[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Increment versus decrement detection


There are for me, at least two parts to this thread (see sig), Al Bregman' comments, and Bruno Repp's question.

I agree with Al on this and see the time when what he would like to see, becomes possible. My experience is that this is a continuation of the decline of the present way of collecting and disseminating information. There is much interesting work that has taken place because the results have been locked into vaults, called publications, whereas a new model of dissemination is found in the Wiki world, and through sites such as MySpace (in the arts) and the kind of more rapid communications found in FaceBook. In the professional world in which I travel, these ideas are frowned upon, but in the educational world -- the world of the growing researcher, these are the models for communication. More on the 'communications' thread later if anyone is interested.

I find Bruno Repp's questions interesting, but am not able to provide either a framework or a context in which to reply. If I read correctly, the question is based around presenting 12 equally spaced piano notes, all of the same dynamic level. I would like to be able to hear what was presented in order to compare my perceptions with the findings. Was the experiment worked in the middle register of the piano where variations in decay times can be detected in the rate of change of the harmonic partials? Was this done with a sampled piano sound?

If it was done with an acoustic piano, then the spectral content of each note changes at the microsecond level, and a well-formed ear works on multiple levels of multiple parameters to detect changes. More likely, this was done with a sampled instrument. Was this a sampled instrument that each sample was transposed up and down one semitone, or was it an 88 sample library?

Was the change of dynamic expressed simply by changing the amplitude of the sound file, or was it a library that contained 88o piano samples, that is, each note was recorded with 10 dynamic levels?

When the audio samples were being selected for the experiment, what was the amount of variation? (.3 dB, 1 dB, 3 dB ...)

What was the period between the notes? What was the amount of change?
These, in my experience, are important factors. Were the note 3 seconds apart, or 350 ms apart? Was the amount of variance 10 ms, 50 ms or 150 ms?

Once the 'other' parameters are set, (and controled for), it might be possible to start the examination of the role of memory in this study.

If all 12 notes were to be equally spaced and of equal amplitude, why would the 'control' version need to be played?

If the control version were played with no variation in amplitude or timing, did the participants report changes in amplitude or in timing? I would see this as being a 'control' on the outcome of any subsequent test. If in the 'control' (unvaried) version, all participants reported that the notes were of equal amplitude and equal timing, then I would consider their ability to detect differences as being 'significant'. But if there were errors in the 'control' version, how would these be controled for in the test.

In response to why it is easier to detect increment than decrement, the term hysteresis comes to mind immediately. http://en.wikipedia.org/wiki/File:B-H_loop.png

In the second part of the experiment I am most interested in the 'false alarms', as the subject has detected something changing when no change has taken place.

In the second part, without knowing anything about the "so-called" melodies, it is hard to comment. The word melody is both language- constrained, and culturally constrained. Were these melodies based upon a twelve-note equally tempered scale that showed 'tonal' features? Was an underlying metric structure implied, or inherent? I think of how the Viennese play waltzes. http://www.youtube.com/watch?v=KEdJ6pQXmoU

For a view of how Sachs and Nettl disagree on aspects of melody, start here:

Sometimes the conclusion drawn is the one which is predicted, but may not be accurate. There was an article from March 2005, now lost, that correlated physical violence in men as being inversely proportional to the length of the first finger (Jupiter). The article went on to point out that this correlation did not exist in women, for some reason.

Most palmists would explain this immediately -- a view that would be rejected by scientists, in that in palmistry, Jupiter is reflective of the individual's self-image regarding superiority / inferiority. Men with feelings of inferiority may act this out in the form of aggression, while in women this is reflected by submission. Violence in women, as reported in the literature on chiromancy, is associated with a long Jupiter, often supplemented by a short, bulbous nailed thumb. But such nonsense is not the domain of serious scientific consideration. Yet, a knowledge of this may strengthen your research.



earth rat


Date:    Fri, 2 Oct 2009 00:02:47 -0400
From:    Al Bregman <al.bregman@xxxxxxxxx>
Subject: Re: Increment versus decrement detection

Hi Bruno,

The query you sent about your new data was just the sort of thing that the AUDITORY list should be used for -- the sharing of ideas that can strengthen research and hasten development. Thanks from all of us for sharing these findings. I hope that the responses will have have
beneficial effects for your research.



On Wed, Sep 30, 2009 at 10:37 AM, Bruno Repp <repp@xxxxxxxxxxxxxxxx> wrote:
Dear colleagues:

I recently conducted some experiments in which participants had to detect a single changed tone in an isochronous melody consisting of 12 successive complex tones (piano tones) of different pitch but equal loudness. The change to be detected was either in intensity or in duration, and it was either an increment or a decrement. These were four separate tasks, not intermixed. The melody containing the change was immediately preceded by the same melody not containing any change. I would be grateful for comments or references that would help me understand the following trends in my data:

(1) Decrements are harder to detect than increments. Is this well known from
previous psychoacoustic research? What is the reason?

(2) The pattern of variation in detection scores (hits and false alarms)
across the 12 tones in the melody, which reflects influences of pitch
contour and other factors, is uncorrelated between increment and decrement
detection. It seems like these two tasks have little in common. Why?

(3) In increment detection, both hits and false alarms tend to increase across the 12 tones in a melody, but decrement detection shows the opposite
trend. It is as if tones were expected to become softer as the melody
progresses. Why?

Many thanks in advance for any helpful replies!


Bruno H. Repp
Haskins Laboratories 300 George Street
New Haven, CT 06511-6624
Tel. (203) 865-6163, ext. 236
Fax (203) 865-8963

Albert S. Bregman, Emeritus Professor
Psychology Department, McGill University
1205 Docteur Penfield Avenue
Montreal, QC, Canada H3A 1B1.
Office:  Phone: (514) 398-6103, Fax: (514) 398-4896
Residence phone & fax: (514) 484-2592


Date:    Sat, 3 Oct 2009 12:57:50 -0400
From:    Fred Herzfeld <herzfeld@xxxxxxxxxxxx>
Subject: Jan Nordmark

Hello list,

Can anyone tell me the whereabouts of Dr. Nordmark. Up to the end of 1993 I was in touch with him (in London and in Stockholm) regarding my new theories of the hearing mechanisms in vertebrates. This work is now "completed" and I cannot locate Jan to let him know the final mathematical model.

Fred Herzfeld, MIT '54
78 Glynn Marsh Drive #59
Brunswick, Ga.31525


End of AUDITORY Digest - 2 Oct 2009 to 3 Oct 2009 (#2009-226)