(These are some notes to help you sort out the introduction to The Phenomenology of Perception. Don't take them as an authoritative secondary source or anything else like that (e.g. don't quote them in your paper). Just make use of them in getting an initial understanding of this section of the book.)
For the empiricist, sensations are simple "raw feels" or "qualia" (Husserl's "hyle", perhaps) which are individuated by their intrinsic character. [pp. 3-4] They lack meaning or intentionality, or anything like the structure of a concept or general type [top-mid p.15]. They are individual bits of unstructured data.
But for Merleau-Ponty, even the most basic and primitive data given to consciousness are intentional/directed; following Husserl and Sartre, he thinks that "all consciousness is consciousness of something" [mid p.5; p.38]. The most "elementary event is already invested with meaning", and carries with it a kind of gestalt -- a differentiation of structure into figure and ground, at least [top p.14]. [A figure/ground inversion, then, would, he thinks, not be "the conscious data given a different attentive interpretation", but would present different basic data -- even a different "world" [top p.16] (at least in the sense of a new ontology and taxonomy available).] He takes the empiricists to be guilty of the "experience error" [mid p.5] -- the fallacy of taking things known as transcendent or external and making their externality secondary to some imagined internality; the error of silently converting perceived external properties into sensed internal ones which we then see as prior. [This is essentially the same thing that Sartre called the "fallacy of immanence", and which U. T. place -- in "Is Consciousness a Brain Process" -- calls the "phenomenological fallacy".]
Merleau-Ponty also rejects the "constancy hypothesis": the claim that the basic inputs to consciousness have a constancy in their correlation with stimuli such that the same stimulus will produce the same sensation. He claims that the basic perceptual qualities are not in this way constant with the proximal stimulus to the organism. They are determined by more than just that stimulus; as he puts it, the sensory (or perceptual) apparatus is "not just a transmitter". [Or maybe "not just a "transducer", in Fodor's Modularity of Mind lingo.] [top p.9, mid p.10]
The empiricist comes up with what it is that the mind "adds" to sensation by "building up from it" by a kind of formula: By subtracting what's actually "in" the proximal stimulus from what's "in" the structured perceptual information, we're left with what it is that the mind "adds" to the stimulus. [mid p.21] But the prestructured nature of perception for Merleau-Ponty places this addition not in consciousness, but in the body.
Not having recourse to pre-structured, intentional perceptual states, the empiricist is, Merleau-Ponty's view, forced into the position of having to make everything other than basic sensation into judgement. [pp.32-3] Anything that goes beyond what's immediately given in the stimulus/basic sensation (like, say, that what I see has a backside, or is a human being) must be seen as some kind of inference, hypothesis, or drawing of conclusions in the mind which take sensation as premises. This transforms the world that we actually live in and care about -- that of human action, the social world of everyday life -- from a reality perceived and acted on to a mere artifact of associations. [pp. 23-4]
Merleau-Ponty claims that the perceptual gestalt cannot be accounted for by the association or projection of memories, but that instead, the structuring of perception as a meaningful presentation must be prior to the association of memories. He claims that we need something like a general property or concept or type to unify the perceptual judgement -- something which makes it into not just a "seeing", but a "seeing-as" [top p.15]. But the associations in memory which percepts get is in virtue of this content they have, and so the unifying content must be prior to and not derivative from these associations [see top of p.20 and p.35].
He also cites some empirical data for this claim. The example of the "double-syllable errors" [pp.17-18] is offered to show that association errors are dependent on the task structure, and not just on bare co-occurrence. And the case of recognizing a "hidden" figure [pp.18-19] is given to support the view that resemblance is more effective when you look for it, but that added association/training needn't make a difference.
Merleau-Ponty sees science as making the mistake of forcing phenomenological or perceptual categories into objective categories [mid p.11]; he thinks that the "indeterminacy of structured perceptions" is not to be replaced by "objective" properties [bottom of p.6]. Thus, the "real" length of the lines in the Muller-Lyre illusion is not a question about perception, but only about the objective world. [top p.6] The phenomenology of perception must move from the categories of the world to the categories under which we perceive it. [mid p.49] By substituting in its own conception of "objectivity", science (and particularly empiricism, as noted above) reinterprets the human world of persons, objects, politics, and everyday life from a reality lived in to an artifact of association. [pp.23-4]
The "scientific" account of perception takes perception as simply another process in the world to be explained [top of p. 17]. But this "relieves it of essential function of `inaugurating' knowledge". It sees the objective reasons for correct perception as prior to perception, while Merleau-Ponty emphasizes the sense in which they come after perception -- as knowledge of the world given in its first instance through our perception of the world. [mid p.17].
Merleau-Ponty sees the constancy hypothesis as implying that attention "illuminates and clarifies" basic given sensations rather than creating some new form or gestalt [p.26]. But on his account, the "normal function of attention" is "a process of composition, not copying" [mid p.9]. On the "intellectualist" view (a view which is close to a kind of Leibnizian rationalism, if that helps; but which might also be taken as a kind of naive direct realism, for our purposes), the object perceived and the sensations it produces already contain all the structure which might be found by attending to it. [mid p.27] Attentive perception then just provides a clearer representation of the structure and properties that were already there. This makes attention "impotent in creation" [top p.28], and leaves no room or indeterminacy for attention to fill in by constructing a form which isn't already there. [mid p.28]
On Merleau-Ponty's "creative" notion of attention, both constancy and "the world presented as reality itself" are left behind. The "world" under whatever we see as a correct objective taxonomy is not presented as such to consciousness through perception. Instead, it's organized and taxonomized in some way that the process of perception constructs [top p.29]. Instead of illuminating pre-given structures, attention "articulates figures out of indeterminate horizons" [btm p.30]; it imposes new structures on data which taken alone underdetermine what gestalt they might be organized into.
