Timothy E. Moore

SubliminalAdvertising:What You SeeIs What YouGet

This paper provides an evaluation of the evidenceand arguments advanced in support of the effec-tiveness of various subiiminal advertising tech-niques. Such practices are purported to infiuenceconsumer behavior by subconsciousiy aiteringpreferences or attitudes toward consumer prod-ucts. Whiie there is some marginai evidence thatsubliminai stimuii may infiuence affective reac-tions, the marketing relevance of this finding re-mains to be documented. The notion that subiim-inai directives can infiuence motives or actions iscontradicted by a large body of research evidenceand is incompatibie with theoreticai conceptionsof perception and motivation.

IN September 1957 sotne unwitting theatre audiencesin New Jersey were invited to *'drink Coca-Cola"and "eat popcorn" in briefly presented messages thatwere superimposed on the movie in progress. Expo-sure times were so short that viewers were unawareof any message. The marketing firm responsible re-ported a dramatic increase in coke and popcorn sales,although they provided no documentation of thesealleged effects. Public reaction was, nevertheless, im-mediate atid widespread:

" . . . t he most alarming and outrageous discoverysince Mr. Galling invented his gun." {Nation, 1957p. 206)" . . . take this invention and everything connectedwith it and attach it to the center of the next nuclearexplosive scheduled for testing." (Cousins 1957, p.20)

Opponents were indignant that unforgiveable psycho-logical manipulations would be visited upon innocentand unknowing consumers. Minds had been "brokenand entered" according to the New Yorker (1957, p.33). There was much talk of Brave New World and1984. But even while laws were being drafted pro-

Timothy E. Moore is Associate Professor of Psychology and Educationand Chairman, Department of Psychology, Glendon College, York Uni-versity, Toronto.

hibiting the use of subliminal advertising on televi-sion, Hollywood was incorporating the idea into twonew movies, and a Seattle radio station started broad-casting *subaudible' messages such as "TV's a bore."

In May 1978 police investigators in an unnamedmidwestem city attempted to apprehend a murdererby interspersing subliminal messages among framesof TV news film describing the murder {New YorkTimes 1978, p. c22). Later that year some departmentstores in Toronto began broadcasting subliminal au-ditory messages whose intent was to deter shoplifters.The "sinister implications" of such practices worriedthe Globe and Mail. Could unscrupulous prime min-isters deliver political propaganda sublimin^ly? {Globeand Mail 1978, p. 6). In British Columbia the fol-lowing year, a Ministry of Human Resources policymanual on child abuse was denied inclusion in a gov-emment-cotmnissioned publication because the man-ual's cover contained "sickening and obscene" sex-ual imagery imbedded in an apparently innocuousphotograph of an adult's hand clasping a child's.

Reports of various forms of subliminal manipu-lation are fairly common. Evidently the practice isstill with us, although a few twists have been addedsince 1957. Given its covert nature and the ethicalconsiderations involved, the prevalence of subliminaladvertising is very likely underestimated by reliance

3 8 / Journal of Marketina. SDrina 1982Journal of MarketingVnl AR ICnrin

upon published reports. At any rate, such techniquesare believed to he widespread by a great many peoplewho can hardly be faulted for vigorously protestingagainst their use. John Q. Public has his hands fulltrying to cope with forms of exploitation of which heis fully aware. Should he also be worried that Mad-ison Avenue is sneaking directives into his subcon-scious through the back door? Such a possibility haspervasive ramifications (Brown 1960). The potentialimportance of the topic has not escaped those in mar-keting (Hawkins 1970, Kelly 1979, Saegert 1979);however, all lament the dearth of empirical research.

There are at least three identifiable means of sub-liminal stimulation for which strong behavioral effectshave been claimed. The first of these involves verybriefly presented visual stimuli. Presentation is usu-ally by means of a tachistoscope, a device for care-fully controlling the exposure duration of a visualstimulus. Directives or instructions are flashed soquickly that the viewer is unaware of their presence.Such stimulation purportedly registers subconsciouslyand allegedly affects subsequent behavior. This methodof stimulus presentation has been used frequently byinvestigators interested in subliminal perception, al-though their purposes have usually been quite differ-ent. As a result, a body of research literature existsthat bears on the claims being made for some kindsof subliminal advertising. Some examples from thisliterature will be described and some studies analyzedin detail. It should be emphasized that stimulationbelow the level of conscious awareness can be shownto have measureable effects upon some aspects of be-havior. The point at issue is whether these effects aresufficient to warrant the conclusion that goal-directedbehavior can be manipulated by such stimulation.

Another means by which behavior control is at-tempted is through the use of accelerated speech inlow volume auditory messages. Here too, the claimis that while the message may be unintelligible andunnoticed at a conscious level, it is nevertheless pro-cessed subconsciously and imparts direction to thereceiver's behavior.

The third procedure consists of embedding or hid-ing sexual imagery (or sometimes words) in pictorialadvertisements. These are concealed in such a waythat they are not available to conscious perusal. Theyhave, however, a subconscious effect or so it is ar-gued.

The effects attributed to these procedures mayconsist of either (1) general, nonspecific, affectiveconsequences that are assumed to have some positivebut unspecified persuasive influence, or (2) a highlyspecific, direct impact upon some particular motiveOf behavior. In what follows, the evidence and ar-guments put forth in defense of the effectiveness ofthese procedures will be reviewed and critiqued.

