Inhibiting Brand Name Recall: a Test of the Salience Hypothesis

Paul W. Miniard, The Ohio State University
H. Rao Unnava, The Ohio State University
Sunil Bhatla, The Ohio State University
ABSTRACT - Two articles by Alba and Chattopadhyay (1985, 1986) demonstrated that having subjects think about a brand can interfere with the retrieval of competitive brand names. This paper reports the findings of a study designed to provide a further test of some conceptual and pragmatic issues germane to recall inhibition.
[ to cite ]:
Paul W. Miniard, H. Rao Unnava, and Sunil Bhatla (1989) ,"Inhibiting Brand Name Recall: a Test of the Salience Hypothesis", in NA - Advances in Consumer Research Volume 16, eds. Thomas K. Srull, Provo, UT : Association for Consumer Research, Pages: 264-270.

Advances in Consumer Research Volume 16, 1989      Pages 264-270

INHIBITING BRAND NAME RECALL: A TEST OF THE SALIENCE HYPOTHESIS

Paul W. Miniard, The Ohio State University

H. Rao Unnava, The Ohio State University

Sunil Bhatla, The Ohio State University

ABSTRACT -

Two articles by Alba and Chattopadhyay (1985, 1986) demonstrated that having subjects think about a brand can interfere with the retrieval of competitive brand names. This paper reports the findings of a study designed to provide a further test of some conceptual and pragmatic issues germane to recall inhibition.

INTRODUCTION

In a series of recent experiments, Alba and Chattopadhyay (1985, 1986) have demonstrated that having subjects think about a particular brand can inhibit the recall of competing brands. Similarly, thinking about a particular product category was found to suppress the retrieval of competitive product categories (e.g., deodorant sprays versus deodorant soaps). Their results provide rather compelling support for the robustness of this effect. Recall inhibition was demonstrated across a number of different product categories and induction techniques. Moreover, inhibition was also observed among brands consumers would consider buying even after a full day had elapsed between the salience induction and brand recall, results which lend further support to the pragmatic importance of this effect.

Alba and Chattopadhyay's "preferred explanation" for why inhibition occurs is the salience hypothesis originally advanced by Rundus (1973). In essence, the probability of retrieving a brand name is determined by the strength of association between the product category and brand name divided by the summed strengths of associations between the product category and all brand names (referred to as the ratio rule). The presentation of a brand name causes an increase in associative strength, thereby enhancing its salience (i.e., the prominence or level of activation in memory) and accessibility at the expense of unpresented brand names.

Our inquiry into this area was motivated by both conceptual and pragmatic interests. As revealed by the discussion in the following section, relatively little progress has been made in establishing the tenability of the ratio rule. Moreover, our analysis of the ratio rule suggests some potential restrictions in the pragmatic value of inhibiting brand name recall.

THE SALIENCE HYPOTHESIS

As noted above, the salience hypothesis prescribes a ratio rule formulation to account for recall inhibition. The particular predictions one derives from the ratio rule depend upon one's assumptions about how the salience induction affects the cue item's strength (Basden, Basden and Galloway 1977). In order to illustrate these varying predictions, two examples are employed. The first example, presented in Table 1, represents the effects predicted when the increase in a cue item's associative strength is proportional to its preexisting strength. Table 1 depicts a product category comprised of four brands ranging in their preexisting associative strength from 1.0 (indicating a maximum association between the brand name and product category) to 0.0 (reflecting the absence of an association). The probability of recalling a given brand is derived from-a brand's associative strength divided by the summed strengths of all brands. In addition to the preexisting associative strengths and recall probabilities, Table 1 also reports how these would change as a function of cueing with each of the brands. In deriving these numbers, we arbitrarily employed a proportional increase of one-third. The critical assumption here is that the change in strength is proportional rather than the particular magnitude at which the proportional change is assumed to occur.

The impact of cueing with a given brand can be derived from a comparison of the average recall probability for noncued brands (recall of noncued brands is the conventional criterion based on preexisting (i.e., analogous to a control condition) associative strengths. For example, when brand A is used as a cue, the average recall probability for noncued brands is 16% [(32 + 16 + 0/3]. In the absence of cueing, the average recall probability for noncued brands is also 16%. In other words, the ratio rule predicts that cueing with brand A should not induce inhibition since its associative strength is already at the maximum level.

