Notes on “The Evidence of the Senses: A Realist Theory of Perception” by David Kelley
These notes paraphrase the author’s points and are not accurate quotes unless in quote tags.
Footnotes are (mostly) not reproduced here.
{ My own added comments are in these curly brakets. }
Chapter 7 Perceptual Judgments
Perceptual judgment is the conceptual identification of what is perceived. Recognizing a percept as belonging to a certain type is the only epistemological issue concerning perceptual judgment.
Distinguishing a percept from a perceptual judgment is impossible for a theory that holds judgment is precisely what distinguishes perception from sensation. In Representationalism experiential states are interpreted as signs of external objects, interpretation being judgment. In Sensationalism the awareness of a whole with invariant properties is constructed from fluctuating sense data, construction being judgment. In the Realist theory judgment occurs after perception.
Perceptual judgment is analyzed in terms of the two parts of the act: reference and predication. The linguistic expression of a perceptual judgment is a statement of the form “x is P”. The questions that can be asked are: Why do we believe (predicate) “x is P” but not “x is R”? and Why do we believe (reference) “x is P” and not “y is P”?
Perception and Reference
For a judgment to be justified by perception, the person judging must perceptually discriminate the object he takes to be an instance of the concept predicated.
A camouflaged soldier may present facing surfaces to the field of view but if he is not isolated as a figure discriminated from the background there is no awareness of a particular and no judgment can be formed. Awareness of various patches of color only justifies judgments about those patches. When there is awareness of the soldier he is recognized as an object, an entity with 3 dimensionality.
Demonstrative reference
“That is a hawk” is a statement that can be made about different birds in different fields by different persons. What makes identical statements be about different things is not the conceptual content but the demonstrative reference to different birds. The reference of the word ‘that’ is determined by the speaker’s perceptual focus. There can be no failure of correspondence between the object of perception and the object of belief in demonstrative reference.
Attributive reference
“The man with the martini is tall” A judgement made of whatever satisfies the description whether or not it is present to perceptual focus is an attributive judgment. Attributive propositions are not perceptual judgments.
Demonstrative and attributive reference can be present at the same time. When looking over the people in a room one forms the judgment “The man with the martini is tall”. But if he actually has water in his glass, the reference is still to the man one is looking at (demonstrative) not a different man across the room (attributive). In this case there were two judgments “This man has a martini” and “This man is tall” and perception determines the referent of both. {Proper names and definite descriptors can both be used in either mode of reference, but if demonstrative reference is present then there is redundant reference and it (demonstrative mode) has priority; the referring expression can be mistaken without disrupting the reference. A referring expression cannot be determined to be mistaken unless we know which referent to compare it with. }
Conceptually guided perception
Discrimination is necessary for justification, but not for speculating about what predicates might be ascribed. In trying to make sense of an unfamiliar view or obscure scene a variety of theories may be tried out. Searching for a pencil on a cluttered desk is a perceptual exploration conceptually guided a criteria for what a pencil looks like. Trying to learn what a cornice is in architecture might require explicit guidance in picking out referents as examples. The conceptual content of the guidance does not constitute the object. There is no definite judgment “Here is a pencil” or “Here is a cornice” until the object has been discriminated perceptually. Perceptual discrimination has an epistemological priority over any judgment or judgment-to-be.
Degrees of discrimination
Perceptual discrimination is not unitary but admits of degrees. Different circumstances and different sense modalities provide awareness of different sorts of structures and there are borderline cases. The factors relevant in determining what (if anything) has been discriminated also determine what (if anything) the corresponding perceptual judgment is about. Cases considered:
Visual occlusion- To perceive a thing as a whole does not require perceiving the whole thing - front, back and inside. (see Ch. 5) A perceptual judgment need not require perceiving front, back and inside to be justified. A cat is partly hidden behind a chair. The edge is experienced as belonging to the chair and the cat as extending beyond the limit imposed by the obstruction.
Majority visual occlusion- A hunter sees a patch of brown amid the brush on a hillside and judges he sees a deer. Two possibilities are compatible with the principle advanced.
