Hi Massimo,
I totally agree with Dick. An adaptive procedure works regardless of the starting values you set – after all, it’s one of the main reasons to use an adaptive procedure that you don’t need to know the “threshold” in advance.
However, for fostering the participant’s understanding of the task and for their motivation, it certainly seems better to start with a condition where they can likely give a correct response.
Just a cautionary note, I would not term this “approach from above” as in a method of limits. In the latter task, we know that the approach from above almost certainly gives results a different measured “threshold” than an approach from below (due to “hysteresis”-like response biases). In an adaptive procedure, regardless of where you start, the relevant stimulus aspect (frequency difference in your case) will at some point arrive at a value where the participant can respond with (in your example) the proportion correct tracked by the adaptive rule. Unless you run too few trials, the tracked point on the psychometric function will be “bracketed” by the adaptive track during a certain number of trials, so that the starting value is very unlikely to play any role.
Greetings from southern Germany, where it’s also rather cold an rainy right now 😉
Daniel
---------------------------------
Prof. Dr. Daniel Oberfeld-Twistel
Johannes Gutenberg - Universitaet Mainz, Experimental Psychology
Wallstrasse 3
55122 Mainz
Germany
Phone ++49 6131 39 39274
http://www.staff.uni-mainz.de/oberfeld/
From: AUDITORY - Research in Auditory Perception <AUDITORY@xxxxxxxxxxxxxxx> On Behalf Of Richard F. Lyon
Sent: Saturday, November 18, 2023 6:31 AM
To: AUDITORY@xxxxxxxxxxxxxxx
Subject: Re: Estimation of auditory sensory thresholds from below: any evidence?
I think the subject needs to feel like they're doing the job well, so starting easy (above threshold) makes sense. And they get to know very well what signal to listen for.
If you approach from below, they don't know what signal to expect, so may not be in an "optimal detector" state of mind.
Dick
On Fri, Nov 17, 2023 at 9:19 PM Massimo Grassi <massimo.grassi@xxxxxxxx> wrote:
Dear all,
I was recently writing a paper about threshold estimation in hearing and at some point I stopped.
Let suppose you are estimating a frequency discrimination threshold with a classic 2-down 1-up staircase rule (Levitt, 1971). Theoretically, you could approach the threshold from below or from above, I repeat this every year to 1st year students when I teach them the method of limits! In practice, however, in psychoacoustics we rarely-to-never approach the threshold from below.
Is there any reason for this "tradition"? I know the psychometric function should be symmetric below and above threshold! But this does not explain the "tradition"!
All the best from "it is almost winter" North East Italy.
m