Similarly, average recall probability is unaffected when brand D is used as a cue. Note, however, that brand D's expected inability to undermine retrieval is dependent upon the assumption that changes in associative strength are proportional to the preexisting level. As will be shown shortly, changing this assumption can lead to very different predictions. [The assumption that cueing increases an item's associative strength at a proportional rate appears problematic for items that are initially unassociated with the category (e.g., brand D in Table 1). It seems inconceivable that strength would remain unaffected by cueing as some increase, no matter how minute or temporary in nature, should occur when a person actively processes a single stimulus. This limitation could be overcome by modifying, this assumption to include a positive constant (i.e., strength increase is equal to a constant plus some increment based on initial strength).]

In contrast, inhibition is anticipated when cueing involves either brand B or C. When brand B is used as a cue, the average recall probability is 20.7% versus an average of 23% in the absence of cueing. The respective probabilities are 26.7% and 23.3% for brand C. Thus, the use of brand B should produce a 10% reduction in recall probability [(23 - 20.7)/-73] versus nearly a 6% reduction for brand C. In other words, brands more strongly (but not perfectly) associated with the product category should induce greater inhibition.

TABLE 1

AN ILLUSTRATION OF THE RATIO RULE'S PREDICTIONS BASED ON PROPORTIONAL INCREASE ASSUMPTION

On the other hand, somewhat different predictions are derived from the ratio rule under the assumption that a cue item becomes, at least temporarily, maximally associated with the item category. Table 2 captures this situation. As before, cueing with brand A should not influence recall. However, a very different pattern emerges for the remaining brands. Whereas associative strength is enhanced more for strongly associated items under the proportional increase assumption, the maximum increase assumption produces a greater increase in strength for items possessing weaker preexisting associations. Consequently, the ratio rule prescribes greater inhibition for brands having weaker associations. Indeed, brand D is now anticipated to produce the strongest amount of inhibition.

We suspect that both assumptions about the increase in item cue strength are valid depending upon the particular recall context. If, for instance, inhibition is encouraged by presenting a single cue item and recall occurs immediately after the cue is removed, it would seem reasonable to believe that the cue's associative strength increases to the maximum level. On the other hand, the use of a large number o' cue items and/or large time intervals between cue exposure and retrieval may favor the proportional increase assumption.

The preceding discussion suggests several fundamental predictions based on the ratio rule. Firs!. regardless of which assumption may hold about the effect of cueing on a cue item's associative strength, the ratio rule predicts that inhibition will not occur when cueing involves an item of maximum strength (e.g., brand A in Tables 1 and 2). Second, the use of cue items with intermediate levels of association (brands B or C) is anticipated to undermine the retrieval of items possessing maximum association. (as suggested by the decline in the retrieval probability of brand A in Tables 1 and 2). Even a cue lacking any association may cause this effect under the maximum increase assumption (e.g., the impact of cueing with brand D on brand A's recall probability in Table 2). Finally, the ratio rule predicts that retrieval should be differentially affected by the preexisting strength of cue items, although the precise pattern depends on how cueing influences a cue item's strength.

TABLE 2

AN ILLUSTRATION OF THE RATIO RULE'S PREDICTIONS BASED ON MAXIMAL INCREASE ASSUMPTION

Support for the salience hypothesis has come from findings that having subjects think about a particular item leads to a reduction in the retrieval of remaining category items. While such evidence is consistent with the salience hypothesis (assuming that the cue item's preexisting associative strength was less than maximum), it is also congenial with alternative explanations that have been offered to account for recall inhibition (Nickerson 1984). Consequently, a more demanding test would entail examination of the ratio rule's predictions about the role of relative associative strength.

In this regard, empirical evidence is extremely sparse. Only two studies have considered the potential influence of a cue's associative strength. In an early study, Karchmer and Winograd (1971) concluded that the preexisting strength of cue items had little effect on the amount of recall inhibition observed in their study. However, the accuracy of this conclusion is debatable. An examination of the reported recall means indicates that subjects exposed to weak cues retrieved 5.7% fewer items in the first experiment and 12.4570 fewer items in the second experiment. In comparison, subjects receiving strong cues recalled 8.9% and 18.4% fewer items in the two experiments. Thus, recall inhibition was around 50% greater in both experiments for strong relative to weak cues. [This reinterpretation of Karchmer and Winograd's results is consistent with the ratio rule only if it is assumed that the strength of cue items increased at a proportional rate. In their study, subjects in the cued recall conditions were given a total of 25 cue items. Given this large number of cues, it seems unlikely that the associative strength of each cue item reached maximum strength. Consistent with this, subjects displayed less than perfect retrieval of the cue items during recall.]

More recently, Alba and Chattopadhyay (1986; Experiment 1) report that variations in a cue brand's preexisting associative strength, operationalized in the form of a brand's retrieval probability during free recall, did not affect the amount of inhibition. There is, of course, the possibility that such null effects are more reflective of methodological than conceptual weaknesses (e.g., failure to use cues possessing "sufficiently" different levels of associative strength).

