I have a question about observational theory-dependency (or theory-ladenness), particularly concerning the issue of perceptual theory-dependency.
It is quite often debated whether perception is theory-dependent by debating whether perception is cognitively penetrable. That is, whether perceptual organs and brain systems are, in a sense, informationally encapsulated from the higher, cognitive areas of the brain, where such background theories and knowledge that could presumably influence perception reside. If brain processes in the higher, cognitive areas do not interfere with those processes in the lower, perceptual areas, then perception is cognitively impenetrable. So, while perception may influence one’s thoughts, the latter cannot influence the former. Like I said, this is a frequently debated issue, but my concern is not with the debate directly but rather with a side issue. (Jerry Fodor and Paul Churchland have famously debated this issue in the past, and more recently Athanassois Raftopoulos and other philosophers have brought up the issue again. Here at the Library, Hugo discusses the issue briefly in his broader essay on Theory-ladenness, where he also explains why the latter is an important issue in philosophy. Links for the Raftopoulos articles, along with other cited references, are at the end of this post. I’d also be happy to point interested readers to any other mentioned or related authors and articles.)
Let us assume that perception is in fact cognitively impenetrable and, thus, is not theory-dependent in the cognitive sense. Nonetheless, it is a widely held view (even by Fodor and many other proponents of cognitive impenetrability) that perception is a computational/information process in which the perceptual systems of the brain compute and process stimulus received at the sensory organs to produce the perceptual output that we ultimately experience. The perceptual system does not simply relay sensory information from the sensory organs. The computations and information processes are quite complex and involve many layers. These computations are based on certain low-level knowledge about regularities in an organism’s perceptual environment, acquired through evolutional and developmental (e.g., child development) mechanisms. This knowledge includes principles such as local proximity (adjacent elements in a scene belong together), colinearity (straight lines remain straight across a scene), cocircularity (curved lines keep a constant radius of curvature across a scene), and numerous other principles and constraints (see Raftopoulos 2001, p. 429; Gilbert 2001 pp. 691-693; Geisler 2008). This low-level knowledge is sometimes referred to as theories. So, regardless of whether perception is cognitively penetrated and theory-dependent in the cognitive sense, it is nonetheless a computational or informational process that is dependent on some kind of low-level knowledge or theories. All of the above is pretty much accepted by Fodor and his defenders. (There are, however, those who would disagree with this computational account of perception. James Gibson and Rodney Brooks, for example, hold a non-representational or ecological model of perception. See also recent discussions on direct realism.)
My question is why are proponents of cognitive impenetrability (and other philosophers who worry about theory-dependency) seemingly unconcerned or less concerned with the computational nature of perception and its low-level knowledge/theory-dependency?
Consider the following passage from a paper by Athanassois Raftopoulos (2001, p. 424) describing Jerry Fodor’s views on the matter:
“Fodor’s (Fodor, 1984) argument is that, although perception has access to these background theories [the low-level theories that perceptual computational processes are based on] and is a kind of inference, it is impregnable to (informationally encapsulated from) higher cognitive states, such as desires, beliefs, expectations, and so forth. Since relativistic theories of knowledge and holistic theories of meaning argue for the dependence of perception on these higher states, Fodor thinks that his arguments undermine these theories, while allowing the inferential and computational role of perception and its theory-ladenness.”
First, I was unaware that arguments concerning relativistic theories of knowledge and holistic theories of meaning necessarily relied on the observer’s higher cognitive states (thoughts, beliefs, etc.). I assumed that they simply relied on the assumption that perception (and observation in general) exists within the context of other theories and knowledge, which even cognitively impenetrable perception does.
Second, even if such arguments did only rely on the observer’s higher cognitive states (e.g., high-level theories), I don’t understand why they wouldn’t work just as well if perception was only dependent on the lower, non-cognitive states (e.g., low-level theories). For example, if two observers whose perceptual systems are wired differently claim to have two different and mutually inconsistent (or incommensurable) perceptual experiences, how do we judge which one is correct and, thus, which perceptual system (which set of low-level theories) is correct? How is this different from judging which of two high-level theories of perception or observation is correct? (I am not claiming that such judgements can’t be made, but only asking how they differ in the low-level and high-level cases. Another way to put it is that, if such judgements can be made with low-level theories, then why not with high-level theories—as indeed some philosophers claim?)
To continue, if competing high-level theories of perception can lead to relativism (as well as skepticism, circular reasoning, etc.), then why not competing perceptual systems? In neither case is there such a thing as empirically pure knowledge that one can use to arbitrate between the different theories or different perceptual systems. Both in the cognitively penetrable and impenetrable cases, the observer is relying on perceptual information that has been processed by the brain, and thus susceptible to bias and errors, to make reliability judgements. (Perhaps one can say the cognitively impenetrable case involves less processing and less variability in processing than the cognitively penetrable case and, thus, is less susceptible to bias and errors and, hence, relativism claims, although not immune to them. And so perhaps it makes the case against relativism (and for realism) a little easier to argue.)
