[From Rick Marken (990417.2050)]
This is a reply to an old post from i.kurtzer (990331.1400).
Isaac sent me a couple of papers that seemed to be examples
of conventional research that represent attempts to test for
controlled variables. I'll just comment on one of them now. It's
called "Guidance of locomotion on foot uses perceived target
location rather than optic flow" by S. Rushton, J. Harris,
M. Lloyd and J. Wann and published in _Current Biology_ (1998).
The aim of this study was to determine what "cue" people use to
"guide" walking to a target. Two possibilities were considered:
1) focus of expansion (FoE) and 2) retinal target location.
When you walk toward an object there is an "optic flow" of the
visual scene outward from the fixation point; this flow is zero
at the fixation point; that point is the FoE. I think the FoE
hypothesis is that people walk towards a preselected target by
keeping it fixated so that it is at the FoE. So it's fairly
weird for Rushton et al to talk about the FoE "guiding"
locomotion; it's really locomotion that guides the FoE.
The other hypothesis is that the retinal location of the target
(relative to the fovea) "guides" locomotion. So people locomote
to a target by keeping the target retinally located on the fovea.
Again, it's really the locomotion that controls target location
on the retina, keeping it foveated; but who's keeping score?
Rushton et al wanted to know whether people are "guided" to a
target by FoE or retinal location. They reasoned that, if people
are guided by FoE, then locomotion to a target would _not_ be
disrupted if people wore prism glasses. Such glasses change
the retinal location of the target but they don't affect optic
flow; an object at the FoE without the prism glasses will still
be the FoE with them on.
What Rushton et al found was that the prisms did disrupt the
ability to walk towards a target. Subjects stood about 10 meters
from a target (a ball held at arms length by the experimenter).
A digital camera recorded their paths to the target (from above).
These paths were curved (they would not be if the subject were
walking toward the FoE) and they usually ended somewhat to the
side of the target (also expected if the subjects are "guided"
by retinal target location rather than FoE).
Superficially, it looks like Rushton et al have done something
like a test for the controlled variable. They came up with some
hypotheses about the variable under control (FoE, retinal
location); manipulated a variable (prisms off/on) that would be
a disturbance if the subject were controlling one variable
(retinal location) but not the other (FoE). They even measured
(indirectly) one of the hypothetical controlled variables
(angle of target relative to fixation; they called it alpha)
and measured the effect of the disturbance on this variable
(they correctly predicted that alpha should be kept close
to 16 degrees -- the displacement angle of the prism glasses --
if alpha is under control; see their Figure 2b).
Rushton et al found that subjects did, indeed, act (by taking a
curved rather than a straight path to the target) as though they
were resisting the prism disturbance to retinal location; they
also presented evidence that the retinal location of the target
is kept foveated (alpha is nearly constant at the prism angle;
see their Figure 3). So they concluded (not in these words, of
course) that FoE was _not_ a controlled variable and that retinal
location _is_ the variable people control when they walk to
a target object.
In fact, there is no test for a controlled variable in this
study. Rushton et al's results show what people _don't_ control
not what they _do_ control. Their results show clearly that
don't control FoE (and probably not retinal location either;
see below) when they walk to a target. At best, the Rushton
et al research _eliminates_ some possible hypotheses about
the variables people might control when they walk to a target.
Unfortunately, these researchers leave the impression that
they _have_ identified_ a controlled variable (or, in their
words, a "cue" that "guides" locomotion): retinal target
location. This is not the case at all.
What Rushton et al have done is just the first few steps
of a test for controlled variables. They guessed at two
possible controlled variables. Then they predicted (correctly,
I assume) how those variables would behave when a person walked
to a target with and without a disturbance (prism displacement).
Those predictions can be described in terms of alpha, the angle
between actual target location and the direction of the subjects
gaze. If the subject is controlling FoE then alpha will be (and
remain) zero as the subject walks to the target, with or without
the prism glasses. If the subject is controlling retinal target
location then alpha will be (and remain) zero as the subject
walks to the target _without_ the prism glasses and 16 degrees
(the prism angle) as the subject walks to the target _with_ the
prism glasses.
Rushton present, in their Figure 2b, an excellent description
of the expected behavior of alpha (the hypothetical controlled
variable if subjects are controlling retinal target location)
as the subject walks to the target with prism glasses on. The
prediction is that alpha will be constant (at 16 degrees) over
the course of the walk. The actual behavior of alpha for
5 different subjects is shown in their Figure 3; in only one
case was the behavior of alpha anywhere close to being a
constant (and in this case alpha was close to zero rather
than 16 degrees, an indication that this subject was _not_
controlling the retinal location of the target). Rushton et al
conclude that the subjects were controlling alpha (at 16 degrees)
because the _average_ value of alpha (over 5 subjects) over
the course of the walk was _nearly_ constant at _nearly_ 16
degrees. So Rushton et al are violating the first commandment
of behavioral research: thou shalt not base conclusions about
individual behavior on group averages.
There are many other problems with the Rushton et al study
from a PCT perspective. Most important, Rushton et al don't
say whether the data in their Figure 3 represent the first,
second or nth try at walking to a target with prisms. It may be
that the poor control of alpha seen with the prism glasses is
just a result of not having given the subjects time to learn to
control with the glasses on. As it sits, we don't know whether
the behavior of alpha shown in Figure 3 represents _poor_
control of retinal target location or _good_ control of some
other variable.
So the Rushton et al study is not even an "accidental" test for
controlled variables (if it were a test then it would have
to have been accidental since Rushton et al have no idea what a
controlled variable is) since they don't show us what variable
_is_ under control when people walk towards a target. Their
research does _eliminate_ a couple of possible controlled
variables (FoE for sure and retinal target position _almost_
for sure) which, I suppose, is better than nothing -- but not
much.
I think that, in order to do the test for controlled variables
properly, one _really_ has to know what a controlled variable _is_.
Best
Rick
···
--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken/
