|Philosophy of Mind
Why Granny Should Have Read French Philosophers:
The Phenomenology of Fodor or
David F Wolf II
ABSTRACT: In 1983, Fodors Modularity of Mind popularized faculty psychology. His theory employs a trichotomous functional architecture to explain cognitive processes, which is very similar to Merleau-Pontys phenomenology of perception. Each theory postulates that perception is a mid-level procedure that operates on transduced information and that perception is independent of our cognitive experience. The two theories differ on whether perception is informationally impenetrable. This difference is essentially an empirical matter. However, I suggest that Merleau-Pontys allowance of cross-modal communication within perception explains our ability to identify features in noisy backgrounds better because his theory offers a more definitive ontology that matches human substantive behavior. Likewise, evidence within cognitive science suggests that Merleau-Pontys phenomenology is a more accurate depiction of human cognitive processes.
Fodors modularity thesis popularized faculty approaches to cognitive psychology. This theory bears a striking resemblance to the phenomenological theory that Merleau-Ponty proposed two decades earlier. Both theories employ a trichotomous functional architecture to explain cognition and view perception as a mid-level processing of information that lies between the world and consciousness. The key feature that differentiates the views is whether that middle level of processing is completely impenetrable by consciousness. If Fodor was to relax his strong position of the impenetrability of information in modules, modules could both be somewhat encapsulated and maintain a general independence from consciousness. Then only the degree of perceptions independance from consciousness would distinguish his theory from Merleau-Pontys. Currently, both theories can account for the substantive, outward, behavior of humans. Only the procedural behavior, the internal process, differentiates the theories. The conundrum of deciding between the theories is resolvable by an empirical critical experiment. While this will require more knowledge of cognitive psychology, current evidence suggests that Merleau-Ponty was correct and the mind is less encapsulated than Fodor's original claim.
The Two Theories and Their Resemblance
Merleau-Ponty distinguishes three aspects of the psychological process; basic sensations, perception, and the associations of memory (Merleau-Ponty, 1994). Basic sensations receive raw information from the world and transduce them for our perceptual processes. Perception unifies the infinite amount of information about our environment, from our environment, into a meaningful structure. Perception is interpretive, but its presentation of the world is as distal and objective. There are three central features of perception for Merleau-Ponty. First, perception is synthesized independently by the body and not by the mind (consciousness). He states,
Perception acquaints the viewer with a directed object, not by creating objects in consciousness qualia, but by being a conduit to the observed object. The final feature of perception is its informational opaqueness. Our bodies supply us with representations of the world and our consciousness is not aware of the qualities within our prestructured perceptual devices. Therefore we cannot make judgments about the fundamental assumptions of the mapping systems implicit within perception. The third aspect of the psychological processes, association of memories, is the thetic aspect of our phenomenological experience. It is responsible for conscious thought, which includes the fixation of beliefs.
Like Merleau-Ponty, Fodor divides psychological processing into a three-part functional taxonomy. These parts are: transducers, modules (input systems) and a central system (Fodor, 1983; 1985). Transducers, like Merleau-Pontys sensations, take external stimuli and transform it for the input systems to use. Modules are rigidly specialized, domain specific, innately specified, autonomous, unassembled input systems that are encapsulated. Encapsulation entails that the inference producing mechanisms have general features of their architecture that constrain both their access to background information and the information that they output. Thus modules are structures so that: i) minimally some information that is accessible to the cognitive process is unavailable to the module and ii) the operation the module performs has access to only the information in the database of that module and the retransmitted stimulus. Finally, central systems globally and isotropically access information. This is typical unencapsulated behavior it is holistic and Quinian (Fodor, 1983, p. 117). Central systems are responsible for the cognitively penetrable psychological behavior, consciousness, and fixes beliefs from representations supplied to it by modules.
There are obvious similarities between Fodors modularity thesis and Merleau-Ponty's theory of psychological processing. First, both transducers and sensations receive information from stimuli and retransmit it to a middle level system to process. Merleau-Ponty refers to this level as perception and Fodor calls the level a function of modules. In each case this level, perception, is prior to the fixation of belief and belief fixation occurs in consciousness (a central system). Second, both theories have perceptual processes create representations of the objective world, instead of proximal reflections of the stimuli. For example, Merleau-Ponty writes, "we make perception out of things perceived (1994, p. 5)." Third, perceptual systems are prestructured and this structure determines how information is presented. Again Merleau-Ponty states, "that within the organism the structure depends on variables such as the biological meaning of the situation (p. 11)." Fodor argues that,
Fourth, the perceptual processes architectural procedures are informationally opaque, i.e., the fundamental assumptions implicit within the perceptual processes are not observable or accessible by consciousness.
The Theoretical Variances and a Schema for Adjudication
The key difference between Fodor and Merleau-Ponty rests on Fodors strong claim about encapsulated systems, i.e., they are informationally impenetrable. To Fodor, modules only employ a preprogrammed theory of mapping distal causes to proximate effects and nothing else. The flow of information in psychological processing is unidirectional, i.e., from transducers to modules to the central system. Consequently the central system cannot influence, in any way, how humans perceive things.
