Meaningless Differentiation Revisited

Susan M. Broniarczyk, University of Texas at Austin
Andrew D. Gershoff, University of Texas at Austin
ABSTRACT - Recent research by Carpenter, Glazer, and Nakamoto (1994) suggests that brands can gain a competitive advantage with "meaningless" differentiation, even if consumers are aware of its irrelevance. Two experiments are presented that investigate the role that the attribute’s label and the decision context play in the consumer inference process associated with meaningless differentiation. Results show that attribute inferences regarding the value of the unique attribute depend on the attractiveness of the differentiated label and its correspondence with the irrelevance disclosure. Additionally, the impact of meaningless differentiation on consumer choice was shown to depend on the degree of meaningful differentiation between the alternatives.
[ to cite ]:
Susan M. Broniarczyk and Andrew D. Gershoff (1997) ,"Meaningless Differentiation Revisited", in NA - Advances in Consumer Research Volume 24, eds. Merrie Brucks and Deborah J. MacInnis, Provo, UT : Association for Consumer Research, Pages: 223-228.

Advances in Consumer Research Volume 24, 1997      Pages 223-228

MEANINGLESS DIFFERENTIATION REVISITED

Susan M. Broniarczyk, University of Texas at Austin

Andrew D. Gershoff, University of Texas at Austin

[Thanks to Joe Urbany and Pat West for their helpful comments on an earlier draft.]

ABSTRACT -

Recent research by Carpenter, Glazer, and Nakamoto (1994) suggests that brands can gain a competitive advantage with "meaningless" differentiation, even if consumers are aware of its irrelevance. Two experiments are presented that investigate the role that the attribute’s label and the decision context play in the consumer inference process associated with meaningless differentiation. Results show that attribute inferences regarding the value of the unique attribute depend on the attractiveness of the differentiated label and its correspondence with the irrelevance disclosure. Additionally, the impact of meaningless differentiation on consumer choice was shown to depend on the degree of meaningful differentiation between the alternatives.

INTRODUCTION

The crux of marketing is to create a niche in the marketplace by differentiating from competitors on an important attribute. A differentiated position increases consumer recall and consideration of the product, resistance to competitive attack, and product preference for market segments that value the distinctive attribute. Recent research by Carpenter, Glazer, and Nakamoto (1994) questions the widely-held assumption that marketers need to differentiate on an attribute that is relevant, meaningful, and valuable. Counter to prevailing marketin strategy (Porter 1985), they find that marketers can benefit even when differentiation is meaningless.

Marketers have long recognized the persuasive powers of advertising. Advertising has been shown to frame consumer decisions such that featuring an attribute in an advertisement increases its importance in consumer product evaluation (Gardner 1983, MacKenzie 1986). Featuring an irrelevant attribute, however, was thought to have long-term negative consequences because consumers would discover during product trial that the attribute did not, in fact, improve product performance. In their review of consumer learning, Hoch and Deighton (1989) questioned this assumption arguing that many product experiences do not provide unambiguous feedback about attribute performance. Furthermore, they asserted that because consumers treat a marketer’s claim as a default hypothesis (Hoch and Ha 1986), confirmatory biases in testing would actually lead consumers to perceive the uninformative product experience as verifying the marketer’s claim.

This malleability of consumer learning led Hoch and Deigton (1989) to suggest that brands could gain long-term benefits by differentiating on attributes that give the appearance of being valuable, but in fact, are irrelevant to improving product performance. For example, Folger’s instant coffee advertises its patented "flaked coffee crystals" although the shape of the coffee crystal is only relevant for ground, not instant coffee.

In two laboratory experiments, Carpenter et al. (1994) (referred to as CGN) found support for meaningless differentiation. Two results are noteworthy. First, consumers valued an irrelevant attribute in the absence of any marketer claim regarding its superiority. Second, CGN found that even after being explicitly informed that the differentiation was meaningless, subjects still positively valued the differentiated brand. Attribute belief measures confirmed that subjects learned that the attribute was irrelevant (i.e., they reported no value for the unique level of the attribute). Nonetheless, meaningless differentiation had a positive impact.

