Subliminally priming arousal : the effect on consumers’ brand personality preference and trait evaluation

(1)

Masters’ thesis: Business Administration – Marketing track (FEB)

Subliminally priming

arousal:

The effect on consumers’ brand personality preference and

trait evaluation.

Student: Alexandra van Gorp, Msc. Student ID: 10555684

Date of submission: 27-01-2017 Supervisor: Tina Dudenhöffer, Msc. Second reader: Alfred Zerres, Dr.

(2)

1

Statement of Originality

This document is written by Alexandra van Gorp who declares to take full responsibility for the contents of this document.

‘I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.’ The Faculty of Economics and Business is responsible solely for the supervision of

(3)

2

Abstract

Introduction. Even though the influence of primed valence on consumer behavior has been

the subject of numerous studies, the effect of primed arousal remains surprisingly under-researched within the domain of affective priming. Current research therefore investigates how subliminally priming arousal influences the way consumers evaluate brands; how it influences brand personality preference and brand trait evaluations. Moreover, current study adds to our understanding of priming effectiveness by investigating the influence of the modality of the subliminally presented stimuli.

Method. Preceded by a pre-test, an online survey was conducted. 239 participants participated in the study, who were all recruited via Amazon Mechanical Turk (mTurk). The subjects were randomly assigned over the research conditions, as the research had an (2 x 5) between subjects research design; resulting in 10 conditions. Logistic regression analyses and analysis of variance’s (ANOVA’s) were conducted to analyze the data.

Results. Our results indicate no significant effect of primed arousal on brand preference, nor

on brand personality trait evaluations. We did however find a significant difference in evaluative response between the two priming modalities. As such, we found that individuals who were primed with pictures, compared to those who were primed with words, evaluated the brand as more sophisticated.

Conclusion. Current study was not able to meet the expectations of primed arousal on brand

personality preference and brand personality evaluations that were based on existing literature. However, we did find a significant effect of modality on brand personality trait evaluation; although it is disputable that the differences in evaluations were due to different levels of experienced arousal as no main effects of primed arousal were found. It is believed that the methodological limitations in current research are held to be responsible for the lack of significant findings. It is therefore strongly suggested to use different measures for future research.

(5)

4

1. Introduction

As branding, advertising and slogans cover the globe in a colorful and intrusive fashion, it almost seems that consumerism is a part of the fabric of our society. However, not all forms of marketing content are designed to be so readily apparent, as research reveals that even information presented unconsciously is also able to affect consumer behavior. Ever since the James Vicary movie theater experiment in 1957, in which Vicary claimed to be responsible for a significant increase in popcorn and Coca-Cola sales by flashing the subliminal messages “Eat Popcorn” and "Drink Coca-Cola” during movie screenings, subliminal messaging has become a notorious advertising strategy. Subliminal (literally "below threshold") messaging refers to exposing individuals to certain stimuli, without them being consciously aware of said stimuli. For example, it might be inaudible for the conscious mind, or it might concern an image that is transmitted so briefly that it’s consciously unperceived - and yet, perceived unconsciously. As such, this unconscious information is proposedly capable of eliciting reactions from consumers, without the consumers having a vast rationale to explain their actions.

Although Vicary later admitted to have forged the results of his experiment in order to put his failing marketing firm on the map (Karremans, Stroebe & Claus, 2006), the thought of being subliminally persuaded led to public outrage and general concern (Pratkanis, 1992). Given the Orwellian “big brother” zeitgeist at the time of Vicary’s experiment, the public’s ethical concerns are quite understandable. After all, what if the government controls our minds, what happens to our free will? In this regards, Key (1989) wrote the following:

“Subliminal indoctrination may prove more dangerous than nuclear weapons. The substitution of cultural fantasies for realities on a massive, worldwide scale threatens everyone in this precarious period of human evolution. Present odds appear to favor total devastation.” (Key, 1989, p. 18).

