own “neural” hypothesis. Today, in SM, white is construed as a separate, independent channel from hues and fitted to the neural color-code hypothesis (although see Webster and Mollon’s suggestion to the contrary apropos their habituation experiments). What I see here is Kuhnian normal science. Once we make a foundational assumption we work out the details accordingly. If one makes a different foundational assumption the details are usually different. This does not take away the need for novel evidence but, on the contrary, underscores its importance. I don’t exactly know how the additional structure or “predictions and explanations” of Schier’s diminish the value of my arguments and conclusion. I also fail to see how the psychophysical data canned these days in spatial representations, “phenomenal color spaces,” would uniquely confirm some specific neural theory or demonstrate its unique epistemological prowess.

    Let me take up one more scientific controversy mentioned by Schier to expose the dubious strategies behind her conclusion about SM’s epistemic riches. She points out that Webster and Mollon’s findings do “not undermine the full version of SM” because instead of refuting the hypothesis under test we can adjust the Euclidean assumption about the nature of color space. Of course: Quine taught us that we could even adjust the “meaning of words” in the face of “recalcitrant observations,” falsifiers. So why not adjust the very space of representations; this kind of adjustment was done by French conventionalists and a few others some hundred years ago, in “space-time physics.” Apropos this (and other possible adjustments I already mentioned in my paper), Schier concludes “the very data that Seppalainen claims falsifies Churchland’s hypothetical identity are explained by the detailed SM which motivates the identity,” and further, that this and other data is “not at odds with the standard model, but actually explained by it.” Explained indeed but at what cost? Those of us who take holistic testing as an axiom will not be surprised by the many ways that empirical anomalies are dealt with in SM. Popper called the maneuvers that Schier proudly parades as advancing explainyness and predictyness “ad hoc modifications.” Kuhn called it normal science. I can’t make up my mind between Popper and Kuhn with respect to names but I do insist on novel predictions with both of them not just accommodations of the old same data together with a cavalier attitude towards protecting the hypothetical identity at all costs including unfalsifiability stratagems. If a model in psychophysics or neuroscience does not even specify the basic dynamical properties (“independence of the channels”), structural features (“number of channels”), or representational framework for data (“phenomenal color space”) we are dealing with hopelessly unfalsifiable posits and stratagems. My debate with Schier is back where we started or at least were a little while back: the (un)falsifiability of the neural identity-hypothesis concerning the opponent color code. I rest my case in light of Schier’s lucid illustration of this very point. I am more convinced than ever that if we turn this scientifically hypothetical identity under these circumstances of methodological controversy and practice into a reduction we are scraping the bottom of the ontological pail.

    Finally it is time for some concluding remarks. For the purposes of this comment, I hope my critics can at least entertain the bold argument. Naturalists turn to science for ontological insights. Reductionism is one kind of naturalism. But the origins of ontological reductionism are in epistemological reductionism. That, in turn, featured classically widely unifying theories. Few of us see such unifying theories in psychology. The type of reduction we are dealing with in the color case is not of this type but a “local” one. And this new reductive notion fits the inductive experimental practices of sensory neuroscience. There the use of isomorphisms is seat-of-the-pants style reasoning by way of analogies across experimental data-models (I recommend Davida Teller’s work to get a sense of this inductive style). That is why predictions, novel ones, are so important. The difficulties in accomplishing such may indicate that the ontology of and from sensory neuroscience should not be forced into a theory-reductionist form of unification. Being open to exploring other options for epistemological and ontological unification is not giving into dualists. We need other options to Kantian proto-dualism. Luckily there is plenty of material in the interdisciplinary experimental practices of sensory neuroscience for materialism – even if the outcome may fall short of identities.

  2. Tom
    I was making a point about Churchland exegesis, not about naturalism more generally. Here’s the passage that persuades me that he his defending what you call the “modest argument”:

    “Withal, how those experiments actually turn out is strictly beside the philosophical issue that opened this paper. We began by confronting the philosophical claim that no physical theory could ever yield specific predictions concerning the qualitative nature of our subjective experience. And yet here we have before us a theoretical initiative that yields precisely the sorts of predictions—in qualitative detail—that were supposed to be impossible.”

    I admit that, relative to this modest argument, physicalism and color reductionism receive no support. I admit that Churchland can’t resist tipping his hand through parts of the paper, conceding that he has expectations for how the experimental work will play out. But that’s not enough to conclude that he’s defending the bold argument. What passages within the paper lead you to believe that it is the bold argument that Churchland is pursuing?

    Second, here’s a quick sketch of why I value explainy and predicty theories: given a choice between a theory that has passed many tests and a theory that has not I prefer the former, other things being equal; given a choice between using empirical data to evaluate theories and ignoring empirical data that could be helpful for purposes of evaluating theories I prefer the former strategy, other things being equal. The preferable strategy cannot be pursued, however, unless the theories we are dealing with are explainy and/or predicty. Explainy, predicty theories are preferred over non-explainy, non-predicty theories, because they are subject to kinds of investigation that their contraries are not. When someone gets round to it and when the technology and funds are available, and so on, we will have data that may help advance a particular debate. We need not try to settle some issue through purely a priori principles. Furthermore, if some believe that a particular debate cannot be influenced by empirical discovery, then evidence that this is false would be important. It might be necessary to argue only that theories are explainy or predicty, if those theories are perceived as not being so.

    I will try and address some of your other points later.