-
The brown patch lies behind the bushes so it is perceived to extend beyond the visible contours. An object of indefinite extent is discriminated, a deer.
-
This is not a perceptual judgment but an inference - that shade of brown is a color deer have but not bushes or earth - and so not a counterexample to the principle.
Hearing- Hearing does not discriminate entities in the sense of solid physical objects but events. Yet hearing still follows the principle because judgments about what is heard are given descriptively in the form “that sounds like … the sound made by the banging of pipes in the walls” or “that sounds like … the sound made by the banging of a gong.” The sounds of music and speech are not described but named {for music instrument/key/note, for speech voice/word}
Touch and sight - What object is judgment about when one can see a tub of water and put a hand in it to feel it is warm? Touch only informs about the water in contact with the hand but the judgment is about all of the water in the tub because the water is seen in its entirety. In this example, sight picks out the object referenced and touch justifies the predicate; different sense modalities are integrated together.
Blind touching- Groping in a dark cave, your hand plunges into something wet and cold. The pool of water is not discriminated as a definite object, but the border between air and water is discriminated. The judgment about the water being wet and cold applies to an indefinite extent of water.
Qualities- Primary qualities apply to whole entities: a whole apple is round, not its parts. A justified perceptual judgment about a primary quality requires discrimination of the entity having that quality. Secondary or continuous qualities apply to parts as well as wholes. A justified perceptual judgment about a secondary quality requires discriminating what the judgment is about: a portion or patch that is red, something in peripheral vision moved.
Hallucinations and dreams- Dreaming is noncognitive, the subject is detached from perceptual contact with reality and not in control of his mental processes. Nothing that happens in a dream needs to be justified, no principles of justification apply.
Hallucinations occur because of inner causes that radically disrupt cognitive functioning. Again the loss of control over the ability to be rational, be objective and judge makes justification impossible.
Even if a hallucination could justify, it would be in virtue of its internal phenomenal similarity to an actual percept. But the reason a percept justifies is not because of its similarity to a hallucination. The asymmetry entails that if there is such a thing as hallucinatory justification it is hierarchically dependent on perceptual justification. {Attacks on perceptual justification from hallucinatory justification cannot avoid the stolen concept fallacy.}
Perception and Predication
What is the justification for stating “x is P” but not “x is R”? Which of an entities attributes must one perceive - and how must one perceive them - in order to justify the predicate of a proposition stating a perceptual judgment?
Recognition
To recognize is to perceive a similarity, as a red object possesses the same quality as other red objects. But a red object will also be similar to shades of orange and purple at the borderlines with red, and similar to other objects in other respects.
Recognition involves:
-
isolation of a relevant dimension of similarity
-
a sense of what degree of similarity along that dimension is necessary
-
a capacity to ignore perceptible differences within the permitted range of similarity
A theory of universals is necessary to know what qualities must be discriminated in the object to make it an instance of the relevant concept of an attribute or type. That is beyond the scope of this book.
A theory of concepts is necessary to know how the dimensions of similarity are isolated and retained as concepts and then combined with percepts to produce recognition. That is also beyond the scope of this book.
A perceptual judgment is conceptual, and has a propositional form. But there are more primitive, nonconceptual types of recognition.
• Children can learn to recognize an object when it reappears after being hidden long before they begin to acquire language.
• Adults can recognize the shape of a blackberry bush even when unable to describe it in words or think of it determinately in its absence.
In the propositional expression of a perceptual judgment the predication involved is direct and immediate, not mediated by any criteria or implicit inference. The predication is experienced with the same sense of immediate familiarity that characterizes preconceptual awareness.
Foundationalists of the Cartesian Empiricist pattern often accept an implicit theory of recognition based on a criterial feature list. Recognition of the object as a kind follows the recognition of the features defining that kind (a counterpart to the epistemological priority of attributes over entities in sensationalism). No explanation is offered for how the features are recognized, so the same problem remains but pushed down one level. It has not even been acknowledged that the statement “This is red” does not directly express the seeing red, but the judging that this is red. It has not been acknowledged that even simple concepts such as “red” are abstractions and that we cannot literally see abstractions.