However, in recent years, scientific research has revealed that when subliminal messaging is at all effective, it often shows very little effects (Bargh, 2002). Therefore, we all still have free will, and are not a mere product of our environment. Then, why use subliminal

(6)

5 messaging? Even though measurable effects are rather small, it’s widely accepted that it is not only our conscious mind that holds the key to our thoughts and behavior. We’re constantly being influenced by sub- and supraliminal stimuli, subtlety shaping our thoughts. And even though the power of subliminal persuasion is too small to change people’s attitude, subliminal messages are able to trigger a prior attitude or disposition, as long as certain conditions are met (i.a. Dijksterhuis & Smith, 2002; Bargh, 2002). One of these conditions, researched by Strahan at al. (2002), is that the subliminal message has to be goal-relevant to the receiver in order to have an effect. In this sense, when we take Vicary’s movie theater experiment as an example, the subliminally presented message “Drink Coca Cola” might have actually had a positive effect on individuals who were thirsty, as the prime they were exposed to (“Drink Coca Cola”) actually fulfilled a goal that they had at the time of being exposed to the prime (in this case: quenching thirst). Individuals who were not thirsty on the other hand, were unlikely to be affected by the subliminal message when we follow the same reasoning.

In recent years, much scholarly attention has focused on the effectiveness of subliminal messaging, and how it influences consumer behavior. As such, we now know that that subliminal messaging is effective in influencing consumers’ choice process by subliminally flashing brand names (Karremans et al., 2006) and – besides leveraging currently desired goals – we also know that subliminal messaging is capable in activating goals that are desired in their own right, such as saving money (Chartrand et al., 2008). Moreover, subliminal messaging is also found to affect consumers’ evaluation process by manipulating one’s emotional state (i.a. Veltkamp, Custers & Aarts, 2011), a phenomenon described as “affective priming”.

In short, affective priming investigates whether the exposure to a first, unconsciously perceived stimulus (the prime), affects the evaluation of a subsequently supraliminal presented stimulus. Based on the notion that individuals constantly feel the need to evaluate their surroundings (e.g. good – bad / safe – dangerous / like – dislike) (Gilette, Huang, Hatcher & Moroz, 2000), affective priming aims to influence this evaluation process by preceding the deliberate cognitive analysis of the subsequently presented stimulus. As our evaluations are not only based on deliberate reasoning but also on our emotions and automatic responses (Storbeck & Clore, 2008), affective priming leverages our automatic, emotional system. Or, metaphorically speaking, when we regard the well-known expression:

(7)

6 ‘to see the world through rose-coloured glasses’, affective priming aims to manipulate the evaluation process by influencing the color of the glasses through which we look.

Regarding emotional responses and factors that might account for the variance in emotional assessments, there appears to be a consensus that emotional responses are direct functions of two independent dimensions. One dimension being affective valence, the other being arousal (Robinson et al., 2004). In this regards, valence refers to the intrinsic attractiveness or evasiveness of the emotion. This ranges from positive to negative, from good to bad, happy to sad and so on (Robinson et al., 2004). The second dimension, arousal, refers to the physiological and psychological state of being reactive to a stimuli. This, ranging from sleepy (un-aroused), to being alert and ready to move and respond (highly aroused). It widely accepted that valence and arousal are important determinants when it comes to emotional processing. As such, these dimensions have been identified as the most important factors of connotative meaning (Osgood, 1969), feeling states (Russell & Barrett, 1999) and emotion related behavior (Lang et al., 1995).

In the process of evaluating brands, consumers use different strategies to determine whether they like or dislike brands. One of these strategies is by taking the perceived brand’s personality into account. Aaker (1997) defines brand personality as “the set of human characteristics or traits that consumers attribute to a brand”. In this sense, a brand’s personality is similar to human’s personality in that the traits are relatively stable and built over time (Pendergrast, 1993). It is however different from a human’s personality in how it is constructed, as it does not only rely on the direct and indirect experiences consumers have had with the brands such as product reviews and brands’ product endorsers, but also on product-related attributes such as its price category and its symbol and logo (Batra et al., 1993). By attributing human characteristics to a brand, consumers are enabled to use brands as a vehicle for self-expression (Belk, 1988). As such, they are able to to reflect their ideal self (Malhotra, 1988) and/or to reflect specific dimensions of the self (Kleine, Kleine & Kernan, 1993). Therefore, brands are encouraged to build strong brand personalities, as it helps them distinguish themselves from competitors. Moreover, it is also suggested that as markets continue to mature and competition within industries intensifies, companies are not exclusively evaluated on the products or services they offer; they are also evaluated on how their personality is perceived. Or, as formulated by Keller and Lehman (2006), “the success of a 21st century business will be defined as much by who it is as what it does.”