In sum, while the ratio rule provides a very precise framework for deriving predictions about the nature of recall inhibition. available evidence sheds relatively little light on these predictions. The impact of cueing with items completely associated with a category has not been considered. Nor has research examined whether inhibition will also occur for items that are perfectly associated with a category. Although research has shown that cueing undermines retrieval, this finding is equally compatible with other explanations beyond the salience hypothesis. Research undertaking a more precise test of the ratio rule by considering the role of a cue's associative strength has yielded mixed support at best.

RESEARCH OBJECTIVES

One objective of this study was to provide a further examination of the ratio rule's predictions concerning the influence of a cue item's relative associative strength. Given the dependency of these predictions on how cueing affects a cue item's strength, it is difficult to unambiguously specify the direction of this effect (i.e., whether cues induce more or less inhibition depending on their strength), with one exception. Regardless of how cueing influences a cue's strength, the ratio rule does suggest that inhibition will be less for cues possessing maximum association than those having intermediate strength levels. Indeed, the ratio rule predicts that inhibition will not occur when cueing involves items of maximum strength. These considerations indicate that manipulations of a cue's strength, which include conditions representing maximum and intermediate levels of associative strength, should provide a more rigorous test of the salience hypothesis.

Accordingly, we employed a manipulation that encompassed the full range of associative strength. Depending on the particular condition, subjects were exposed to a cue brand possessing either a maximum, intermediate, or nonexistent level of association. Such a manipulation would undermine attributing any null effects of a cue's relative strength, as reported by prior investigations, to methodological factors (i.e., failing to use a manipulation of "sufficient" strength). Moreover, this manipulation should also provide evidence concerning any potential restrictions on a given brand's ability to evoke inhibition. Confirmation of the ratio rule's position that inhibition cannot be evoked by a cue that is perfectly associated with the category, for instance, would represent one such restriction.

Another objective was to examine the ratio rule's prediction that the recall of items possessing maximum associative strength can be inhibited. In the context of product categories and brand names, it would seem reasonable to believe that brand names may often achieve a perfect associative strength with a product category. Massive advertising budgets sustained over time may lead to a brand becoming synonymous with the product category. A consumer's preferred brand, which is likely to serve as an exemplar for the product category (cf. Alba and Hutchinson 1987; Cohen and Basu 1987), should also be maximally associated with the product category.

We are, however, skeptical about the potential for inhibition to encompass items having maximum association. Indeed, it seems quite inconsistent to have an item that is completely associated with a category and then expect less than total recall of the item. If in fact items are perfectly associated with a category, then we anticipate that activating a category will also trigger the activation of these items regardless of the presence or absence of cue items.

This issue was explored in the present study by testing whether inhibition would also occur for preferred brand recall. Beyond its implications for the soundness of the ratio rule a test of preferred brand inhibition possesses considerable pragmatic significance. If inhibition effects do extend to the consumer's preferred brand, then the practical value of inhibition is enhanced considerably.

METHOD

Design

The influence of a cue brand's associative strength was tested in the context of a single factor design with four treatment conditions plus a control condition. Depending on the treatment condition, subjects were exposed to a print ad for their preferred brand or a brand drawn from either their consideration, familiar, or unfamiliar sets. Subjects in the control condition were exposed to randomly chosen brand names from the product category.

Subjects and Procedure

A total of 178 subjects recruited from undergraduate marketing classes at a major midwestern university participated in the study for course credit. The study consisted of two phases. In the first phase, a booklet containing a list of brand names for five product categories (soap, toothpaste, deodorant, TV sets, and shampoo) was administered to subjects in a classroom setting. Only two of these categories (deodorant and toothpaste) were of interest. The rest were included so as to obscure the relationship between the first and second phases.

The list for each product category appeared on a separate page with the brand names in alphabetical order. As the experimenter read the list aloud, subjects placed a "1" next to all of the brand names that were unfamiliar to them. Next. subjects were asked to write other brand names that they were aware of but which did not appear on the list. Blank lines were provided for this purpose at the bottom of the page. Subjects were then instructed to place a "2" next to those brand names that they would consider buying for their own use. Finally, subjects were asked to indicate the brand they preferred the most.

This initial phase of the study enabled us to operationalize the inhibition treatment conditions. After randomly assigning subjects to conditions, an ad was developed using an appropriate brand name. Subjects in the preferred brand cue condition, for example, received an ad for their preferred brand. In the remaining treatment conditions, there were typically several alternative brands from which to choose in developing the stimulus ad. In such cases, the brand was selected randomly. Thus, each treatment condition employed a variety of brand names which shared the property of belonging to the same brand classification category.