This latter finding is particularly interesting because it suggests that the uniqueness of a meaninglessly differentiated attribute can result in inferences of alternative superiority apparently without inferences of attribute level superiority. That is, an alternative differentiated by an irrelevant attribute can receive enhanced evaluation without subjects generating positive inferences about the attribute. Given the significance of this finding for marketers and public policy makers, we believe that this study deserves replication and extension in two directions. First, we believe that it is important to examine factors that may impact the likelihood that consumers will draw positive inferences from meaningless differentiation. Experiment 1 examines the moderating effects of the differentiated attribute label’s attractiveness and its correspondence with the irrelevance disclosure on meaningless differentiation. Second, we believe it is important to understand the contexts in which meaningless differentiation will affect consumers’ decisions relative to other meaningful information. Specifically, a second experiment examines whether an irrelevant attribute can be used to counter a competitive weakness on a relevant attribute.

MEANINGLESS DIFFERENTIATION PROCESS

Several psychological processes may account for meaningless differentiation leading to a meaningful brand advantage. CGN posit that uniqueness may create value because it leads to greater weight in evaluation (Kahneman 1973), prominence in inter-brand comparison (Houston, Sherman, and Baker 1989), and simplification of decision-making (Fiske and Taylor 1984). Although differentiation leads to greater attention directed towards the meaningless attribute level, it is not obvious that the consumer will necessarily infer a positive value. CGN suggest thatconsumers may believe that the distinctive attribute causes better product performance (McGill 1989) especially if they make pragmatic inferences (Harris and Monaco 1978) that a brand would not offer a unique attribute level unless it added value.

Nonrevelation of Irrelevance

In the CGN study, subjects were not provided with any claims regarding the differentiated attribute’s benefits, merely its identifying label and a brief description. Examination of their stimulus materials reveals that meaningless differentiation was achieved by having one brand of jackets described as having "alpine class" down fill whereas the other seven brands in the competitive set were described as having "regular" down fill. Down fill was described as the type of bird feather used with "alpine class" being goose feathers whereas "regular" was a mixture of goose and duck feathers.

Several aspects of the stimuli are worth noting with respect to the likelihood of a positive inference to the meaninglessly differentiated attribute. First, the label "regular" has a baseline connotation suggesting that any differentiation is positive. Second, consumers may prefer a pure to a mixture filling. Third, the attribute labels are nondescript regarding specific attribute performance allowing for multiple interpretations (Hoch and Ha 1986) and a potential likelihood of consumers inferring greater value than warranted given the descriptive attribute information (Preston 1977). Fourth, consumers may possess prior beliefs that goose feathers are softer than duck feathers. Additionally, they may also use their intuitive beliefs regarding the relationship between "alpine class" and mountains and skiing to generate inferences that compare favorably to the generic "regular" label in the product class of winter jackets (Broniarczyk and Alba 1994).

Thus, consistent with traditional models of consumer evaluation (Fishbein and Azjen 1975), the favorableness of differentiation is expected to be a function of the benefits associated either directly or by inference with the attribute. Therefore, our replication of CGN will include a manipulation of the previously unexamined effect of attribute label attractiveness. Meaningless differentiation is hypothesized to depend on the attractiveness of the differentiated attribute’s label relative to the competitive set. Specifically, when the differentiated attribute is described by a more attractive label, it will be positively valued, whereas when it is described by a less favorable label it will be negatively valued. Uniqueness in and of itself is not expected to be sufficient to generate positive value.

Revelation of Irrelevance

The finding that led CGN to ascribe value to uniqueness per se was that subjects positively valued the meaninglessly differentiated alternative, even when pre-warned of the irrelevance of the differentiation. Prior to judging the jackets, one group of subjects was informed that for down fill "the age of a bird matters, the type of bird it comes from does not make a difference." Manipulation checks confirmed the success of the irrelevance instruction with these subjects claiming indifference between alpine and regular down fill yet their product judgments revealed a positive bias towards the alternative with the "alpine class" label. Moreover, CGN still found a positive benefit for meaningless differentiation when analysis was restricted to only subjects who were indifferent or had lower preferences for "alpine class."