(8)

7 Although the connection that people feel with brands and their personalities are to a great extent based on deliberate, conscious thought-processes (e.g. what do I want to reflect when I associate with a brand, what does this brand says about my self-image?), the way consumers perceive and evaluate brands and their personalities is also largely based on emotions and automatic responses (Sung & Kim, 2010). Sometimes, consumers are attracted to certain brand personalities because they just feel good. In current research, the focus lies on how subliminal messaging is able to influence this emotional, automatic response; and how affective priming affects the way consumers’ perceive and evaluate brand personalities.

1.1 Problem statement

When reviewing the existing literature on affective priming, it became evident that past studies mainly focussed on the effect of primed valence on consumer behavior and evaluations. In this regards, much research has been devoted to priming individuals with different facial expressions (i.a. Murphy & Zajonc, 1993) and by subliminally presenting negative versus positive picture stimuli, such as images of snakes versus kittens (Hermans, Spruyt, De Houwer & Eelen, 2003). However, it is strongly suggested that emotional evaluations are not only driven by valence, but also by experienced arousal (i.a. Robinson et al., 2004). Regardless of this widely accepted notion, literature on the influence of primed arousal on consumer behavior remains surprisingly scarce. Therefore, it would be interesting to find out how subliminally priming individuals with stimuli that defer in terms of arousal, affects their evaluation process. Moreover, as brand-personalities are playing an increasingly important role in the eyes of consumers (e.g., Keller & Lehmann, 2006), current research focuses on how subliminally priming arousal influences how consumers perceive and evaluate brand personalities. In conclusion, the following research question is proposed:

How does subliminally priming arousal influence the consumers’ evaluation of a brand’s personality?

1.1.2 Subquestions

In order to answer the main research question, there are several sub-questions formulated.

As discussed earlier, consumers tend to attribute human characteristics to a brand, in order to evaluate them and compare them to other brands. Therefore, the first sub-question is as follows: ‘how does priming arousal influence consumers’ brand personality preference?’

(9)

8 and, ‘how does priming arousal influence the consumers’ perception on brands’ perceived personality traits?’. Moreover, as valence has previously been identified as an important driver of the decision-making process: ‘what is the effect of valence on subliminally priming arousal?’. The last sub-question concerns the modality of the prime, as it is expected that the way in which the prime is presented, might influence its effectiveness: ‘what is the effect of using picture primes versus word primes?’

1.2 Theoretical Contributions

As affective priming is able to affect consumers’ perception and preferences (Bargh, 2002), it is an important concept to research from a marketing perspective. And, as the influence of primed arousal remains underexposed in relation to the impact of valence -although they appear to be just as important in the evaluation process- current research adds to our knowledge on the impact of arousal on affective priming. In other word, current research extends to the literature on affective priming by researching a phenomenon

(arousal,) what to our knowledge remains under-researched. Moreover, current research

investigates how priming arousal affects consumers’ brand personality preferences. Therefore, results of current study could provide greater insight in how consumers evaluate brands, and could lead to practical implications for influencing consumers’ perceptions of brands.

1.3 Structure

Firstly, in order to paint a broader picture of current topic, an overview of the literature on subliminal messaging, affective priming and brand personality will be provided. The literature review will conclude with the hypotheses formation, that will function as the foundation for current research. The literature review will be followed by the methods section, in which the used measures will be discussed. Subsequently, the analyses and the results of current research will be provided. This thesis will be concluded with a discussion, in which the results of current research will be discussed, as the limitations of current study and directions for further research.

(10)

9

2. Literature review

2.1 Subliminal messaging

2.1.1 Defining subliminal messaging

Subliminal, derived from the Latin sub limen, literally translate to “beneath threshold”. Therefore, subliminal messages are messages that are presented in such a fashion, that it’s not possible to be picked up by the conscious mind; they are processed without (below) our conscious awareness (Smith and Rogers, 1994). In regards to one’s perceptive threshold, Hull (1943) suggests the following: “the term threshold implies in general a quantum of resistance or inertia which must be overcome by an opposing force before the latter can pass over into action.” In other words, it is proposed that a threshold is an obstacle that information has to overcome in order for it to pass over into action. Moreover, according to Hull (1943), this threshold is subjective; as it each individual might maintain a different threshold to his or her consciousness awareness. Therefore, the most suitable way to test if information has been consciously perceived appears to ask the recipient afterwards whether they can recall what they witnessed; a method frequently used in previous studies (i.a. Bargh & Chartrand, 2000; Karremans, Stroebe & Claus, 2006).

