Journal of the Academy of Marketing Science

, Volume 38, Issue 4, pp 456–470

Brand related information as context: the impact of brand name characteristics on memory and choice

Authors

    • Indian School of Business
  • H. Shanker Krishnan
    • Marketing Department, Kelley School of BusinessIndiana University
Original Empirical Research

DOI: 10.1007/s11747-009-0175-8

Cite this article as:
Samu, S. & Shanker Krishnan, H. J. of the Acad. Mark. Sci. (2010) 38: 456. doi:10.1007/s11747-009-0175-8

Abstract

Consumer exposure to new brand names can occur in contexts with or without brand information being present. Whereas previous research has examined the effects of brand name characteristics (association set size and word frequency) on memory in the presence of brand information, this paper also assesses brand name effects in contexts without brand related information and extends it to brand consideration and choice. Two different processes are found to be operating as consumers retrieve brands and make a choice. Recall seems to benefit from a distinctiveness based route, which also guides consideration and choice. In contrast, implicit memory is influenced by familiarity, with corresponding consideration and choice effects. The implications are that firms’ choice of brand names and media needs to consider the combination of brand name characteristics that will lead to desired outcomes vis-à-vis distinctiveness or familiarity based processes.

Keywords

Brand namesMemory and recallImplicit memoryBrand associationsBrand considerationBrand choice

Introduction

Traditional models acknowledge that when a new brand is introduced in the market, consumers become aware of it primarily through advertising, where the brand name and other brand related information (e.g., features) are presented together (Orth and Marchi 2007; Campbell and Keller 2003). However, there is ample evidence that consumers avoid commercials and other forms of communication (Roehm et al. 2004) and even resist such persuasion attempts (Kirmani and Zhu 2007; Friestad and Wright 1994). To counter such behavior, brand managers often use non-traditional environments to gain exposure for their brands. In such contexts, the brand name is more prominent and there is a minimal amount of other information related to the brand. Such contexts include:
  • ➢ Reading a news report about the introduction of a new brand

  • ➢ Mere exposure to a brand name in a store while shopping

  • ➢ Exposure to a brand name that is used by other consumers

  • ➢ Reading about a brand name as part of an Internet chat room discussion

  • ➢ Exposure to a brand in a movie or TV serial (product placement)

Previous research has largely focused on how brand related information from advertising messages is processed, stored, retrieved from memory, and used in the consideration/choice process (Grimes 2008; Lurie and Mason 2007; Lee and Labroo 2004; Roehm et al. 2004; Krishnan and Shapiro 1996). However, there has been limited focus on consumer responses when exposure to brand names occurs outside the context of brand related information (Ferraro et al. 2009; Yang and Roskos-Ewoldsen 2007). This is an important area to examine as firms continue to innovate with new forms of communication when introducing brands. Hence, this paper seeks to examine the impact of context (brand related information versus no-brand information) along with brand name characteristics on memory, consideration, and choice.

We do know that brand names play an important role in communicating product benefits, image, or identity of the product (Sinn et al. 2007). Characteristics of brand names may also guide consumer choice. These effects may transpire in accordance with two paths, depending on the characteristics of the brand name. The first path suggests that brand name characteristics that make the brand distinctive impact memory and choice, with benefits to brand equity and the parent company (Klink 2003; Myers 2003). The second path identifies familiarity as a major driving force in some conditions, as familiarity can make it easier to retrieve the brand name and guide choice (Hoyer and Brown 1990). These two paths can in fact be observed for high-equity brands such as Coca-Cola, Campbell’s, and Disney that have specific characteristics that lead to consistently high performance on familiarity, uniqueness, and image (Myers 2003; Del Rio et al. 2001). For a new brand, the characteristics of the name may play an even more critical role as consumers may not be aware of its attributes. Rather, memory for the brand and subsequent choice may be based on initial impressions about the name. Hence, understanding desirable characteristics of brand names has important implications for all brands, especially new ones.

Importance of brand information as context

Many marketing studies have used advertising stimuli containing brand information to examine the effects of brand name characteristics on memory performance (e.g., Campbell and Keller 2003; Meyers-Levy 1989). For example, imagery (Lutz and Lutz 1977), associations (Del Rio et al. 2001; Meyers-Levy 1989), appropriateness (Pavia and Costa 1993), character and radical level suggestiveness (Lee and Ang 2003), semantics and sound symbolism (Klink 2001), and distinctiveness (Kohli and Hemnes 1995) have all been identified as desirable characteristics for brand names.

Increasingly, consumers (especially generation X and Y) are skeptical of persuasive forms of communication because of their manipulative nature, and rely on other non persuasive sources of information (Coates 2004; Roehm et al. 2004). The Internet has also made it easier for consumers to weaken the marketer’s control over information by allowing them access to independent sources (Ward and Ostrom 2003). Incidental brand encounters, leading to processing of brand information may also be widely prevalent (Ferraro et al. 2009). Research has documented other contexts in which consumers are exposed to brand information such as in-store displays (Allenby and Ginter 1995), word-of-mouth (Kiecker and Cowles 2001), product placements (Yang and Roskos-Ewoldsen 2007; McKechnie and Zhou 2003), and newer sources such as Internet chat rooms (Boush and Kahle 2001).

