I'm doing a preliminary research on auditory sensory memory, especially differences due to distictive neural pathways as compared to the olfactory sense neural pathway.
As you know, some of the fundamental psychophysical testing procedures used in auditory testing today have been originally tested and honed for/with the olfactory senses (see for example: ' Sensory Evaluation Techniques').
However, the current neurological research shows a distinct pathway for smell/taste as compared to hearing. The coupling with sight and hearing appears to be more tighter than with the olfactory senses.
I would be interested in any tips for references from the fields of neurology, psychophysics or psychoacoustics that deal with these issues and try to address these in the test methodologies used currently for auditory testing.
As a teaser, I'll point you to an interesting MMN research article that suggests that guided attention can enhance auditory perception, but has no measurable effect on the long term auditory memory as currently tested for single stimulus detection tasks. I think this poses interesting questions for some of the testing auditory methodologies used, especially considering the time intervals, and affirmation of correct answers in auditory discrimination tests. There are other similarly interesting research articles, mainly from the field of neurology.
Does Auditory memory depend on Attention?, L.Demany, S. Clément, C. Semal. Journal of the Acoustical Society of America, 109, 686-700. 2001
Is this the paper?
If I may paraphrase and summarize: if you know what type of difference to listen for, your sensitivity is increased.
In the experiment, trained listeners were randomly presented with one of three types of differences and asked to discriminate between A and B. Their sensitivity was increased if they were also shown a visual clue as to what type of difference was being presented.
It may have implications for codec testing. For instance, should the identify of the codecs be made known for a particular test? If so, an experienced listener who knows his codec artifacts might know what to listen for and thereby gain an advantage in discrimination ability (which may or may not be desirable, depending on the goal of the test).
Yes, that was the paper indeed. And the conclusion they made for single stimulus auditory discrimination was as suggested.
I'm also interested in papers on habituation and sensitation for auditory nerve firings, which may or may not have an effect on how many repetitions and at which duration each should be used in a auditory test.
Initial hypothesis suggest sensitation to differences with prolonged exposure (i.e. becoming intimate with the test signal through repeated exposure and coded transfer to long term auditory memory) to identical stimulus, but relatively rapid lowering of late auditory evoked potential levels with repeated similar signals (i.e. the differences vanish).
All references are welcome and I'd like to hear about people's personal experiences as well.
What is the maximum number of single session test repetitions with perceptually similar samples you can do, before the audible differences between the samples (minor impairments) become increasingly difficult to discern or become impossible to notice at all?
In a 16 sessions ABX listening test, the differences vanish near the 6th one for me. That must be between 10 and 20 repetitions.
However, I got the opposite effect on the Astral Projection sample for MPC, I must listen to it repeatedly for about 10 times in order to be able to start ABXing it. Only after several tens of repetions do the difference vanish to my ears.
It depends. If your ability to hear differences diminishes as you do more repetitions, is because you or your auditory system gets tired. In my experience, when that happens, if I rest a little bit, I can detect differences again. In my experience, it also depends on how tired I (or my auditory system) is on that moment/day. Being stressed doesn't usually help, either.
Anyway, I think there's people here who has more experience in ABXing and could give their experience on these matters.
I think what Pio2001 is describing is essentially sensitisation and habituation.
I have noticed with my own tests that even 5-6 repeats in a row can be way too much on hard to spot differences. I'm by no means a trained ear in many of the things I've tested, so maybe the number can be higher for more experienced listeners?
Habituation (or non-selective adaptation) starts to kick in and I lose the ability to discern the differencs.
Just as KikeG also explained, baselining or de-adapting by listening to nothing for a little while usually helps.
If I do a series of 6 three times, I can get better scores than doing one series of 18 in a row.
Then of course there are the really hard to spot differences that usually are hard, because the difference is within a new category of sensory attention that I have not consciously trained before or because the difference is so small that my treshold for distinguishing is barely met.
Usually then sensitisation helps. I can become increasingly aware of a small difference by repeating only the reference or the comparison stimulus (not going back and forth between both, but only listening to one of them in a row). This may require long amount of time, but when it happens, the ability to discern between the reference and the comparative stimulus is hightened.
Until adaptation starts to kick in again.
I also have a feeling, based not on any plotted data and could be totally off the base, that the difference between initial difficulty and difficulty after several repeats does not develop linearly.
What I mean is that, a stimulus pair that I can barely distinguish will become indistinguishable with only a few repeats, but a big initial difference does not seem to become indistinguishable at all, when a certain amount of sensitisation/learning has occured.
Hence, for very small impairments I would personally try a very small number of repeats (say four at a time), then de-adaptation through a break in the test (no stimulus at all) and another repeat of four again, continuing this until a desired number of statistical certanity is met.
Of course, this is only practical if one is interested only in are there audible differences rather than what usually noticeable differences there are that do not take extreme amount of attention and abnormal situations to discern.
Why it happens like this, I don't know yet. Sensory attention (becoming tired mentally) probably attributes to this, but is not perhaps the whole explanation. What are the the neural processes behind this? Also, is the level of late auditory evoked potentials linked to this?
Also, it's interesting to note that smell/taste seems to behave rather differently in some points. Especially in terms of adaptation rates and to some degree in long term sensory memory. There might be an evolutionary explanation for this already out there, for all I know (still haven't found one yet though).
Why do I think this is interesting? I think it's interesting because one can put forth a hypothesis that if we rob the test subject of reference stimulus completely, we can slowly but surely diminish the quality of the stimuli without the test subject noticing anything that has changed. Surely there must be a limit after which the subject would notice the diminishing of the quality even without a reference stimulus, but the difference between the diminished and reference could at that point be so big that this kind of difference would be spotted always and with 100% certainty in situation where the reference was constantly available and used.
It makes you think about the test methodologies and what it is exactly that one wants to test. Audible differences under what kind of repeat stimulus conditions? In reference to what (and when last heard) reference?
maybe the number can be higher for more experienced listeners?
I think I can experience the mental thing.
After having listened for some repetitions, I often find myself thinking about something else, and I must concentrate on the sample again.
It's like mentally adding numbers, or playing chess, I must draw all my mental attention to the problem. This is not natural, because we usually listen to music for entertainment. But when you study music and learn to play an instrument, you also learn to listen to every note played in a piece, and to follow them on the score. It requires a mental effort.
ABXing is the same. The difference that must be heard is at the edge of the perception threshold, thus a mental effort is needed to hear it perfectly, and it's very difficult to keep this effort constantly over a long period of time. There is always a moment when we get distracted by something... or by nothing.
So making mental efforts should help increasing the number of repetitions until which the difference vanishes.
That's what I experienced during the few ABX tests I succeeded.
After a second thought, I wouldn't trust the first comments I made, because they result of all the ABX tests I tried... and I failed about 3 out of 4. Thus they would rather apply to the "number of repetitions until which I become aware that there is actually no difference at all !"