To put it yet another way, even if it turned out that perception is not theory-dependent (not cognitively penetrable), are we not still left with the original epistemic problem of showing that our perceptual experiences or beliefs are true or accurate? The latter being a problem because we don’t have direct, independent access to the external world (unless one believes in direct realism). Though we may not be dependent on theories, we are still dependent on our bodies and mind, our perceptual faculties. How do we know that our perceptual faculties accurately represent the external world when our only access to that world is through our perceptual faculties? This is part of the “problem of the external world” (see Stanford Encyclopedia of Philosophy entries on “Epistemology” and “Epistemological Problems of Perception”). Attempts to resolve it have led to skepticism, idealism, and other such views, not unlike the potential consequences if it turns out that perception is in fact theory-dependent (cognitively penetrable). (I’m not claiming here that such pessimistic views are the only conclusion. It is a continuously debated issue, and realism views have also been expressed via various arguments: (abductive) inference to the best explanation, success semantics, reliabilism, etc. I was only trying to point out how the epistemic problem of perception and the issue of cognitive penetration share similar worries.)
I will now briefly list some reasons that seem to be (or could be) expressed in the literature as to why philosophers are seemingly unconcerned or less concerned with the computational nature of perception and its low-level knowledge/theory-dependency? (As a consequence of this attitude, however, these reasons—and perhaps others—are not often discussed or defended, to the best of my knowledge.)
(1) Some philosophers are just unaware or not fully aware of the computational nature of perception. Others hold a non-representational, ecological, or direct realism view of perception and, thus, reject that perception is a computational process. Either way, these philosophers believe that perception, even if cognitively impenetrable, delivers unmediated, direct factual information about the environment. I will put aside these group of philosophers, as we are interested in why philosophers who recognize and accept the computational nature of perception are, nonetheless, unconcerned.
(2) One reason seems to be that the low-level theories that perceptual computations rely on appear to represent “general truths” about our world, about “general reliable regularities about the optico-spatial properties of our world” (Raftopoulos 2001, p. 429). My question is how is such a conclusion reached. Are we not relying on perceptions produced by processes that are based on these very “truths” and “reliable regularities” to judge that the processes themselves are truthful or reliable? Is there not the possibility of circularity and relativism here as well, as with high-level theories? (Again, I am not claiming that judgements can’t be made in such a way—many philosophers offer strategies for averting or dealing with circularities when it comes to high-level theories of observation. I’m asking why in the low-level case such judgements are often portrayed as being prima facie unproblematic, unlike in the high-level case.)
(3) Another reason seems to be that low-level theories are generally fixed and hardwired, and so are not affected by our desires, beliefs, or expectations, as high-level theories can be (*). And moreover, the low-level theories are generally consistent from one human observer to another, unlike high-level theories (*). But fixedness of one’s perceptual experience and agreement (consistency) with others does not necessarily imply correctness, truthfulness, or accuracy. After all, a change in theories can sometimes be for the better, and a consensus can be mistaken. (*Recent neurological evidence on perceptual plasticity, learning, and variability may undermine these reasons to some greater or lesser extent. See, for example, Gilbert 2001.)
Some philosophers argue that the flexibility associated with the cognitive penetrability of higher level organisms evolved as a way to serve the individual needs of different agents within their lifetime (Goldstone 2003; Macpherson 2012, pp. 32-33). Otherwise, the perceptual system would have to come ready-made to represent an enormous amount of perceptual possibilities, which would have rendered it too large and unwieldy. For example, the perceptual system may come ready-made to represent faces, but not necessarily any individual face. In this sense, we can look at the high-level theories as customizable “software” extensions of an underlying perceptual “hardware”. On such a view, it makes less sense to draw a truth bearing distinction between the two; it is not the form (hardware or software) that matters, but the content and how accurately it serves to represent the environment—but see (4) below. (It is also possible that individual perceptual needs could be served via some form of low-level perceptual learning; see Goldstone 2003 and Gilbert 2001. But then that would imply that the low-level theories are not entirely fixed or hardwired. Also, bear in mind that some proponents of cognitive impenetrability would deny that individual faces or even faces, in general, constitute perceptual content; rather, it is the low-level constituent features of the face that constitute perceptual content. See Macpherson 2012, pp. 31-34 and also the Stanford Encyclopedia of Philosophy entry on “Contents of Perception”.)
(4) Some may consider low-level theory-dependency to be more objective and reliable than its high-level “evil” cousin. Low-level theories have evolved over millennia in the perceptual systems of biological organisms to be highly reliable. On the other hand, high-level theories are created by us imperfect human beings and can be fallible. But this seems somewhat unconvincing to me. After all, many high-level theories of perception (and observation more broadly) can be reliable, and the low-level theories can be mistaken at times, as in the case of visual and other perceptual illusions.
