How Preferred Brands Relate to the Self: The Effect of Brand Preference, Product Involvement, and Information Valence on Brand-Related Memory

Feng, Rui; Ma, Weijun; Liu, Ruobing; Zhang, Miao; Zheng, Ziyi; Qing, Ting; Xi, Juzhe; Lai, Xinzhen; Qian, Cen

doi:10.3389/fpsyg.2019.00783

ORIGINAL RESEARCH article

Front. Psychol., 26 April 2019

Sec. Organizational Psychology

Volume 10 - 2019 | https://doi.org/10.3389/fpsyg.2019.00783

How Preferred Brands Relate to the Self: The Effect of Brand Preference, Product Involvement, and Information Valence on Brand-Related Memory

$\r\nRui Feng&#x;$ Rui Feng^1†

Weijun Ma^2*†

Ruobing Liu^2*

Miao Zhang²

Ziyi Zheng²

Ting Qing²

Juzhe Xi²

Xinzhen Lai²

Cen Qian²

¹Department of Marketing, School of Management, Shanghai Business School, Shanghai, China
²Shanghai Changning-ECNU Mental Health Center, School of Psychology and Cognitive Science, East China Normal University, Shanghai, China

This study adopted the paradigm of the self-reference effect to explore how brand preference, product involvement, and information valence affects brand-related memory by three experiments. Experiment 1 examined memory differences between positive/negative information of self-/other-preferred brands. Results showed increased memory of positive words (i.e., the effect of information valence) in the self-preferred brand group, yet memory of self-preferred brands was poorer than that of other-preferred brands. Experiment 2 examined effects of degree of brand preference and information valence, and revealed a positive association between degree of preference and memory of brand-related positive words. Experiment 3 explored the effects of brand preference and product involvement. Results showed that the memory of high-preference brands was stronger in the high-involvement group. Additionally, product involvement demonstrated a significant positive correlation with memory. The observed effects of information valence, especially in self-preference (Experiment 1) and high-preference (Experiment 2) conditions, can be explained by self-schema and mnemic neglect theories. The increased memory of highly preferred brands in a high-involvement condition can be explained by intimacy and self-expansion models (Experiment 3).

Introduction

At the mention of two different cellphone brands, such as Apple and Samsung, two different adjectives are likely to come to mind that describe the brands. The reason why different memories of various brands form, and exactly how these memories differ, remains to be elucidated. The positive or negative memories consumers hold regarding various products and brands strongly contribute their final purchase decisions. As self-schemas effectively influence memory, the choice making process of consumers is actually a process of seeking psychological identity and self-expression. This is similar to a concept presented by Aaker and Fournier (1995), stating that when people build a relationship with a brand, they actually view the brand as a character, a partner, and a person, and that they choose to build relationships with brands because such relationships help them develop a sense of who they are (Aaker and Fournier, 1995; Fournier, 1998).

The “self” is a remarkably important topic in social psychology. This is especially true of the self-reference effect (SRE; Rogers et al., 1977). While this effect has been widely applied in studies of interpersonal relationships, it has not been applied to research examining self-brand relationships. The current manuscript uses the Remember/Knowledge (R/K) paradigm of the SRE to determine how products and brands influence memory, exploiting application domains for the SRE and providing marketing strategy insight.

Self-Reference Effect (SRE)

The SRE (Rogers et al., 1977) refers to the phenomenon of better memory being observed in self-referential tasks than other tasks. The purported mechanism responsible for this effect is the more elaborate and organizational processing of self-related information when compared to other information. Previous research of this effect has followed two different paradigms, specifically, the classical SRE paradigm, and the R/K paradigm (e.g., Liu and Zhu, 2002). In the classical SRE paradigm, subjects rate adjectives on four tasks designed to force varying kinds of encoding: structural, phonemic, semantic, and self-referential, and then are asked to freely recall the rated words (e.g., Rogers et al., 1977). The R/K paradigm is more sensitive than the classical (e.g., Conway and Dewhurst, 1995), asking subjects to first identify old items in a recognition test, and then indicate which items are accompanied by a recollective experience (“Remember” responses) and which are recognized on some other basis (“Know” responses). The SRE is primarily shown in “R” responses, as previous research (Zhang et al., 2005) has verified that “R” responses reflect the SRE more effectively and more sensitively than “K” responses.