To Merleau-Ponty, there are instances where the consciousness can communicate with perception to effect how perception interprets information from basic sensations. In the case of sexuality, Merleau-Ponty suggests that perception cannot account for a complete sexual context, "perception has lost its erotic nature (Merleau-Ponty, 1994, p. 159)." Instead, erotic perceptions and experiences are a product of inter-communications with his trichotomous functional processes. He states that,
Merleau-Ponty also suggests cross modal communication in aspects of tactile and visual perception. Thus, to Merleau-Ponty perception is informationally penetrable.
To determine if perception is informationally impenetrable, we need to show that some other system can affect a particular modules output. McGurke and Mcdonald (1979) document behavior that supports such as case. They revealed that visual stimuli directly affect phonetic perception. Their research suggests the most plausible explanation of the observed phenomena is the existence of inter-module communication. Presumably the visual module sends information to the auditory module to determine the sound of particular phonemes. While this phenomenon is not a top-down instance of consciousness influencing perception, it does give evidence for cross modal, i.e., intermodular, communication. This would disprove the strongest conception of the encapsulation of modules, such as the version proposed for language research by Forster (1979).
A more interesting case of penetrable information flow is top-down aspects of cognition, viz., consciousness affecting perception. This instance is most noticeable when faced with noisy stimuli. The famous picture of a Dalmatian on a dappled background illustrates how we can discern images from seemingly chaotic fields by consciously restricting the domain that we search. If we do not see the image and are informed of its presence, we look for and often discover the canine. Phonetic identification probably best illustrates this type of phenomena. Often we may listen to a song and perceive an incorrect lyric, such as, "let Milo open the door." Later when you learn that the lyric is actually, "let my love open the door," you from that moment on hear the lyric correctly. Samuels (1981) work on phonetic restoration rigorously identifies how the context of a sound contributes to how we hear it. If we play a pseudo-word and a blocking sound, e.g., basi and a beep, the context surrounding the noisy stimulus will determine what phoneme the subject hears. In the above case, the context could determine if the subject hears basis or basic.
It appears that in these instances consciousness is somehow priming perceptual devices, modules. The prime instructs the module to restrict its processing to a particular context or domain. It acts as a heuristic given to the module that allows the module to identify a particular representation. Thus, this priming of a perceptual device, or module, greatly prunes the inappropriate potential search options and makes the problem of identification tractable. Coincidentally the perceiver can notice the pattern faster.
The above instances suggest evidence contrary to the claim that perceptual modules are synchronically impenetrable to background cognitive knowledge. Fodor (1992) admits that such a schema is implausible. Instead, he allows a lesser degree of informational penetrability to modules, i.e., they are diachronically penetrable. In effect, training and experience can affect the availability of certain contexts or background theories to modules. Fodor states,
Likewise, Fodor reluctantly accepts top-down instances of perception. He writes, "perception is diachronically penetrable only within strictly maybe endogenously defined limits (Fodor, 1992, p. 248)." Fodors theory then allows for the central system to train modules along a rigidly defined range. This level of cognitive influence can train a module much in the same way a training set for a back-propagation connectionist machine molds the overall machine. Simply, the central system can create a limited set of feedback that will develop the modules capability, but these instances are not common. This level of cognitive influence on modules, however, is not as interactive as Merleau-Pontys theory and does not conform with the aforementioned evidence for the cognitive priming of modules.
Nevertheless, Fodor does not have to accept that the above cases confirm top-down processing. Variations or plasticity in output from noisy environments may not necessarily be a feature of the informational penetrability of modules (Karmiloff-Smith, 1992). A serial feedforward, i.e., unidirectional and encapsulated, modular architecture could account for instances of processing information about ambiguous contexts (Borland & Cutler, 1996). The modules could be producing two, or more, representations for the central system to select. Alternatively, Jordan and Jacobs (1992) proposed that with divergent computation multiple modules could be acting on a single input source to offer variant outputs for evaluation. The priming feature actually acts as a contextual cue within consciousness, i.e., the identification of the representation influenced by the cue occurs in the central system and not in the modules (perception).
While this answer appears to ignore interesting perceptual identification tasks, the claim simply states that the filtering process doesnt occur within the modules. Instead, discerning phenomena occurs in consciousness or in a preconscious state within the central system. Therefore, the aforementioned noisy stimuli cases can be consistent with Fodors modularity theory provided that the central system is sometimes performing an unconscious selection or filtering process that chooses representations in light of their appropriateness.
One should note that keeping modules informationally impenetrable reduces the role modules play in perception. Now the central system is responsible for performing particular acts of perception. Mainly, modules supply the central system with a set of representations and the central system must in turn apply environmental cues to use as a filter for selecting the perceived representation. This modification or filtering process directly affects the later process of the central system, i.e., the fixing of a belief to the representation. Thus this taxonomy blurs the clarity of separating conscious behavior from subconscious manipulation of information that presents the image to consciousness. Failing to crisply make this distinction between these processes, ceteris paribus, makes maintaining informational impenetrability a less desirable psychological taxonomy.
