[visionlist] Forced Choice – please educate me

[Frederick Kingdom, Dr.]

Michael

Here’s my two-pennyworth. I think there are two interrelated issues here.  First even with the classic nAFC or nIFC task, in which on each trial your n stimuli are presented in different locations or in different intervals you can still have location or interval biases. In principle however sensitivity and bias can still be decoupled in these tasks using signal detection theory via a reformulation of the data into hits and false alarms.  Second it depends on what you mean by “forced choice”; forced choice in terms of the location or interval of the target or forced choice in terms of the available decisions.  For example in your example of the pattern being right/left there are two ways you can do the experiment: present one or other orientation on a trial and the observer has to decide “left” or “right”, or present left and right in two intervals and the observer has to choose the interval with “left” (or “right”).  Some would argue that the first  “single interval” method is not true forced-choice, in spite of the fact that it involves two alternative decisions.    Irrespective of what you mean by forced-choice however the first method is undoubtedly the more prone to bias, but in both cases you can in principle use SDT to decouple sensitivity from bias.

Fred Kingdom


From: visionlist <visionlist-bounces at visionscience.com> on behalf of Michael Bach <bach at uni-freiburg.de>
Date: Tuesday, January 21, 2025 at 9:33 AM
To: visionlist at visionscience.com <visionlist at visionscience.com>
Subject: [visionlist] Forced Choice – please educate me
[Some people who received this message don't often get email from bach at uni-freiburg.de. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

Dear Colleagues:

I have the impression that the general understanding of a psychophysical forced-choice paradigm has changed over the decades.

To my understanding, bias/criterion effects can only be effectively mitigated in an nAFC (n-alternative forced choice) design where the alternatives are _equivalent_. Examples: the pattern is on the right/left (balanced, of course), the stimulus was in the first/second interval, or the gap of a Landolt C is in one of 8 directions. “Equivalence” may be challenging to achieve, particularly in the latter case, where the oblique effect may interfere.  Alternatives such as “seen/not seen” or “go/no” are not equivalent and necessitate measures (d’, …) to address bias/criterion.

For me, nAFC implicitly implied equivalent alternatives, but was this ever the case? Whatever, young scientists do not seem to understand it this way.

While this might be just seen as a nomenclature trifle, the real problem arises due to the prevalent “scientific folklore” that a forced-choice paradigm eliminates the effects of bias / criterion change (which, in particular, occur through perceptual learning in longitudinal studies). However, if forced-choice does not employ _equivalent_ alternatives, this assertion is invalid.

I have encountered opinions suggesting that the addition of a “not seen” button constitutes forced choice (yes…, but:) and also eliminates bias…


Looking forward to your advice, best, Michael
--
https://can01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmichaelbach.de%2F&data=05%7C02%7Cfred.kingdom%40mcgill.ca%7Cf143d16344ba407c557308dd3a285af7%7Ccd31967152e74a68afa9fcf8f89f09ea%7C0%7C0%7C638730668395103597%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C60000%7C%7C%7C&sdata=UGBO2QHCCyBCSB2ajgJBYNZHtjrwysHKYrgZPW55WFU%3D&reserved=0<https://michaelbach.de/>


_______________________________________________
visionlist mailing list
visionlist at visionscience.com
https://can01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fvisionscience.com%2Fmailman%2Flistinfo%2Fvisionlist_visionscience.com&data=05%7C02%7Cfred.kingdom%40mcgill.ca%7Cf143d16344ba407c557308dd3a285af7%7Ccd31967152e74a68afa9fcf8f89f09ea%7C0%7C0%7C638730668395122612%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C60000%7C%7C%7C&sdata=wDCuyns0lalMA%2Blzh61VkLhJu2onh6NpB7raRiKa6Ag%3D&reserved=0<http://visionscience.com/mailman/listinfo/visionlist_visionscience.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://visionscience.com/pipermail/visionlist_visionscience.com/attachments/20250121/a4048234/attachment.html>