The attributes of a concept can be understood as automatically predicated of anything identified as an instance of a concept, or understood as arrived at separately and inferentially from the premise that a thing is an instance of a concept. There is no epistemological difference between these positions; both commit one to the truth of propositions about the attributes and any doubt about whether an attribute is actually present also equally undercuts the justification for the conceptual identification.
Two Views of Concepts
Construction theory - Concepts are constituted by their definitions. Statements expressing a meaning given by the definition are analytic. Statements identifying attributes not specified in the definition are synthetic (whether in universal or particular form).
[footnote- a representative construction theory is Lewis, Analysis of Knowledge and Valuation, Book II, 1946]
Discovery theory - Concepts are constituted by the kinds of things picked out, those things may have open-ended number of properties in common. A definition is given in terms of the most essential common properties, but what is most essential is subject to revision without changing the concept. No analytic-synthetic distinction can be made.
[footnote- a discovery theory is developed by Ayn Rand, Introduction to Objectivist Epistemology 1979. A discovery theory is also implicit in the views of Saul Kripke and Hilary Putnam on the semantics of natural kind terms. See Kripke, Naming and Necessity, 1980 and Stephen Schwatrz (ed.) Naming, Necessity and Natural Kinds, 1977 esp. “Introduction” p 13-41 and Putnam’s “Is Semantics Possible?” and “Meaning and Reference”, 102-118, 119-132.
On either theory of concepts there is some collection of facts attributed to each referent due to being an instance of the concept, and some of those facts will be imperceivable or not perceived in a particular case. How can a perceptual judgment committing one to facts not perceived be justified?
Construction theory will claim that identifying an object as an instance of a concept only predicates the defining features of the concept. If all defining features are observable then predicating that concept is justified conclusively by perception. This leads to a structural foundationalism where all knowledge rests on a foundation of predicates that are strictly perceptual.
Discovery theory denies there is a difference in status between defining and non-defining features, so a predication cannot be limited to only some of the features of a concept. No concept can be predicated without committing to facts not perceived. Whatever inductive evidence one has that kind K has property P must justify the judgment that this K has P. All adult perceptual judgments rest in part on prior inductive knowledge. Foundationalism can only be true in a temporal {emphasis added} or genetic sense, there was a first time K was perceived and subsequent knowledge builds on that original encounter.
There is a compatibility between the construction model and representationalism, and also between perceptual realism and the discovery model. Subsequent analysis of concepts of sensory qualities will be on the basis of the discovery model.
Concepts of Sensory Qualities
Do concepts of sensory qualities designate the intrinsic attributes or the forms in which we perceive them? The question poses a false dilemma. Discovery theory holds that concepts evolve over time by integrating new knowledge to them. The development of a single concept over time has three distinct stages:
Naive level - there is no knowledge of illusions or perceptual relativity, no appearances.
Commonsense level - takes into account perceptual relativity, the possibility of illusion and uses the idea of appearances
Scientific level - incorporates knowledge of causal mechanisms of perception, both on the side of the sensory apparatus and of the objects perceived.
At the naive level what is perceived is attribute-in-a-form, a unitary content experienced as wholly external. Only background knowledge permits isolating the relational aspects of the identity perceived, knowledge which is lacking at that level. A concept of a sensory quality formed at the naive level will not be a concept of the attribute as opposed to the form, nor will it be a concept of the form as opposed to the attribute. A naive perceiver implicitly takes anything which looks red to be red, but will later learn this is a mistake and will reclassify certain objects under a new concept of appearance formed for the purpose - “looks red.” Under normal conditions what looks red is red, so there will be a continuity of the concept in picking out the same kinds of things.