However, our bias to have a psychology that distinctly identifies our phenomenological experiences is not a definitive tool for determining the degree of informational opacity of modules. Essentially the claim about the informational impenetrability of modules is empirical (McClamrock, 1989). Cognitive psychology needs to determine more about information structure of the mind to determine the degree of top-down behavior that influences perception. With more information about the procedural behavior of the psychological process of cognition, we could create a critical test to verify the level in which modules are informationally impenetrable.
I have shown that Merleau-Ponty and Fodor have a number of similarities between their taxonomies of cognition. The key feature that differentiates the two lies in determining the particular information flow into perception and consciousness. Merleau-Ponty argues that cross modalities can affect various perceptual processes. On the other hand, Fodor argues that the information flow is generally unidirectional and modules are informationally impenetrable. With instances of noisy stimuli, people do appear to have some ability to consciously effect their perception (Wolfe, 1990). Fodor can explain this phenomenon by creating a taxonomy that can produce the same substantive behavior without implementing bi-directional or recurrent modules. Thus, determining which theory is better is a matter of gaining more information about cognitive behavior and setting up a critical experiment. If modules could be slightly influenced by particular processes, either through rigid inter-module connections or similar communicative links between consciousness and modules, then the modules could be generally informationally impenetrable (to a degree) and explain priming. Until Fodor allows for this, or drops the requirement of informational impenetrability completely, his theory will differ from Merleau-Pontys in a procedurally interesting way.
Currently, any reference to encapsulation within cognitive science is remarkably subtle (Dell, 1996). While the underlying cortical structures of the brain display a certain degree of independence (Shallice, 1988) support their possible modularity and independence, there is compelling evidence that these structures have recurrent connections (Sejnowski & Churchland, 1989). Further, evidence also suggests the computation of vastly distributed representations within cognition (Georgopoulos, Taira, & Lukashin, 1993). Consequently, the most fecund interpretation among today's research in modularity rests on an interpretation that allows for partial or occasional bi-directional processing of information (Spivey-Knowlton & Eberhard, 1996). This interpretation runs contrary to Fodor's informational impenetrability requirement for modules and is remarkably close to Merleau-Ponty's phenomenology.
Notes(1) I would like to thank Ron McClamrock for insightful comments.
Borland, J. & A. Cutler (1996) Interaction with Autonomy: Multiple Output Models and the Inadequacy of the Great Divide. Cognition. 58: 309-320.
Dell, G. (1996) Models and Modularity in Langauage Processing. Proceedings of the Eighteenth Annual Conference of the Cognitive Science Society. Earlbaum: Hilldyle, NJ. 47.
Fodor, J. (1992). Observation Reconsidered. A Theory of Content. MIT Press: Cambridge.
Fodor, J. (1985). Precise of the Modularity of Mind. Behavioral and Brain Science, 8: 1-42.
Fodor, J. (1983). Modularity of Mind. MIT Press: Cambridge.
Forester, K. (1979). Levels of Processing, in Sentence Processing. Cooper, W. & E. Walker (eds.). Erlbaum: Hillsdyle, NJ.
Georgopoulos, A., M. Taira, A. Lukashin (1993) Cognitive Neurophysiology in the Motor Cortex. Science. 260: 47-52.
Jordan, M. & R. Jacobs (1992). Modularity, Unsupervised Learning, and Supervised Learning, in Connectionsim: Theory and Practice. Davis, S. (ed.). Oxford University Press: New York.
Karmilnoff-Smith, A. (1992) Beyond Modularity: A Developed Perspective on Cognitive Science. MIT Press: Cambridge.
McClamrock, R. (1989). Holism Without Tears: Local and Global Effects in Cognitive Processes. Philosophy of Science. 56: 258-274.
McGurke, S. & R. McDonald. (1979). Hearing Lips and Seeing Voices. Nature. 264: 746-748.
Merleau-Ponty, M. (1994). Phenomenology of Perception. Routledge: London.
Samuel, A. (1981). Phonemic Restoration: Insights From a New Methodology. Journal of Experimental Psychology. 110, 4: 474-494.
Sejnowski, T. & P. Churchland (1989). Brain and Cognition. in Foundations of Cognitive Science. Posner, M. (ed.). MIT Press: Cambridge.
Shallice, T. (1988) From Neuropsychology to Mental Structure. Cambridge University Press: New York.
Spivey-Knowlton, M. & K. Eberhard (1996). The Future of Modularity. Proceedings of the Eighteenth Annual Conference of the Cognitive Science Society. Erlbaum: Hilldyle, NJ, 39-40.
Wolfe, J. (1990). Three Aspects of the Parallel Guidence of Visual Attention. Proceedings of the Cognitive Science Society. Erlbaum: Hillsdale NJ, 1048-1049.