Trouble with multiple RTs when using sound files (case 436)

[dorimedini]

i'm trying to time responses to certain notes in a song.

to make it simple for testing, i took a midi file at exactly 60 bpm (homemade),
and made an experiment that plays the song in the background while
the participant press spacebar every second (onbeat).
after running this a few times, if i add up the times in the 'Time' columns
(the time between every press of the spacebar)
after~40 notes the response time deviates ~500ms.

this makes it very hard for me to automatically determine whether or not
the participant responded quickly enough to a specific note - the participant
has a window of 1000ms to respond, and if the delay around not #40 is 500ms
the participants' window closes to about 500ms.

when i try this expermient out on myself, i may be off a couple of millisecs on each beat, but not half a second...

attached is the midifile, the input trial file and the output log.




is there any way to calculate the response times in such a way that corresponds to the midi file? or any other workaround?




thanks for any help


[Note from moderator -- second post begins here]

first of all, this is similar to my previous post,
but with a wav file instead of a midi file (because i thought maybe the non-native media-player needed to play midi files caused bad timing), so sorry for the duplicity..

anyway, i wrote an experiment where the participant listens to a melody and presses space every tie he/she hears a wrong note.
i need to be able to test the response times, to see if the participant responded in under a second, so the timing is critical.

i found out that the closer to the end of the melody i get, directrt 'adds' more milliseconds to my response time - about 1 second every 2 minutes.

(if anyone is skeptic about my conclusions, read the parenthesized paragraphs:



(
after doing some test runs on myself, i found that even though i new exactly when the notes would appear and i reacted in under 0.5secs, if the note was near the end of the melody directRT said it took me over one second to react.

i didn't think it was my fault, so i tried a new test: a created a midi file with a single note every second exactly, converted it to a wav file and ran a test run - i reacted every note.

what i expected to see is a reaction time of (approximately) 1000ms on each stim, and a total of around 60,000 ms by the last stim - that is, if i sum up the response times of every stim (60 notes, around 1000ms response time per note) i should get around 60,000 ms.
but i get around 500ms of deviation - around 65,000 ms.)
)




is this a known issue? is there anything i can do to eliminate the inaccuracy?


attached is the output log, the midi file (not used in the experiment)
and the log_tweaked contains the result of summing up the RTs:
1. column M sums up the RT column (column L)
2. column N sums up all the "Time N" columns (should be more accurate, but still isn't enough)

i couldn't upload the wav file, no trick with zipping it or anything worked... if someone replies asking for it i'll give it another shot.

please help me out, i'm on a deadline... even a simple "yeah, it's a problem and can't be solved, here's why:" would be okay.

a solution would be better, though.


thanks for any time anyone can spare