Subliminal PerceptionMeasurable responses of one kind or another cansometimes be shown to be contingent upon stimula-tion that the perceiver is unaware of. Pierce and Jas-trow (1884) demonstrated that subjects could makereliable discriminations among stimuli differing inweight, even though they reported that the stimuliwere not discriminably different. In this classic study,subjects indicated the degree of confidence in theirjudgments concerning very slight differences in pres-sure applied to the subjects' fingers. In those instanceswhere no confidence at all in perceived variation ofpressure was reported, subjects were neverthelessobliged to say which of the two pressures was greater.Their judgments were correct 60% of the time.The Threshold ConceptToday, the notion that people can respond to stimuliwithout being able to report on their existence is ac-cepted and well documented (Bevan 1964a, 1964b;Dixon 1971; Erdelyi 1974). Taken literally, sublimi-nal means "below threshold." However, there existsno absolute cut-off point for stimulus intensity belowwhich stimulation is imperceptible and above whichit is always detected. When stimuli of varying inten-sities are presented over several trials, the minimumsignal strength that is always detected is much higherthan the one that is almost never detected. If someabsolute threshold existed, then there ought to be adeterminable stimulus intensity above which the re-ceiver always responds and below which there is noresponse. Instead, a particular stimulus is sometimesdetected and somtimes goes unnoticed. As a result,an individual's perceptual threshold is usually defmedas that stimulus value that is correctly detected 50%of the time. The threshold, or limen, is, therefore, astatistical abstraction.

For a given individual, this threshold may varyfrom day to day or from minute to minute. Moreover,thresholds differ rather widely between individuals.Many studies of subliminal perception are flawed be-cause the investigators assumed that some specificexposure duration or stimulus intensity automaticallyguaranteed that the stimulus would be sufficiently be-low threshold that its presence would be undetectedfor all the experimental subjects on all the trials.Often this assumption is unwarranted. Stimuli belowthe statistical limen (which itself fluctuates) may benoticed as much as 49% of the time. As a result, stud-ies that make little or no effort to determine a thresh-old for individual subjects are at risk because stimuliare presented that are effectively supraliminal forsome subjects on some trials. The results may thusbe due to the effects of weak (but not subliminal)stimulation.

Subliminal Advertising: What You See Is What You Get / 39

Obviously the notion of a perceptual limen is oflimited usefulness. For present purposes we may usethe term subliminal perception to refer to the follow-ing situations (Dixon 1971, p. 12):

(a) The subject responds to stimulation the energyor duration of which falls below that at which heever reported awareness of the stimulus in someprevious threshold determination.

(b) he responds to a stimulus of which he pleads totalunawareness.

(c) he reports that he is being stimulated but deniesany awareness of what the stimulus was.

In these instances the subject cannot recognize thestimulus. "These situations define subliminal percep-tion, and are to be distinguished from those where theindividual, though unaware of the stimulus responsecontingency, is either not necessarily unaware of thestimulus, or, alternatively, could be made aware ofthe stimulus if his attention were drawn to it" (Dixon1971, p. 13). People are often unaware of stimulationor of the processes mediating the effects of a stimuluson a response (Nisbett and Wilson 1977). This is aseparate issue from subliminal stimulation, wherein(he subject cannot identify the stimulus.

Some Illustrations of Subiiminal PerceptionThere is ample evidence that weak stimuli that are notreportable can be demonstrated to influence behavior,For example, a number of studies by Bevan and hisassociates (Bevan 1964b) have shown that subliminalstimuli can alter judgments of perceived intensity ofsupraliminal stimuli when the former are interpolatedinto the presentation series. In one of these studiessubjects were asked to judge the intensity of weakelectric shocks delivered to their wrists. Betweentrials subliminal levels of shock were also adminis-tered. Careful control procedures ensured that thesestimuli were not detected. The effect of these inter-polated stimuli was to elevate the judged intensitiesof the detectable shocks. A controi group that re-ceived no subliminal stimulation routinely estimatedtheir shocks to be less intense than the experimentalgroup. A similar effect was found for judgments ofthe perceived loudness of tones. Apparently the sub-liminal stimuli trigger physiological activity that af-fects the perception of similar supraliminal stimuli.

Signal detection research provides another exam-ple. In a signal detection task, weak stimuli are pre-sented; some are detectable, some are not. If subjectsare asked to provide confidence ratings of their judg-ments about the presence or absence of a signal, theirratings are highly correlated with the stimulus inten-sity. This is true even for signals that were reportedlynot detected (Green and Swets 1966, Swets 1961).

Perceptual defense literature provides yet another

sort of illustration. Many studies have shown that ta-boo or emotionally loaded words have higher recog-nition thresholds than do neutral words. That is, ittakes a longer exposure duration for whore to be iden-tified than for shore. At first this may appear illogical.How can something taboo be defended against unlessit is first recognized as being taboo? The paradox isresolved if it is assumed that "perception" is by nomeans a discrete experiential event that is automati-cally determined by some particular stimulus pattem.Rather, perception is treated as a multiprocess chainof events that begins with stimulus input and tenni-nates (subjectively) with conscious recognition of anobject or event. However, not all input is subjectedto the same sequence of mental processing. Stimuliare selectively filtered, transformed and attended toaccording to a variety of factors that are independentof the particular input. These include memory, ex-pectations, attention, affect and other variables. Per-ception then, as we conventionally use the term, rep-resents " . . -the conscious terminus of a sequenceof nonconscious prior processes" (Erdelyi 1974).Conscious recognition need not be and often is not theend point for many sorts of input. Some stimuli mayinitiate mental activity of one sort or another withoutbeing available to conscious reflection or report. Thisis what is typically meant by the term subliminal per-ception. In the case of taboo items, some kind of de-fensive selectivity operates to bias the processing ofemotionally charged inputsuch selectivity having itsimpact prior to a conscious recognition of the input.

Recently Zajonc (1980) has reviewed evidencefrom several studies showing that under some circum-stances, unattended stimuli can be processed to a de-gree that is sufficient to elicit a subsequent affectivereaction (i.e., like/dislike) without their being rec-ognized as having been previously encountered."Affective reactions can occur without extensive per-ceptual cognitive encoding. Reliable affective dis-criminations (like/dislike ratings) can be made in thetotal absence of recognition memory (old-new judg-ments)" (Zajonc 1980, p. 151). While unattendedstimuli are not necessarily subliminal, one study pur-ports to show that affect can be influenced by stimulithat are truly subliminal (Kunst-Wilson and Zajonc1980). That some behavioral processes may be influ-enced by stimuli whose presence is not consciouslynoticeable by the receiver is not at issue here. Thepreceding examples testify to the validity of sublim-inal perception as a phenomenon. The important ques-tion is whether the subliminal effects obtained justifythe claims made for subliminal advertising. This ques-tion is critical because what must be posited in orderto support such a proposition is not merely an effect,but specific, (relatively) powerful and enduring ef-fects on the buying preferences of the public.