In regards to how subliminal messaging works, the dual process theory by Stanovich and West (2000), provides a useful theoretical bridge to explain its underlying mechanism. In their study, the authors discuss the two-process theories of reasoning by the use of two distinct cognitive systems. They refer to this as “system one” and ‘system two”. The first system is described as being fast and automatic, it responds to emotions and is based on basic instinct. Contrarily, the second system is described as being slow and effortful. System two uses logic and consciously focuses on incoming information to come to conclusions. Subliminal messaging therefore, merely involves system one – as subliminal stimuli are per definition invisible to system two. In other words, subliminal messages and subliminal priming unconsciously leverage system one; and thereby making individuals rely more heavily on their instinctive impulses. As system two does not know what system one is reacting to, the subliminal message recipient has no rationale of why an idea pops into their head, or why they suddenly feel a certain craving.

(11)

10

2.1.2 Subliminal message modality

Looking at the general concept of subliminal messaging, two different modalities of subliminal persuasion can be distinguished. First, there’s the visual category, in which visual stimuli are presented for a very brief amount of time. Secondly, there’s the audio-category, which involves sub-audible messages. This category involves messages that are played at a very low volume, as messages that are played backwards or accelerated. For current thesis the focus lies on the visual category, as evidence for the effectiveness of subliminal auditory stimuli remains weak (Egermann, Kopiez & Reuter, 2006).

Within the visual category, we can subcategorize visual stimuli into picture targets, and word targets. As the category name suggests, the first category implies the presentation of pictures; often used to elicit emotional responses. As such, a lot of scholarly attention has focused on the subliminal presentation of different facial expressions (e.g. Yamada & Decety, 2009; Dimberg, Thunberg & Elmehed, 2000; Sweeny, et al., 2009) and on images that are expected to trigger strong emotional and arousing responses, such as images of snakes and dead bodies (Hermans et al., 2003). The second visual category, word targets, does not involve any images but text. An example of the use of word targets is James Vicary’s movie theatre experiment, in which he subliminally presented the messages “eat popcorn” and “drink Coca Cola” to the audience.

It is conceivable that the use of word targets versus picture targets differ in terms of effectiveness, as they rely on different cognitive systems. In this sense, Glaser and Glaser (1989) propose a distinction between a semantic system and a lexical system. Pictures are believed to primarily rely on the semantic system, whereas words are believed to have privileged access to the lexical system. Houwer et al. (1994) investigated how these different systems influence the processing of word- and picture stimuli. In their research, subjects had to name the valence of the picture- or word stimuli as quick as possible: whether they were positive or, negative. They found that individuals were able to evaluate picture stimuli more rapidly compared to word-stimuli. Therefore, they suggest that the semantic network, to which pictures have privileged access, contains affective information.

2.1.3 Subliminal messaging and consumer behaviour

Whereas initial research on subliminal messaging mainly focused on the question whether subliminally presented stimuli could be recognized and discriminated (e.g. Collier,

(12)

11 1940; Coyne et al., 1943), James Vicary’s Movie Theater Experiment caused a shift in research focus. In his experiment, Vicary claimed to be responsible for a substantial increase in popcorn and Coca Cola sales after subliminally flashing the messages “eat popcorn” and “drink Coca Cola” to the audience in a movie theater. Even though his findings were not replicated – and he later admitted his experiment was a mere publicity hoax for his failing marketing firm (Karremans et al., 2006), the potential influence of subliminal messages on consumer behaviour and attitudes has since then been the focus of many studies. After all, it’s possible effectiveness on persuading consumers, harnesses interesting potentials for marketers.