The Effect of Intimacy on SRE

A number of studies have consistently demonstrated that intimacy has an important effect on SRE. Specifically, referring to highly intimate others promotes a memory effect that is almost as strong as self-reference (Symons and Johnson, 1997; Zhou and Su, 2008; Xiong et al., 2012). This can be reasonably explained by the self-expansion model proposed by Aron and Aron (1986). The model suggests that human beings have a motivation to include other people, including their resources, concept, and identity, into the self-concept in an effort to increase their own self-efficacy (Aron and Aron, 1986; Aron et al., 1991; Jia and Shi, 2012), and that, to the degree that a relationship is intimate, it is likely to overlap with one’s self-representation. In other words, the higher the degree of intimacy, the higher the degree of self-inclusion, which leads to better memory effect.

The Effect of Information Valence on SRE

Rogers et al. (1977) pointed out that people had better memory of self-related trait adjectives than other words. Kuiper and Derry’s (1982) further study found that, in self-referential conditions, people’s memory of positive adjectives was enhanced when compared to negative adjectives. A study by Argembeau et al. (2005) also supported this notion, in that positive trait information was typically better recalled than negative trait information when encoded in reference to the self, but not when encoded in reference to someone else or when processed for general meaning. Green and Sedikides (2004) discovered that information that threatened positive aspects of the self (e.g., “I would be unfaithful in an intimate relationship”) was less likely to be remembered in self-referential conditions, but not under other conditions.

Self-schema theory can explain the effect of enhanced memory of positive self-referential adjectives. The self-schema theory raised by Markus (1977) points out that self-schemata are cognitive generalizations about the self, derived from past experience, that organize and guide the processing of the self-related information contained in an individual’s social experience. For example, one study found that self-schemata facilitated the processing of information about the self, contained easily retrievable behavioral evidence, and provided a basis for the confident self-prediction of behavior on schema-related dimensions (Markus and Wurf, 1987). This also explains the general mechanism of SRE to some extent, because individuals are more sensitive to self-related information. Liao and Chen (2006) found that self-schemata influenced mental health. While patients with depression had underlying negative self-schemata, psychologically healthy people demonstrated rather positive self-schemata, such as positive self-concept and high self-esteem. Self-esteem was also found to influence memory, as people with high self-esteem tend to show greater preference for self-related positive information (He and Qin, 2009). Since self-related positive information results in elaborate processing from one’s self-schemata, the schemata activate an information network to more effectively extract the memory in recognition phases, leading to enhanced memory.

In addition, the mnemic neglect model (Sedikides and Green, 2000; Sedikides and Green, 2004) can also explain differences in memory of positive and negative words in the SRE. The model postulates a two-stage sequence of processing self-referential feedback. In Stage 1, the individual appraises the plausibility of enacting the behavior based upon general self-knowledge. That is, the individual compares the behavior to semantic information about the self. Since the self is usually a positive, complicated, and motivated structure (Gaertner et al., 2002; Sedikides et al., 2003), if the feedback is threatening to the self, processing is largely confined to Stage 1 and does not proceed to Stage 2. In Stage 2, the behavior is further compared to relevant and specific self-knowledge. This leads to negative information “neglected” at Stage 1. From the point of view of self-protection, the model is essentially selective memory that ignores negative self-related information and pays more attention to the positive information (Green and Sedikides, 2004, 2006; Chen and Zhao, 2009). Therefore, according to the model, subjects find it difficult to recall negative words because they have been exposed to less processing.

Brand Preference and Product Involvement

While many empirical studies have been performed in the field of interpersonal memory, very little research has been conducted examining brand memory. It is possible that SRE paradigms can also be applied to the study of brand-related memory, in which brand preference and product involvement can play the same role intimacy plays in the SRE.

Crites et al. (1994) defined preference as revealed attitudes and tendency toward certain objects. Brand preference can then be understood as the positive evaluation, degree of preference, and purchase predisposition of a specific brand (Zhang and Lei, 2011). This can be influenced by the self, as the self ultimately shapes attitudes. For example, Ji (2008) found that the higher the degree of matching between a brand personality and the consumer personality, the higher the degree of involvement and the higher the degree of brand preference. Although such studies have combined conceptualizations of the self with brand preference, they have remained narrow in scope, focusing solely on the domain of the impact of the self on brand preference, but lacking more profound and extensive exploration.

The previously discussed concept of involvement can be defined as the perceived relevance of an object based on inherent needs, values, and interests (Zaichkowsky, 1985). Any type of product-related information can cause involvement between consumers and products, such as advertisements or product packaging. Consumption behaviors are intricately interwoven with involvement, and, once consumers have formed a stable involvement relationship with a product, they will perform both purchasing and impulsive consumption behaviors.