40 / Journal of Marketing, Sprinq 1982

Subliminal AdvertisingCould subliminally presented stimuli have a market-ing application? Can advertising effectiveness be en-hanced through subliminal stimulation? Before re-viewing the few laboratory studies that have addressedthis question directly, it will be useful to considerwhat sorts of subliminal influences would be neces-sary in order to obtain some marketing relevance. Ata minimum, we might hypothesize that a subliminalstimulus produces (or increases) some positive affec-tive reaction to that stimulus. Whether or not such anaffective response, if obtained, could have any rele-vant motivating influence is another question. It isprobably safe to assume that positive affect would notdo any harm and could conceivably influence a prod-uct's attractiveness. A much stronger prediction forsubliminal effects would be one that hypothesizessome direct behavioral consequence (i.e., purchas-ing). Since the former prediction does not necessarilyentail any interesting marketing implications, and thelatter prediction clearly does, these hypotheses willbe referred to as weak and strong claims respectively.

Practical DifficultiesRegardless of which claim is under investigation,there are some profound if not insurmountable oper-ational constraints associated with presenting sublim-inal stimuli in a typical marketing context. One prob-lem has to do with individual differences in threshold.There is no particular stimulus intensity or durationthat can guarantee subliminality for all viewers. Inorder to preclude detection by those with relativelylow thresholds, the stimulus would have to be soweak that it would not reach viewers with higherthresholds at all. Lack of control over position anddistance from the screen would further complicatematters. Finally, without elaborate precautions, su-praliminal material (i.e., the film or commercial inprogress) would almost certainly wash out any poten-tial effects of a subliminal stimulus. In order to du-plicate the results of laboratory studies that haveshown subliminal effects, it is crucial to duplicate theconditions under which tbe effects were obtained.From a practical standpoint, this is virtually impos-sible. Nevertheless, it could be argued that if 1% of10 million viewers are influenced by a subliminal adthat completely misses the other 99%, the subsequentbehavior of that 1% might make the exercise cost ef-fective.

Does the relevant research indicate that some pos-itive affect could become associated with a particularproduct through the use of subliminally presentedstimuli? The evidence is not strong. The Kunst-Wil-son and Zajonc (1980) study referred to earlier usedirregular, randomly constructed octagons as stimuli.

The stimuli themselves were first presented at one-millisecond durations and filtered so that recognitionwas at chance level. Subjects were instructed to payclose attention to the screen, even if nothing was dis-tinguishable. The same stimuli were subsequentlypresented for one-second intervals, paired with newstimuli. Subjects' recognition of old versus new stim-uli was reported to be at chance; however, the oldstimuli were judged to be preferable to the new ones60% of the time. The effect was subtle but statisticallyreliable (p < .01, 2-tail).

It is tempting to speculate that repeated subliminalexposures could bring about an increasingly strongeraffective reaction, with the stimuli themselves re-maining unrecognized. A study by Shevrin and Frit-zier (1968) does not support such a notion. These au-thors demonstrated a differential effect of two differentsubliminal stimuli upon evoked potentials (EEG) andfree word associations in the absence of a consciousdiscrimination between the stimuli. The effect was afleeting one, however: *'the subliminal verbal effectsappeared only in the first .001-second condition, sug-gesting that, beyond a certain point, multiple expo-sures of stimuli work against subliminal influences"(Shevrin and Fritzler 1968, p. 298). Two points aboutthe Kunst-Wilson and Zajonc study are worth empha-sizing.

First, the stimuli themselves, consisting of (rela-tively) meaningless geometric shapes, were subjectedto subliminal exposure levels; this exposure seems tohave had a subsequent effect upon preference. Sec-ond, the experimental subjects were actively attendingto the stimuli throughout the subliminal viewing con-dition; during this time, no other stimulation was pres-ent that could distract attention or mask the subliminalstimuli.

Could this procedure be utilized in an advertisingcontext? It is possible that a display's attractivenesscould be subliminalJy enhanced by having that samedisplay exposed for subliminal durations prior to itssupraliminal presentation. Whether the magnitude ofthe resuhant effect could have any practical impor-tance is not known. Moreover, it is not obvious howthe subliminal exposure could be accomplished. Su-perimposing the subliminal display on top of supra-liminal material is not a good bet:

" . . . Ongoing supraliminai stimulation to whichattention may be directed almost certainly will swampany effect by a simultaneous stimulus below theawareness threshold . . . at a peripheral level, lat-eral inhibition and contour suppressing mechanismscould well block any neural transmission from theweaker of two stimulus arrays . . . a similar effectof restricted channel capacity would almost certainlyoperate centrally as well. The potential effects of onestimulus may be completely negated by the presenceof another" (Dixon 1971, p. 175-76).

Subliminal Advertising: What You See Is What You Get / 41

splicing or somehow integrating the subliminalstimulus into ongoing supraliminal material (even iftechnologically possible) is not too promising either,because unless a sufficient blank interval is includedbefore and after the insert, supraliminal material willmask the subliminal stimulus (Kahneman 1968). Ifsuch intervals are provided, the viewer will mostprobably be aware of an interruption, even though thestimulus itself may not be detectable. At least 100milliseconds of "clean" background on either side ofthe target stimulus would be necessary to preclude amasking effect. As a result, subjects could infer thepresence of a stimulus. If complete unobtnisivenessis a priority, the stimulus and surrounding intervalwould have to be carefully located at naturally oc-curring breaks or cut points. Even then, completelydisguising the fact of stimulation may not be possible.