We argue that the presence versus absence of brand information as a contextual cue may lead to significant differences in how brand names are perceived. One key aspect of traditional ad-based communication is that the brand name is generally encountered in the presence of other brand information which enhances the persuasive nature of the message (e.g., pictures, headlines, claims). Without these, the consumer is exposed to the brand name in isolation or with limited brand information (Ferraro et al. 2009; Klink and Smith 2001). Brand information as a context is expected to moderate the effect of brand name characteristics on consumer responses. Specifically, we compare effects of brand name characteristics (frequency and association set size) on memory and choice in contexts where brand information is present with contexts where brand information is not available.

Theory and hypotheses

Brand recall

Assessing the role of frequency and association on memory has a long history (Reynolds and Besner 2004; Nelson and Xu 1995; Gregg 1976). Meyers-Levy (1989) examined the effects of these factors on memory using a radio ad. Frequency (or distinctiveness) was operationalized as the frequency of occurrence of a word in the English language while association set size was operationalized as the total number of different responses to the word.

Meyers-Levy’s (1989) argument runs as follows. For low frequency brand names, there is no effect of association set size on memory, because the brand stands out during processing due to being distinctive (e.g., Polo). In other words, because this is an ad context with other brand information also present, the associations do not have a differential effect on making the brand more distinctive. When a high frequency word is used as a brand name (e.g., American or Soft N Gentle tissue), the brand name gets only limited focus, as processing and encoding occur with minimal effort due to the common nature of these words. Consumers have diverse non-brand associations for the word because of its common usage (Meyers-Levy 1989). When such brands are featured in an ad, both brand and non-brand information are encoded together. When consumers try to retrieve the brand name, these two sets of associations also get activated. The logic can then be extended to consider the impact of the number of associations linked to the word. Any word used as a brand name may have large (e.g., H.I.S.) or small association set sizes (e.g., Dawn). For large set sizes, the higher number of non-brand associations creates interference for retrieval of brand information, thus lowering brand name recall. For small set sizes, on the other hand, fewer non-brand associations are retrieved, making the brand more distinctive and thereby improving brand recall. Meyers-Levy’s results (1989) support this expectation that for high frequency words used as brand names, small (versus large) set size leads to higher brand recall. Thus, the brand info condition would replicate previous research (Meyers-Levy 1989) using a different type of ad (print instead of radio). Specifically,

H1a: In the brand info condition, brand name recall will be higher when the brand name has a small versus a large association set size for brand names that are high frequency words. Brand name recall will not differ between small and large association set sizes for brand names that are low frequency words.

These predictions are in the context of an ad, where brand names are processed with other brand information such as the brand image and claims. We posit that in many real-world contexts, the brand name may be seen in the presence of information not related to the brand (e.g., in a TV serial/movie placement, or observing a friend using a brand) (Ferraro et al. 2009). In such contexts, high frequency words used as brand names will receive superficial processing, with primarily non-brand associations encoded and subsequently activated. Such processing occurs for words with both large and small association set sizes. Hence, because of the superficial processing and the emphasis on non-brand associations, the two conditions should not differ on interference. Consequently, for no-brand info conditions, association set size will not lead to differences in brand recall for high frequency words used as brand names.

Let us now consider words that occur with low frequency that are more distinctive (e.g., Kenmore). They produce “processing aimed at distinctively and meaningfully encoding the words in relation to contextual information” (Meyers-Levy 1989). More intense processing occurs because low frequency words are relatively unusual and may even create difficulty in understanding their meaning. In a no-brand info context, the absence of brand information leads to activation of associations specific to the situation where the exposure takes place (e.g., a rainy day). Such associations may be more idiosyncratic in nature. For brand names with a large set size, the activation of a large number of idiosyncratic (possibly secondary) associations increases the probability that they will interfere with recall of the brand, because the brand name is not strongly connected to them. For small association set size, because there are fewer associations that could potentially interfere, the brand name continues to be distinctive due to the processing benefit bestowed by the low frequency brand name. Accordingly, the nature of the interaction in the no-brand info context is expected to be different. The specific predicted pattern is:

H1b: In the no-brand info condition, brand name recall will be higher when the brand name has a small versus a large association set size for brand names that are low frequency words. Brand name recall will not differ between small and large association set sizes for brand names that are high frequency words.

Overall, the differences between these two contexts predicted in the two-way interactions in H1a and H1b lead to an overall three-way interaction for recall.

H1c: There will be a significant three way interaction between association set size, frequency, and context on brand name recall.

Implicit memory

Advances in the field of memory suggest that consumers possess implicit memory, i.e., use of information without conscious recollection of its source. The encoding process is the same whether implicit memory or recall is being used. However, the process through which information is retrieved differs for implicit memory and recall. In recall, there is a conscious process of retrieving a brand name; hence, engaged processing of brands facilitates this process. In contrast, implicit memory occurs without intentional retrieval; rather the brand name seems familiar and is automatically activated in memory (Lee 2002; Holden and Vanhuele 1999). In this case, retrieval is considered implicit because there is no conscious awareness of where the brand name was encountered previously. Such retrieval processes are likely to prevail in many shopping situations where consumers are pressed for time (Bogart and Tolley 1988) or have low motivation. The marketing literature documents implicit memory retrieval as measured by “indirect” tests (e.g., Krishnan and Chakravarti 1993, 1999; Duke and Carlson 1993). The primary feature of indirect tests is that there is no reference to the brand names seen earlier; rather facilitation on task performance is used to indicate memory. For example, a commonly used indirect test uses stem-completion (e.g., IMP______ for a brand name IMPERIAL margarine). If subjects who were recently exposed to the brand show a higher propensity to complete the stem completion task than those who were not exposed, this is taken as evidence for implicit memory.