It is regrettable that the effects of brand preference and product involvement on the brand-related memory remain unexplored. Hence, the current study aimed to fill the research gap by adapting the R/K SRE paradigm to study brand-related memory.

Research Purposes and Hypotheses

Research Purposes

In modern life of consumption, the self and brands have a close connection (Angle and Forehand, 2016). The SRE has been applied to many interpersonal relationship studies, such as research of the mother-reference effect (Symons and Johnson, 1997; Zhou and Su, 2008; Xiong et al., 2012), but it rarely appears in research regarding memories of brand information. Zhou et al. (2012) his found that free recall scores of self-owned objects were significantly higher than other-owned objects, showing the ownership effect (i.e., a virtual ownership reaction that occurs merely by use of pronouns such as “my” and “his”), impacted memory. In the present research, the self was combined with brand preference and product involvement to explore the relationships among these factors.

The overall goal of the current research was to provide evidence that brand preference, degree of brand preference, and product involvement would each demonstrate an impact on the memory of relevant product information by using the R/K paradigm of the SRE. Additionally, the current series of studies attempted to test differences in memory of brand-related positive and negative information under the influence of brand preference and product involvement.

Research Hypotheses

Memory of Brand-Related Information

According to the self-expansion model (Aron and Aron, 1986; Aron et al., 1991), the self has a tendency to include preferred brands and other high-involvement products into the self-concept. Analogous to the SRE, individuals should demonstrate a memory advantage for information of preferred brands and high-involvement products. Further, memory of positive and negative words associated with these, and other, products should differ.

The effect of brand preference on memory

This research predicted an impact of brand preference on memory of related information. According to Aron’s self-expansion model, people include other people into the self (Aron and Aron, 1986; Aron et al., 1991; Jia and Shi, 2012). It has been suggested that brands are also likely to be included within this framework (Reimann and Aron, 2009). The more self-expansion people feel toward a brand, the more loyal they will be to it. As a result, based on the self-expansion model, highly preferred brands will receive more self-expansion, thus leading to the better memory of relevant information. This effect would be similar to that observed in intimacy.

The effect of product involvement on memory

Ji (2008) found that when consumers perceived that a product helped to achieve their ideal image, involvement with the product would increase. Similarly, Chen (2007) also pointed out that if the connection between people and a product became stronger, involvement levels would also rise, leading to various behaviors, such as caring about the product and searching for relevant information. This then results in a stronger emotional investment with the product. With high-involvement products, consumers spend more time comparing among different markets, and when it came to low-involvement products, the time was saved (Clarke and Belk, 1979). High-involvement products are important to consumers and require a large amount of information processing, while low-involvement products are not very important, thus requiring less information processing. Therefore, the elaborately processed information of the former will be more easily extracted during recognition phase than the information processed for the latter.

The interaction between brand preference and product involvement

Since intimacy, according to the self-expansion model, has a strong effect on the SRE in relationships among different people, it is possible that brand preference and product involvement will have a similar effect on memory in the relationship between people and objects. Based upon the self-expansion model (Aron and Aron, 1986), high-involvement products will be more deeply included into the concept of the self, leading to more elaborate processing of relevant information. This then suggests that brand preference will have a significant influence on memory under this condition, with the memory of high-preference brands being better than the one of low-preference brands.

Similarly, low-involvement products will be more weakly included into the self-concept. Because of this, relevant information will receive less processing, weakening the influence of brand preference on memory under this condition, presenting as no memory difference between the information of high- and low-preference brands.

The effect of information valence on memory

Being analogous to the difference in memory effects between positive and negative words in the SRE (Rogers et al., 1977), the current study suggests that individual’s memory of positive and negative words of products with different levels of product involvement and brand preference will be different, and that this effect can be explained by self-schema theory and the mnemic neglect model. In high-preference brands, individuals will form positive schemata and selectively ignore negative information, thus, in this condition, the memory of positive information for high-preference brands should be significantly better than the memory of negative information. In low-preference brands, memory of positive and negative information should demonstrate no difference. Furthermore, this result should be more noticeable in high-involvement products, as the relevant information will likely receive more complex processing.