Evidence Involving the Weak ClaimIn addition to Kunst-Wilson and Zajonc (1980), twoother studies report subliminal effects relevant to theweak claim. Byrne (1959) flashed the word "heef"for successive five millisecond intervals during a six-teen-minute movie. Experimental and control subjectsdid not differ in their verbal references to beef, asmeasured by word association tests. Nor did experi-mental subjects report a higher preference for beefsandwiches, when given a list of five alternatives.Experimental subjects did, however, rate themselvesas hungrier than control subjects. This difference heldup when ratings were co-varied with hours of fooddeprivation. Byrne offered no explanation for thisfmding. It is not obvious why the word " b e e fshould induce hunger particularly when it failed toinfluence semantic associates. Moreover, the methodof presentation involved superimposing the stimuluson the movie. For reasons outlined earlier, such aprocedure is likely to interfere with rather than en-hance any potential subliminal affects.

In a similar study, Hawkins (1970) flashed theword "coke" for 2.7 millisecond-intervals during thepresentation of other supraliminal material. Subjectivethirst ratings were higher for the "coke" group thanfor a control group that received a subliminal non-sense syllable. Hawkins concludes that "a simplesubliminal stimulus can serve to arouse a basic drivesuch as thirst." (p. 324). As Saegert (1979) haspointed out, "Hawkin's results may simply be a TypeI error, especially in view of the fact that other trieshave been made" (p. 55). The fact that Hawkins per-formed five independent 1-taii statistical tests vi'hereone analysis would have sufficed lends support toSaegert's position. There are methodological short-comings in both of these studies. Even if the resultsare taken at face value, their relevance to advertisingis minimal.

Evidence Involving the Strong Claim

The strong claim for subliminal advertising posits spe-cific behavioral consequences as a result of a sublim-inal directive. A study by Zuckerman (1960) requiringstudent nurses to write stories describing the contentsof a series of pictures that were projected onto ascreen in front of them is pertinent to this issue. Un-known to the subjects, the instructions "write more"and "don't write" were tachistoscopically superim-posed on the pictures at successive points during thepresentations. A control group was treated in a similarfashion but received blank slides in place of thosecontaining the subliminal directives.

The study was composed of three successive con-ditions: (1) baseline, during which no subliminal mes-sages were presented, (2) "write more," duringwhich subjects in the experimental group received a"write more" instruction for .02 seconds, concur-rently with the picture they were asked to describe,and (3) "don't write," during which the experimentalsubjects received a "don't write" directive, again su-perimposed for .02 seconds on the picture being pro-jected. During each condition, pictures were pre-sented for 10 trials each. After each trial, subjectswrote a description of what they had seen. Zuckermanfound that nurses in the experimental group wrotemore during condition 2 ("write more") than theyhad during baseline. Furthermore, he noted a slightdrop in output between condition 3 ("don't write")and condition 2, and interpreted this as evidence thatthe subliminal instructions were effective.

Unfortunately, there is a strong possibility thatthese results were due to a methodological artifactwhich psychologists call a "ceiling effect." This oc-curs when performance reaches an asymptote and can-not be further improved upon. The slight drop thatZuckerman observed may not have been a real de-crease in performance but rather, a levelling off. Thisinterpretation is supported by a comparison of the per-formances of experimental and control subjects. Forsome reason the students in the experimental groupwere enthusiastic writers. They wrote much more dur-ing baseline than did the controls. They wrote stillmore during condition 2 ("write more"), and the con-trols increased their output as well. By condition 3 theexperimental subjects may have reached asymptote.They were all "written out," and a slight drop wasobserved in the number of words written.

Because of time constraints (and possibly writer'scramp), experimental subjects may already have beenwriting as much as could reasonably be expected bythe end of condition 2. The slight drop during con-dition 3 may be due to statistical artifact. When vari-ability is possible in only one direction (in this casedown), a slight decrease in performance is predict-

42 / Journal of Marketing, Spring 1982

able. Controls were still increasing their output duringcondition 3 and by the end, their output had barelysurpassed tbat of the experimental subjects' perfor-mance during baseline. When differences betweengroups are large prior to any experimental manipu-lation, it is risky to attribute some subsequently ob-served differences to that manipulation. In this study,the preexisting difference between experimental andcontrol subjects was as great or greater than any othersubsequently observed difference between or withingroups. Zuckerman has little to say by way of ex-plaining the finding, but submits that "the subject'soperant behavior is supposedly brought under controlby suggestive cues of which he is not aware" (p.404). This is not an explanation but rather a descrip-tion of the outcome couched in operant terminology.

Dixon (1971), commenting on Zuckerman's re-sults, speculates that "it may be impossible to resistinstructions which are not consciously experienced"(p. 177). Again, this is more an assertion than an ex-planation, but it does reflect an apparently prevalent(although not articulated) notion that instructions, di-rectives and/or slogans are intrinsically compelling.When the instruction is delivered supraliminally thereceiver can counter-argue or derogate the source,thereby diminishing the stimulus' influence. How-ever, if the instruction is presented subliminally, therecipient is unaware of its presence and is conse-quently unable to counter-argue.

Several researchers have investigated and de-scribed some of the cognitive processes that may me-diate acceptance of advertising claims (Harris et al.1979, Wright 1973). Wright analyzed the responsesof 160 women who were exposed to a target adembedded in other surrounding material, and subse-quently queried about their reactions to the argumentscontained in the advertising message. Counter-argu-ing by the receiver was identified as an important pro-cessing strategy. Neither the rehability nor validity ofthis finding is being disputed. However, it would bea mistake to assume that "resistive cognitions" arean inevitable consequence of advertising. Such a po-sition is reminiscent of a behavioristic view of peopleas passive receivers of inputs to which they respondin automatic and stereotyped ways.