Initially, the effectiveness of subliminal persuasion was well disputed (i.a. Moore, 1982; Greenwald et al., 1991). Nowadays however, there appears to be a consensus that, although the effects are usually fairly small, subliminal messaging carries the potential to affect consumer behaviour (Bargh, 2002). Its effectiveness however, depends on several factors. In this regards, Strahan et al. (2002) researched the effectiveness of subliminal messaging on consumer-choice. In their research, they found that subliminal messaging is capable in influencing consumer’s beverage consumption; but only in individuals who are thirsty. Individuals that reported a lower level of thirst were not influenced by the subliminal message. Hence, the authors state that subliminal messaging is effective in influencing consumer choice, but only when the individual has a certain goal intention (e.g. quenching thirst) that is related to the presented stimulus (e.g. thirst related key words).

This finding, that the subliminal message must fulfill a certain goal the consumer has, has been replicated many times. As such, it was found that consumers who reported higher levels of tiredness were more likely to be affected by a subliminal message about energy evoking dextrose pills compared to individuals who weren’t tired (Bermeitinger et al., 2009), and that individuals who were primed with sadness preferred to listen to a CD that put them in a better mood that was described as “energetic” and “upbeat”, rather than to a CD with music that was described as “strong” and “powerful” (Strahan at al., 2002).

More recent work by Veltkamp et al. (2011) and Chartrand et al. (2008) found that subliminal messages also has the potential to be effective in absence of a specific deprivation (e.g. being thirsty) or goal (e.g. restoring sad mood to a better mood). As such, Chartrand et al. (2008) researched the effectiveness of priming consumers with brands that are associated with a thrift goal, and found that the primed subjects subsequently preferred lower priced

(13)

12 goods over more luxurious and expensive alternatives. Given these results, the authors concluded that certain behaviours (such as for this instance: saving money) can be influenced by subliminal messages because they are desired in their own right. Therefore, the authors state that priming can also be effective without a specific deprivation; or as they put it, it’s possible for subliminal messaging to “strike when the iron is cold” (Veltkamp et al., 2011).

2.2 Affective Priming

2.2.1 The influence of affect on consumer perception and behavior

Accepted throughout the literature, is the notion that individuals feel the need to constantly evaluate different aspects of their environment. This is likely to have an evolutionary function, as even animals use a hedonic decision mechanism that weighs the need for e.g. food versus the risk from predators (Gilette et al., 2000). One of the strategies we use as an information heuristic, is by the use of our affect. This phenomenon is referred to as the affect-as-information model (Schwarz, 1990; Schwarz & Clore, 1983; Wyer, Clore & Isbell, 1999).

Breaking down the construct of affect, existing literature seems to agree that affect consists out of two dimensions: valence, and arousal. In this regards, valence refers to the subjective positive or negative evaluation of an experienced state. Examples of positive valence are happiness and joy, whereas examples of negative valence are anger and fear. Arousal on the other hand, is described as “the state of being activated, either physiologically

or psychologically” (Niven & Miles, 2013). In this sense, arousal is associated with feelings of vigor, energy, and tension. This is in line with the description of arousal as provided by Russell and Mehrabian (1977), who researched the influence of arousal on consumer behaviour. In their work, they define arousal as “the subjective experience of energy mobilization, which can be conceptualized as an affective dimension ranging from sleepy to frantic excitement”.

As the information provided by one’s affect is found to serve as a basis for judgments and decisions and as a guide for cognitive processing (Storbeck & Clore, 2008), it has been well researched within a consumer-behavior context. In this regards, research on the influence of affect on consumer choice found that when individuals face a choice, they are likely to decide on a product or brand that is congruent to their state of arousal (Di Muro &

(14)

13 Murray, 2012). In other words, it is suggested that consumers are likely to prefer more-arousing products and brands if their experienced arousal-level is high, and less more-arousing products if their current arousal is low. This phenomenon is referred to as the “arousal maintenance effect” (Di Muro & Murray, 2012). However, it was found that the arousal maintenance effect only occurs when the valence of the experienced affective state is positive: subjects only chose products that were congruent to their experienced level of arousal when they were in a good mood. Contrarily, when the valence of their affective state was negative, it was found that they showed a preference for products that were opposite to their level of arousal. As such, consumers who were in a bad mood preferred low arousing products when they felt highly aroused, and high arousal products when they were not or little aroused. The authors propose that this is caused by a process that they refer to as “mitigation”: people who are in a bad mood try to “feel better”, by changing their arousal level. With their findings, Di Muro and Murray (2012) suggest that the consumers’ choice-process is affected by the interplay between their level of arousal and the valance of their affective state.

