Courtesy of the Drudge Report, a story purporting to demonstrate media bias by showing video of George Bush in front of just part of a sign containing the word "families". Seen in a still photo, it looks like it's been framed to say "I lie". The story is fairly ridiculous, for the same reasons that taking a few frames out of a commercial with the word "bureaucrats" in the background didn't really demonstrate that Al Gore was being accused of consorting with rats.
One good debunking of subliminal messages comes from the urban legend gurus at Snopes.com. While it deals with the falsity of the experiments that purported to demonstrate subliminal effects, it doesn't actually explain why these messages are unlikely to work. SC's certainly got some things to say on the subject.
While the Snopes article claims that the original experiments displayed the messages for 1/3000th of a second, this is almost certainly not correct. Film is projected at 24 frames per second. It wasn't always like that; early films were run at 16 frames per second, and as much as 48 frames per second could be simulated with appropriate manipulation of the shutter. At no time in the history of film could an image have been displayed for as little as 1/3000th of a second -- and it's still not possible with modern video, which ups the frame rate to 29.97 frames per second in the U.S. (25 fps in Europe, which uses a different standard). A more plausible technical claim is advanced in the 1985 G.I. Joe episode, "Flint's Vacation", in which the second-baddest dude in pop culture history is inserted into every 20th frame, which is at least technically credible. As is clear from watching the episode, but not so clear from the script, Cobra Commander's voice is similarly spliced into the audio of the "news", and it's only by watching extended clips that the residents of Pleasant Cove receive their instructions.
Of course, the real problem with these claims isn't really that their proponents are unsophisticated about video projection. For the most part, word recognition just doesn't happen on the scales being discussed. One good summary of literature (through '94) on spoken word recognition is Gareth Gaskell's thesis, which includes measurements of 500-700 milliseconds for recognizing single words (although it's important to note that this is total reaction time; the actual time to hear and recognize the word is shorter, but it's still not on the order of the 30 milliseconds that a single video frame gets). Other papers mention throwing out measurements of less than 200 milliseconds for response times. SC could go dig up a whole bibliography of measurements from the course in psycholinguistics he took as a graduate student, but the point is merely to demonstrate that word recognition takes place on a much longer time scale than video frame refreshes. Assuming an audio track synchronized to the video, it is more or less impossible for a signal to be reconstructed in the brain with snippets that are far too short being played back about 0.8 seconds apart.
Of course, there's no claim of subliminal audio in the political stories here. But the numbers work out similarly for visual word recognition, and the same principle applies -- seeing something for one or two frames isn't going to be enough to recognize a word. Even in the cases where these things are on screen for a full second -- and "rats" wasn't, although SC can't be sure how long "ilie" was -- the stimulus is competing with a far more perceptually salient signal, which is to say the actual news presentation or the surface content of the commercial. It simply defies everything we know about word recognition to take seriously the idea that subliminal messages are meaningfully perverting an otherwise pristine political process ([that's OK, we've got plenty of other ways to pervert it -- ed.]).
(Edited on 10/6/04 at 1:59 a.m. to include note about response vs. recognition; 2:25 a.m. to correct math error.)
Comments