Perhaps the single most important lesson to belearned from cognitive psychology in the last decadeis that the meaning of a stimulus does not reside inthe stimulus itself. Meaning is constructed by the re-ceiver in active, complex and often specialized ways.With respect to advertising the selectivity of attentionand the active control over subsequent processing ofthe input means that stimulation is not a sufficientcondition for any response at all, let alone some par-ticular response. We are constantly subjected to a bar-rage of external and internal stimuli, of which only

a fraction acquire phenomenal representation. Someneural activity is no doubt provoked by stimuli thatare not consciously processed. But to attribute to asubliminal stimulus a strong influence, which it can-not be shown to have when supraliminal, is not jus-tified by any theoretical rationale. For this reason itis appropriate to insist on especially clear well-repli-cated empirical evidence before accepting such aproposition. To the author's knowledge, Zuckerman's(1960) finding has not been replicated, and the studyitself is vulnerable to an important methodologicalcriticism.

There is an additional problem with proceduresthat attempt subliminal persuasion through the use ofwritten directives. In order for a subliminal messageto exert a behavioral effect (the "strong" claim), thefull and precise meaning of the message would haveto be extracted from it. Dixon (1971) has reviewedmany subliminal perception studies showing thatwhen words are used as stimuli, "the stimulus tendsto elicit responses from the same sphere of meaning"(p. 102). Since competitors' products may well becontained in this sphere, it would be essential that thefiill meaning of the stimulus words be identified. Aneffusion of mere semantic associates would be insuf-ficient. There are no published studies that demon-strate that people educe tbe full meaning of a sublim-inal word stimulus, and there are at least two studiescasting some doubt on the possibility (Heilbrun 1980,Severance and Dyer 1973). For this reason, it is dif-ficult to construe a subliminal directive as an argu-ment that cannot be consciously resisted.

SummaryBefore turning to other methods that attempt sublim-inal persuasion, it will be useful to summarize theevidence reviewed. Research supporting the null hy-pothesis is much less likely to find its way into printthan that which demonstrates some potential influ-ence. The paucity of evidence may simply be a re-flection of its lack of availability. On the basis of whatlittle data are available, one could tentatively con-clude that subliminal presentation of a stimulus mayproduce a positive affective response to that stimulus(Kunst-Wilson and Zajonc 1980). This positive affec-tive response was obtained with subjects who wereattending only to the subliminal stimuli. Whether thisfinding could be utilized successfully in a marketingcontext remains to be seen. Apart from the questionof the magnitude of the effect, not to mention its va-lidity (Bimbaum 1981, Mellers 1981), there are somepractical difficulties associated with achieving a real-world application.

The evidence that subliminal directives can exertany control over behavior is much less compelling(Zuckerman 1960), although there has been ample

Subliminal Advertising: What You See Is What You Got / 43

opportunity for replication. Moreover, this strongclaim for subliminal influence is not accompanied bya coherent explanatory rationale. Previous reviews ofthe strong claim have reached similar conclusions.One of the first rigorous scrutinies of this issue wasdescribed by McConnell et al. (1958), no doubt pre-cipitated by the furor generated by the popcorn ad inNew Jersey. These authors were sceptical that any butthe simplest forms of behavior could be affected bystimulation below the level of conscious awareness.Bevan (1964a) concluded that the "influences of sub-liminal stimulation upon preference and choice, ifthey occur at all, are highly subtle, and the possibilitythat they could constitute an effective means of con-trolling consumer behavior or political opinion ishighly unlikely" (p. 91). Equally strong misgivingswere expressed by Goldiamond (1966) and Anastasi(1964). Empirical documentation has remained elu-sive: "all things considered . . . secret attempts tomanipulate people's minds have yielded results assubliminal as the stimuli used" (McConnell 1977, p.231).

Subaudible MessagesThe eye is capable of receiving far more informationin a short period of time than is the ear. Thus moststudies of subliminal perception have involved visualstimulation because the investigator can attempt todetermine what particular features of a display are re-sponsible for various sorts of neural activity that mayoccur below the level of conscious awareness. In con-trast, studies addressing auditory reception have beenconcemed primarily with signal detectiondeterminingthe presence versus absence of a weak signal. Becauseauditory information is, perforce, temporally ex-tended, it is particularly vulnerable to loss throughlack of attention or auditory masking.

This probably accounts for the total absence ofpublished studies investigating possible effects of sub-audible messages. While the eye is sensitive primarilyto spatial information, the ear is basically a processorof temporal information, especially in the case ofspeech perception. The difference is an importantone. A great deal of information can be presented si-multaneously in a visual display. An auditory stimu-lus is more extended in time; information arrives inconsecutive bits. A speech stimulus may be thoughtof as a sound pattern whose acoustic features fluctuateover time. Consequently, there is no procedure forcreating tachistoscopic-like auditory stimuli. Control-ling the exposure duration of a visual stimulus doesnot change the stimulus itself; it merely limits the timeavailable for processing it.

If speech is compressed or telescoped in time, thesignal itself is altered. While the speech stream canbe subjected to a surprising amount of mutilation

without intelligibility being affected (Licklider andMiller 1951), there is a limit to the amount of distor-tion that can be tolerated without a loss in compre-hension. Information is transmitted at the rate of about150 words per minute in normal speech. Studies haveshown (Foulke and Sticht 1969) that comprehensiondeclines fairly rapidly at rates beyond 300 words perminute. There are two reasons for this. The first in-volves signal degradation. When playback speed isincreased, component frequencies and pitch are bothaltered. The intelligibility of the signal consequentlysuffers. Secondly, channel capacity is taxed when acritical word rate is reached. Speech comprehensionrequires the continuous registration, encoding andstorage of information. These operations take time.When the word rate is too fast, not all the input canbe processed as it is received. The result is that somespeech information is lost. Reducing the volume ofaccelerated speech will only compound these diffi-culties. Mass media accounts of subaudible messagesreport presentation rates of greater than 2,300 wordsper minute (Toronto Star 1978, p. cl ; WashingtonPost 1979, p. c4; Time 1979, p. 63). The message issimply repeated 8- or 9,000 times an hour. Becauseof the fast rate, what may once have been a messageis rendered an unintelligible scratching sound. Thatsuch stimuli could have any influence on behavior(except to annoy) is a claim totally lacking empiricalsupport. Since the stimulus has no apparent meaning,presenting it at a supposedly subaudible level does notthereby confer any added significance.