Overview of the Current Experiments

To test the impact of brand preference, product-involvement, and information valence on memory, three experiments were performed. Experiment 1 used the R/K paradigm of SRE as the method and shampoo as the material to compare the memory effects of self-preferred and other-preferred brands. Additionally, the impact of information valence was also measured. It was predicted that the memory of self-preferred brands would be enhanced and that positive information of the brand would demonstrate enhanced memory when compared to negative information.

Experiment 2 also used the shampoo as material, exploring the impact of information valence and the different impacts of high, medium, and low brand preference on memory. It was predicted that a higher degree of brand preference would lead to better memory, and that memory differences between relevant positive and negative information would be even more obvious.

Experiment 3 added the variable of involvement to explore the impact of involvement and brand preference on memory. Specifically, this experiment used shower gels and computers as materials of low and high product involvement, respectively. It was predicted that information regarding the preferred brand with higher product involvement would demonstrate the best level of memory, with a greater difference between positive and negative words.

Experiment 1: The Effect of Self- and Other-Preferred Brands and Valence of Relevant Information on Memory

This experiment intended to examine whether memory of a self-preferred brand would be enhanced when compared to the memory of an other-preferred brand. Additionally, this experiment explored whether the valence of words (i.e., positive or negative) would show differing effects depending on the preference category.

Experimental Hypotheses

Individuals have a tendency to incorporate preferred brands into their self-concepts (Aron and Aron, 1986). Therefore, considering this information and the SRE, it was hypothesized that individuals would demonstrate a memory advantage for information of these preferred brands (i.e., brands incorporated into the self) through elaborate and organizational processing.

It was also hypothesized that the memory of positive and negative information of these preferred brands would differ. According to self-schema theory (Markus, 1977), one of the most common self-schemata is maintaining positive self-image. This schema activates a “net” to extract positive information when forming memories. In addition, the mnemic neglect model (Sedikides and Green, 2000; Sedikides and Green, 2004) indicates that people will ignore self-related negative messages. As a result, the memory of positive trait adjectives is better than the memory of negative adjectives when referencing the self, but not when referencing others. Therefore, people will likely also hold a positive schema for preferred brands and ignore their negative aspects, leading to enhanced memory of positive words describing self-preferred brands.

In summary, the hypotheses of Experiment 1 were:

Hypothesis 1a: The overall recognition and R rates of self-preferred brand information will be significantly higher than that of the other-preferred brand;

Hypothesis 1b: Subjects’ brand preference will demonstrate a significant positive correlation with the memory of the product’s relevant information;

Hypothesis 1c: The recognition and R rates of positive information will be significantly higher than negative information in the group of self-preferred brands, and no difference will appear in the group of other-preferred brands.

Methods

Participants

A total of 33 participants (17 males, M_age = 21.9 years, SD_age = 1.24) with normal or corrected to normal vision were recruited from universities in Shanghai.

Experimental Design

The experiment was a 2 (referential task: self-preferred brand vs. other-preferred brand) × 2 (information valence: positive vs. negative) within-subjects design. Dependent variables were recognition and R/K rates.

Materials

The materials of Experiment 1 consist of questionnaire and adjectives. Please see Appendix A for details.

Questionnaire

Shampoos are commonly used objects in daily life, which are familiar to people and exhibit a high degree of market concentration (Gu, 2008). Therefore, a questionnaire targeting shampoo brands (e.g., Pantene, Rejoice) was designed to collect data. The questionnaire contained two parts. The first was to confirm and evaluate a self-preferred brand. In this section, subjects were asked to choose their favorite shampoo brand and evaluate it on two dimensions (preference and similarity), with each item ranked on a seven-point, Likert-type scale (1 = strongly disagree, 7 = strongly agree). The second part was to determine and evaluate an other-preferred brand on the same two dimensions. Specifically, after finishing the first part, subjects were told that their preferred brand was different from the former participant’s choice, and were asked to guess the former’s favorite brand and evaluate it on the same two dimensions.

Adjectives

All adjectives used in the paradigm to describe brands were chosen from the Research group of Modern Chinese Glossary of Common Words (2008), half of which were positive and the other half were negative. Words were divided into four groups, balanced in terms of the valence, frequency, and phonetic sequence. Each group had 28 words (14 positive and 14 negative), totaling 112 words in all four groups. Two groups of words were randomly chosen to describe the self- and other-preferred brands in the learning phase, and the other two groups of words were defined as new words in the recognition phase. Word order of appearance was randomly generated by E-prime (E-prime is a psychological experimental operation platform which is an advanced graphic design environment and a cross-platform system to realize computerized behavior research).