Relevant to our research is the observation that implicit memory may be very relevant in situations where consumers are exposed to a brand name without other brand information. Of particular interest, specific factors seem to affect implicit memory differently than recall, partly because there is less potential for interference (Yang and Roskos-Ewoldsen 2007; Lee 2002; Shapiro and Krishnan 2001; Krishnan and Trappey 1999). Despite the fact that a brand name’s distinctiveness (word frequency) and associations appear to be important characteristics influencing brand name effectiveness, there have been very few attempts to study its impact on implicit memory (see Yang and Roskos-Ewoldsen 2007; Zinkhan and Martin 1987; Vanden et al. 1987, for related research). This is especially critical as these factors are used as important screening characteristics for new brand names; hence, their effects on memory are important to establish within and outside brand information contexts. Thus, a key objective is to expand the set of memory measures, and consider response patterns for implicit memory.

For implicit memory, the expectations were guided by previous research (e.g., Lee 2002; Shapiro and Krishnan 2001; Krishnan and Shapiro 1996). High frequency brand names have been found to perform better on an implicit memory measure of stem-completion (Krishnan and Shapiro 1996; Holden and Vanhuele 1999). This is because a high frequency word is encountered more frequently, with consumer memory getting activated with each occurrence. Based on these recurring activations, increased familiarity leads to facilitation in implicit memory. The following hypothesis suggests that this main effect should be replicated, irrespective of the context in which the brand name is exposed.

H2a: Implicit memory for the brand name should be higher when brand names are high (versus low) frequency words.

The basis for the superior performance of high frequency words is familiarity. With larger set sizes, familiarity is further heightened due to the links with other concepts as such links help activate the brand name periodically (Shapiro and Krishnan 2001). Interference between multiple associations is likely to be lower on implicit memory tests (Lustig and Hasher 2001) because there is no effort at intentional recollection (see Schacter 1987). Hence, for high frequency words used as brand names, the prediction is that a familiarity based process for large (versus small) association set size should lead to better implicit memory for the brand names.

In contrast, for low frequency words used as brand names, differences are not expected between the set sizes. As explained earlier, low frequency words are encoded after elaborative processing aimed at distinctively and meaningfully encoding the words in relation to contextual information. However, literature suggests that such elaborative processing does not influence implicit memory (Shapiro and Krishnan 2001; Graf and Schacter 1985) because the nature of retrieval is automatic. Hence, for both set sizes, the low frequency words will have low implicit memory. This leads to the following interaction, irrespective of the brand info context:

H2b: For brand names that are high frequency words, implicit memory for the brand name should be higher when the brand name has a large versus a small association set size. For brand names that are low frequency words, implicit memory for the brand name should not differ between large and small association set sizes.

Study 1

Method

Study 1 was designed to test H1 and H2 and used a 2 (brand info versus no-brand info) X 2 (low versus high word frequency) X 2 (small versus large association set size) between subjects factorial design, replicated across two product categories as a within subjects factor. Student subjects were assigned to the same condition for both product categories. Products were chosen that were relevant to the student target segment (anti-perspirant and cereal).

Pretest determined stimuli

The study used words to represent fictitious brand names in order to control for prior familiarity. An initial set of forty-four words was selected from the list provided in Paivio et al. (1968) based on the following criteria: high frequency > 100 and low frequency < 10, large association set size > 7 and small association set size < 4. Paivio et al. (1968) selected words from the word frequency norms published by Thorndike and Lorge (1944), which is the same source used in other marketing studies, including Meyers-Levy (1989). In these norms, the authors count the frequency of words occurring in the English language (books, newspapers, magazines, etc.) and, for example, a word frequency of 2 means that it occurs twice in a million English words. Meyers-Levy’s (1989) recommendation was followed in classifying words having ten or fewer occurrences per million words as low (she used 15) and words having 100 or more occurrences as high.

For association set size, Paivio et al. (1968) was again used as a starting point, and then the set size for each of the words in the study was independently estimated in a pretest using the production method (Paivio et al. 1968). In their study, they used a total of 950 words randomly arranged into booklets of 50 with each noun appearing 15 times down the left margin in each page. Eighteen subjects were given instructions and allowed 30 seconds to associate to each word. We followed a similar procedure in the pretest to measure association set size: subjects were instructed to write associations to the one word appearing multiple times (15 times) on each page. The time available for subjects was not limited, but they were instructed that they should take 1 minute per page. During debriefing, it was found that subjects did complete all the associations they had for a particular word within the given time period. The number of associations for each word was averaged across subjects. These words were subsequently pre-tested for appropriateness to the two product categories. Eighteen subjects rated each name on a 7-point scale (1 = extremely inappropriate, 7 = extremely appropriate); several brand names were chosen that were equal on appropriateness for each product.

Based on these pretests, four target brands were chosen for each product category—one for each of the four conditions (Frequency x Association Set Size). The association set sizes were significantly different across the high and low conditions (means of 7.5 for high association set size and 2.5 for low association set size, t = 14.5, p = 0.01) for all products. For anti-perspirant the brand names ANSWER (large set size) and RESPECT (small set size) were high frequency words while the brand names ELEVATE (large set size) and RETTRO (small set size) were low frequency words. Similarly, for cereal, the brand names EARTH (large set size) and TRUTH (small set size) were high frequency words while the brand names GALAXY (large set size) and ESSENCE (small set size) were low frequency words. In addition, for each product category, two control brand names were included to make the task more realistic (Queen and Monarch for anti-perspirant, and Stock and Facts for cereal); these had moderate scores on both association set size and word frequency. In summary, the pretests ensured that target brand names differed on word frequency or association set size, but were equivalent on product appropriateness.