The accelerated nature of subaudible messages isperhaps a tangential issue. Could such messages havean influence if the presentation rate were normal, butthe volume at a subthreshold level? Relevant evidencemitigates against such a notion. Weak auditory stimuliare very susceptible to auditory masking. Moreover,there is some experimental evidence that attentionalfocus can effectively prevent weak auditory stimulifrom receiving any processing at all (Broadbent 1958,Eriksen and Johnson 1964, Peterson and Kroener1964). Studies in dichotic listening reveal that verylittle of the content of an unattended message is pro-cessed when attention is focussed on another concur-rent message (Kahneman 1973, Moray 1969, Treis-man and Geffen 1967). Moreover the unattendedstimuli used in these investigations are by no meanssubliminal in strength.

Speech sounds are different in principle from otherauditory inputs (Liberman et al. 1967). Because ofspeech's temporal dimension, a certain minimal amountof attention may be essential for comprehension. Thiswould make subliminal presentation of auditory mes-sages not just difficult but impossible. In fact, it isdifficult to conceive of a means by which speechcould be rendered subliminal according to the con-

44 / Journal of Marketing, Spring 1982

ventional definition outlined earlier (see Dixon 1971,1981). Neither accelerating the message nor reducingits volume seems to provide appropriate analogs to themethods used in the visual modality. At any rate, theprocedures tried to date do not appear promising.

Whether or not subliminal effects could be ob-tained from auditory messages under more carefullycontrolled conditions remains to be seen. At the pres-ent time there is no evidence that sucb influence ispossible, let alone any practical application. It shouldalso be emphasized that a change in modality does notprovide a defense against some of the objectionsraised earlier regarding the subliminal effects of visualstimuli. The assumption that behavior can be auto-matically triggered by the presentation of some par-ticular stimulus is as unwarranted for auditory mes-sages as it is for visual ones.

Embedded StimuliA different kind of procedure for achieving subliminaleffects has been described by Key (1973, 1976,1980). In these books the author alleges that variouserotic images or words have been surreptitiously con-cealed in magazine, newspaper and television adver-tisements. High-speed photography and airbrushingare among the techniques wbereby subtle appeals tosubconscious sex drives are hidden. Their use is ubiq-uitous. Ritz crackers have the word sex baked intothem; a Gilbey's Gin ad is full of microscopic erotica.None of these are visible to the naked eye. In fact,it apparently requires weeks of analysis for many ofthem to be discovered, and sometimes they areembedded upside down.

According to Key, " . . . humans can be assumedto have at least two sensory input systems, one en-coding data at the conscious level and a second op-erating at a level below conscious awareness" (Key1973). A concealed word or symbol, " . . . usuallyinvisible to consciousness appears instantly perceiv-able at the unconscious level" (Key 1976). He goeson to claim that visual or auditory stimulation whosespeed and/or intensity are beyond the range for nor-mal sensory reception can nevertheless be transmitteddirectly into the unconscious, whence subsequent be-havior is manipulated. Precisely how these implantedcues affect a given product's desireability is not tooclear, but Key assures us that they are very effective.The Ritz crackers, in fact, are reported to taste betterbecause they have the word sex stamped onto them.Key provides no documentation for the effects that heattributes to embedded stimuli. For this reason, hisassertions should be regarded as hypotheses awaitingempirical investigation. Key also describes some psy-chological mechanisms through which embeddedstimuli purportedly operate. These latter claims in-

volving perception, memory and the subconscioushave probably rendered his speculations quite unpal-atable to research psychologists. Man's sensory ap-paratus has been studied extensively for many years.There is no evidence for more than one class of sen-sory input systems, as Key claims, nor is there evi-dence of unconscious perception of stimuli that falloutside the functional range of our receptor organs.Key appears to invent whatever features of perceptionand memory would be necessary to achieve the resultsimputed to embedded stimuli. The notion of a sepa-rate super-powerful sensory system serving the sub-conscious (exclusively) cannot be accommodated byany theory of perception, past or present. It is notsurprising that Key's books have not been favorablyreviewed by the scientific community (Schulman1981). They are mentioned here because while theycontain the least scientific substance, these books areprobably largely responsible for the promulgation ofa belief in the power of subliminal manipulation.

Whether or not erotic imagery has been deliber-ately planted is not relevant to a consideration of theimagery's alleged effects. A diligent search for a phal-lic symbol will probably be successful. How its pres-ence and reiationship to an advertised product mightbe interpreted is another matter, but the consequenceis by no means predictable. The amount of informa-tion available from a purposeful scrutiny of a displayis limited only by the viewer's imagination. Holdingadvertisers responsible for one's erotic musings is an-alagous to accusing Rorschach of insinuating partic-ular themes into the inkblots. A cursory glance yieldsfar less information than a careful inspection. Undertypical circumstances, the ad's most salient charac-teristics will receive the lion's share of perceptual ac-tivity (Hochberg 1978), if they receive any attentionat all. Completely ignoring a stimulus is an option thatpeople frequently exercise. If you do not activelysearch for hidden extras, what you see is what youget, and there is nothing subliminal about such pe-rusal. The fine print near the bottom of an ad is likelyto be far more important than any concealed genitaliacould be.

