Subject: Re: [AUDITORY] Maximum temporal window for pattern recognition and the perception of redundancy From: "Roy D. Patterson" <rdp1@xxxxxxxx> Date: Fri, 3 Feb 2017 09:23:22 +0000 List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>Dear Dario, Bob and Auditory List, The person who has done the most work on the perception of repeated noise and what it says about temporal integration is Christian Kaernbach. Just put 'Christian Kaernbach and repeated noise' into a good scholarly search engine and you will get a list of his publications on the topic, and they include references to the other people who have worked on the phenomenon. A good place to start is a paper that appeared in JASA Vol. 94.1 (details below). Bobs informal listening provides a good summary of what you will discover with a little listening. The auditory system finds sequences of random numbers which it hears as auditory features and having heard one it listens for it to occur again and emphasizes that the feature in the perception we hear. It takes about 4 cycles of a random noise to get the process going and it you listen long enough you can do it with noises 2 seconds and more in duration. Regards, Roy Patterson Temporal and spectral basis of the features perceived in repeated noise Christian Kaernbach DOI: http://dx.doi.org/10.1121/1.406946 On 02.02.2017 16:52, Bob Masta wrote: > Dario: > > Regarding my initial response (which doesn't seem to > have appeared yet on the listserv), it turned out to be > trivial to create a variable-length noise sequence listening > test, but much harder to actually *take* the test. The > upshot is that this is probably not what you want to do. > > I used the Daqarta Generator to create continuous white > noise, and used the DDisk option to record 60 seconds of > that. Then I switched the Generator 'Wave' type from > simply generating white noise, to using the file as the > source in Play mode. Play mode lets you specify start > and end points and automatically loops, so you can get > any sequence length you want, and you can change the > specific sequence by picking a different section of the file. > > One second is trivially easy to hear patterns in, but as you > go up in time it depends a lot on the particular section of > noise: If there are "features" that stand out, you can often > hear the sequence of features as something like > 'dadaDAdaDA'. But since it depends so much on the > chunk of noise you choose, it isn't a good way to tell > anything about human limits on pattern recognition... > there might not be a pattern in the chunk you choose. > > So I'm guessing that you are probably more interested in > something like a random series of musical notes. > Daqarta can easily do this using its MIDI options. (Let me > know if you are interested. It's also free.) But here the > testing would be pretty tedious, since humans seem to > have a pretty good ability to distinguish musical > sequences, again depending on the sequence. The more > 'random' it sounds (less 'musical') the harder it is. After > all, music perception is all about pattern recognition! > > Best regards, > > Bob Masta > > ====================== >> Thank you and best wishes. >> Dario