This result is counter to a large body of research showing that consumers are discriminating regarding product benefits. In fact, consumers have been found to reduce their likelihood of purchasing products that included unwanted positive features or premiums (Simonson, Carmon, and O’Curry 1994). Notwithstanding the solid evidence of its success, the irrelevance revelation of CGN warrants special considertion because the uniqueness explanation is dependent on subjects believing the differentiation to be meaningless.

Examination of their irrelevance disclosure shows that it corresponded to the attribute description (i.e., goose and duck feathers) but not the attribute label (i.e., alpine and regular). In their principle of correspondence, Ajzen and Fishbein (1977) posited that an equivalent level of specificity in measurement is needed for attitudes to be highly predictive of behavior. Analogously, we propose that an equivalent level of specificity between the irrelevance disclosure and the differentiated attribute’s label is necessary for the revelation to be maximally effective in counteracting subjects’ inferences. [Thanks to Joe Urbany for suggesting this literature.] Thus, the irrelevance disclosure of CGN most likely eliminated inferences of "alpine" related to the superiority of goose over duck feathers but not other favorable inferences related to its appropriateness for winter activities.

In related research on deceptive advertising, qualifications of expanded advertising claims have been shown to have partial success in negating some but not all of the positive inferences implied by the expanded claim (Burke et al. 1988). Thus, when the set of positive associations of the attribute label is sufficiently large, the meaningless differentiation effect may remain robust even when subjects are informed of the irrelevance because discounting is only applicable to those positive associations specifically related to the disclosure. [We appreciate a reviewer for making this point.] Thus, our second research proposition is that as the correspondence between the the irrelevance disclosure and the label on the differentiated attribute increases, the likelihood of subjects maintaining their inferences decreases, and consequently, the effect of meaningless differentiation decreases.

STUDY ONE: ATTRIBUTE ATTRACTIVENESS AND CORRESPONDENCE WITH REVELATION

Experimental Design

Overview. We had three objectives in Study 1: to replicate CGN, to examine the role that the attractiveness of the unique attribute’s label plays in meaningless differentiation, and to examine the role of its correspondence with the irrelevance disclosure. The between-subjects design had three factors: 2 (Attractiveness) X 2 (Correspondence) X 2 (Revelation) and a control condition. Attractiveness was a factor that varied whether the label on the the differentiated attribute was higher or lower in attractiveness relative to the other brands in the choice set. The Correspondence factor compared whether the correspondence between the irrelevance disclosure and the differentiated attribute’s label was low or high. Revelation was a factor that varied whether subjects were explicitly informed or not informed that the differentiated attribute was irrelevant.

Pretest. A set of fifty undergraduate subjects participated in a pretest to determine if the down fill attribute levels differed in preference on a 9 point scale ranging from 1=Like to 9=Dislike. As predicted, "alpine" was rated as more preferred than "regular" [M=3.68 versus M=4.92, t(49)=3.42, p<.001] and "goose" was more preferred than "duck" [M=3.88 versus M=5.40, t(49)=3.83, p<.001]. Additionally, there was no preference difference between the high attractiveness labels of "alpine" and "goose" [t<1] nor the low attractiveness labels of "regular" and "duck" [t(49)=1.07, p>.29].

Attribute labels were then selected that varied in correspondence to the irrelevance instruction manipulation. Recall that the irrelevance revelation of CGN was that "the age of a bird matters, the type of bird it comes from does not make a difference." The attribute labels of "goose" and "duck" were deemed to have a high direct correspondence with the irrelevance instruction whereas the labels of "alpine" and !regular" were deemed to have a low direct correspondence. [A pretest also confirmed that subjects were likely to have a more diverse set of associations to the labels of "alpine" and "regular" than to "goose" and "duck". Fifty three different undergraduates rated the specificity of the down fill labels on a 9-point scale (1=Abstract, 9=Concrete). "Goose" (M=7.62) and "duck" (M=6.96) were both viewed as more specific information for down fill than "alpine" (M=4.36) and "regular" (M=6.96).]