While Key appears to have misjudged the efficacyof embedded stimuli, it would be a mistake to dismissout of hand all of his remarks concerning the latenteffects of advertising. Ads may influence us in someways which have nothing to do with consumer be-havior per se. For example, ads help to transmit var-ious cultural stereotypes. If women are consistentlyportrayed in insignificant or demeaning roles, theviewer may develop an attitude towards them that isultimately prejudicial and harmful to women as agroup. Moreover, these attitudes are not consciouslyformed. The rich literature on observational learninginvestigates how such learning takes place (Comstock

Subliminal Advertising: What You See Is What You Get / 45

et al. 1978). While the acquisition of such attitudesmay occur subconsciously, there is nothing subliminalabout the presentation of the role models. On the con-trary, they are distressingly conspicuous. This kindof implicit learning can have important and pervasiveconsequences (Poe 1976; Rush 1980; Walstedt, Geisand Brown 1980).

ConclusionA cenmry of psychological research substantiates thegeneral principle that more intense stimuli have agreater influence on people's behavior than weakerones. While subliminal perception is a bona fide phe-nomenon, the effects obtained are subtle and obtain-ing them typically requires a carefully stmctured con-text. Subliminal stimuli are usually so weak that therecipient is not just unaware of the stimulus but is alsooblivious to the fact that he/she is being stimulated.As a result, the potential effects of subliminal stimuliare easily nullified by other ongoing stimulation in thesame sensory channel or by attention being focussedon another modality. These factors pose serious dif-ficulties for any possible marketing application.

A second major problem pertains to the psycho-logical mechanism through which a subliminal stim-ulus could in principle influence behavior. The prop-osition is appropriate only if one characterizes aperson as a static organism who processes stimulusinput passively and responds in automatic predictableways. In fact, psychological research has generateda large body of evidence that such a characterizationwould be false. There is substantial evidence for theimportance of centralized control and mediating pro-cesses and good reason to believe that humans havehighly mobile selective attention. The sheer volumeof constant sensory stimulation implicates a construc-tive, synthetic model of focal attention and perceptionrather than a purely receptive one. As Broadbent(1973) said, " . . . the brain is made of unreliablecomponents, so that it is very unlikely that any par-

ticular impulses in any particular nerve cells will oc-cur predictably and consistently whenever a particularstimulus strikes our senses. In addition, we are beingbombarded all the time by a very large quantity ofinformation; and in relation to this large quantity ofinformation we are all, like Winnie the Pooh, bearsof very little brain" (p. 31).

Empirical support for subliminal influences of apragmatic nature is neither plentiful nor compelling.On the basis of research evidence accumulated todate, the most one could hope for, in terms of mar-keting application, would be a potential positive af-fective response to a subliminal stimulus. Whethersuch an effect could actually be obtained in a realisticviewing situation, and whether the magnitude of theeffect would make the exercise worthwhile is still anempirical question. There is no empirical documen-tation for stronger subliminal effects, such as inducingparticular behaviors or changing motivation. More-over, such a notion is contradicted by a substantialamount of research and is incompatible with experi-mentally based conceptions of information process-ing, learning and motivation.

None of this is to deny the existence of motivesof which one may be unaware, nor to deny that sub-liminal stimulation can be used to investigate differ-ences between unconscious and conscious processes(Carr and Bacharach 1976, McCauley et al. 1980,Shevdn and Dickman 1980). The point is simply thatsubliminal directives have not been shown to have thepower ascribed to them by advocates of subliminaladvertising. In general, the literature on subliminalperception shows that the most clearly documentedeffects are obtained only in highly contrived and ar-tificial situations. These effects, when present, arebrief and of small magnitude. The result is perhapsbest construed as an epiphenomenona subtle andfleeting by-product of the complexities of human cog-nitive activity. These processes have no apparent rel-evance to the goals of advertising.

REFERENCESAnastasi, A. (1964), "Subliminal Perception," in Fields of

Applied Psychology. A. Anastasi, New York: MeGraw-Hill.

Bevan, W. (1964a), "Subliminal Stimulation: A PervasiveProblem for Psychology," Psychological Bulletin, 61 (no.2), 89-99.

(1964b), "Contemporary Problems in AdaptationLevel Theory," Psychological Bulletin, 61 (no. 3), 161-187.

Bimbaum, M. (1981), "Thinking and Feeling: A SkepticalReview," American Psychologist, 36 (no. 1), 99-101.

Broadbent, D. E. (1958), Perception and Communication.New York: Pergamon.

(1973), In Defence of Empirical Psychology, Lon-don: Camelot Press.

Brown, K. C. (1960), "Hemlock For the Critic: A Problemin Evaluation," Journal of Aesthetics and Art Criticism, 18(no. 3), 316-19.

Byrne, D. (1959), "The Effect of a Subliminal Food Stimuluson Verbal Responses," Journal of Applied Psychology, 43(no. 4), 249-251.

Carr, T. and V. Bacharaeh (1976), "Perceptual Tuning andConscious Attention," Cognition, 4 (no. 3), 281-302.

Comstoek, G., S. Chaffer, N. Katzman, M. McCombe andD. Roberts (1978), Television and Human Behavior. NewYork: Columbia University Press.

Cousins, N. (1957), "Smudging the Subconscious," SaturdayReview, 40 (October 5).

Dixon, N. F. (1971), Subliminal Perception: The Nature qf

4 6 / Journal of Marketing, Spring 1982

a Controversy, London: McGraw Hill.(1981), Preconscious Processing. London: Wiley.

Erdelyi, M. H. (1974), "A New Look at the New Look: Per-ceptual Defense and Vigilance," Psychological Review, 81(no. 1), 1-25.

Eriksen, C. W. and H. J. Johnson (1964), "Storage and DecayCharacteristics of Nonattended Auditory Stimuli," Journalof Experimental Psychology, 68 (no. 1), 28-36.

Foulke, E. and L G. Stieht (1969), "Review of Research onthe Intelligibility and Comprehension of AcceleratedSpeech," Psychological Bulletin. 72 (no. 1), 50-62.

Globe & Mail (1978) (October 13), 5.Goldiamond, L (1966), "Statement on Subliminal Advertis-

ing," in Control of Human Behavior, Volume I, R. Ulrich,T. Stachnik and J. Mabry, eds., Glenview, IL: Scott Fo-resman.