Roderick Firth argues that a naive perceiver (a child) has an early form of the concept “looks red”. He takes this to be a property which objects have independently of him but will later learn this is a mistake and will form the new concept “red” for the property that objects have independently of him. A second argument is that a child implicitly uses “looking red” as a criteria for predicating red, so the concept is really that of the criteria “looking red”. [Firth, “Coherence, Certainty, and Epistemic Priority,” Journal of Philosophy, LXI (1964) 547.]
Refutation: Given the mechanisms of color constancy, there are objects that an adult would call red that a child would not. Therefore the child’s concept of “red” is not coextensive with the adult concept of “looks red”.
The criteriological argument is of the form: we are aware of A by means of or in virtue of B, therefore we are directly aware of B not A. This is the Representationalist pattern that was rejected in deciding what we really perceive, it is no more valid in deciding what we conceive.
The commonsense level allows the attribute to be isolated. The scientific level allows quantitative judgments about qualia to be made (“that looks like a 600nm red”).
Perceptual relativity as noted and understood at the commonsense level is important to perceptual judgment, but the deeper relativity discovered at the scientific level is not. Advancing from the naive to the commonsense level may require some reclassification of objects in terms of the concept, advancing from the commonsense level to the the scientific level will require none. Secondary qualities predicated on the basis of perception (red, warm, sweet) are concepts which do not distinguish the attribute from the form. Red, warm and sweet are forms jointly caused by microscopic structural and chemical properties of the objects perceived and the physiological means of perception. {Knowing that red light is electromagnetic radiation at 600nm wavelength does not make a difference in its appearance.}
Illusions and Perceptual Relativity
“Being F” vs. “Appearing F” can only be distinguished conceptually. Illusions are a conceptual phenomena, misclassifications based on real appearances.
3 facts jointly make illusions possible:
• Form varies with conditions.
• A concept of secondary quality is a concept of an external feature of objects, in respect of which they are similar independently of us.
• A concept of secondary quality is formed and predicated on the basis of similarities in the forms with which we perceive its instances.
For an illusion to occur one must perceive an attribute not in the usual form but another form in which one normally perceives a different attribute and for which one possesses a different concept. (An entirely novel appearance won’t have a corresponding concept so no misclassification will occur thus no illusion.)
Only in relation to our concepts can we identify any form of perception as illusory. Abnormal conditions are those in which we perceive objects in illusory forms.
Definition of Normal Condition- any condition of perception within a range that allows discrimination of the similarity to other objects subsumed by the same concept.
One must perceive the object in a form which is normal for the perception of F objects (F a concept of a sensory quality).
One must take into account any evidence one has that the conditions of perception are abnormal.
The conceptual override - Using background knowledge of what F looks like in abnormal conditions to make a judgment makes that judgment an inference.
Justified error - An object which is Not-F may be perceived in the form and normal conditions for the perception of F. One can be perceptually justified in judging a Not-F is F.
Two concepts of justification:
• “Being in a position to know” is what justifies - meaning in contact with reality. Knowledge is the correct identification of things as they are independently of our beliefs. By this theory an hallucinator is not in a position to know what he asserts, and neither is the subject of an illusion.
• “Reasonableness” What justifies is what makes it reasonable to think so. Justification is normative, a standard of what ought to be cognitive conduct. But “ought implies can”, a person cannot be held accountable to a standard impossible to apply in a given case. By this theory the subject of an illusion is reasonable in forming the judgment to which his experience prompts him, and so is the hallucinator.
Holding to either theory of justification in disregard of the other is another manifestation of issues discussed in Chapter 1. The first theory disregards the process of knowing {intrinsicism}, the second theory discards reality as the standard {subjectivism}. Objectively, a percept, even an illusory percept, arises from the interaction of object, senses and conditions. A subject takes an object to be F on the basis of similarities that are the real product of perceptual contact with reality. Hallucinations can be reasonably interpreted in certain ways, but there can be no perceptual justification without perceptual contact with reality. The subject of an illusion can be justified, an hallucinator cannot.
Does the third principle require positive knowledge that conditions are normal? - No. The healthy functioning of one’s own nervous system cannot itself be verified on the perceptual level. Also, checking the verifiable conditions of perception involves further perceptual judgments, those judgments would need to be checked, etc. in an infinite regress.
