A little over ten years ago, I discovered a program that shipped with my screensaver package called xsublim. This program (which no longer ships with xscreensaver, and in fact no longer builds) takes a text file, splits it into individual words, and then displays each word on random place on your screen for a tiny fraction of a second. Experimenting with it led me on an interesting path through the backwoods of cognitive science.
XSublim was written as a lark, and by default it will display slogans from 1984 and They Live. However, in the half century since pop-culture’s ideas about subliminal messages crystallized, research has progressed: contrary to the media depictions of frightening hypnotic mind control, subliminal messages have only weak and difficult-to-predict effects, and complicated messages (longer than one or two words) are barely processed at all.
I’m mostly interested in the mind-expanding (rather than mind-controlling) potential of subliminal messages — an area they are ultimately better suited for. The point of greatest interest is the way that subliminal messages bypass conscious awareness and therefore do not require conscious attention: subliminal learning is a little bit like earning royalties or interest, in that the returns may be small but the effort-to-return ratio is nearly zero. Replacing a week of active study with a year of passive study is a no-brainer.
Mechanism of action
When we talk about subliminal messages, what we’re really talking about is what cognitive scientists call ‘priming’: a weak nudge that inclines us toward particular associations. Priming creates a sense of familiarity and attractiveness — flash faces on a screen, and those faces will be rated as more attractive during a survey. Priming also encourages similar ideas: people who see the word “water” flashed on a screen will be slightly more likely to interpret “bank” as “river bank” than as “financial institution” and those who see “doctor” will respond slightly more quickly to medical-related prompts — a concept known as “spreading activation.” It does this just by seeding a pattern in working memory — ‘priming the pump’ and making related thoughts flow more easily.
This makes it not terribly useful for advertising. A subliminal ad can make people who are currently thirsty slightly prefer Coca Cola, but it can’t get them to buy a case when they already have Sprite in the fridge. Similarly, the impulse is so weak that it can’t be used to get people to act out of character (you can’t be subliminally instructed to kill the president, although if you are playing a first person shooter video game it might be able to get you to shift your target one in a thousand times). If we want to squeeze utility out of subliminal messages, we need to be a little more creative.
Subliminal exposure can produce a sense of familiarity, which can be used to aid learning. Even though seeing the contents of a math textbook subliminally a million times will not let you skip the math class, it will make the terms and patterns seem more familiar, less imposing, and easier to remember — so later study will be easier. Seeing the contents of the textbook after class can help keep the ideas in memory. I typically begin to display the ebook versions of books I intend to read months in advance, continuously — making the actual task of reading them much faster, when I finally get around to it. I have had dense philosophical texts in circulation for years.
While sentence-scale patterns aren’t easily absorbed, the emotions associated with words certainly are — so you can slightly boost your mood by priming yourself with happy words or positive affirmations, or slightly boost your sex drive by displaying erotic fiction in this way.
The most interesting use, for me, is simply scrambling normal responses. Our environment is full of cues (shared by our peers) that subtly nudge our behavior using the same mechanisms as subliminal exposure; we are biased toward repeating the same behaviors by staying in the same environment. But, we can take control of our cognitive environment and add a random element, canceling out some of the weak impulses by adding new ones that pull in different directions. This can be very useful in of itself, without planning: how much of what we call genius is just seeing a familiar view through a slightly warped perspective? There is a greater possible range of perspective-warping through subliminal nudges than through our go-to methods — drugs, sleep deprivation, psychoanalysis, improv comedy games — even if the intensity is much lower. At the very least, a little bit of randomness can keep us from indulging too much in our worst biases.
There’s one other application I should mention — a social one. (I haven’t tested this directly since around 2006, but the theoretical justifications are solid.) Watching the same subliminal messages as someone at the same time is a great bonding experience: sharing the same slight behavioral biases encourages similar behaviors on a small scale (body language, word choice), which is interpreted as mirroring. So, the shared biases from shared exposure leads to a feeling of emotional connection and being ‘on the same wavelength’ — something that can kickstart a feedback loop ending in an experience of ‘pseudo-telepathy’ where the two of you are thinking the same thing at the same time in response to many various things.
Notes on Priming Research
The recent ‘replication crisis’ in psychology has hit the subfield of ‘social priming’ hard, with some popular papers failing to have their results replicated, and with some even getting called fraudulent. So, the whole subfield is being looked at with a skeptical eye, and some of that dubiousness is spilling over into priming proper by nonscientists. Social priming papers have made suspiciously strong claims, and depend on a whole variety of largely-untested assumptions. The replication failures of social priming do not extend to priming itself, which has been robust under testing & forms the foundation of other research (like the stroop test and the implicit bias test).
That said, my own experiments are the work of an enthusiastic amateur and are not properly controlled, although they rely on what I consider to be obvious small extrapolations of reliable known effects. I would love to see serious research in this domain.
Some applications I’ve mentioned have been directly tested in a controlled environment. Specifically, the effects in the ‘Mechanism of Action’ section above are all from serious published research.
Social priming research tends to claim specific primed impulses have certain predictable responses — such as (in one famous experiment) seeing words related to old people resulting in a slower walking speed. We could not reasonably expect this to be true, for the same reason that we can’t expect subliminal advertising to work: priming is weak, and our environment is full of priming, so complex actions with many steps (like walking through a hallway) would be affected by many nudges and would behave unpredictably.
Notes on Implementation (or: Try this at Home!)
As I mentioned earlier, xsublim is obsolete and doesn’t work on newer computers. I have written two programs that duplicate the important parts of xsublim and add useful features. They are available here. They run only on Linux, like the original xsublim. (If you find or create an implementation for another platform, I’d love to hear about it.)
To make things easier, I have included a script called sublimWrap, which randomizes the font and color.
These programs will read a stream of text and display it, word by word, until they run out of text to read. So, normally, I write some code to keep them displaying text indefinitely. Here are some examples:
# display dictionary words in a random order — good for scrambling responses
(while : ; do shuf < /usr/share/dict/words ; done) | sublimWrap
# display random wikipedia articles
(while : ; do w3m -dump http://en.m.wikipedia.org/wiki/Special:Random ; done) | sublimWrap
# I’m good enough, I’m smart enough, and doggone it, people like me!
(while : ; do fortune daily_affirmations ; done) | sublimWrap
(Originally published on Mondo 2000)