It is believed that affect does not only influence preferences, it also as an impact on how we perceive things. As such, a substantial amount of research has suggested that one’s level of experienced arousal may impact the affective and evaluative responses of consumers (Reisenzein, 1983). In this regards, the evaluation of products or brands might be more extreme if the consumer experiences higher levels of arousal. The excitation-transfer theory by Schachter and Singer (1962) offers a possible explanation of this polarized reaction. According to this theory, the residual excitation from one stimulus will amplify the excitatory response to another stimulus, regardless of their difference in valence. In this sense, it is suggested that arousal is non-specific; it merely amplifies the reaction that is provoked by the subsequent stimulus (Schachter & Singer, 1962; Zillman, 1971). As such, White, Fishbein and Rutstein (1981) asked their participants to evaluate the attractiveness of a female target. They found that arousal, physical as well emotional arousal, caused a polarization in the participants’ judgements. Their research suggests that participants misattributed the residual arousal that was induced by either exercise or emotional material as part of their affective response to the female target. Therefore, their results offer support for the excitation-transfer-theory (Schachter & Singer, 1962).

(15)

14

2.2.2 Priming Arousal

Although the impact of valence has been the subject of numerous affective priming studies, studies on priming arousal remain scarce, even though it is evident that the arousal dimension plays an equally important role in affective processing (Lang et al., 1995; Russell & Mehrabian, 1974).

Over the recent years, neuroimaging techniques are increasingly being used to advance our knowledge on cognitive processes. There are several neuroimaging studies dedicated to how our brain responds to subliminal stimuli that defer in terms of arousal (high versus low). One of these studies, by Hinojosa et al. (2009), compared the processing of stimuli with the same valence but different levels of arousal. They primed individuals with high arousal primes (e.g.: victory, sex, salary, euphoria) versus low arousal semantics (e.g.: massage, bed, peace). Outcomes of their study suggests that arousal and valence influence affective priming in distinctive ways. Therefore, they pose that arousal plays a singular part in the affective priming process. Moreover, they also found that one’s level of arousal can be manipulated through the process of priming. This last notion has been confirmed by Balconi and Lucchiari (2008), who used gamma band activity (GBA) to study cortical activation patters during subliminal information processing. They found that GBA activity is enhanced more by high arousal images (angry and fearful facial expressions) compared to low arousal images (happy and sad facial expressions). Schupp et al. (2000) also studied neural processes during subliminal information processing by the use of event-related potential (ERP). Their study reveals that affective pictures of high arousal elicited larger late positive potential (LPP) compared to pictures that were low in arousal; meaning that our brain responds to different arousal-levels in distinctive ways.

2.3 Brand Personality

2.3.1 Defining brand personality

When consumers evaluate and compare different brands, it’s not uncommon they take the perceived brands’ personality into account. A company’s brand personality – not to confuse with brand identity – is defined by Aaker (1997) as "the set of human characteristics or traits that consumers attribute to a brand”. In this sense, the humanization of a brand enables consumers to relate to the brand; it gives them something more vivid and more complete than the functional benefits that are associated with the brand (Fournier, 1998). It is

(16)

15 suggested that brand personalities enables consumers to project a personal aspect of him- or herself that might be desirable for relationships the consumer seeks (Aaker 1997; Wallendorf & Arnould, 1988) and that brand personalities function as a vehicle of consumer self-expression: it aids consumers in expressing their actual self, ideal self, or specific aspects of the self (Belk, 1988). Also, a brand identity might offer a sense of comfort to the consumer, by providing a “fit” with their self-concept (Aaker, 1999).

