Section A – Prof. Maund
Merlin, you write that Maund “says we don’t perceive sense data, but we are aware of it. Maybe he means ‘aware of it’ from a third-person perspective.”
No, he is here staying with the first-person perspective of the perceiver. He is saying that in the perception of an object, one is aware of, but does not perceive, one’s sensory representation of the object. We often use the word perception in a more general sense to mean simply any sort of awareness, but Maund is here using perception to refer strictly to sensory perception. Our awareness of the sensory representations instrumental in a sensory perception is a type of direct awareness, but it is not itself a sensory perception.
Maund has us perceiving physical objects always by being aware of intermediary sensory representations. We do not perceive the intermediaries. When we hear the sound of the bell at the rail crossing or see the moving barrier, we are caused to have sensory representations. We are aware of both the phenomenal sensory items and the physical objects that caused them. And we are aware of the latter through being aware of the former, on Maund’s view. “In conscious attentive perception, one perceives physical objects and their qualities by becoming aware of sensory representations, that is, of a set of quality-instances (tropes) that are natural signs for the objects in question” (71).
The sensory representations activated in a perception are not the object of which the perceiver is aware (66). “The perception of physical objects is direct (and immediate) but . . . nevertheless [it] operates through awareness of intermediaries” (68). Because of the latter element in perceptual experience, Maund calls his theory a hybrid direct-indirect theory of perception.
Perhaps Maund’s theory should be taken to be purely direct realism. I’ll leave that question for future scrutiny. Be that as it may, Maund’s theory and the theories of all those other philosophers I listed in post #4 are in a class of theories known representational realism (direct or indirect or hybrid).
Section B – Stephen
Is it legitimate to call some component or aspect of perceptual processing representational (or representative)? Is this use of representation only an analogy with our representations in drawing a stick-man or in making a statement?
Yes, it is legitimate. No, it is not an analogy, not any more than chemical affinity is an analogy with human affinity.
Component (3) in Merlin’s model of representation is “the interpreter of the representation, who can distinguish between, and relates” (1) the referent of what is represented and (2) the representation itself.
In order to make clear an important point on which I agree with Merlin’s view, I want to magnify, in the following statement, the elements in his model: The successful interpreter of the representation must be able to recognize the referent when she detects and discriminates it, she must be able to recognize the representation when she detects and discriminates it, and she must be able to recognize the representation as a representation of that referent. [I nod to J. Haugeland’s “Pattern and Being” on this (1993).]
Biological systems that are incapable of recognition are incapable of being interpreters of representations. Can the preattentive conscious elements in our episodes of attentive perception recognize something? That question is highly pertinent for an assessment of Professor Maund’s representative realism. But I want to address a more radical question. Can preconscious elements in our perceptual processing recognize something?
I say not. There are pattern-recognition systems, machines we have invented or the natural neuronal pattern-recognition systems. This use of the term recognition is an analogy. Neuronal pattern-recognition systems are part of what makes our genuine recognition possible. Genuine recognition is attained, however, only when those subsystems are hooked up in the right ways in a high-level animal brain.
Is it possible that activity patterns in some of the neuronal subsystems subserving our perceptions are representations? I agree with Merlin that such neuronal subsystems’ activity patterns cannot be an interpreter of representations, for they cannot genuinely recognize anything. Can they be representations?
They meet Fred Dretske’s requirement for a representational system. They are systems having functions “to indicate how things stand with respect to some other object, condition, or magnitude” (1988, 22). The sense of indication here is simply causal. Rabbit tracks naturally indicate a rabbit whether or not they are recognized as indicators of a rabbit. As a result of plant evolution, the collumella region of the cap of a gravitropic root has the function of indicating the direction of gravity. The collumella region of such a root of such a plant is a representational system, on Dretske’s view.
To count as a representational system, I think we need Dretske’s requirements, but we need more besides. Animals such as us, and even birds and insects, have more indirect and more complex reactions to their external environment than gravitropism. Behavior of these organisms is very active compared to plant tropisms. Such animals actively search out information from the world, information that is needed for them to perform successful actions.
In many mammals (and especially in humans), the retina is nonuniform, with a fovea that provides particularly detailed information about the visual scene. Thus, the successive fixations must be integrated, with the information acquired with the eyes in one position remapped as the eyes move. Other sensory modalities, such as audition and touch, also must be integrated into this evolving representation of the visually perceived world. (37)
In the saccadic system, the brainstem saccade generator converts space (retinotopic code) to time (firing of the excitatory burst neurons). The working memory holds a “plan” (the retinotopic position of targets) until it is executed, whereas dynamic remapping updates the plan as [saccadic] action proceeds. (67)
Neural Organization: Structure, Function, and Dynamics
Arbib, Èrdi, and Szentágothai
Saccades are essential to the formation of our integrated percepts of visual scenes. These coordinated actions in the visual system can perform automatically for us, without our conscious direction of the steps. In this context, we find the scientists quoted above talking of information, remapping (so mapping), representation, and retinotopic code. These are legitimate concepts in this context. They are not concepts merely analogous to representation in Merlin’s sense, the sense entailing conscious action. (And they are not ‘stolen concepts’.)
Neuronal subsystems subserving our perceptions do have indicating functions that feed into our perceptions. So they meet Dretske’s condition for being representational (sub)systems. The representations subserving perception are neither iconic nor language-like. They are distributed computational representations. The sense of representations here is the mathematical one (cast perhaps very generally, say perhaps the one in terms of objects and morphisms in mathematical category theory), but instanced in the operational setting of functioning neuronal systems. So the morphisms, or mappings, of the mathematical representations are actually being performed by physical neuronal activity.
I do not have an argument to the effect that all of our conscious representations—imagination, memory, conceptualization, planning—must be held by neuronal processing that is representational in the mathematical sense. But one notable research program progressing under the hypothesis of such a dependency is displayed in this book:
The Harmonic Mind
From Neural Computation to Optimality-Theoretic Grammar
Smolensky and Legendre