The second phase of the study involved testing, for recall inhibition. In order to divorce the two parts f.cm one another as well as minimize any potential carryover effects, six weeks elapsed between the first and second phase. Similarly, different experimenters conducted each part.

Subjects participated in small groups of two to eight. Subjects in the treatment conditions were told that the study focused on factors that make advertisements effective. They then opened a booklet in front of them and examined a print ad for one minute. If they finished before the time elapsed they were to begin reading the ad again. After one minute, the experimenter instructed subjects to turn the pa,e and indicate the.r evaluation of the ad on a set of semantic differential scales. This measure was employed simply for maintaining the cover story. On the following page, subJects encountered the recall measure asking them to list all of the brands from the product category they could remember. Subjects were given four minutes to perform this task. Very little recall was observed after three minutes, which suggests that the time allotted was adequate.

Subjects assigned to the control condition were exposed to a print ad for an unrelated product category (i.e., either a magazine or a wristwatch). However, a brand name from the test product category did appear immediately below the recall question. This brand was randomly determined for each subject. Thus. Doth treatment and control subjects were exposed to a cue brand, but differed in the amount of time they spent thinking about the brand prior to recall.

This procedure was then repeated a second time for a different product category. The only restriction imposed was that a subject not be assigned to the same experimental condition for both products. Toothpaste and deodorant comprised the two product categories. Product order was counterbalanced across subjects.

Stimulus Materials

Stimulus materials for treatment subjects were specially prepared ads, typed on white paper and devoid of illustrations. The copy for each ad contained statements which were mainly puffery. This was done to minimize the possibility of subjects engaging in elaborative processing and consequently being reminded of any other brands while reading the ads (see Alba and Chattopadhyay 1986). The copy for both product ads was approximately equal in length (135-145 words) and possessed the same number of brand mentions (seven). For the ads used in the treatment conditions, only the brand name was changed while the copy was held constant. The ads for the control group employed the same copy and brand names.

RESULTS

Table 3 summarizes the mean number of noncued deodorant and toothpaste brands recalled by experimental condition. An overall test of recall inhibition was first performed by comparing the average recall of the four treatment conditions versus the control condition. Inhibition was not detected for deodorant as the comparison failed to attain significance (p>.1). The contrast was significant (p<.05) on the other hand, for recall of the toothpaste brands. Treatment subjects (M=7.94) recalled fewer brands than control subjects (.191=8.50). Consequently, subsequent analyses focused on the results for toothpaste.

A 2(Order) X 5(Experimental Condition) ANOVA was then undertaken to test for potential order effects. Of particular concern was the possibility that he initial product task sensitized subjects such that inhibition may have Seen greatly reduced, if not eliminated, for the subsequent product tasks. This concern was not supported as neither the order main Effect (p>.9) nor the interaction (p>.4) achieved significance. Finally,-we tested for gender effects with a 2 (Gender) X 5 (Experimental Condition) ANOVA. Although a significant (M<.02) main effect was observed such that females (M=8.20) recalled more brands than males (M=7.73), gender did not interact with the experimental conditions (p>.4) The following analyses are therefore pooled across product order and subject gender.

Before testing whether the amount of inhibition would vary with the associative strength of a cue brand, it was first necessary to determine whether associative strength differed as expected across the brand classification categories. Consistent with Alba and Chattopadhyay (1986), associative strength was inferred from the probability of a brand being retrieved during a free recall test. Strong support for the presumed differences was found in a separate study involving the same subject population. None of these subjects recalled a brand they had previously classified as unfamiliar, while 98% of them recalled their preferred brand. Recall rates for brands From the familiar (69%) and consideration (84%) sets tell between these two extremes. A very similar pattern was also observed in the control condition of the main study: unfamiliar (0%), familiar (60%), consideration (79%), and preferred (100%).

Pairwise comparisons were undertaken using the Tukey-HSD procedure. These analyses revealed hat, relative to the control group, inhibition occurred only for those subjects cued with their preferred brand (p<.05). Comparisons among the cue conditions indicated no significant differences.

A test of whether recall for the preferred brand could be inhibited required estimating the proportion of subjects who recalled this brand. Recall of the preferred brand was nearly perfect across the conditions: control (100%), unfamiliar (100%). Familiar (92%), and consideration (97%). Results for the preferred brand treatment condition are excluded since the preferred brand served as the cue brand. A contrast between the three appropriate treatment versus control conditions was not significant (p>.25). However, the strength of this finding is undermined by the fact that, for the treatment conditions in which it was appropriate to estimate preferred brand recall (i.e., where the cue brand was something other than the preferred brand), inhibition was not observed for total brand recall.