Procedure and measures

One hundred and fifty-four introductory marketing students participated in the study to fulfill course requirements, resulting in a cell size of approximately 19 for each of the eight conditions. Subjects were assigned randomly to one of the conditions. All subjects received instructions that the purpose of the study was to evaluate potential brand names and that such tests were routinely performed by firms during the brand name selection process. The entire sequence took 40 minutes.

For the brand info context, ads were created for each of the three brands (1 test and 2 filler brands) within each product category, and the experimental condition was manipulated by changing the brand name (Answer, Respect, Elevate, and Rettro for anti-perspirant, and Earth, Truth, Galaxy, and Essence for cereal). Thus, all subjects in the brand info context were exposed to a total of six ads (three for anti-perspirant and three for cereal) placed in a separate folder. Subjects were told that these were new ads and asked to view them as they normally would if they were in a magazine and were given a total of 2 minutes (see “Appendix 1” for sample ad). Each ad contains other supporting information (picture, headline, claims) that would normally be seen in a typical ad.

All subjects in the no-brand info condition were provided two exposures to each of the six brand names (three in each category—one target brand and two filler brands). In the first exposure, they evaluated the appropriateness (1—not appropriate, 7—extremely appropriate) of the brand name. In the second exposure, subjects evaluated whether they thought the brand would succeed in the marketplace (1—would not succeed, 7—would definitely succeed). The two evaluations were performed for both product categories. The purpose of these tasks was to ensure that subjects in the no-brand and brand info conditions had approximately equal exposure. Following these exposures, all subjects followed the procedure described below.
  1. 1

    Filler task: Subjects completed a 15 minutes filler task where they evaluated brand name extensions. The brand names used in the filler task were real brand names unrelated to this study. This task ensured that their short-term memory was cleared.

     
  2. 2

    Implicit memory test for brand names: The implicit memory test was based on previous research and made no reference to the brand names of products to which subjects were initially exposed (Krishnan and Chakravarti 1999; Schacter 1987). Subjects were given a word-completion task and asked to complete each of the word-stems (e.g., ESC_____) with the first word that came to mind. The first three letters of the four targets were mixed in with twelve distracter stems to disguise the purpose of the task. The targets were placed in the third and tenth positions out of sixteen possible positions. This ensured sufficient spacing between the target words.

     
  3. 3

    Check for test awareness: In order to ensure that subjects were not using voluntary retrieval strategies (Lee 2002), subjects were asked the following specific questions to assess whether they noticed any link between the stem-completion task and any of the previous tasks: (1) What do you think was the purpose of the tasks that you just completed? (2) What was your general strategy in completing these tasks? (3) Did you notice any relation between the brand-product rating task and the word-completion task?

     
  4. 4

    Recall test: Subjects were given a word-completion task and asked to complete each of the word-stems (e.g., ESC_____). The instructions asked subjects to think back to the brand names they had seen earlier, and complete the stems with brand names they remembered. The first three letters of the four targets were mixed in with twelve distracter stems to disguise the purpose of the task. The targets were placed in the third and tenth positions out of sixteen possible positions. This ensured sufficient spacing between the target words.

     
  5. 5

    Brand association: Similar to the pre-test, subjects generated as many associations as possible for each of the target brand names (manipulation check). They were given specific instructions to think of the brand name after writing each association to avoid response chaining effects. Response chaining is the process by which subjects might generate linkages from earlier associations rather than from the brand name; it is generally addressed by having subjects think of the brand name after each association has been generated.

     
  6. 6

    Manipulation check: Manipulation check for brand name frequency asked subjects to evaluate the distinctiveness of brand names using a 1 (not distinctive) to 7 (extremely distinctive) scale, while, “Initial exposure to the {brand name} was similar to a typical print ad,” on a 1 (strongly disagree) to 7 (strongly agree) was used to check the context manipulation.

     

Results

Manipulation checks

The data were initially checked to see whether any of the subjects had indicated awareness of the implicit memory measure. This check was based on the three questions immediately following the implicit memory test (step 3 above). All of the responses, except one, indicated that subjects completed the word stems with the first response they could think of when they read the word stem. Only one subject thought that there was some relation between the two tasks. This subject was dropped from all further analyses.

Analysis of the manipulation check measures showed that all three manipulations were successful. For both products, ANOVA with association set size as the dependent variable revealed only a main effect for association set size (Anti-perspirant: Mlarge = 6.04, Msmall = 5.56, F1, 146 = 4.86, p = 0.029; Cereal: MLarge = 6.45, Msmall = 4.65, F1, 146 = 17.47, p = 0.00). Similarly, low frequency words used as brand names (Mlow = 4.67) were perceived as more distinctive (on a scale of 1 = not distinctive and 7 = extremely distinctive) than high frequency words used as brand names (Mhigh = 2.32; t1, 152 = 14.89, p = 0.00). Finally, subjects in the brand info condition (Mbrand info = 5.47) had higher agreement compared to those in the no-brand info condition (Mno-brand info = 1.95; t1, 152 = 23.87, p = 0.00) that the exposure was similar to a typical ad.