One can assume conditions are normal in the absence of evidence to the contrary, but one must be alert to that evidence. Evidence of abnormal conditions may be perceived as external phenomena, revealed in the way that everything appears, or only known conceptually. The scope of the abnormality can be restricted to a single feature such as color, or all judgments altogether. The strength of the evidence of abnormality can vary so as to render judgments completely unjustified or merely probable.
Arguments for requiring positive evidence of normal conditions:
- Skepticism - We cannot be certain of any conclusion unless we know our faculties are operating reliably. We cannot know our faculties are functioning reliably unless conditions are normal. Therefore we cannot be certain of any conclusion reached by perception until we know that conditions are normal.
Refutation: First premise can be supported only if there is a real possibility that our faculties might be operating in fundamental disconnection with reality and constituting their own contents; this was dismissed in Chapter 1.
The skeptic presupposes the knowledge that conditions can affect perception, knowledge which can only be gained through perception. The skeptic uses that knowledge while attempting to deny its source.It is impossible to misidentify an object as F unless one already possesses the concept F, but the concept could not have been formed unless conditions were sufficiently stable to perceive the referents of F.
- Wilfrid Sellars - Claims the not only must conditions be normal the subject must know they are normal. [sellars, “Empiricism and the Philosophy of the Mind” in Science, Perception and Reality (1963) 147-148] This rests on his view of knowledge as propositional. Prelinguistic non-propositional justification precedes any reflective ability to explain and defend our perceptual judgments to others. (see Chapter 6)
Positive evidence of normal conditions is not required. If there is no evidence of abnormal conditions then a perceptual judgment is justified directly by perceptual awareness and does not require propositional knowledge of other facts.
The growth of a subject’s knowledge through the commonsense into the scientific level makes one aware of more of the variables constituting normal conditions, so the constraint of the second principle will apply more often. There are no universal rules about what judgments are justified in what conditions that can hold for all knowers.
The Autonomy of Perception
Perception can be influenced by psychological factors - expectations, needs, interests, biases, preconceptions. Interest directs the amount attention and the choice of what to focus on. What contradicts expectations will get more attention. In general, purposes and background knowledge provide a context for perceiving. Psychological factors do not threaten the objectivity of a percept or judgment unless they alter the percept or cause one to make judgments not supported by perception. Non-objective thinking is possible across the spectrum of all cognitive activities because objectivity is not automatic. So long as subjective biases can be identified and overcome then objectivity is still possible and a virtue.
Are subjective biases causally determining in a way that makes objectivity impossible?
Short answer - No.
• The universal form of the claim “all objectivity is impossible” is itself impossible and cannot claim to be objective.
• Statistical evidence that more people are nonobjective under certain motivations does not establish nonobjectivity as inescapable.
• Objectivity in perception is easy to achieve compared to introspection or scientific analysis.
Long answer - various arguments in favor of “yes” are considered and refuted.
Philosophical arguments
Thomas Kuhn and H.R. Hanson have argued perception is “theory-laden” influenced by the the very theories that are supposed to be tested by perception.
A scientist familiar with an X-ray tube will notice more and different things about it than a layman, but the percept of the scientist does not go beyond that of the layman much less distort it. Scientific and conceptual knowledge in general can enable the extraction of more information from a given stimulus. Distortion can only be demonstrated when opposing theories lead to incompatible percepts for the same stimulus.
Hanson claims that Tycho Brahe will see the sun rising over the Earth while heliocentrist Johannes Kepler will see the horizon dropping below the unmoving sun. Hanson claims this is a perceptual difference bit this is not credible. Anyone with a little imagination can see both perspectives, and while we are all heliocentrists today it is still more common to see the sun rising. Anyone can see the walls moving past while walking down a hallway.