Experimental Procedure

After finishing the questionnaire, subjects completed the R/K paradigm experiment. The experiment was conducted in three phases: learning phase, interference phase, and testing phase. The first and third phases were completed on E-prime 2.0. The instructions of learning phase and testing phase are showed in Appendix A. The entire experimental protocol lasted for approximately 30 min.

Learning phase

Subjects were told that this was an adjective evaluation experiment, and they were to complete two kinds of judging tasks by answering the questions “Is the word xxx (e.g., ‘popular’) appropriate to describe the xxx (self/other-preferred brand chosen by participant)?” Each task involved 28 adjectives (14 positive and 14 negative), for a total of 56 adjectives. According to each participant’s own choices of self- and other-preferred brands in the questionnaire, brand names were modified in E-prime by the experimenter and appeared in the experimental instructions. The appearances of these two tasks were balanced. At the beginning of each task, there were detailed instructions, telling that after the presentation of a fixation point and a blank screen lasting for 500 ms, each adjective would be presented for 2 s. Participants’ reactions were not recorded, as the purpose of this phase was to allow them to learn the words.

Interference phase

At the end of the learning phase, subjects were given a 3-min break. Then they were asked to perform 64 mathematical calculations to avoid repetition of the words that appeared in the learning phase. This phase lasted for 7–9 min.

Testing phase

After the interference phase, subjects were asked to complete recognition tasks. They were randomly presented 56 old and 56 new words (half positive and half negative). First, subjects were asked to determine if words were “new” or “old.” When words were deemed “old,” participants were asked to judge whether they exactly “Remembered” the words or just “Knew” the words. There was no fixed presentation time for each word, as words were switched when participants pressed a button. Before each word, a fixation point and blank screen was presented for 500 ms.

Measures

Dependent variables (i.e., memory results) were the “Recognition rate” (the number of words correctly judged as “old”)/56 (the total number of words in the learning phase), the “R rate” (number of “Remember” words)/56, and the “K rate” (number of words subjects “Know”)/56.

Results

This study used SPSS 23.0 to analyze data in the questionnaire, recognition rates, and R/K rates.

Manipulation Checks

A paired-sample t-test was performed to test the preference and familiarity of self-/other-preferred brands. Results indicated a significant difference in degrees of preference [t(32) = 5.82, p < 0.001, www.frontiersin.org = 0.51] ( is the effect size reflecting the percentage of variation in the dependent variable which can be explained by the independent variable in t-test). The preference of the self-preferred brand (M = 6.12, SD = 0.78) was much higher than the other-preferred brand (M = 4.85, SD = 1.13). The familiarity of the self-preferred brand (M = 5.12, SD = 1.14) was also higher than the other-preferred brand (M = 4.73, SD = 1.33), but not significantly [t(32) = 1.51, p > 0.05, www.frontiersin.org = 0.07]. Therefore, the self-preferred brand and the other-preferred brand were well distinguished on degree of preference.

Moreover, recognition rates of new words were calculated by using the number of the new words that were misjudged as “old” by participants. A within-subject one-way analysis of variance (ANOVA) performed on the recognition rates of new words and old words revealed a significant effect [F(1,35) = 433.22, p < 0.01]. Recognition rate of old words (M = 0.60, SD = 0.15) was higher than new words (M = 0.24, SD = 0.11), indicating an effective manipulation of old and new words.

Recognition Rate

Descriptive statistics of recognition rates and R/K rates of different referential task and different information valence are presented in Table 1. A 2 (referential task: self-preferred brand vs. other-preferred brand) × 2 (information valence: positive vs. negative) repeated measures ANOVA on recognition rates revealed a significant main effect of referential task [F(1,32) = 10.90, p = 0.002, η² = 0.25]. The recognition rate of self-preferred brands (M = 0.56, SD = 0.17) was lower than that of other-preferred brands (M = 0.63, SD = 0.15), a result in direct opposition of hypothesis 1a. The main effect of information valence was not statistically significant [F(1,32) = 0.73, p = 0.40, η² = 0.02].

TABLE 1

Table 1. The M (SD) of the recognition rate and the R/K rate of different referential tasks and information valence.

The interaction between referential task and information valence was significant [F(1,32) = 4.46, p = 0.04, η² = 0.12]. In the self-referential condition, there was a marginally significant difference between the memory of positive words and negative words [F(1,32) = 2.98, p = 0.09, η² = 0.09], indicating that the memory of the former was enhanced. In other-referential condition, there was no statistically significant memory difference between two kinds of words [F(1,32) = 0.13, p = 0.72, η² = 0.004].

