What I would probably do in that case is create multiple sound files with variable delays before the tone occurs. Randomly select one of those sound files on each trial *before* the images appear and specify a 0ms time value so there is no delay between the onset of the sound file and the onset of the images. Then use "RT:any" (or whatever) for each image and apply a min-max value of 0-1000! for the trial's style. That way, each image will only be on the screen for 1000ms, but will record the RT to the sound file if it occurs during its interval. I just made that up--does it make sense?