The consensus of current research is that, within certain limits, the response to a stimulus is proportionate to the intensity of that stimulus. The fainter the stimulus, the lesser is the response tendency.
In response to the scandal created when a few gullible advertisers were taken in by the claims of those who stood to make a buck by hawking subliminal perception techniques, some jurisdictions passed laws against their use. But paranoia is not the same thing as proof. (After all, the Puritans also had laws against witchcraft.)
In the American Psychological Association's searchable data base on publications in psychology, there was an article published in January of this year in the Journal of Experimental Social Psychology, vol. 48 number 1, pp. 258-360 by Legal, Chappe, Coiffard, and Villard-Forest, entitled, "Don't you know that you want to trust me? Subliminal goal priming and persuasion." Before being presented with a persuasive message about the consumption of tap water, the experimenters subliminally primed one group of subjects, with the goal "to trust," and did not prime others. Then the subjects were given a questionnaire about their perception of the persuasive message, the source of the message, and their intentions to consume tap water. The results indicated that the primed subjects had a better evaluation of the message, and expressed a greater intention to consume tap water.
Of course, one swallow does not a summer make. And time will tell whether or not this study is replicable. Jason Nier, in an article entitled, "What Every Skeptic should Know about Subliminal Persuasion," in The Skeptical Inquirer, vol. 16, 1992, forcibly argued that research into the area of subliminal persuasion was "either fraudulent or flawed."
In establishing statistically significant results of an experiment, one calculates the likelihood that the results which you obtain could have been obtained by chance alone; and if the odds against chance are high enough, you accept your hypothesis. But in order to be accepted as an established scientific finding, the name of the game is replicability, or the extent to which a given experimental finding can be repeatable at will under the same controlled conditions. In the fifty years since the claim about subliminal perception was first made, several studies on this topic have been conducted. Despite the occasional positive finding amidst the considerable effort which has been devoted to its pursuit, the goal of scientific replicability has to date not been achieved. Thresholds (or Limens) of perception do vary, as you can easily verify yourself when you hear your name spoken at the next table in a crowded restaurant, but this is not the same thing as subliminal perception. We are definitely attuned to pick up meaningful stimuli more easily than stimuli which are not meaningful. This has been studied ever since the days of Wundt in the late 1800s, and it forms a big chunk of the experimental literature on perception. But subliminal perception? its mostly a hoax, folks!