Brand recall

A full factorial ANOVA with association set size, word frequency, and brand info as independent variables and recall as the dependent variable showed a marginally significant main effect for context (Mbrand info = 1.49 for and Mno-brand info = 1.30 for, F1, 146 = 3.31, p = 0.07, eta squared = 0.02), a significant main effect for word frequency (Mhigh = 1.51 and Mlow = 1.28, F1, 146 = 4.95, p = 0.028, eta squared = 0.03), and for association set size (Msmall = 1.57 and Mlarge = 1.22, F1, 146 = 11.41, p = 0.00, eta squared = 0.07). None of the two-way interactions were significant (p > 0.8). The main effects were qualified by a significant three-way interaction (F1, 146 = 8.21, p = 0.005, eta squared = 0.05), supporting H1c. Consistent with predictions (H1a), the interaction patterns in the brand info condition replicated previous research (Meyers-Levy, 1989); small association set size (M = 1.90) led to significantly higher recall than large set size (M = 1.30, F1, 146 = 8.87, p = 0.003) for high frequency words used as brand names, while there were no differences for low frequency words used as brand names (Msmall = 1.40, Mlarge = 1.35, F1, 146 = 0.06, p = 0.80). The interaction patterns in the no-brand info condition supported H1b (see Fig. 1); small association set size (M = 1.53) led to significantly higher recall than large set size (M = 0.84), F1, 146 = 10.95, p = 0.001) for low frequency words used as brand names, while there were no differences for high frequency words used as brand names (M small = 1.39, M large = 1.44, F1, 146 = 0.07, p = 0.79).
https://static-content.springer.com/image/art%3A10.1007%2Fs11747-009-0175-8/MediaObjects/11747_2009_175_Fig1_HTML.gif
Figure 1

Study 1: recall scores.

Implicit memory

All three main effects were significant. Implicit memory was higher in the brand info condition (Mbrand info = 1.34 and Mno brand info = 1.09, F = 5.21, p = 0.024), for high frequency words (Mhigh = 1.45 and Mlow = 0.97, F = 19.49, p = 0.00), and large set size (Mlarge = 1.39 and Msmall = 1.03, F = 10.81, p = 0.001). All three main effects are consistent with a familiarity based process, because large set sizes, high frequency words, and brand information can all enhance the familiarity of a brand name. These main effects were qualified by a significant interaction between word frequency and set size (F = 8.01, p = 0.005). Consistent with H2b predictions, for high frequency brand names, large set size (M = 1.79) led to higher implicit memory than small set size (M = 1.12; t = 4.97, p = 0.00); for low frequency brand names no differences emerged between large and small set sizes (t = 0.291, p = 0.772) (see Fig. 2). Note that there were no differences in the interaction patterns between the brand info and the no-brand info groups, consistent with expectations. None of the other interactions were significant.
https://static-content.springer.com/image/art%3A10.1007%2Fs11747-009-0175-8/MediaObjects/11747_2009_175_Fig2_HTML.gif
Figure 2

Study 1: mean value for implicit memory.

Discussion

Study 1 examined the role of context as a moderator of the impact of brand name characteristics on memory. The study is highly relevant considering the diverse kinds of exposure to brand names afforded by nontraditional media. The results show that the context in which brand name exposure occurs can lead to differences in recall. When exposure takes place in a brand info context, we replicate previous research (Meyers-Levy 1989) using print media, thus reinforcing its generalizability. The results from this context suggest that words used as brand names with high frequency and small association set size perform better. Yet, when exposure occurs in a no-brand info context, brand names with low frequency and small association set size perform better on recall. Our theory suggests that this combination of factors makes the brand stand out in these conditions, consistent with a distinctiveness path. Therefore, when consumers are exposed to brand names in nontraditional media, this combination of characteristics is more valuable from a manager’s standpoint. Also, the results for implicit memory where high frequency and large set size lead to better retrieval underscore the relevance of a familiarity path.

Thus Study 1 contributes by (a) replicating a previous finding of brand name characteristics on memory, (b) developing theory and demonstrating empirically that these effects are different in a nontraditional context, (c) extending the research into implicit memory effects, wherein a familiarity based process appears to be prevalent, and (d) identifying and empirically demonstrating two separate paths to higher memory retrieval. However, Study 1 focused purely on memory outcomes and did not consider how consumers use the information. From both a theoretical and practical standpoint it is important to consider the effects of brand name characteristics on processes closer to actual purchase such as brand consideration and choice. Study 2 was set up to replicate H1 and H2 and to test effects on brand consideration and choice.

Study 2

Hypotheses

Consumer choice behaviors appear consistent with a multi-stage choice process (Erdem and Swait 2004; Chakravarti and Janiszewski 2003; Desai and Hoyer 2000; Shapiro et al. 1997). Brands accessed from memory are pared into a smaller set of brands (consideration set), which are further evaluated during choice (Nedungadi 1990). Variables that affect memory (recall and implicit memory) may influence brand consideration and choice differently (Lynch et al. 1988). Therefore, another key objective of this research is to determine how brand name characteristics affect brand consideration and choice.