Procedure. Two hundred and forty undergraduate students from a southwestern university participated for payment of 5 dollars. The procedure and stimuli are replicated from CGN. All subjects evaluated 8 hypothetical down jackets described on 3 binary attributes of fill rating, cover material, and stitching. The description of the fourth binary attribute, down fill, varied between conditions. In the high attractiveness/low correspondence condition, the differentiated brand had a down fill label of "alpine" whereas the 7 other brands were described as having "regular" down fill. This condition is a replication of CGN. In the low attractiveness/low correspondence condition, the reverse scenario was present with the differentiated brand having a down fill label of "regular" whereas the other 7 brands were described as having "alpine". The high correspondence conditions consisted of replacing the ambiguous attribute labels of "alpine" and "regular" with the bird-specific attribute labels of "goose" and "duck", respectively. Thus, the high attractiveness/high correspondence condition compared a brand differentiated by "goose" down fill to 7 other brands described as having "duck" down fill. The low attractiveness/high correspondence condition compared a brand differentiated by "duck" down fill to 7 other brands described as having "goose" down fill.

In the nonrevelation condition, subjects were provided with basic attribute descriptions but no explanations about the preference ordering of the attribute levels. In the revelation condition, subjects were provided with attribute descriptions detailing the preference ordering of the attribute levels and explicitly informed that for the down fill attribute the "age of the bird matters, the type of bird does not." The control condition received no information about the fourth attribute of down fill.

The rating task conformed to CGN’s where subjects anchored a blank ten-centimeter rating scale with their least and most preferred jacket. For each of the remaining jackets, they placed a mark on the line corresponding to their relative preference for the jacket. The range of ratings was from 0 (least preferred) to 100 (most preferred) with a ruler used to assign ratings to intermediary markings.

After evaluating the jackets, subjects’ preference ratings for the meaningless attribute of down fill were assessed on a 9-point scale where (1) indicates preference for "regular", (5) indicates indifference between alpine and regular, and (9) indicates preference for "alpine." Inadvertently, preference ratings for "goose" relative to "duck" down fill in the high correspondence conditions were not collected.

Results

Manipulation Check. T-tests were performed against the midpoint of the rating scale (5) to see if subjects were indifferent to the attribute level of down fill. Confirming pretest results, when irrelevance was not disclosed, subjects preferred alpine to regular down fill [M=5.85, t(26)=2.26, p<.04]. Consistent with CGN, when irrelevance was disclosed and subjects’ choice did not involve the differentiated attribute (i.e., the control condition), the irrelevance revelation was successful with subjects claiming indifference to alpine versus regular [M=5.36, t<1, p>.39]. However, our results show that revelation of irrelevance in the low correspondence conditions was able to reduce but not eliminate the likelihood of inferences [high attractiveness: Mrevealed=5.95, (t(24)=2.47, p<.03) versus Mnonrevealed=6.75, (t(24)=4.53, p<.01) and low attractiveness: Mrevealed=7.18, (t(27)=6.82, p<.01) versus Mnonrevealed=7.79, t(24)=10.38, p<.01).

Additionally, these results suggest that the post-choice measure of attribute inference was sensitive to the choice context. For instance, in the low attractiveness condition, the majority of brands ossessing "alpine" down fill appeared to convey pragmatic information that was consistent with subjects’ intuitive beliefs and thus led to more extreme beliefs about the irrelevant attribute.

Overall Means. The overall mean ratings are presented in Table 1. Comparison of the control conditions revealed differential preference for a nontarget attribute between revelation conditions that was attributable to subjects’ unfamiliarity with down jackets. Specifically, our southwestern subjects in the nonrevealed condition preferred cotton to synthetic cover material whereas subjects in the revealed condition were informed that synthetic was preferred to cotton. Thus, because it was not appropriate to compare differences from the control between revelation conditions, we ran separate analyses for each revelation condition. The data were analyzed using ANOVA with independent variables of attribute attractiveness and attribute correspondence. The dependent variable was the difference in a given subject’s rating on the meaninglessly differentiated jacket from the mean rating for that jacket in the control condition.