(5) Last but not least, low-level theories may avoid certain problems of reference associated with high-level theories. Raftopoulos (2008, pp. 78-80; 2012, pp. 14-15) acknowledges that perception is dependent on certain low-level theories or knowledge, but adamantly argues that they are non-conceptual in nature, unlike high-level, cognitive theories. (The non-conceptual nature of perception, in general, is a debated issue, which very much depends on what we mean by “concept”, a debated issue in itself. See Tacca 2011 and the Raftopoulos 2012 reference above and also the Stanford Encyclopedia of Philosophy entries on “Contents of Perception”, Section 6, and “Nonconceptual Mental Content”, Section 4.1.)
Now, in the non-conceptual case, just as in the conceptual case, we still face the problem of circular justification (epistemic problem of perception): how do we know that our perceptual faculties based on these low-level theories or knowledge accurately represent the external world when our only access to that world is through these very same perceptual faculties? Typically, in the conceptual case, philosophers argue around the circularity by appealing to pragmatic or aesthetic considerations. A pragmatic argument (a la success semantics) might be that observations based on some theory or perceptual system allow an organism’s interactions with its environment (to find food, reproduce, etc.) to be more successful than observations based on some other theory or perceptual system, and therefore the more successful theory or perceptual system must be the truer one. But such pragmatic (or even aesthetic) arguments are not without their problems. Particularly, some philosophers claim that pragmatic arguments, like success semantics, are plagued with certain problems of reference (see Raftopoulos 2008, pp. 66-76). And some philosophers, like Raftopoulos—and this is the point here—claim that these problems of reference can be avoided if perception is non-conceptual rather than conceptual. (Though this is debated; see SEP entry on “Nonconceptual Mental Content”, Section 4.1.) Then barring other problems with pragmatic arguments, they can be availed to circumvent the problem of circular justification when it comes to perception.
Thus, given that perception is a computational process, there seems to be some legitimate benefit to denying that perception is also cognitively penetrable (a contentious move), but only if one is also willing to deny that the remaining low-level theories of perception are conceptual in nature (possibly also a contentious move). While these moves do not eliminate the original problem of circular justification associated with theory-dependent (cognitively penetrable) perception, they may make appeals to pragmatic considerations more viable, assuming that non-conceptual perception can solve certain problems of reference that conceptual perception presumably cannot. Additionally, it is possible that non-conceptual perception may help solve other problems in the philosophy of perception. Still, at least with regards to circular justification and perception, the aforementioned and only contingent benefit somehow does not seem to be worth all the fuss surrounding cognitively penetrable perception, given that perception is already a computational process.
Perhaps some of you know of more significant reasons, but I will leave it that. I apologize if the post is too long. I hope some of you find the query interesting and worthwhile. Maybe I’m just missing something simple, and someone can enlighten me in two lines. (I will kick myself but will be happy to lay the matter to rest.)
Geisler, Wilson S. “Visual perception and the statistical properties of natural scenes.” Annu. Rev. Psychol. 59 (2008): 167-192.
Gilbert, Charles D., Mariano Sigman, and Roy E. Crist. “The neural basis of perceptual learning.” Neuron 31.5 (2001): 681-697.
Goldstone, Robert L. “Learning to perceive while perceiving to learn.” Perceptual Organization in Vision: Behavioral and Neural Perspectives (2003): 233-278.
Macpherson, Fiona. “Cognitive penetration of colour experience: rethinking the issue in light of an indirect mechanism.” Philosophy and Phenomenological Research 84.1 (2012): 24-62.
Raftopoulos, Athanassios. “Is perception informationally encapsulated?: The issue of the theory-ladenness of perception.” Cognitive Science 25.3 (2001): 423-451.
Raftopoulos, Athanasios. “Perceptual systems and realism.” Synthese 164.1 (2008): 61-91.
Raftopoulos, Athanassios. “The cognitive impenetrability of the content of early vision is a necessary and sufficient condition for purely nonconceptual content.” Philosophical Psychology ahead-of-print (2012): 1-20.
Tacca, Michela C. “Commonalities between perception and cognition.” Frontiers in Psychology 2 (2011).
Stanford Encyclopedia of Philosophy Entries:
Bermúdez, José and Cahen, Arnon, “Nonconceptual Mental Content”, The Stanford Encyclopedia of Philosophy (Spring 2012 Edition), Edward N. Zalta (ed.), URL = <http://plato.stanfor...onconceptual/>.
BonJour, Laurence, “Epistemological Problems of Perception”, The Stanford Encyclopedia of Philosophy (Winter 2012 Edition), Edward N. Zalta (ed.), URL = <http://plato.stanfor...ion-episprob/>.
Siegel, Susanna, “The Contents of Perception”, The Stanford Encyclopedia of Philosophy (Spring 2013 Edition), Edward N. Zalta (ed.), forthcoming URL = <http://plato.stanfor...ion-contents/>.
Steup, Matthias, “Epistemology”, The Stanford Encyclopedia of Philosophy (Winter 2012 Edition), Edward N. Zalta (ed.), URL = <http://plato.stanfor...epistemology/>.