After a brand has been retrieved from memory, the consumer needs to evaluate the brand based on how much he/she knows about it. Two specific processes are particularly relevant, as identified in Study 1. First, based on previous exposure to the brand name as well as the amount of information that the consumer knows about the brand, consumers have enhanced feelings of familiarity. It is a well-established finding in the literature that familiarity based processes enhance liking, and are thus the basis for consideration and choice (Sinn et al. 2007; Heilman et al. 2000; Hoyer and Brown 1990). A second process is more relevant in cases that invoke more processing. In such cases, distinctive aspects of the brand may make it stand out from other brands; distinctiveness can help consumers justify why the brand should be considered and chosen over competitors. Overall, these two processes are expected to manifest in specific conditions as drivers of consideration and choice.

Let us now consider which of these processes will operate in various conditions. Low frequency words used as brand names are more distinctive, which can make them more memorable. Nedungadi’s (1990) research suggests that more memorable brand names, because of their ease of retrieval, have a higher probability of being included in the consideration set. When low frequency words used as brand names are coupled with a small association set, the ease of retrieval is enhanced and makes the brand more likely to be included in the consideration set. Further, consumers will attribute the ease of retrieval to the brand being distinctive and are more likely to choose the brand.

In contrast, high frequency words used as brand names are more familiar to consumers (Sinn et al. 2007). Brands with larger association set sizes can build on this familiarity even further by leveraging the existing associations. These familiarity processes are expected to work with lower levels of processing, potentially through implicit memory. The benefit to the brand is that enhanced familiarity leads to liking, which not only increases the probability of consideration but also impacts choice. Taken together, we can expect a brand frequency X association set size interaction as follows:

H3: Brand consideration will be higher when the brand name has a large versus small association set size for brand names that are high frequency words; also, brand consideration will be higher when the brand name has a small versus large association set size for brand names that are low frequency words.

H4: Consumers are more likely to choose a brand when the association set size is large versus small for a high frequency word; consumers are more likely to choose a brand when the association set size is small versus large for brand names that are low frequency words.

Method

Study 2 used a 2 (brand info versus no-brand info) X 2 (low versus high word frequency) X 2 (small versus large association set size) between subjects factorial design, replicated across four product categories as a within subjects factor. Student subjects were assigned to the same condition for all four product categories. Products were chosen that were relevant to this target segment (anti-perspirant, cereal, jeans, and chewing gum).

Pretest determined stimuli

Study 2 used the same brand names from Study 1 along with a similar procedure to select names for the two new product categories (jeans and chewing gum). The study used words representing hypothetical brands names as the stimuli in order to control for prior familiarity.

Based on similar pretests, four target brands were chosen for each product category—one for each of the four conditions (Frequency x Association Set Size). For jeans, the brand names STREET (large set size) and FORMS (small set size) were high frequency words while the brand names GLACIER (large set size) and EQUITY (small set size) were low frequency words. For chewing gum, the brand names OCEAN (large set size) and MOMENT (small set size) were high frequency words while the brand names NECTAR (large set size) and GLAZEE (small set size) were low frequency words. In addition, for each product, two control brand names were included to make the task more realistic (Sultan and Chief for jeans, and Effort and Winter for chewing gum); these had moderate scores on both association set size and word frequency. In summary, the pretests ensured that target brand names differed on word frequency or association set size, but were equivalent on appropriateness.

Procedure and measures

Two hundred and forty-five marketing students participated in the study to fulfill course requirements, resulting in a cell size of approximately 30 for each of the eight conditions. Subjects were assigned randomly to one of the conditions. All subjects received instructions that the purpose of the study was to evaluate potential brand names and that such tests were routinely performed by firms during the brand name selection process. The entire sequence took about 40 minutes.

In contrast to Study 1 which used print ads, Study 2 used press releases to manipulate brand info context. In the brand info context, for each product, press releases were created for each of the three brands (1 test brand and 2 filler brands) and the experimental condition was manipulated by changing the brand name. All subjects in the brand info context were exposed to a total of 12 press releases (three each for each of the four products) placed in a separate folder. Subjects were told that these were new press releases and asked to view/read them as they normally would if they were in a magazine and were given a total of 2 minutes (see “Appendix 2” for sample). As can be seen, the press release contains other supporting information (e.g., product attributes) that would normally be seen in a typical press release.

The procedure (exposures, recall/implicit memory measurement, manipulation checks for association set sizes and word frequency) were exactly identical to Study 1 and are not repeated here. Study 2 had additional measures for brand consideration (e.g., I am likely to consider EFFORT when I am making a CHEWING GUM purchase, on 1 (strongly disagree) to 7 (strongly agree) scale), and brand choice for each product category (subjects were required to allocate 100 points to the target brand and the two filler brands), as well as manipulation check measures for the context manipulation.

Results

Manipulation checks

The data were initially checked to see whether any of the subjects had indicated awareness of the implicit memory measure. This check was based on the three questions immediately following the implicit memory test (step 3, Study 1). All of the responses, except five, indicated that subjects completed the word stems with the first response when they read the word stem. The five subjects who thought that there was some relation between the two tasks were dropped from all further analyses.

Analysis of the manipulation check measures showed that all three manipulations were successful. As there were no differences across the four products used in the study, the average number of associations were computed and used as the independent variable with association set size, word frequency, and context as independent variables. Only a main effect for set size (Mlarge = 4.33, Msmall = 3.56, F1, 187 = 20.09, p = 0.00) was significant. Similarly, low frequency words used as brand names (Mlow = 5.25) were perceived as more distinctive (measured on 1 = not distinctive and 7 = extremely distinctive) than high frequency words used as brand names (Mhigh = 4.01; F1, 232 = 47.19, p = 0.00). Finally, subjects in the brand info condition (M = 5.31) had higher agreement (1 = strongly disagree, 7 = strongly agree) compared to those in the no-brand info condition (M = 2.88; F1, 223 = 62.20, p = 0.00) that “I was given information about the attributes of the brand.”