Perception, for Merleau-Ponty, pre-consciously unifies stimuli in meaningful structure. It both "gives to" and "finds in" data a meaningful structure or gestalt [btm p.36]. It does so pre-consciously: "the perceived object presents itself as a totality and a unity before we have apprehended the intelligible law governing it" [top p.42]. It's a process which is interpretive [top p.37] and structured far beyond the empiricist's notion of sensation. It presents things as objective and distal (unlike sensation, which presents as subjective and proximal) [p.40]. The empiricist's basic sensations are then taken as things we don't actually perceive, but which we make some kind of inference to in our scientific reasonings [top p.37]. The perceptual effects (illusions?) generated in the "moving eye" and "intervening objects" examples [btm p.48] are structured for us preconsciously in our perception, and not generated by some kind of conscious reason.
It would, for Merleau-Ponty, be a mistake to run together this notion of perception with the general notion of judgement [p.34]. If perception (the way things look) was generated by whatever we know in the way overall judgement (what we believe) is, we'd lose the ability to account for many phenomena: The persistence of illusion [also see p.21], the idea of a gestalt shift, and making an "appearance/reality" distinction all rely on being able to distinguish a salient sense of how something looks that is opposed to how we take it (judge it/believe it) to be. We need this separation to make sense of the general idea of a "field of vision" [btm p.5] as opposed to a realm of judgement. It's also central to the notion in ordinary experience of the distinction between sense experience and judgement [mid p.34]. (It's perhaps also critical to the psychological distinction between perception and cognition.) There are then two kinds of mistakes avoid making about perception: don't mistake it for sensation (or transduction), and don't mistake it for the general notion of the fixation of judgement or belief [p.5]. [There are clear parallels between Merleau-Ponty's notions of sensation, perception, and association of memories and Fodor's Modularity view that we should adopt a "trichotomous functional taxonomy of psychological processes... which distinguishes transducers, input systems, and central processors". [Modularity of Mind, p.41] For both, perceptual states are (1) prestructured rather than transduced [Fodor, top p.41; Merleau-Ponty, top p.9, top p.11]; (2) intentional representations of the distal world rather than reflections of the proximal stimuli [Fodor, mid p.42 and p.45; Merleau-Ponty, btm p.4 and p.5]; and (3) separated from and prior to the full processes of the fixation of belief [Fodor, btm p.42 through top p.43, top p.46; Merleau-Ponty, top p.21, mid p.19].]
To see and understand perception itself, we must "step back from it" so that we no longer are seeing "through" it to objects in the world, but are making perception itself an object of reflective consciousness. [Recall Sartre on "thetic" and "non-thetic" consciousness of the self in Transcendence of the Ego.] [p.37] When we use or "see through" perception to the world without making perception's "positing" of objects explicit (as opposed to, as in its normal occurrence, implicit and preconditional), we can't examine the perception itself, but only its object [top p.41]. To thus focus on perception itself will mean putting aside its facticity: We must "put out of court" any philosophical "realism" which takes as basic starting datum some eventual results of consciousness -- the objective characterization of the world, or the scientific description of it -- and start instead from the structures of perception themselves [top p.47].
This "making explicit" the idea of perception as "positing" rather than simply "reflecting" or "illuminating" is at the heart of Merleau-Ponty's (somewhat restricted, compared to Husserl's) notion of bracketing. We reject, for our methodological purposes, any account which takes perception as simply starting from and reflecting the properties of some "objectivity". An intellectualist (or "direct realist") account of perception "loses the contingency and ambiguity of perception" [top p.39] by starting from the assumption of a "reality beyond appearance" which perception directly apprehends; it makes the basic data of input necessarily reflect some actual pre-existing structure in the world, rather than examining the processes by which it might constitute its perceived world [btm p.39 - top p.40]; it takes for granted "uncritical perception" rather than critically examining that process.
The idea of a "motive" in Merleau-Ponty's account of perception is a critical but somewhat ambiguous (or as he says, "fluid") one; but also one which he thinks we must uncover [top p.50]. Objects are seen as "motives" rather than causes [top p.31] or reasons for perceptions. He glosses this "phenomenological notion of motivation" as "one phenomenon releasing another, not by means of some objective efficient cause, like those which link together natural events, but by the meaning which it holds out." (A question: What does "release" mean here if it's neither `cause' nor `reason'?) [top p.50, emphases mine]. Objectifying perception in science will "lose sight of the original relationship of motivation" [mid p.50]. Attentive vision might of course have a different appearance "motivated" [p.31] by forming a new meaning from (and thus perceiving new objects in) the data.
The point of going through the psychological phenomena in parts 1-3 of the introduction was to raise the problem of the constitution of the world for the subject. Merleau-Ponty believes that starting instead from the "natural attitude" (as Husserl does) avoids facing this critical issue. By taking the entire scope of natural experience and then just placing a kind of "marker" on everything to indicate its new status as "bracketed", we never leave behind the constituting relationships between perceptions, judgements, objects, and the world which are the central focus for Merleau-Ponty's phenomenology. But by focusing on the break between perception and reality (given reality, in both the sense presupposed the notion of stimuli or sense-data and that presupposed by the notion of true judgement), we break off perceptual consciousness from the simple reflection of the real, and are thereby forced to make explicit and examine the process of reconstruction of the world from this consciousness. This then brings us to the central question of a transcendental phenomenology like Merleau-Ponty's: What is the structure of the preconditions on our taking the world in its various aspects as real? How do we as perceiving (and acting) subjects constitute it?