6
$\begingroup$

As a newbie to psychophysics experiment, I have a small project to do visual field testing on healthy, adult, naive subjects. The experiment stimulus is a small sine grating (0.3 degree, 6 cycles/degree) embedded in one of 25 possible locations in a noise background picture (18 degree). The orientation of the grating, and the contrast of background noise is fixed, and I need to measure the sensitivity (d') at 25 locations given a series of (probably 3) grating contrasts, and a series of stimulus exposure time (50,100,200,300,400 ms).

In a single trial, subject knows where the grating will be located, and maintains fixation at the center of noise background. Then the stimulus will be briefly presented (once in Yes/No and 2AFC, twice in 2IFC) for one of the stimulus exposure time, and subject need to judge whether the grating is present (Yes/No), or is located in which stimulus (2IFC and 2AFC). For a each unique combination of grating location, grating contrast and stimulus exposure time, I plan to do 16-20 repetitions to measure d'. So there are LOTS of trials.

I think this should be a mature experiment, but I am stuck on choosing between the following 3 experiment paradigms. The major problem is the time required to complete the experiment. I have run pilot experiments on all of them.

  1. 1-interval yes/no task. This task runs fast, but suffers from changing response bias. Is the response bias going to negatively affect d'? In signal detection theory, d' is insensitive to bias only when the variance of response to signal and signal+noise is equal. However, I don't know whether there is firm experimental evidence to support this.

  2. 2AFC task. This task comes from a suggestion in textbook. Instead of presenting one cue of target location on each trial, I can present two cues that is symmetric around fixation point, and subject need to judge which cue location has the grating. The task assumes that human visibility map is central symmetric, and effectively cut the number of locations to half, so it runs even faster than the Yes/No task. However, it is rarely used in literature about visual field testing, what might be the potential drawback?

  3. 2IFC task. I see many literatures doing this, but it is the slowest one (more than double of the time for Yes/No). What's worse, because the stimulus is brief presented, and the interval between them is short (400 ms), sometimes subjects report that they "see the stimulus but forget which interval has it". So in some of our pilot data, the sensitivity at fovea is lower than in near-peripheral, because subjects had made an error. How do psychophysical researchers overcome the problem of subject error, and the problem of long, boring experiment?

Thanks for any help.

$\endgroup$
  • 1
    $\begingroup$ Thank you for a very well researched question (although I can encourage you to also add sources where applicable). Could it potentially be of benefit to ask the three points in three separate questions? I'm unfamiliar with this type of research, so I'm not the greatest judge on this, but they seem like questions which could stand on their own. $\endgroup$ – Steven Jeuris Oct 25 '18 at 16:27
1
$\begingroup$

I am unsure as to the exact purpose of your experiment; the title seems to point toward a visual field test, whereas the question body seems to point to a visual acuity (VA) test? Are you interested in testing grating VA across the visual field? In the latter case, a subject needs to focus on a particular point and visual scanning is absolutely forbidden. The VA will be determined according to the last visible grating across a number of visual angles.

On way or the other, I think you answered your question yourself.

A yes/no task is prone to bias, as the threshold to say 'yes' may change over time. It's also prone to investigator influence, as the % false-positive rate (which you should keep track of online) may exceed a certain threshold (often set at 25%). At that point the researcher should intervene and tell the subject to elevate their threshold, after which the data are likely substantially different from previous trials. However, it's fast, as only one stimulus per trial is needed.

I have personally always used serial 2AFC tests. A subject definitely needs time to do the test. As you say yourself, 400 ms inter-stimulus time is simply too short. I think I would opt for a presentation time of a few seconds, a downtime of at least a second, and then the second stimulus.

The problem with the simultaneous 2AFC task is one of definition - if one stimulus is the target, then the other one is the no-stimulus. If they are up simultaneously, then isn't this identical to the yes/no task? The subject either perceives a grating, or not (yes/no). As wikipedia states:

Two-alternative forced choice (2AFC) is a method [where] the subject is presented with two alternative options, only one of which contains the target stimulus, and is forced to choose which one was the correct option.

A true 2AFC task is thus when they need to choose between A or B. That's only possible when doing it serially in your case as far as I can see, as a simultaneous task can basically be reduced to a yes/no task (they see the grating, or they don't).

$\endgroup$
  • $\begingroup$ Hi, thanks for the reply. I will add a few clarification here. 1. "a subject needs to focus on a particular point and visual scanning is absolutely forbidden": Yes I do it by online monitoring with an eye-tracker. Subject need to fixate at screen center throughout the trial. 2. "That's only possible when doing it serially in your case": In my 2AFC task, I present two cue locations (equal distance, opposite direction from fixation point) at a time, the grating will be in one of the two locations, and subject need to judge which location has it. Isn't it a true 2AFC task? $\endgroup$ – Cloudy Oct 26 '18 at 0:27
  • $\begingroup$ @Cloudy I'm not entirely following what your goals are, but keep in mind that fixation and attention aren't the same thing, and if you have multiple targets then attention will almost certainly wander and might have different patterns for different subjects. $\endgroup$ – Bryan Krause Oct 26 '18 at 0:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.