Can anybody expand on that? Does it mean "sound" rather than "music"? Can anybody point me to some sources?

Where's the quote from?

Well, I have to disagree with the quote. I can hear my favourite songs in whole and with all the details in my head, if I concentrate. And it can pass some time from when I have listened to them. So...

I agree with psycho. Certain pieces are such favorites of mine that I have listened to one version many many times. When I hear a version by a different orchestra the subtle differences are so jarring that in my head I keep hearing the favorite version over top of the new version. It makes it pretty hard to listen to.

The quote is from this thread.

I agree with Mr. Krueger. Yes, I guess you could say it's remembering the sound (i.e. parts of the actual waveform or its spectrum) rather than the content. Which is exactly what you want to do when doing a blind listening tests to assess, for example, lossy coding artefacts. For me, the time frame of remembering - or at least being able to compare - two waveforms is about 3 seconds. Which is why I usually create 3-second loops in blind tests.

Chris

P.S.: Yes, for two versions played by different orchestras, the differences in the spectra and time envelopes is actually huge, so there's a high chance that you can identify such differences even after months.

Yes, it implies minute details about sound quality -- not music. I often hear various numbers like this quoted but would be interested in knowing the scientific studies on which these numbers are based.

Years ago, someone asked me the point at which the time gap between switching among speakers reduces the reliability of listeners' ratings due to limitations in auditory memory I didn't know the exact answer so I wrote a software program that measured listeners' absolute detection threshold of added resonances with a parameter that allows you to vary the time gap between presentation of A/B where A or B can be the same (resonance or no resonance condition) or different (either A or B contains the resonance). I could vary the the time gap between a fraction of a second up to several seconds. I only used myself as a subject - but I recall that the detection threshold of the resonance tended to be higher when the time gap was either very short (< 1 sec) or very long (> 4 sec) and relatively constant between 1-4 seconds. When it's too short the stimulus A tends to have masking or residue effect on stimulus B, and vice versa. When the time gap is too long auditory memory comes into play. I never completed the experiment -- but I wish someone would.

Cheers
Sean Olive
Director of Acoustic Memory
R&D Group
Harman International

Thanks for your replies.

I assumed the statement didn't relate to our ability to remember things like our favourite songs, but I wanted to be more clear on what it does or doesn't relate to, under what conditions and so on.

I have no idea what it means. Primary partial loudness memory has about a 200 millisecond persistance at best. Featurememory is a few seconds. Auditory object memory can be persistant.

Songs we don't even like. I have heard this called a "song wedgie". Can't get it out of your head even if you want to.

I once remarked to a friend in college (30 years ago) that there was always something playing in my head. He tested me on this once by asking me at random what I was playing in my head. It was a commercial jingle.

I think I can probably play the entire album Yessongs in my head in complete detail.

I used to be able to do that with a lot of albums I loved. Now I keep myself so submerged in new music that it would probably take a bit of refamiliarization to do that.

Source?

Several books I read, the most recent of which was "This is Your Brain on Music" by Daniel Levitin. Seems well-documented, there are about 30 pages of Bibliographic notes.

When you say "I can hear my favourite songs in whole", does that mean that you could ABX good MP3 coders with high reliablity based on just this kind of memory?

Both the original ABX TTL comparator, and the PCABX software comparator had a feature for adjusting the mute time between samples.

Many of us found that you could take mute time out to 1 second or a little more without hurting listener performance for small differences appreciably. Someplace 5 or 10 seconds out, listener sensitivity dropped into a hole when the comparison related to small differences that were audible under ideal conditions.

IME, if the differences are pretty gross like a 1 dB level shift, then the time delay would matter far less or not at all.