Paul Churchland advances an argument [Churchland, Scientific Realism and the Plasticity of the Mind, 1979 Chap. 2) that perceptual judgment is theory-laden. His extended analysis of our concepts for heat argues that the concepts contain and implicit commonsense theory about the nature of heat, a theory called into play whenever we recognize something is hot or cold but which goes beyond what is actually perceptible. His argument is convincing but poses no difficulty for the temporal foundationalism described earlier. Adult perceptual judgments do go beyond what is given due to background knowledge, but that background knowledge was itself discovered perceptually at an earlier time. This admits that at one time the cognitive content of our concepts of hot and cold was exhausted by what was perceptible but Churchland argues against that by a thought experiment.
An alien race can see hot and cold because their respond to infrared radiation. They see it in the form of sensations of black and white. They have no other color sensations or color vocabulary. Should we translate their words used for their sensations as hot and cold or white and black? Churchland argues correctly that the translation should be hot and cold. (It is correct because the meaning of a concept is what it picks out in the world, what it refers to.) Even perceivers who discriminate in radically different ways than our own senses do can possess the same concepts we do. But Churchland goes on to conclude that this is because the meaning of concepts is set semantically. By assuming perception is non-cognitive, and providing an example showing a concept referring to something beyond the bare sensation, he takes as proven the coherence theory of meaning as determined by the network of acceptable sentences that use the concept. (He follows the pattern of proving Linguistic Idealism by attacking Cartesian Empiricism in a false dichotomy.) But perception is cognitive as argued earlier, it is the direct awareness of external objects and their attributes and concepts of sensory qualities are about the intrinsic attributes of those objects. The argument does not address this view so proves nothing.
Psychological arguments
The experimental data will not bear the interpretation given them.
Motivational effects on perception Jerome Bruner and other proponents of the “New Look” approach tested the influence of motivation on perception.
Children aged 3-5 turned a crank and were rewarded with candy. The control group received the candy directly while the experimental group received a poker chip they could trade for candy. After a sufficiently long exposure to this little economy the children were asked to make size estimates of a poker chip by adjusting a variable standard until it looked the same size to them as the chip. Children in the experimental group set the standard larger than children in the control group. It is concluded that value placed on an object can affect its perceived size. [Lambert, Solomon, Watson, “Reinforcement and Extinction as Factors in Size Estimation,” Journal of Experimental Psychology, XXXIX (1949) 640.]
Comments
• Such young children (often used in these experiments) are more likely to be influenced by need or desire than adults.
• The overestimation was 6 percent - statistically significant but not justifying the perjorative word distortion.
• Numerous perceptual mechanisms aid discrimination by sharpening contrasts, tuning out irrelevant stimuli, etc. An increase in apparent size of valued objects may be one of these mechanisms.
• The method used for size estimation is the best possible evidence in psycho-physics used to measure dimensions of the forms in which objects are perceived. This is the strongest possible evidence that the effect was on the percept itself not the perceptual judgment.
• The mechanism interfered with the cognitive task of making accurate size comparisons, but children are not as flexible as adults in shifting their mental sets for new tasks.
• Bruner’s own retrospective judgment [see Bruner, “The Functions of Perceiving: New Look Retrospective,” in Jeremy Anglin (ed.) Beyond the Information Given (1973) 116-120] is that the experiments failed to show regular distorting effects on perception which were predicted by the New Look theory. Motivational factors influenced the direction and degree of attention not the cognitive content of percepts.
Cognitive Effects on perception Bruner and others also studied how expectations, sets, and background knowledge influenced perception. Experiments showing that practice at a task makes one better at it are not a problem. What we are looking for is evidence that cognitive factors can give rise to judgments in conflict with the stimulus information.
Bruner and Postman showed subjects playing cards and were asked to identify them. Each card was shown for a progressively longer time from 10 to 1,000 milliseconds until it was identified correctly. Some trick cards were introduced that had the wrong color for their suit (a red 3 of clubs, a black 6 of hearts). Normal cards could be identified after an average exposure of 28 milliseconds, the trick cards required an average of 114 milliseconds. [bruner, Postman “On the Perception of Incongruity: A Paradigm,” Journal of Personality, XVIII (1949)]
• The experiment is sometimes said to show that expectations can prevent us from seeing objects which violate our expectations, but the measured effect was on the perceptual judgment not the percept.