TABLE 3

MEAN NONCUED BRAND RECALL BY CONDITION

DISCUSSION

We-should first acknowledge that the recall inhibition effects observed here are relatively weak compared to those reported by Alba and Chattopadhyay. Inhibition was not detected for deodorant, and the magnitude of inhibition for toothpaste was far from overpowering. This difference may simply reflect the influence of methodological factors. Whereas Alba and Chattopadhyay processed subjects individually, our sessions consisted of multiple participants. In subsequent discussions with Alba and Chattopadhyay, they indicated that they had found greater inhibition when subjects are processed individually. Presumably, subjects are more likely to faithfully undertake the induction task when placed in a one-on-one setting.

Properties of the ads might also be a contributing factor. Our ads, unlike those employed by Alba and Chattopadhyay, lacked pictorial stimuli. The presence of such stimuli may facilitate efforts to enhance the salience of the advertised brand. Finally, the cover story may have inadvertently undermined the strength of the treatment induction. Subjects might have devoted some effort to critiquing the ad in anticipation of subsequent questioning, thereby detracting from their concentration on the brand itself.

Given this apparent limitation in the strength of the treatment induction, it is not possible to draw any strong conclusions about the ratio rule's prediction concerning the influence of a cue item's relative associative strength. Failure to detect recall differences between treatment conditions may simply be an artifact of a "weak" induction. This same constraint prevents any meaningful conclusion about the potential to inhibit the recall of items possessing complete association, such as the preferred brand.

More challenging to the ratio rule, however, is the finding from the preferred brand cue condition. Whereas the rule predicts that inhibition should not occur for cues having maximum associative strength, recall was in fact significantly lower for this condition. Even the validity of this result, however, is threatened by differences across the experimental condition in the effect of deleting the cue brand from subjects' recall protocols. Given that the likelihood of recalling a particular brand varies across the various brand classification categories, one might argue that deleting the cue brand penalizes some conditions more than others. For example, because the preferred brand enjoys virtually perfect recall, subjects cued with their preferred brand are more heavily "penalized" by deleting the cue brand than subjects in the remaining conditions. In the typical inhibition study, this problem is eliminated by deleting all of the cue brands from all of the conditions. Unfortunately, our particular design, in which different subjects received different cue brands even within a condition, renders this solution impractical (i.e., it would essentially eliminate all brands).

Thus, the strength of our evidence is greatly qualified by the above concerns. Obviously, further investigations are necessary to address the conceptual and pragmatic issues raised in this paper. Indeed, evidence concerning the potential for reducing retrieval of the preferred brand is clearly needed for clarifying the potential usefulness of efforts to induce recall inhibition. It is our hope that the present paper has raised so ne important issues that may help guide future work aimed at understanding the process and value of recall inhibition.

REFERENCES

Alba, Joseph W. and Amitava Chattopadhyay (1985). "The Effects of Context and Part-Category Cues on the Recall of Competing Brands," Journal of Marketing Research, 22 (August), 340-349.

Alba, Joseph W. and Amitava Chattopadhyay (1986), "Salience Effects in Brand Recall," Journal of Marketing Research, 23 (November) 363-359.

Alba, Joseph W. and Wesley Hutchinson (1987), "Dimensions of Consumer Expertise," Journal of Consumer Research, 13 (March), 411-454.

Basden, David R., Barbara H. Basden, and Bets,v C. Galloway (1977), "Inhibition with Part-List Cueing: Some Tests of the Item Strength Hypothesis," Journal of Experimental Psychology, 3 (1), 100-108.

Cohen, Joel B. and Kunal Basu (1987), "Alternative Models of Categorization: Toward a Contingent Processing Framework," Journal of Consumer Research., 13 (March), 455-473.

Karchrner, Michael A. and Eugene Winograd (1971), Effects of Studying a Subset of Familiar Items on Recall of the Remaining Items: Tne John Brown Effect," Psychonomic Science, 25 (November 25), 224-225.

Nickerson, Raymond S. (1984), "Retrieval Inhibition from Part-Set Cueing: A Persisting Enigma in Memory Research," Memory and Cognition, 12 (6), 53 1 -552.

Rundus, Dewey (1973), "Negative Effects of Using List Items as Recall Cues," Journal of Verbal Learning and Verbal Behavior, 12, 43-50.

----------------------------------------