Brand recall

A full factorial ANOVA with association set size, word frequency, and context as independent variables and recall as the dependent variable showed a significant main effect for context (Mbrand info = 2.49 and Mno brand info = 2.22, F1, 232 = 7.96, p = 0.005, eta squared = 0.03), a significant main effect for word frequency (Mhigh = 2.47 and Mlow = 2.25, F1, 232 = 5.12, p = 0.025, eta squared = 0.02), and association set size (Msmall = 2.52 and Mlarge = 2.20, F1, 232 = 11.23, p = 0.001, eta squared = 0.05). None of the two-way interactions were significant (p > 0.17). The main effects were qualified by a significant three-way interaction (F1, 232 = 16.05, p = 0.00, eta squared = 0.07), supporting H1c. Consistent with predictions (H1a), the interaction patterns in the brand information condition replicated Study 1 and previous research (Meyers-Levy 1989); small association set size (M = 3.03) led to significantly higher recall than large association set size (M = 2.31), t58 = 3.59, p = 0.001) for high frequency words used as brand names, while there were no differences for low frequency words used as brand names (Msmall = 2.27, Mlarge = 2.38, t60 = 0.501, p = 0.62). The interaction patterns in the no-brand info condition supported H1b (see Fig. 3) and replicate Study 1; small association set size (M = 2.53) led to significantly higher recall than large association set size (M = 1.83), t 56 = 4.18, p = 0.000) for low frequency words used as brand names, while there were no differences for high frequency words used as brand names (M small = 2.26, M large = 2.28, t 58 = 0.10, p = 0.92).
https://static-content.springer.com/image/art%3A10.1007%2Fs11747-009-0175-8/MediaObjects/11747_2009_175_Fig3_HTML.gif
Figure 3

Study 2: recall scores.

Implicit memory

All three main effects were significant. Implicit memory was higher in the brand info condition (Mbrand info  = 1.62 and Mno brand info = 1.33, F1, 232 = 5.14, p = 0.02, eta squared = 0.02), for high frequency words (Mhigh = 1.69 and Mlow = 1.26, F1, 232 = 11.30, p = 0.001, eta squared = 0.045), and large set size (Mlarge = 1.60 and Msmall = 1.35, F1, 232 = 3.98, p = 0.04, eta squared = 0.02). These main effects were qualified by a significant word frequency X set size interaction (F1, 232 = 4.05, p = 0.045, eta squared = 0.02). Consistent with H2b predictions and replicating Study 1, for high frequency brand names, large set size (M = 1.95) led to higher implicit memory than small set size (M = 1.44; t 118 = 3.08, p = 0.003), whereas for low frequency brand names no differences emerged between large (M = 1.29) and small set sizes (M = 1.25, t 118 = 0.184, p = 0.855) (see Fig. 4). None of the other interactions were significant.
https://static-content.springer.com/image/art%3A10.1007%2Fs11747-009-0175-8/MediaObjects/11747_2009_175_Fig4_HTML.gif
Figure 4

Study 2: mean value for implicit memory.

Brand consideration

The main effect for context was significant (F1, 219 = 13.86, p = 0.00). When brand information was present (M = 5.16), consumers were more likely to consider the brand compared to when brand information was not present (M = 4.61). The frequency X set size interaction was significant (F1, 219 = 38.64, p = 0.00). When frequency was high, large set size (M = 5.21) led to significantly higher consideration than small set size (M = 4.34, t 112 = 3.81, p = 0.00). When frequency was low, small set (M = 5.45) led to significantly higher consideration than large set size (M = 4.53, t 111 = 4.60, p = 0.00). Hence H3 was supported (see Figs. 5 and 6).
https://static-content.springer.com/image/art%3A10.1007%2Fs11747-009-0175-8/MediaObjects/11747_2009_175_Fig5_HTML.gif
Figure 5

Study 2: mean value for brand consideration.

https://static-content.springer.com/image/art%3A10.1007%2Fs11747-009-0175-8/MediaObjects/11747_2009_175_Fig6_HTML.gif
Figure 6

Study 2: mean value for brand choice.

Brand choice

Association set sizes had a significant effect on brand choice. Brand names with large set sizes (M = 44.46) were significantly more likely to be chosen than those with small set sizes (M = 37.65, F1, 223 = 19.08, p = 0.00). The frequency X set size interaction was significant (F1, 223 = 19.95, p = 0.00), consistent with H4. When frequency was high, brand names with large set size (M = 47.65) was significantly more likely to be chosen than those with small set size (M = 33.78, t115 = 5.75, p = 0.00). When frequency was low, there were no differences between brand names with large (M = 40.87) and small set sizes (M = 41.25, p > 0.8), contrary to expectations.

Discussion

Study 2 replicates the effects found in Study 1 for implicit memory/recall using a different brand information context, namely press releases. This study adds to the external validity of the findings for implicit memory and recall. Study 2 also extends our understanding of the impact of brand name characteristics to brand consideration and choice. It can be seen that brand consideration and choice are affected due to different processes in two conditions: low frequency/small set size, and high frequency/large set size. The former conditions parallel the recall effects and suggest that enhanced distinctiveness drives consideration and choice. In contrast, the combination of high frequency/large set size impacts consideration and choice via familiarity processes, thereby validating the importance of the implicit memory effects.