Replication of CGN. Recall that the high attractiveness/low correspondence condition was identical to the CGN manipulation (1 alpine/ 7 regular). Our findings in this condition replicate the CGN results such that the irrelevant attribute increased brand evaluation regardless of the disclosure of the attribute’s irrelevance (nonrevealed: D=+38.12, p<.02 and revealed: D=+18.25, p<.05). Thus, consistent with CGN, the revelation of irrelevance did not eliminate the benefit of meaningless differentiation. However, unlike CGN, we can not attribute the observed effects to the uniqueness of a meaninglessly differentiated attribute alone. Examination of those 14 of 24 subjects in the revealed condition who reported an equal or lower evaluation of the differentiated level of "alpine" downfill in the manipulation check showed no significant change from the control condition (D=+1.53, p>.89).

Attractiveness. The first hypothesis stated that the benefit resulting from meaningless differentiation depended on the attractiveness of the irrelevant attribute’s label. As hypothesized, the nonrevealed condition showed a significant main effect of the attractiveness of the differentiated attribute’s label [F(1,92)=18.30, p<.0001]. An alternative differentiated by an attribute with an attractive label (D=+28.58) was preferred to the control condition whereas it was less preferred when it had an unattractive label (D=-3.19).

The revealed condition showed a similar pattern of results for attractiveness [Dhigh attractiveness=+13.71 vs. Dlow attractiveness=-14.44, F(1,92)=14.43, p<.001]. Thus, uniqueness alone is not sufficient to create positive worth and in some cases may decrease product evaluation. These results suggest that the effect of meaningless differentiation is due to subject inferences about the value of the differentiated attribute.

Correspondence. The second hypothesis stated that for the disclosure of irrelevance to neutralize these inferences that there needed to be a high correspondence between the revelation of irrelevance and the label on the differentiated attribute. Analysis is therefore restricted to the revealed condition only. Results showed no main effect for Correspondence (F<1) but the predicted Attractiveness X Correspondence interaction was significant (F(1,92)=5.08, p<.03). As hypothesized, simple effects show that there was a considerable effect of attractiveness at the low correspondence level [Dhigh attractiveness=+18.25 vs. Dlow attractiveness=-21.70, F(1,47)=26.88, p<.001] whereas this effect was reduced, although still significant at the high correspondence level [Dhigh attractiveness=+9.17 vs. Dlow attractiveness=-6.54, F(1,45)=4.40, p<.05].

A stronger picture of correspondence emerges when each cell mean is compared to the control. At low correspondence levels, having an attractive (p<.05) or unattractive (p<.05) label resulted in meaningless differentiation significantly increasingor decreasing alternative evaluation, respectively. However, at the high correspondence level, neither the high (p>.17) nor low (p>.54) atractiveness label resulted in an effect for meaningless differentiation.

TABLE 1

MEAN RATINGS FOR EXPERIMENT 1

In summary, uniqueness alone did not lead to a positive valuation. In the high attractiveness/low correspondence condition, we replicated CGN results such that meaningless differentiation had a positive effect on consumer evaluations, but these results were insignificant when analysis was restricted to only those subjects satisfying the manipulation check. Furthermore, our results showed a significant effect for the attractiveness of the irrelevant attribute such that an attractive label led to a favorable effect whereas an unattractive label led to an unfavorable effect for meaningless differentiation. However, a high correspondence between the irrelevant attribute’s label and disclosure of its irrelevance nullified the inferences that consumers may have generated regarding the attribute’s benefits and hence resulted in no effect for meaningless differentiation.

However, the test used by CGN and replicated here, set a high standard because the eight jackets in Study 1 provided a full factorial representation of the three relevant attributes. The meaninglessly differentiated attribute was included in the description of one of these alternatives and hence the relative rating scale effectively required subjects to value the meaninglessly differentiated attribute over a meaningful one. In Study 2, we contrast the effect of meaningless differentiation in a choice context where products are identical on meaningful attributes to a context where the differentiated alternative is inferior on the meaningful attributes. Additionally, Study 2 was run to further examine the strength and completeness of the revelation of the differentiated attribute’s irrelevance and to gain a better understanding of consumer inferences related to the irrelevant attribute.