R Rate

A 2 (referential task: self-preferred brand vs. other-preferred brand) × 2 (information valence: positive vs. negative) repeated measures ANOVA on R rates also revealed a significant main effect of referential task [F(1,32) = 14.13, p = 0.001, η² = 0.31]. The R rate of self-preferred brands was lower than the one of other-preferred brands. The main effect of information valence in this analysis was also non-significant [F(1,32) = 2.23, p = 0.15, η² = 0.07].

The interaction between referential task and information valence was significant [F(1,32) = 5.33, p = 0.03, η² = 0.14]. For a further simple effect analysis, we fixed the referential task and analyzed information valence. In the self-referential condition, memory of positive words was significantly better than the memory of negative words [F(1,32) = 5.85, p = 0.02, η² = 0.16]. In the other-referential condition, there was no significant memory difference between the two kinds of words [F(1,32) = 0.04, p = 0.84 η² = 0.001].

K Rate

A 2 (referential task: self-preferred brand vs. other-preferred brand) × 2 (information valence: positive vs. negative) repeated measures ANOVA on K rates found no significant main effects of referential task or information valence [F_{referentialtask} (1,32) = 0.54, p_{referentialtask} = 0.47, η²_{referentialtask} = 0.02; F_{informationvalence} (1,32) = 0.31, p_{informationvalence} = 0.58, η²_{informationvalence} = 0.01]. The interaction between referential task and information valence was also non-significant [F(1,32) = 0.09, p = 0.77, η² = 0.003].

Correlations Between Brand Preference and Memory

Correlations among the degree of brand preference, recognition rate, R rate, and K rate are reported in Table 2. The degree of brand preference had a marginally significant negative correlation with the overall recognition rate (r = -0.21, p < 0.10) and a significant negative correlation with the R rate of negative words (r = -0.26, p < 0.05), indicating that the higher the brand preference, the lower the overall recognition rate, and the R rate of negative words decreased along with an increase of brand preference. Additionally, the memory of positive words was not significantly related to brand preference.

TABLE 2

Table 2. Correlations (r) among preference, recognition rate, and the R/K rate of positive and negative words.

Discussion

The results of Experiment 1 showed that memory effects of shampoo brand information were different under two different referential conditions, an effect in which information valence plays a role. ANOVA revealed that the memory of self-preferred brands was worse than that of other-preferred brands on both overall recognition and R rates. Correlation analysis showed that brand preference demonstrated a negative correlation with R rate, which was consistent with the ANOVA results, and failed to support hypotheses 1a and 1b. Overall, in recognition and R rates, the memory of positive words was better than negative words in the self-referential condition and not in other-referential condition, supporting hypothesis 1c.

Hypotheses 1a and 1b were not supported; the overall recognition and R rates of the information of the self-preferred brand was worse than that of the other-preferred brand. This may be because the familiarity of self-preferred brands was not significantly higher than that of the other-preferred brands, despite the fact that the former demonstrated higher preference than the latter. Liu (2015) found that similarity of materials had a large impact on working memory spans. Therefore, similarity might influence memory levels of different referential tasks (Symons and Johnson, 1997). Correlation analysis also showed that preference was negatively correlated with the R rate of negative information, which meant that the higher the brand preference, the worse the negative information was remembered, and the memory of negative information was so poor that it made the preference negatively correlated with all of the information of the brand. In other words, it was not the self-preferred brand, but rather the negative information of the brand that resulted in poorer memory. Nevertheless, there was no memory difference observed in the K rate, providing evidence in support of assertions made by previous researchers that there is no SRE in K responses.

Experiment 1 also found that, in the self-referential condition, the memory of positive information was enhanced when compared to that of negative information. However, in the other-referential condition, the difference disappeared. This result provides support for hypothesis 1c. Previous studies have suggested that it is difficult for people to remember negative words which may threat their positive self-image, and easy for them to remember positive words which may improve their self-image (Argembeau et al., 2005). The observed difference was not observed in K rate, once again supporting the notion that K rate is not sensitive enough a measure of the SRE.

Since different memory effects of positive and negative information of self-preferred brands were found in Experiment 1, Experiment 2 would remove the other-referential group and divide the referential conditions into the groups of high, moderate, and low levels of self-preference, examining the main effect of the degree of brand preference on memory, and its interaction with information valence.