There are philosophers who claim that there's such a thing as “non-conceptual experience”. Christopher Peacocke, for example, says that there are experiences “with non-conceptual contents” (1992). That is, concepts don't wholly or partly determine all mental contents.
Perhaps Kant was correct about experience. He argued that we apply certain “a priori concepts” or “categories” automatically: i.e., non-cognitively. It can be added here that we may also apply non-Kantian contingent concepts in such a way. It may follow from this that there's no such thing as non-conceptual experience.
One may consequently ask what a baby experiences or sees. The Kantian “manifold”? Not according to Kant himself. The reason being that the baby’s “cognitive faculties” (in contemporary terms), or its “transcendental a priori structures” (in Kantian terms), are already determining the baby's experience of the world according to necessary concepts or categories.
The Kantian scheme is broadly correct; although we shouldn’t adhere too zealously to the letter.
Take this un-Kantian way of putting things.
The innate neurophysiological structures determine the mental concepts and the mental concepts (applied non-cognitively) determine both perceptual experience and linguistic categories (such as subject/object/substance, cause and effect, temporal sequence, spatial ordering etc.).
What if we aren’t a Kantian in the above sense?
Take the following as a piece of a priori theorising about the experiences of a baby.
Babies must have non-linguistic concepts and categories. That of an object and event for a start (this is still a Kantian position).
Take the mothers of babies.
Of course babies don’t at first use the word “mama”, for example; though they may well have a concept of the object they will later call “mama”. That is, mama is individuated, differentiated and perhaps identified (somehow) as the object that will be called “mama” in the future. Babies may even have non-linguistic conceptual criteria of identity for mama. Indeed they must have conceptual criteria of identity in order to individuate mama from, say, the chair on which she sits. They don’t, after all, smile at the chair.
So babies must have the concept [mamma] before they can use the name “mama” otherwise they couldn’t use “mama” to refer to mama. The word 'mama' itself doesn’t individuate mama. The word 'mama' refers as much to the concept [mama] as it does to the object mama. Without all this, there would be an object to the baby; though it wouldn’t be mama.
It may, of course, be difficult to accept non-linguistic conceptual criteria of identity. Can we make sense of non-linguistic concepts and criteria of identity (therefore, non-linguistic individuation)? Of course we can. Take the case of my parents’ dog.
This dog, Joe, knows who I am. It doesn’t have a language. How does it recognise me? By individuating me from other persons and objects by applying conceptual criteria of identity to the object humans call “Paul Murphy”. The possible conceptual criteria of identity Joe applies to Paul Murphy may be [a particular smell], [particular clothes], [facial features] and [a particular voice].
Joe, the dog, has also individuated and applied criteria of identity to events.
For example, (my) picking up the dog chain must be a specific individuated event for the dog. This event “means”, to Joe, that it will be going for a walk (itself an individuated event). Possible conceptual criteria of identity for the dog chain event for Joe could be [the noise of the dog chain], [my picking it up] and [my going into the hallway].
This is a quasi-Kantian position on experience. Kant himself, in his Prolegomena To Any Future Metaphysics, wrote:
“…concepts…under which every perception must be first of all subsumed and then by their means changed into experience.”
“…perception can never work as experience without the law that, wherever an event is observed, it is always referred to some antecedent…”
Kant’s passages are, of course, referring to a priori necessary concepts of experience (e.g. [cause-effect], [substance] [object], etc.), which are automatically or non-cognitively applied to sensations. But the passages could equally apply to a posteriori contingent concepts; which are also sometimes non-cognitively “applied” - though often cognitively applied too.
For example, I cognitively apply the new concept [lad] to a young man of a certain kind; though I don’t always cognitively apply the concept [crash] to a crash. Although the concept [crash] is a posteriori and contingent, a crash is an [event], which is a Kantian concept, and therefore a priori and necessary.
Jennifer Church writes:
“To those of us with Kantian sympathies…it seems that a state…cannot count as an ‘experience’…unless it is a certain way for, or to as a subject.” (1995)
This isn’t just a pedantic conceptual point that the word 'experience' means what Kant says it means. It's to say that there's no such thing as nonconceptual… whatever you want to call it.
One can take a quasi-Kantian position on experience without using Kant’s own terms or indeed philosophical terms at all. For example Michael Tye writes:
“…[in] very early vision…categorisations are automatic [non-cognitive]. They do not demand that the creature have beliefs or thoughts…” (Tye, 1990)
However, Tye goes on to say, about animals, that
“some…further categorizations…do so without the essential involvement of concepts (understood as involving a public language or thought…”
Surely Kantian a priori concepts wouldn’t need a “public language”; and neither would the kind of animal concepts acknowledged later.
Perhaps Tye precludes such “categorisations” as conceptual because they are “automatic”: i.e., non-cognitive and therefore a priori. Though there's no antecedent reason why concepts shouldn’t be automatic and non-cognitive. Indeed we even apply contingent non-Kantian concepts non-cognitively, such as the concept [crash], which is based on the Kantian concept [event]. In any case, an animal is still seeing an object as a particular object or an event as a particular event. And this is where Tye does indeed use a near-Kantian term – “categorization”. Through the Kantian categories, minds/persons individuate objects and events.
A person could have experienced a murder and not experienced it as a murder. This doesn’t mean that no concepts were applied to that event or experience. He certainly experienced it as something. A dog that experienced the murder wouldn't have experienced it as a murder. It would, however, have applied concepts to it. It would have experienced it as an individuated and therefore a determinate event. (It may have been frightened by it.)
The upshot is that (Kantian) experience needn't be propositional or linguistic; though it must be conceptual.
Note: the above was written in 2005/6. Hilary Putnam, for example, radically contradicts the position I adopted at that time. For example, in Putnam's paper 'What Is Innate and Why: Comments on the Debate' (1980), he writes:
“... to have a concept is to have mastered a bit of theory, that is, to have acquired the characteristic uses of such expressions as... and some key beliefs, expressed by sentences involving such expressions, or equivalent symbolism.”
However, later he writes:
“I don't claim that all concepts are abilities to use symbolism; an animal that expects the water to reach the same height when it is poured from a pot back into the glass might be said to have a minimal concept of conservation, but I claim that anything like the full concept of conservation involves the ability to use symbolism with the the complexity of language in certain ways.”
Church, Jennifer, 'Fallacies or Analyses?' (1995)Kant, Immanuel, A Prolegomena to Any Future Metaphysics (1783/2004)
Peacocke, Christopher, A Study of Concepts (1992)
Tye, Michael, 'A Representational Theory of Pains and Their Phenomenal Character' (1990)