Imputing meaning to nothing

Lately I’ve been going through the oeuvre of Brian Dunning, specifically his Skeptoid podcasts. As I’ve previously noted, I think he probably does the highest quality skeptical stuff on the web, and a lot of his podcasts are on things new to me.

The concept of his ‘When People Talk Backwards‘ (episode 105) isn’t new to me, but some of his examples are startlingly good demonstrations of a theme I bang on about here quite often: our physical senses and their neurological support mechanisms are not objective recorders designed to capture the perfect truth, but practical devices which produce a workably accurate sense of what’s happening in a timely manner. Even if perfect accuracy were indeed possible, the processing time would be such that you would be eaten by a predator before you had a chance to realise it was there.

But back to Skeptoid. You can download the whole episode or read the transcript here. The item being discussed is claims by some that if you play recordings of human speech in reverse, you will hear what the speakers really meant. Silly, of course. In the course of demolishing it, Dunning notes that the human hearing mechanism searches out language amongst all kinds of noise and offers an example on the form of a clip of some manipulated sine waves. As he notes, these sound like someone speaking, but you cannot understand what they are saying. I certainly couldn’t, anyway.

He plays this clip three times. Then he plays a clip of a female voice saying a phrase at the same tempo as the noise clip. Then he plays the original clip again, twice. Both times it is perfectly clear what the clip is saying. As Dunning says, ‘it’s almost impossible not to hear the words that you’ve been preconditioned to hear’.

Hey, it’s as good as a circus trick you’d pay real money to see, so go download it and have a listen. This section starts at about 3:40 into the podcast, but it’s definitely worth listening to the whole thing (and the other ones as well).

As soon as I heard that demonstration, I decided to do a blog post. But certain deadlines got in the way so it turns out that I heard it two days ago and am only writing about it now. So, curious, I played the clip again through its three repeats. The first time through it was voice-like noise, the second time the noise began to resolve into words in my mind, and the third time it was clear as day what the phrase was. Every single word.

Now the phrase isn’t particularly interesting or memorable. It never occurred to me to try to see if I could remember it, but I am pretty confident that no matter how hard I wracked my brain I would have been unable to do so, even though it is only 13 words long. Indeed, in order to count the words just now I had to play back the noise clip to remind me of the phrase.

But by the third playing of some manipulated sine waves I had it perfectly, two days after having first heard it. Was it in my brain as a memory, and the noise clip helped retrieve it? Or had I learned the particular ‘accent’ of the ‘voice’ in the noise, allowing me to reinterpret it correctly? Or a bit of both. Or (of course) something I haven’t thought of.

So there we have it, perhaps the closest audio version of an optical illusion I’ve ever come across.

This entry was posted in Imperfect perception. Bookmark the permalink.

4 Responses to Imputing meaning to nothing

  1. Mark says:

    Did you see the doco on SBS last night, Is Seeing Believing? In particular the McGurk effect here: http://www.youtube.com/watch?v=ypd5txtGdGw

    (can see the entire episode here: http://player.sbs.com.au/programs#/programs_08/fullepisodes/fullep_documentary/playlist/Is-Seeing-Believing-Full-Ep/ )

  2. Stephen Dawson says:

    Hadn’t seen the show, but that McGurk thing is excellent. It’s funny just to take your eyes away from the ‘f’ video segments and the sound immediately returns to ‘b’.

  3. Rod Williams says:

    SKEPTOID / Brian Dunning is excellent. Have just about listened to all of the archives.

    Dr. Karl on JJJ is also recommended. His Thursday call-in show is posted afterwards as a podcast sans music. Ditto STAR STUFF, if you like spacey things.

    -Rod-

  4. Brad (msgohan) says:

    I was a bit freaked out when I could make out 80% of the first clip and the end of the second clip the first time he played them, but as noted in the comments on that page, they aren’t just any old sine waves. It seems to be a form of synthesized speech.

    Actually, the abstract from the ‘Science’ article that he links to notes that the study itself found that people understood the sounds: “A three-tone sinusoidal **replica of a naturally produced utterance** was **identified by listeners, despite the readily apparent unnatural speech quality** of the signal. The time-varying properties of these highly artificial acoustic signals are apparently sufficient to support perception of the linguistic message in the absence of traditional acoustic cues for phonetic segments.”

Leave a Reply

Your email address will not be published. Required fields are marked *