• A judgment which conflicts with background knowledge should require more evidence than a judgment conforming with background knowledge, and staring at a card longer is more evidence. Perceiving, thinking and judging extend over time.
• At the very briefest exposures the subjects showed dominance reactions: they just consistently went with the suit or the color and did not have time to reconcile two features.
• Given enough time the subjects were always able to correctly identify the cards.
• Just before recognition occurred many subjects experienced extreme confusion, the sense that something was wrong with the card, the background beliefs about cards, or their compromise hypotheses (“that’s a six of clubs in a red light”). Such confusion is a typical response to anomalies, and often found in scientific discovery and other forms of creative thought.
This is all perfectly compatible with rationality.
There are two problems with these kinds of experiments. First, they usually fail to isolate the effects on perception and are actually about the perceptual judgment. Even the most hasty judgments are partly volitional so epistemological explanations are as valid as psychological explanations. Second, the stimuli are typically impoverished compared to the normal case, i.e. they are abnormal in the sense of the second and third principles of justified perceptual judgment given above. Judgments made in abnormal conditions where the abnormality is the impoverishment are often reached by supplementing the missing information. There is a crucial difference between supplementing and distorting.
To avoid those two problems what is sought is a clear case where cognitive factors clearly influence the judgment and not the percept and in which the stimulus is clearly adequate for normal perceptual judgments, yet the judgment conflicts with the stimulus. The only clear case (and most cited because it is the only evidence available) is the case of phoneme restoration.
Phoneme restoration is a class of auditory illusion in which a recorded spoken sentence is edited to replace one phoneme with a cough sound, but listeners report hearing the missing sound. The illusion continues even when fully informed listeners know what is on the tape. The significance is not the illusion, but that the subjects’ syntactic or semantic grasp of the word or sentence influences what is heard.
Nothing quite like this happens in other perceptual domains, not even other sorts of auditory perception. Generalization about perception as a class is not justified when all the evidence comes from a single sub-class with highly distinctive features.
As biological systems the senses are adapted to discriminate and identify natural objects. Symbolic objects are adapted to the senses. A good symbol minimizes the perceptual resources needed to recognize it. A symbol with a rich identity as a physical unit invites exploration, inefficiently drawing attention away from the meaning it bears. An ideal symbol is as transparent as possible and has the fewest perceptible features that enable identification. Symbols bear meaning only in context, so it makes sense to use context in discriminating symbols especially when the physical signal is obscured or degraded. Trading off some autonomy for a gain in speed and transparency of meaning makes sense for symbols.
Conclusion: The perception of speech is a special case that does not generalize to perception as a class. Needs and values influence attention and background knowledge influences conceptual identification. Perception is autonomous in respect of its cognitive content and capable of providing independent and objective evidence for our beliefs. There exist unusual circumstances where objectivity is not child’s play, such as during a crime where things happen fast and emotions may be running high and expectations are violently upset, but they dramatize by contrast the reliability of judgment in normal cases.
Epistemology with a Knowing Subject
Twentieth century epistemology has been studied abstractly in terms of theories of knowledge, inference structures, logical relations among concepts, and logical justification of knowledge. This has all only worked to the extent it has because propositions and concepts can be studied in abstraction apart from any particular subject possessing those abstractions. This approach does not work at all for perception where the cognitive content possessed by the subject cannot be abstracted apart from the language used to express it.
Acknowledging the subject permits regarding perception as cognitive and makes possible the three principles of justified perceptual judgment. Without a full theory of concepts we cannot specify the way in which concepts must be held in order for predication to be justified by perception but we do know the principle involved would not be another belief but a skill. Justification depends in part upon the exercise of an intellectual virtue, objectivity. Skills and virtues alike with the contents of nonpropositional awareness cannot be abstracted away from the subject and captured in a logic of propositions.
[susan Haack “Epistemology with a Knowing Subject,” Review of Metaphysics, XXXII (1979) 309-336]