Green, D. M. and J. A. Swets (1966), Signal Detection The-ory and Psychophysics, New York: Wiley.

Harris, R. J., T. M. Dubitsky and S. Thompson (1979),"Learning to Identify Deceptive Truth in Advertising," inCurrent Issues and Research in Advertising, J. H. Leighand C. R. Martin, eds., Ann Arbor: U. of Michigan Grad-uate School of Business Administration, Division of Re-search.

Hawkins, D. (1970), "The Effects of Subliminal Stimulationon Drive Level and Brand Preference," Journal of Mar-keting Research, 8 (August), 322-26.

Heilbrun, K. S. (1980), "Silverman's Subliminal Psychody-namic Activation: A Failure to Replicate," Journal of Ab-normal Psychology, 89 (no. 4), 560-566.

Hochberg, J. (1978), Perception, 2nd ed., Englewood Cliffs,NJ: Prentice-Hall.

Kahneman, D. (1968), "Method, Findings and Theory inStudies of Visual Masking," Psychological Bulletin. 70(no. 6), 404-425.

(1973), Attention and Effort, Englewood Cliffs, NJ:Prentice-Hall.

Kelly, J. S. (1979), "Subliminal Embeds in Print Advertising:A Challenge to Advertising Ethics," Journal of Advertis-ing, 8 (no. 3), 20-24.

Key, W. B. (1973), Subliminal Seduction, Englewood Cliffs,NJ: Signet.

(1976), Media Sexploitation, Englewood Cliffs, NJ:Prentice-Hall.

(1980), The Clamplate Orgy, Englewood Cliffs,NJ: Prentice-Hall.

Kunst-Wilson, W. and R. Zajonc (1980), "Affective Discrim-ination of Stimuli That Cannot Be Recognized," Science.207 (no. 1), 557-558.

Liberman, A, M., F. S. Cooper, D. P. Shankweiler and M.Studdert-Kennedy (1967), "Perception of the Speech Code,"Psychological Review, 74 (no. 6), 431-46L

Licklider, J. and G. Miller (1951), "The Perception ofSpeech," in S. Stevens, ed.. Handbook of ExperimentalPsychology, New York: Wiley.

McCauley, C , C. Parmelee, R. Sperber and T. Carr (1980),"Early Extraction of Meaning From Pictures and Its Re-lation to Conscious Identification," Journal of Experimen-tal Psychology: Human Perception and Performance, 6(no. 2), 265-76.

McConnell, J. V. (1977), Understanding Human Behavior.

2nd ed.. New York: Holt, Rinehart & Winston., R. Cutter and E. McNeil (1958), "Subliminal

Stimulation: An Overview," American Psychologist. 13(no. 3), 229-42.

Meliers, B. (1981), "Feeling More Than Thinking," Ameri-can Psychologist, 36 (no. 7), 802-803.

Moray, N. (1969), Attention: Selective Processes in Visionand Hearing. London: Hutchinson Ltd.

Nation (1957), "Diddling the Subconscious: Subliminal Ad-vertising," 185 (October 5), 206.

New York Times (1978) (May 18), c22.New Yorker (1957), 33 (September 21), 33.Nisbett, R. E. and T. O. Wilson (1972), "Telling More Than

We Can Know: Verbal Reports on Mental Processes,"Psychological Review, 84 (no. 3), 231-59.

Peterson, L. R. and S. Kroener (1964), "Dichotic Stimulationand Retention," Journal of Experimental Psychology, 68(no. 2), 125-130.

Pierce, C. S. and J. Jastrow (1884), "On Small Differencesof Sensation," Memoirs of the National Academy of Sci-ences, 3, 73-84.

Poe, A. (1976), "Active Women in Ads," Journal of Com-munication, 26 (no. 4), 185-92.

Rush, F. (1980), "Child Pornography," in Take Back theNight: Women on Pornography. L. Lederer, ed.. NewYork: Morrow.

Saegert, J. (1979), "Another Look at Subliminal Perception,"Journal of Advertising Research, 19 (no. 1), 55-57.

Schulman, M. (1981), "The Great Conspiracy," Journal ofCommunications. 31 (no. 2), 209.

Severance, L. J. and F. N. Dyer (1973), "Failure of Sublim-inal Word Presentations to Generate Interference to Color-naming," Journal of Experimental Psychology, 101 (no.I), 186-89.

Shevrin, H. and S. Dickinan (1980), "The Psychological Un-conscious: A Necessary Assumption for All PsychologicalTheory?" American Psychologist, 35, 421-34.

and D. Fritzler (1968), "Visual Evoked ResponseCorrelates of Unconscious Mental Processes," Science,161 (no. 19), 295-298.

Swets, J. A. (1961), "Is There a Sensory Threshold?" Sci-ence. 134 (no. 3473), 168-177.

Time (1979), 114 (September 10), 63.Toronto Star (1978) (October 23), c l .Treisman, A. M. and G. Geffen (1967), "Selective Attention:

Perception or Response?" Quarterly Journal of Experi-mental Psychology, 19 (no. 1), 1-17.

Walstedt, J. J., F. Geis and V. Brown (1980), "Influence ofTelevision Commercials on Women's Self-Confidence andIndependent Judgment,'' Journal of Personality and SocialPsychology. 38 (no. 2), 203-210.

Washington Post (1979) (May 27), c4.Wright, R. (1973), "The Cognitive Processes Mediating Ac-

ceptance of Advertising," Journal of Marketing Research,10 (February), 53-62.

Zajonc, R. B. (1980), "Feeling and Thinking: PreferencesNeed No Inferences," American Psychologist, 35 (no. 2),151-175.

Zuckerman, M. (I960), "The Effects of Subliminal and Su-praliminal Suggestions on Verbal Productivity," Journalof Abnormal and Social Personality, 60 (no. 3), 404-11.

Subliminal Advertising: What You See Is What You Get / 4 7