STUDY 2: CONTEXT EFFECTS AND INFERENCES

Experimental Design

Overview. A robust effect of meaningless differentiation would be evidence that an inferior alternative differentiated by an irrelevant attribute would be preferred over an alternative with a superior level of a meaningful differentiation. However, in this age of parity products, marketers are also concerned about the ability of meaningless differentiation to affect choice for one alternative over another when all other attributes have equal values.

Study 2 was a two-factor between-subjects design: 2 (Choice context) X 2 (Correspondence). The factor of Choice Context varied whether the meaninglessly differentiated attribute was placed with an alternative that was identical or inferior to its competitor. The Correspondence factor was the same as Study 1 with the bird-specific labels of "goose" and "duck" representing high correspondence and labels of "alpine" and "regular" representing low correspondence. Additionally, detailed process measures were collected to identify the inferences subjects made regarding the irrelevant attribute.

Procedure. Fifty-three subjects participated for course extra credit. Examination was restricted to differentiation of an irrelevant attribute that had high attractiveness and all subjects were informed of the irrelevance of the down fill using the same manipulation as Study 1. In a training phase, subjects evaluated six down jackets all with the common level of "regular" ("duck") down fill. In a subsequent test phase, subjects then choose between Brand X and Y that had identical fill ratings, cover material, hoods, and cleaning requirements. However, Brand X was meaninglessly differentiated with the unique level of "alpine" ("goose") and rand Y shared the common level "regular" ("duck") with the other six jackets. In the inferior choice context, Brand X also had a lower stitching rating than Brand Y whereas in the identical choice condition, both brands had equal stitching ratings. Note that stitching was chosen because Study 1 revealed it to be the least important of the meaningful attributes. After making their choice, subjects provided open-ended responses describing their choice process and answered close-ended questions regarding the benefits associated with the irrelevant attribute.

Choice Results

Manipulation Check. The same 9-point scale measuring preference for the levels of the down fill attribute was used as in Study 1 where (1) indicates preference for "regular" ("duck"), (5) indicates indifference between "alpine" and "regular" ("goose" and "duck"), and (9) indicates preference for "alpine" ("goose"). Consistent with Study 1, the results show that the manipulation check was affected by the decision context. An ANOVA with independent variables of choice context and attribute correspondence revealed a significant effect of choice context (F(1,49)=4.02, p=.05). The manipulation check was successful for both inferior conditions of "goose" (M=5.31, F<1) and "alpine" (M=5.00, F<1) but unsuccessful in the identical condition where choice necessitated that subjects discriminate on the basis of the meaningless differentiated attribute [("goose" M=5.93, F=2.14, p=.053) and ("alpine"M=6.23, F=1.92, p<.08)]. Cognitive responses and inference beliefs revealed a similar pattern and will be discussed in more detail later. However, the choice results will show that an irrelevant attribute was not able to overcome a competitor’s superiority on a relevant attribute. Thus, it appears that the irrelevance manipulation was sufficient in all cases, but the decision task in the identical context forced some subjects to counteract this disclosure.

TABLE 2

STUDY 2 ATTRIBUTE BELIEFS

Correspondence. Choices were analyzed using a loglinear model with the SAS CATMOD procedure with factors of choice context, attribute correspondence and their interaction. Results show a significant main effect for choice context (c2=17.32, p<.01) and no significant effect for correspondence (c2<1) nor its interaction with choice context (c2<1).

Choice Context. In the inferior condition, 13 out of 14 subjects in the low correspondence condition chose the superior Brand Y mentioning its better stitching and 2 subjects explicitly mentioning the irrelevance of down fill. Of the fourteen subjects in the high correspondence/inferior condition, thirteen also chose the superior Brand Y due to its superior stitching with 6 subjects explicitly mentioning the irrelevance of down fill.