  3. David
    Thanks and I agree fully with what you gave as a decision procedure for what kind of research projects to pursue. But in my response, I claimed (and defined) explainy and predicty to be without verification: truth was detached from explanation and prediction. That’s the only way I see that the paragraph you quoted from Churchland leads to your and Schier’s generalized claim that “truth” is not the issue for him. Churchland is careful in the quote to say that “how THOSE experiments actually turn out…” But this disregard of verification for this “one” experiment/prediction cannot be extended to others viz., generalized. For predicty and explainy to matter, somewhere along the way SM must have real predictions – ones that could be false but turned out to be true upon review of data.
    You defined above explainy and predicty with verification, truth. Churchland cannot have it both ways: the modest argument rests on verifications viz., the bold argument.

    Per what Churchland explicitly claims about the bold argument: how about the claim that he is testing the original identity theory in one of its guises (in addition to my analysis that he cannot have modest without bold)? It was somewhere in the beginning – sorry I don’t have access to it now.


  4. Sorry to be so slow in responding to the preceding conversation. I’ve been on a ski holiday in the mountains above Salt Lake City for the past several days, and my email connection from there was suddenly wiped out by a two-day snowstorm of biblical proportions. But I’m now safe and back in San Diego, so let me add just a few comments to the many already in place.

    My original reaction to
    Seppalainen’s critique was profound confusion, for I had difficulty recognizing the Standard Model (or my interpretation of it) in his criticisms of it. I then read Harker’s and Schier’s lucid responses (as the snowstorm was gathering force) and I decided that they had jointly responded to his piece at least as well as I possibly could have (my thanks to you both), and so I would simply stand by and let others evaluate the overall situation. The snowstorm overdetermined that decision, of course, and I had to wait a few days to reconnect, and to discover Seppalainen’s long follow-up.

    I will not enter into that dense discussion. But I will confess to believing that the SM outlined in my original paper is *at least* a first approximation to the truth of chromatic processing and color coding in humans, and that its many predictions (*hundreds* of them) about the chimerical-color sensations are both true AND novel (the dark ones, anyway). The ‘modest’ position is indeed of real importance, given the foolish things that past philosophers have said about the limitations on physical theories, but I’m quite prepared to defend the ‘bold’ position. It is what motivated my paper in the first place.

  5. Please excuse the lateness and (hopefully only slight) incoherence of this response. I blame it on the sleep deprivation caused by my ten month old’s cold. I also appologise for the typos – much of this was written one-handed while he slept on me. So what follows is just a few random thoughts.

    I wanted to say thanks for returning the favor and making me think a lot about scientific and philosophical methodology. As it is not my area of expertise (I am a philosopher of cognitive science/mind) I haven’t ventured to say anything much explicitly about it, but you certainly have got me thinking.

    I continue to be confused by your talk of the ‘ isomorphism-model of the evidentiary relationship’
    Isomorphism is not meant to capture the relationship between evidence and theory. Rather, it is, in this case, a (hypothesized) relationship between two things in the world; namely phenomenal color space and neural activation space. Evidence that they are isomorphic would support the identity. Of course such data by itself would be far from decisive. Many things are isomorphic but not identical.
    Given that isomorphism does not capture the evidentiary relationship, what matters is that the hypothetical identity (within the context of SM) gives rise to a prediction about how to elicit apparently “conceptually incoherent” color experiences. Novel structural predictions are not necessary.
    I certainly don’t think that isomorphisms are explainy or explanatory. Rather they are (if discovered) data to be explained.
    The hypothesis that they are identical leads to the hypothesis that the spaces are isomorphic and therefore motivates looking for the isomorphism. The work of de Valios et al suggests that the place to look for the neural states that are identical to phenomenal colors is not in the thalamus but in the cortex. They do not suggest we need to give up on opponency altogether, but rather that what is required is not the “cone” opponency found in the LGN but “perceptual opponency” found in the cortex.

    It seems to me that your overall point is that Churchland is not being a good naturalist. I think that this is because there is a real sense in which Churchland is not using naturalistic methodologies in this paper, instead he is using the decidedly non-naturalistic methodologies of his opponents.
    For example you say that‘Detaching truth … from some other “cognitive” and maybe even epistemological goals and achievements of science, is for many philosophers of science and all scientists highly suspect.’
    I and as you point out, Churchland in the past (and I am sure also in the present), agree. The problem is that our dualistic opponents don’t. They think that conceivability is a guide to possibility and more importantly in this instance that inconceivability is a guide to impossibility. For them it is inconceivable and hence impossible that science can make any explanatory contact with phenomenal qualities. Chuchland’s goal is to show that it is conceivable and hence possible.
    But you are right that this means that naturalists such as Churchland seem to be accepting some very non-naturalist claims. No self-respecting scientist would be concerned by the conceivability of zombies. But I think it is a mistake to actually pin these methodological views on Churchland. Rather I think that he has effectively granted these claims for the sake of argument, in order to have some common ground on which the debate can be resolved. The alternative is to challenge the underlying methodological principles, a much harder task. To use an analogy, he thinks their game is wrong, but instead of trying to show them they are playing the wrong game, he aims to take them on at their own game and beat them. This is not a totally alien methodology for Churchland. It is clearly what he has done in the paper for this conference.
    This is all the more complicated by Paul’s post where he says he also wants to defend the bold claim ….

    Finally I just wanted to say that we obviously disagree quite a lot as to what is in the SM. What do you take to be part of SM and what not? I don’t think SM is in tension with Helmholtz. It is very difficult to find clear statements of SM, but I think Palmer does as good a job as anyone (Palmer, S. E. (1999). Vision Science: Photons to Phenomenology. Cambridge, MA, MIT Press. chpt 3). He says current model combines Helmholtzian three channels (at first stage) with Herringesque opponent processes and finishes with a third whose dimensions are hue saturation and lightness.

%d bloggers like this: