Dear Al and List,
Repeating a recording of any clearly pronounced word produces illusory words or syllables that has been called the "verbal transformation effect." Googling this term will produce a list of articles dealing with the effect, including the link to "Warren Perception Lab - Auditory Demonstrations," which produces examples of both repeating diotic and dichotic stimuli.
Incidentally, making the stimulus indistinct by the addition of noise reduces the rate at which illusory changes occur.
All the best, Dick
Hi Massimo & List,
I believe that in Skinner's book, "Verbal Behavior" (1957), he reported an experiment he had done getting people to listen to a machine that played a tape of a babble of voices. He called the machine a "verbal summator". I don't remember whether it was a loop or not. He reported that people heard all kinds of words, with different people hearing different words, and thought it tapped people's verbal habits (including thoughts). I recall that he thought it could be used as a projective test. You'd have to check the original to be sure. Also Google finds a number of entries for "Verbal summator", so there may have been follow-up work.
I myself frequently hear words, usually short incomplete phrases, as part of irregular environmental sounds, usually involving splashing water in a resonant space, as when I am washing clothing or dishes in a sink. I can't make it happen voluntarily; it always happens as a double-take (e.g., Did I just hear "count the others"?). The stress patterns of the words in the phrase always correspond with loudness variations in the sound.
Dick Warren and colleagues experimented with recycling white noise in the 1980s (e.g., Warren & Bashford, 1981, Perception of acoustic iterance: Pitch and infrapitch. Perception and Psychophysics, 29, 323-335.) but I don't think words were ever embedded.