4
$\begingroup$

Firstly, I need to declare that I am not psychologyst, however I am currently working on visual recognition research. To be precise, I working on face recognition deficits in Autism Spectrum Disorder. I am in charge of doing an experiment but I feel I come to a dead end even though I didn't even start with it.

As I mentioned above, I investigate face recognition deficits in ASD. I will use faces and objects. For this task I will use 2-AFC match-to-sample task. I need to morph faces and objects and for this purpose I need to use a QUEST staircase for measure morph distance at which participants could discriminate between items with 75% accuracy.

Here is the problem: I don't know how to use a QUEST staircase. Or better say, I don't know how can I decide how many trials should I have, which morph distance should I use between items, how should I know which morph distracter should I use, should I use in the beginning easy test pairs and what exactly does it means, is 85%-morph distracter is an easy pair, etc.? As you can see I don't know to do an experiment, it is almost completely new field for me.

I am trying to find my answers in: Psychophysics: A Practical Introduction by Kingdom & Prins; Experimental Design by Cunningham & Wallraven; Visual Psychophysics by Lu & Dosher just to mention a few. However, I wish to find some book or article where I can see an example step by step how to make a similar experiment, i.e. AFC. I know there is bunch of information but I am just overwhelmed and lost.

What literature/ web sites/ blogs etc. can you suggest to me? How/ where to start?

$\endgroup$
  • 1
    $\begingroup$ I will try to offer information on answers about detection of signals and tests but I must say and it is not my intention to offend, I do not know why it depends or you direct this investigation if it is not your field of work or investigation, perhaps you should get in links with some a university nearby to provide you with contacts with doctors or students who will finish. $\endgroup$ – hexadecimal Jul 15 '17 at 5:53
  • $\begingroup$ Thank you for your effort! I understand your point of view and I agree with it. Before I got to the part of doing an experiment I indeed have studied the literature, I got some basic grasp on the topic and I saw that is very common method/ experiment. I knew that I will need help in one moment and I am surrounded with psychologists so I took the challenge. However, I found out they are not familiar with the 2AFC, they just didn't studied it in detail. They are clinical psychologists so I guess I should expect it. $\endgroup$ – Enasi Jul 15 '17 at 11:06
  • $\begingroup$ The thing is, I am interested in this topic and I knew it I will need to put more effort than someone else from this field. While I was reading literature it seemed to be more or less fine. I got idea how the experiment should be but when I really started to plan it in a detail I just have gotten lost. $\endgroup$ – Enasi Jul 15 '17 at 11:15
  • $\begingroup$ Note that the researcher should always be able to justify the methodology used, and we talk about advanced methodology (Bayesian analysis and inferences) in ASD, beyond the answers offered here, I recommend you to follow my advice and contact with studenten from a nearby university. $\endgroup$ – hexadecimal Jul 15 '17 at 11:40
  • $\begingroup$ I am indeed trying to find some student or researcher, still I wish to gather more information in this or similar way at the same time. $\endgroup$ – Enasi Jul 16 '17 at 4:26
6
$\begingroup$

There is a lot going on in your question. QUEST, as well as many other adaptive procedures, is well suited for a task like estimating morph distance in a 2-AFC paradigm. There are, however, a couple of points in your question, that make me think QUEST is not a good paradigm.

The QUEST paradigm is designed around setting the signal level to find the the "sweet point" of the psychometric function (something like the signal level that maximizes the Fisher Information). I am not aware of any psychometric function that has a sweet point of 75% correct (most are 90+ percent correct).

The second issue, is you mention a fixed number of trials. Typically, QUEST is used with a stopping rule that specifies the expected variance of the estimate (i.e., you run enough trials until the CI on the estimated threshold is small enough).

The easiest way to run these types of experiments is to use software written by someone else (that you trust). Pelli was involved in the early development of both the QUEST procedure and the Psych Toolbox for MATLAB/Octave. The Python based PsychoPy also provides the QUEST procedure and is commonly used.

$\endgroup$
4
$\begingroup$

Start here: Adaptive Psychophysical Procedures (Treutwein, 1995) [PDF]

If you want to get fancy, I suggest a more contemporary approach for the staircase called accelerated stochastic approximation. You just need to figure out how to implement the equation and how to compute the scores, d-prime, etc.

Another classic, more general (ask a librarian): https://www.amazon.com/Signal-Detection-Theory-Psychophysics-Marvin/dp/0932146236

$\endgroup$
  • 1
    $\begingroup$ Thanks you for the links. It would be great if you could also describe the papers briefly, to introduce the papers a bit and make it clear what the gist is of these papers. $\endgroup$ – Robin Kramer Jul 15 '17 at 22:30
  • $\begingroup$ Good answer. I would like to point out that although in this case the sensitivity functions of psychophysics are historically related to the functions of response to stimuli have little to do with the likelihood function. $\endgroup$ – hexadecimal Jul 17 '17 at 9:17
  • $\begingroup$ @hexadecimal I am not sure what your point is? Fisher developed the MLE method. It is applied in QUEST in 1983, when the QUEST method was first published. $\endgroup$ – noumenal Jul 17 '17 at 17:08
  • $\begingroup$ @RobinKramer The 1995 paper is a classic that provides an overview over adaptive staircase procedures. The textbook is one of the perhaps two most popular in psychophysics. $\endgroup$ – noumenal Jul 17 '17 at 17:10
4
$\begingroup$

There are numerous studies that reveal contradictory data about face-to-face in autism. Continues the research paradigm that contrasts faces and objects. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2174902/

I think the best text I've found to get started is: this:https://psych.nyu.edu/pelli/pubs/watson1983quest.pdf

Software: http://www.mas.ncl.ac.uk/~ndjw1/bookmarks/Stats/Software-Statistical_computing/Bayesian_software/index.html (Some links do not work but you can find this software by google)

Differing sources indicate that differences or difficulties in looking at the face in autism are to a greater extent related to communication processes.

Frith, Uta. (1989). Autism: Explaining the enigma. New York: Blackwell. Page 199 et seq.

(Essential book which I recommend starting a task about autism.)

This can constitute a powerful critic to this type of investigations today.

In relation to the above it is considered that the investigation of these factors should be considered in a communication context that on the other hand is not incompatible with research on basic processes.

$\endgroup$
  • $\begingroup$ Thank you. I will check the softwere. I have Pelli's article so I ought to read it better. Thank you for the part about ASD as well, although I am quite okay with it. $\endgroup$ – Enasi Jul 18 '17 at 5:56

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.