QUEST for measures other than intensity

1
Dear all,

We are trying to obtain thresholds for peoples' ability to perceive
biological motion in noise. So the measure of difficulty is "number
of noise dots". The threshold for normal controls seems to be on the
order of 20 (for 82% correct performance). However the QUEST
procedure (Watson & Pelli, 1983) assumes that the thing being
measured is log intensity.

Can anyone help with the following:

(1) Is it valid for us to just use log(number-of-noise-dots) or not?

(2) Does it matter that with intensity, the task becomes easier as
intensity increases, whereas with number of noise dots, the task
becomes harder as the number of noise dots increases?

(3) Is it appropriate to use the QUEST procedure with brain-damaged
patients? Should we use a higher delta to account for the fact that
they make more spurious errors?

PS: There is a paper that uses QUEST for biological motion
thresholds (Neri et al, 1998, Nature 395, pp. 894-6) but they don't
provide full details on exactly how they did it.

Thank you very much,

best,

Stephen Wilson
UCLA

[Many thanks to Denis Pelli and Michael Shadlen for answers to my
previous question.]
2
Dear Denis Pelli,

I am looking for the chapters you mentioned in the reply below. Would
it be possible to post them on your website?

Many thanks,

Céline Gillebert
KULeuven

--- In psychtoolbox@yahoogroups.com, Denis Pelli <denis.pelli@...>
wrote:
>
> dear stephen
>
> there are several different levels at which one can address these
> questions. one can speculate about the nature of underlying
processes
> and so on. however, often we just want to measure threshold,
without
> getting into philosophy.
>
> for that i suggest noting that psychometric functions for a wide
> variety of things are sigmoidal and that the sigmoid varies in
> position along some axis (ie. threshold varies), and that the
> steepness of the sigmoid may vary too. The exact shape of that
> sigmoid is usually not important to measuring threshold. Any
> procedure that assumes a sigmoid with adjustable steepness will
model
> your psychometric function well enough to do a good job of
measuring
> threshold.
>
> empirical tests have shown that QUEST (and probably other
algorithms)
> is quite robust with respect to assuming the wrong steepness of the
> psychometric function, you just lose some efficiency, converging
more
> slowly.
>
> however, it is good to use a scale "log intensity" along which the
> psychometric function has an approximately fixed steepness. in your
> case i'd use "log number of dots" as the "log intensity". it would
be
> a good idea to measure a couple psychometric functions for your
task
> at low and high dot numbers to provide estimates of the parameters
> for QUEST.
>
> good luck
>
> denis
>
> p.s.
> bart farell and i wrote a couple chapters with advice on just this
> issue of how to frame the question to get ready for measurement:
>
> Farell, B., & Pelli, D.G. (1996). Psychophysical methods. In: J.G.
> Robson, & R.H.S. Carpenter (Eds.), A Practical Guide to Vision
> Research (New York: Oxford University Press.
>
> Pelli, D.G., & Farell, B. (1994). Psychophysical methods. In: M.
> Bass, E.W. Van Stryland, D.R. Williams, & W.L. Wolfe (Eds.),
Handbook
> of Optics, 2nd ed., I (pp. 29.21-29.13). New York: McGraw-Hill.
>
> they're not on the web now, but i'll try to add links to my web
page
> later this week.
> http://psych.nyu.edu/pelli/
>
> >Dear all,
> >
> >We are trying to obtain thresholds for peoples' ability to perceive
> >biological motion in noise. So the measure of difficulty is "number
> >of noise dots". The threshold for normal controls seems to be on
the
> >order of 20 (for 82% correct performance). However the QUEST
> >procedure (Watson & Pelli, 1983) assumes that the thing being
> >measured is log intensity.
> >
> >Can anyone help with the following:
> >
> >(1) Is it valid for us to just use log(number-of-noise-dots) or
not?
> >
> >(2) Does it matter that with intensity, the task becomes easier as
> >intensity increases, whereas with number of noise dots, the task
> >becomes harder as the number of noise dots increases?
> >
> >(3) Is it appropriate to use the QUEST procedure with brain-damaged
> >patients? Should we use a higher delta to account for the fact that
> >they make more spurious errors?
> >
> >PS: There is a paper that uses QUEST for biological motion
> >thresholds (Neri et al, 1998, Nature 395, pp. 894-6) but they don't
> >provide full details on exactly how they did it.
> >
> >Thank you very much,
> >
> >best,
> >
> >Stephen Wilson
> >UCLA
> >
> >[Many thanks to Denis Pelli and Michael Shadlen for answers to my
> >previous question.]
> >
> >
> >
> >http://psychtoolbox.org
> >POST a message to: psychtoolbox@yahoogroups.com
> >UNSUBSCRIBE by sending a blank message to:
> >psychtoolbox-unsubscribe@yahoogroups.com
> >
> >Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/
>