Perception – direct vs. representational

by Neil Rickert

The two most important theories of perception are representationalism on the one hand, and direct perception on the other.  There are probably many versions of each of those, and there are some other theories which I see as less important.  By far, the dominant theory — the one most widely accepted — is representationalism.  However, as mentioned in the previous post in this series on perception, I happen to prefer the idea of direct perception.

In this post, I plan to do to things:

  • I will briefly describe both representationalism and direct perception, and their disagreements;
  • I shall try to address some of the misconceptions about direct perception that seem to crop up.

After reading this post, you might still be confused as to what direct perception is.  I will be going into more detail in future posts, so what I write here serves only as an introduction.

Representationalism (indirect perception)

Representationalism is the view that we do not actually perceive the world.  Rather, we see things in the world only indirectly.  Perhaps what we are really seeing is an image of reality that is projected onto our retinas.  Or perhaps the brain is building some sort of internal model of the world, and we are really seeing that model.

Marr’s theory of vision is representationalist.  Most researchers in artificial intelligence are representationalist, and tend to think along the lines of Marr’s theory.  Steve Lehar is an AI researcher who supports representationalism and argues that direct perception is untenable.  You can find his online pages at The Representationalism Web Site.  If you follow the various links on Steve’s page, you will find his arguments for representationalism and against direct perception.  You will even find part of a debate from 2005, on the now defunct “psyche-d” mailing list, where Steve and I crossed swords on the issue.

Phenomenalism, the view that our only access to the world is via perceptual phenomena, is representationalist.  The sense-data theory of perception is representationalist.  In that theory, our sense organs deliver sense-data, and we must infer what the world is like on the basis of that sense data.

Direct perception

The Gibson position is that we directly perceive the environment.  In particular, Gibson is clear that we do not copy details of the world, and that perception does not involve inference.  In Gibson’s view, the perceptual system uses tranducers that are tuned to particular features of the world.  His account depends on the idea of perceptual learning, which would be where those transducers are tuned to specific features.  Note that what Gibson considered a transducer is different from what engineers mean when they use that word.

The Wikipedia page on Gibson is brief, but has useful links and references to various accounts of direct perception.  A paper by William Warren gives a reasonably good account of direct perception, and will take less time than reading Gibson’s several books.  The SEP page on Embodied Cognition might be useful.  And a youtube video (just over 4 minutes in length) gives a pretty good introduction.

I will be going into more detail in future posts.  For that, I will also have to say more about perception and about information.

Some misconceptions

The remainder of this post will be about some misconceptions that I have come across.

It is not just semantics.  I have seen some discussions where the participants seem to think that it is just a matter of how we talk about perception.  The idea would be that both direct perceptionists and indirect perceptionists are talking about the same processes, but just talking about them in different ways.  However, Steve Lehar clearly sees that there is more than semantics involved, when he argues (in pages linked above) that direct perception is untenable.  I see a clear difference at the implementation level.

As a teaser, to be discussed in a future post, I will mention the bar code scanner at the supermarket.  The way that works is the way that I expect direct perception to work.  The scanner itself can be thought of as a transducer that is tuned to bar codes.

Relevance to epistemology.  The first time I came across a mention of direct perception, it was in a book on epistemology.  The author took the traditional view that knowledge is justified true belief, and he argued that direct perception solves the problem of showing that perceptual beliefs are justified.

That is just a mistake.  Worse still, a proper understanding of direct perception actually tends to undermine both traditional epistemology and traditional philosophy of mind.

I should perhaps mention that the author of that book was a proponent of Objectivism, the philosophy of Ayn Rand.

Representation.  Some behaviorists seem to take an almost dogmatic position that opposes any mention of the idea of representation.  I am not sure whether Gibson would have agreed with that.  However, I find it useful to talk of representation, in conjunction with direct perception.  Gibson’s books contain a lot of talk about information, and information is representational.  However, I do not use representation in the way that representationalists want to use it.

Relation to science.  Throughout my theorizing about human learning and human cognition, I have looked to how science is done as an illustration of methods that are known to work.  Humans learn, but the details of how that works are mostly private.  Science allows us to gain scientific knowledge, and much of what is involved there is done in public.  So I see the growth of science as, at least, analogous to human learning.  Studying how science works can illuminate possible methods of learning, even if the two approaches are not identical.

As perceivers, we gain a lot of information about the world.  Scientists, using their scientific instrumentation, make observations of the world.  The obtaining of information via scientific obvservation is, in some ways, analogous to our acquiring information through perception.  Science acquires information with specialized instruments.  We can think of a thermometer as a kind of tranducer that is tuned to temperature information.  We can think of a clock as a kind of transducer that is tuned to time information.  So, if we look at scientific observation as analogous to perception, then it seems to be analogous to direct perception.


That’s my introduction for now.  I have attempted to contrast direct perception with representationalism, and have briefly discussed the main ideas.  I will be going into more detail in future posts.

5 Responses to “Perception – direct vs. representational”

  1. As somebody unfamiliar with the the subject, it would seem to me, at least on the surface, that both could be true, depending on the object perceived and the transducer utilized. Those things we can directly perceive, whether through the sense or with the assistance of tools (e.g., a microscope) would fall into the realm of direct perception, while others (e.g., atoms) would necessarily fall into the realm of representationalism.

    Of course, I could be completely missing the point and talking out my rear. 🙂 I’m sure you’ll clarify in future posts. I’ll be watching for them.


  2. I think that both direct and indirect perception are approaches which will sooner or later be moved to the background. Both reflect the input-process-output model, which I see gradually becoming outdated in this regard. Presently I see embodied cognition as the model which may unite both aspects into the action-experience cycle.

    As a small historical detail, I see Kant’s concepts of space and time as a first approach towards embodied cognition. They are not properties of objects, neither of reasoning – what’s left are the constraints of our physical actions in the world.


    • Both reflect the input-process-output model, …

      Actually, no, direct perception does not reflect the I/O model, so is far more congenial to ideas of embodied cognition. I didn’t say much about that in this post, because I’m planning to get to that in future posts. But that’s why I said that it tends to undermine traditional epistemology and traditional philosophy of mind.



%d bloggers like this: