Perception and stimulus statistics: any good experiments?

As part of a series on consciousness I am writing, I am presently writing up some notes on the Necker Cube. I have a question perhaps some of the more historically minded readers could answer.

Most of you are familiar with the Necker Cube (see one to the left), a line-drawing that can appear as a cube facing down to the left, or up to the right. People’s experience of the stimulus switches every few seconds between these two percepts.

There are quite a few studies of the effects on perception of the Necker cube when you first show subjects a disambiguated picture of a cube. If the disambiguated version is shown for a fairly short time, then they will see the same perspective when you switch over to the ambiguous Necker. However, if shown a disambiguated version for a long time (e.g., minutes) then the opposite perspective is seen after the switch (this is usually interpreted as some kind of low-level adaptation effect).

My question is, has there been research on much longer
time scales, with extensive exposure to disambiguated cubes? For
instance, show subjects the same disambiguated cube over weeks or
months, every day (for example, pictures with many down-left cubes and
zero up-right cubes). Perhaps the disambiguated cubes could be used in
some attention task unrelated to bistable perception. After N days of such training what happens on day N+1 if we show the
ambiguous cube? Will perception be biased toward one or the other perspective?

During training, would the brain change its statistical estimation of the frequency of
the two cubes in the world? If so, what would our prediction be for the percept? Would
subjects be more likely to see the Necker cube as the disambiguated version shown during training? Or would the brain be primed
to see the less probable cube, the
cube that offers a higher surprise value?

Does anyone know if this has been done with Necker cubes, or similar long-term studies have
been done with other ambiguous stimuli? I know there is tons of work on adaptation effects in which stimuli bias perception in the short term (e.g., show a bunch of
lines slanted to the left, then vertical lines will appear
slanted to the right). I am asking about longer-term effects.

For the research on short-term effects of disambiguated Necker-cube like stimuli, see Gerald Long’s work such as:
Long G M, Moran C J, 2007, “How to keep a reversible figure from
reversing: Teasing out top – down and bottom – up processes” Perception
36(3) 431–445 .


  1. gualtiero

    I don’t know if the studies you suggest have been done, but if not, it sounds like they should be.

    Thanks for linking to your posts on consciousness. (And feel free to cross-post here.)

  2. Eric Thomson


    I had a little email exchange with Dale Purves (the ideas here were partly inspired by his work on perception). He says he is not aware of any such work on longer time scales. His emails convinced me it might be better to use other ambiguous stimuli because we see cuboid shapes all the time during the day, and we might bias the statistics more using unfamiliar stimuli.

    Probably one reason it hasn’t been done is that the experiments would take a lot of time.  Getting subjects to agree to do an experiment for multiple sessions is always hard. Then you have to worry about controls (e.g., to be really tight, you might want to have other subjects do the same task without cubes (or on ambiguous cubes) for the same amount of time and then test their perspective on the Necker cube).

    Methodologically, I have been wondering when the best time would be to train the subjects in the ambiguous Necker task (i.e., before, during, or after training in the “nonsense” task whose point is to expose them to the disambiguated cubes). There are problems with each. However, I’m sure all these questions could be answered and the experiment could be done well enough to make the results interesting. Plus, even the work of Long and others that look at priming/adaptation effects after short-term exposure to disambiguated cubes must have similar concerns.

    I found one more researcher doing work on this (Sundareswara): one abstract includes
    Sundareswara, R., Kallie, C. S., & Schrater, P. R.
    (2006). Perceptual bistability modulated by priming. Journal
    of Vision, 6(6):53

    He tends to look at it from a Bayesian perspective, so would expect the Necker to be biased toward the disambiguated cubes. I will email him my question.

  3. Graham Hamilton

    Hi – I know of no study of the sort you are looking for – one problem which may not be a problem with priming is that the stimulus is not bi-stable see the abstract of my paper (in press) below. Cheers Graham

    The Necker cube is not (only) Bi-stable
    Graham Hamilton
    Faculty of Health and Life Sciences, York St John University, Lord Mayor’s
    Walk, York, 31 7EX, UK: email:
    Paul Goddard, Steve Wilson
    Department of Psychology, University of Lincoln, Brayford Pool, Lincoln,
    LN6 7TS, UK
    Following Borsellino, DeMarco, Allazetta, Rinesi & Bartolini, (1972) an attempt was made,
    using four observers, to utilise the Gamma distribution as a quantitative model with which to
    characterise the durations of alternating percepts of the Necker cube. By adopting a modified
    method of collecting observer’s reports of different percepts separately the fits of the distribution
    to the data showed some promise. However, the most important finding was that though changes
    of depth perception (3D) were present, so were changes associated with two-dimensional
    percepts (2D). Most prevalent was a percept likened to a “toffee wrapper” (TW). Interestingly,
    the fits of the Gamma distribution to this data were also good. 2D and TW percepts are discussed
    in relation to contemporary and historical research.

Comments are closed.