Discussion

Research contributions

The goal of this research was to: (a) study the effects of brand name characteristics on memory when brand information is present versus absent; (b) extend these brand name effects to implicit memory; (c) examine their effects on brand consideration and choice, and d) identify and empirically demonstrate two separate paths to superior memory retrieval and choice.

First, the results support the basic premise of differences between the two brand exposure contexts. The pattern for recall in the no-brand info context differs from that found in the brand info context, suggesting that consumers react differently to brand names in a no-brand info situation. In a brand info context, recall was higher for high frequency brand names with small (versus large) association set sizes. In contrast, in the absence of brand info, recall was higher for low frequency brand names with small (versus large) association set sizes. These results suggest that a brand name needs to be distinctive in its specific context, in order to be remembered by consumers. Hence, if a firm has chosen a low frequency brand name with small association set sizes, it may be appropriate to launch the brand with a stronger emphasis on non-traditional forms of promotion (Internet and point of purchase displays).

Second, the implicit memory results extend understanding to memory tests that do not rely on intentional retrieval processes. Given the different patterns for recall and implicit memory in this research, these seem to be tapping different memory processes. Such effects may be particularly important in low involvement situations where implicit memory may play a larger role. In these retrieval situations, it appears that larger association set size is actually desirable for high frequency brand names because it adds to the familiarity level of the brand. This implies that brand name selection should also consider the nature of the choice process for a particular category and situation. If consumers make quick judgments based on familiarity of brand names, greater association set sizes may actually be desirable to prompt automatic retrieval processes. This provides an interesting contrast to the recommendations for recall, where larger association set sizes interferes with retrieval.

Third, the impact of the brand characteristics extends to brand consideration and choice, underlining the importance of these factors in screening potential brand names. There seem to be two processes driving these results. First, unique and brand-relevant information engages a higher level of processing, thereby making the brand more distinctive and facilitating consideration and choice. Second, in the absence of a higher level of processing, familiarity based processes are at work (Hoyer and Brown 1990), which influences consideration and choice via an automatic process. Understanding these two processes is critical given that they not only influence retrieval but also impact the brand choice process.

These memory and choice related findings have important implications for managing high versus low frequency brand names. For product categories where low-involvement retrieval, consideration and choice may be the norm (e.g., paper towels), it may be important to focus on some form of classical conditioning where the brand name is associated with a specific usage situation or attribute. Brand names that are high frequency words may be more desirable because familiar brand names "jump out" (Alba et al. 1991). In order to maintain high levels of familiarity and awareness, managers may focus on a set of consistent associations for these brands. On the other hand, when consumers have high involvement for a product (e.g., cars), managers should consider highlighting the distinctive qualities of the brand by using relevant information about the brand. In such a situation, a brand name that is a low frequency word may motivate consumers to process it in depth. This is likely to be important in product categories where deliberative recollection precedes choice (Krishnan and Shapiro 1996).

Limitations

This paper replicates and extends previous research on brand names (Meyers-Levy 1989) to two different contexts but with different stimuli (products and brand names) and different media (advertising and press release). Given that the results for the new context differed from earlier research, this highlights the importance of extending previous results to other contexts to enrich understanding of brand name effects.

The conclusions must be tempered by the following limitations. Both studies used student subjects in lab experiments. While this is appropriate for theory testing, further research with other populations and field settings may be necessary after the core relationships have been understood. Further, hypothetical brand names were used in the study. Hence, generalizations should be limited to new brand names and results may not be applicable to established brands. Finally, the study results show that the effect sizes are fairly small. This is possibly due to the fact that the brand name is only one piece of information that affects consumer memory and choice, albeit an important one.

Managerial implications

The findings have important implications for managers regarding brand name selection. First, when advertising with product information is the primary means of planned communication for a new brand, names should be selected such that they have high word frequency and small set of associations. This will lead to higher recall of the brand name. Alternatively, if a brand name with high word frequency and small association set size has already been chosen, advertising with supportive brand related information should be the primary means of communication to increase brand recall. This would be similar to typical ad campaigns that provide product information before product launch. On the other hand, if a brand name with small association set size and low frequency has been chosen, managers have the option of designing the advertising communication with or without supporting information. This gives them the option of using either a typical introductory ad with product information or image based ads with limited product information, which can also lead to high brand recall.

Second, when brand managers have decided to introduce the brand with limited information or non-traditional forms of brand exposure (for example, Body Shop using word-of-mouth), the selection of the brand name can have a significant impact on consumer recall and implicit retrieval. Brand names with small association set sizes will be recalled better and so should be chosen in tasks where the consumer engages in effortful recall, typically the situation for a high involvement product. However, brand names with large association set sizes and high frequency should be selected for situations where consumers do not engage in conscious effort to retrieve the brand, typically the situation for low involvement products.

In conclusion, this research shows support for both distinctiveness and familiarity based processes in driving memory, consideration, and choice, as a function of brand name characteristics and exposure context. The prescriptions offered for managers in guiding selection of brand names vary depending upon which of these processes is likely to be implicated.

Acknowledgements

This research was made possible by a grant from the Indiana University Kelley School of Business to the second author. We acknowledge the helpful comments of Richard Olshavsky, Carol Pluzinski, and Anand Kumar.

Copyright information

© Academy of Marketing Science 2009