Conversely, in the identical conditions, the majority of subjects chose the alternative differentiated by a meaningless attribute. In the low correspondence/identical condition, 10 out of 14 subjects chose the meaningless differentiated Brand X even though 6 subjects explicitly mentioned that down fill was irrelevant. Similarly, in the high correspondence/identical condition, 9 out of 12 subjects chose the meaningless differentiated Brand X with 2 subjects explicitly mentioning the irrelevance of the down fill. A fairer test in the identical condition is to see if choices differed from chance (proportion=.50). Results showed that choices of the meaningless differentiated alternative in the identical condition were greater than expected by chance (proportion=.73, t(26)=2.60, p<.02).

Thus, the results suggest that even when revealed to subjects that the differentiation of an attribute is meaningless, that subjects will still show preference for the alternative with a irrelevant attribute when there are no other available attributes on which to differentiate the alternatives. However, when subjects can make their choice on the basis of the relevant attributes, a meaninglessly differentiated attribute does not impact their ecisions.

Inference Results

Subsequent attribute belief measures confirmed that consumers did infer benefits to the meaninglessly differentiated attribute. Attribute beliefs for the differentiated irrrelevant attribute relative to its competitors were measured on a 5-point scale. Two inferences suggested from Study 1 were inferences about the attribute performance of softness and quality [Softness:1=Rougher to 5=Softer; Quality: 1=Lower to 5=Higher]. Results are presented in Table 2. These attribute beliefs were collected at the end of the experiement (post revelation and post choice) and thus they represent inferences that the irrelevance disclosure did not eliminate.

Correspondence. Although correspondence had no effect on choice, it did affect the number of inferences consumers retained about the differentiated attribute. Specifically, subjects in the low correspondence condition reported higher softness to the meaninglessly differentiated attribute (M=3.84) than subjects in the high correspondence condition (M=3.32), F(1,46)=4.10, p<.05. The inference of higher softness to "alpine" but not "goose" is especially pertinent because the irrelevance instruction dealt with whether the down fill affected the softness of the jacket. Although the mean levels of inferences about quality were directionally consistent with more inference being made in the low (M=3.52) than high (M=3.44) correspondence condition, this effect was small and nonsignificant, F<1. These findings provide further insight into the results of Experiment 1 because they provide more direct evidence that a high correspondence between the disclosure of irrelevance and the label on the irrelevant attribute is more effective in counteracting inferences related to the attribute. However, these inferences were only utilized in the choice decision if the context required it.

Choice Context. The choice context displayed some effect on inference measures, albeit less, than the choice measure. Specifically, although the mean levels of inferences about softness were directionally consistent with more inferences being made in the identical (M=3.44) than inferior (M=3.72) context, this effect was nonsignificant, F(1,46)=1.14, p>.29. For the quality inference, subjects in the identical context reported significantly higher quality to the meaninglessly differentiated attribute (M=3.68) than subjects in the inferior context (M=3.27), F(1,46)=2.86, p<.10

Note that these inferences were collected after the choice task, so we cannot be certain of the effect of inferences about the meaninglessly differentiated attribute on the choice decision. [We appreciate a reviewer for making this point.] Thus, subjects may have had an equivalent level of inference making initially but because subjects in the inferior context did not need to utilize the inference to make a decision, they reported lower final levels of attribute beliefs. Alternatively, subjects in the identical context may not have formed any spontaneous inferences but later reported making them to justify their choice.

In summary, Study 2 showed that meaningless differentiation only provides "meaningful" differentiation if subjects are forced to choose between two brands that are identical except on the irrelevant attribute. A marketer can not use differentiation on an irrelevant attribute to overcome a competitive deficiency on a relevant attribute if subjects are aware of the attribute’s irrelevance.

SUMMARY AND DISCUSSION

Our results show that a meaninglessly differentiated attribute can lead to alternative preference, even if subjects are aware of the differentiation’s irrelevance, but only under certain circumstances. CGN claim that positive valuation of meaningless differentiation is primarily driven by uniqueness. We found that attribute inferences regarding the value of the unique attribte depend on the attractiveness of the its label and its correspondence with the irrelevance revelation. There is a limit to the impact of inference-making such that even if consumers make inferences regarding the meaningless differentiated attribute they may not affect choice in the presence of more diagnostic information (Feldman and Lynch 1988).

