Dan's Blackbox Make-A-Thon Project Documentation

I discovered quickly that just having 2 syllables does not make an audio illusion. I tried a few without the desired success, including "good girl" and "bad boy" without success. Words with emphasis on the first syllable worked better for me. And I found that female voices, especially one with an accent, makes better illusions. By pitch shifting "red shoe" a bit, it sounded much better.

I recorded each word or phrase several times, then chose the one set of two syllables that seemed to work best. There is only one set. It is repeated end to end for nearly 2 minutes. That block (the stereo left channel) was then converted to a single cut and aligned on a second track (the stereo right track) for an approximate one syllable delayed. I shifted the second track back and forth until I had what I thought to be the best image. Millisecond shifts made major differences.

A disadvantage I had as the producer of the tracks was that I knew what the syllables were. For the benefit of the audience, and in the fashion of Diana Deutsch, I applied a 10-second fade-in and a 10-second fade-out. This would prevent the listeners from hearing the leading word on the left track and the final word on the right track.

I asked the ACTLab participants to return their listings of some of the words they reported from the 3 samples:

Sample 1 - "hobby": popping, bobby, bee, poppy, be pa, pumping, ping, copy, hop, puppy, pumping, beat, humvee, mom, be, bean, pain, coffee.

Sample 2 - "oh, boy": bobbing, pop, ying, coffee, poppy, bubble, oboe, hubble, hobo, ugly, ohbala, poke, ovula, popo, copland, pogol, puebla.

Sample 3 - "red shoe": red shoes, red shirt, rancher, wrench, stretcher, texan, wretching, stretch, sweatshirt, black shoe, cheers, next scene, rat shit, gretchen, red shin.

Listen to the sequences again. Now what you hear?