However, the choice context itself affects the very inferences that consumers draw. Other information about the alternative such as its relative price (CGN), the relative quality of its other attributes, and the relative quality of its competitors are also likely to affect the evaluative inferences regarding meaningless differentiation.

Meaningless differentiation exerted its strongest effect when there was no meaningful differentiation between the alternatives. However, this research only examined high involvement situations where inference-making is more likely to occur (Stayman and Kardes 1992). Future research should examine whether the salience of a unique attribute can overcome meaningful differentiation in low involvement situations. Additionally, uncertainty associated with a meaningful attribute (Brown and Carpenter 1996) may lead to stronger effects for meaningless differentiation.

REFERENCES

Ajzen, Icek and Martin Fishbein (1977), "Attitude-Behavior Relations: A Theoretical Analysis and Review of Empirical Research," Psychological Bulletin, 84, 888-918.

Broniarczyk, Susan B. and Joseph W. Alba (1994), "The Role of Consumers’ Intuitions in Inference Making," Journal of Consumer Research, 21 (December), 393-407.

Brown, Christina L. and Gregory S. Carpenter (1996), " When Are Irrelevant Attributes Relevant? A Strategic Inference Model," working paper, New York University.

Burke, Raymond R., Wayne S. Desarbo, Richard L. Oliver, and Thomas S. Robertson (1988), "Deception by Implication: An Experimental Investigation," Journal of Consumer Research, 14 (March), 483-494.

Carpenter, Gregory S., Rashi Glazer, and Kent Nakamoto (1994), "Meaningful Brands From Meaningless Differentiation," Journal of Marketing Research, 31 (August), 339-350.

Feldman, Jack M. and John G. Lynch (1988), "Self-Generated Validity and Other Effects of Measurement on Belief, Attitude, Intention, and Behavior," Journal of Applied Psychology, 73 (August), 421-435.

Fishbein, Martin and Icek Ajzen (1975), Belief, Attitude,Intention and Behavior: An Introduction to Theory and Research, Reading, MA: Addison-Wesley.

Fiske, Susan T. and Shelley E. Taylor (1984), Social Cognition, New York: Random House.

Gardner, Meryl Paula (1983), "Advertising Effects on Attributes Recalled and Criteria Used for Brand Evaluations," Journal of Consumer Research, 10 (December), 310-318.

Harris, Richard Jackson and G.E. Monaco (1978), "Psychology of Pragmatic Implications: Information Processing Between the Lines," Journal of Experimental Psychology: General, 107, 1-27.

Hoch, Stephen J. and John Deighton (1989), "Managing What Consumers Learn From Experience," Journal of Marketing, 53 (April), 1-20.

Hoch, Stephen J. and Young-Wan Ha (1986), "Consumer Learning : Advertising and Ambiguity of Product Experience," Journal of Consumer Research, 13, 221-233.

Houston, David A., Steven J. Sherman, and Sara M. Baker (1989), "The Influence of Unique Features and Direction of Comparison on Preferences," Journal of Experimental Social Psychology, 25, 121-141.

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

MacKenzie, Scott B. (1986), "The Role of Attention in Mediating the Effect of Advertising on Attribute Importance," Journal of Consumer Research, 13 (September), 174-195.

McGill, Ann L. (1989), "Context Effects in Judgments of Causation," Journal of Personality and Social Psychology, 57, 189-200.

Porter, Michael E. (1985), Competitive Advantage, New York: The Free Press.

Preston, Ivan L. (1977), "The FTC’s Handling of Puffery and Other Selling Claims Made 'By Implication’," Journal of Business Research, 5 (June), 155-181.

Simonson, Itamar, Ziv Carmon, and Suzanne O’Curry (1994), "Experiemental Evidence on the Negative Effect of Product Features and Sales Promotions on Brand Choice," Marketing Science, 13 (Winter), 23-40.

Stayman, Douglas M. and Frank R. Kardes (1992), "Spontaneous Inference Processes in Advertising: Effects of Need for Cognition and Self-Monitoring on Inference Generation and Utilization," Journal of Consumer Psychology, 1 (2), 125-142.

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