When breaking down the construct, we see that – just as it is the case for human personality traits – a brand personality is both distinctive and enduring (Pendergrast, 1993). In other words, a brand personality is built over time: it’s the weighted average of consumers’ previous impressions of the brand. However, although human and brand personality share a similar conceptualization (Epstein, 1977), they are different in how they are constructed. Whereas an individuals’ personality is inferred on the basis of a persons’ behavior, thoughts and beliefs (Park, 1986), a brands’ identity relies on the direct and indirect contact consumers have with the brand (Plummer, 1985). In this regards, a brand identity does not only rely on the direct experience a consumer has with the brand (e.g. the brands’ product endorsers and the brands’ users imagery) but also on indirect features such as product-related attributes, price and the brands’ symbol and logo (Batra et al., 1993). As brand personality affects brand trust (Sung & Kim, 2010; Hess et al., 2007), brand attachment and commitment (Louis & Lombart, 2010), brand loyalty (Brakus et al., 2009) and is found to positively affect overall brand equity (Valette-Florence, Guizani & Merunka, 2011); it is of great importance for organizations to create meaningful and distinctive brand personalities in the minds of consumers (Siguaw, Mattila & Austin, 1999).

2.3.2 Brand personality dimensions

In order to close the gap between consumer behaviour and brand personality, Aaker (1997) developed a theoretical framework in which five different personality dimensions were identified. In her research, Aaker asked the respondents to evaluate brands on 114 personality traits, by using a 5-point liker scale. Subsequently, factor analysis categorized the personality traits into five different personality dimensions. As this framework was found to be reliable, valid and generalizable, said framework will be used as a basis for current research. The five dimensions as distinguished by Aaker (1997) are the following: sincerity, excitement, competence, sophistication and ruggedness. These dimensions are each reflected by underlying traits. For example, sincerity is considered the umbrella term for brands that

(17)

16 are described as being “down-to-earth”, or “cheerful”. Competence on the other hand, is based on the traits of being “Reliable” and “Intelligent”.

When we connect brand personality to the previously discussed literature on arousal, Batra Seifert and Brei (2015), identified two brand personality dimensions that are associated with arousal. According to their research, the Sophistication dimension is associated with a low level of arousal, whereas Excitement is associated with higher levels of arousal (Batra et al., 2015). For current study, we will therefore exclusively focus on these two dimensions. For an overview of the complete model and its dimension, please consider figure 1.

Figure 1: Brand Personality Dimensions (Aaker, 1997)

2.4 Hypotheses formation

1. Researching the influence of valence and primed arousal on brand personality preference

When reviewing existing literature on the effect of subliminal priming, we have established that the state of subjective experienced arousal is capable of being influenced by a priming condition. As such, priming individuals with high-arousal associated words elicits a higher level of experienced arousal, whereas low-arousal words lead to a lower level of subjective arousal (Hinojosa et al., 2009). When we connect the literature of arousal with consumer preference, existing literature suggests that when consumers are faced with a choice, they tend to base their decision on a product, brand or service that is congruent to

(18)

17 their experienced level of arousal in order to maintain their aroused state – a phenomenon described as “the arousal maintenance effect” (Di Muro & Murray, 2012). However, arousal is not the only factor that comes into play when consumers are faced with choice; it is suggested that consumer choice is interplay between subjective arousal and affect valence. In this sense, the arousal maintenance effect only occurs when individuals experience a pleasant affective state. When they are in an unpleasant affective state on the other hand, they are likely to choose a product, brand or service that changes their arousal level in order to reduce their unpleasant affect. This phenomenon is described as “the mitigation effect” (Di Muro & Murray, 2012). Given that the sophistication brand personality dimension is associated with lower levels of arousal, and the exciting personality dimension is associated with higher levels of arousal (Seifert & Brei, 2015), it is expected that when individuals are subliminally primed, they will show a preference for the brand that is congruent to their experienced level of arousal. Therefore, it is expected that individuals in the high-arousal priming condition will show a preference for brands that are associated with the excitement dimension over brands that are associated with the sophistication dimension. Moreover, individuals who are in the low-arousal priming condition are expected to prefer brands that are associated with the

sophistication dimension over brands that are associated with the excitement dimension.

However, based on the premise that the arousal maintenance effect only occurs when the recipient is in a pleasant affective state, we only expect this outcome to occur when the recipient is in a good mood. When the individual experiences an unpleasant affective state on the other hand, we expect the mitigation effect (Di Murro & Murray, 2009) to occur, meaning that they will prefer a brand that is opposite to their level of arousal (sophistication when arousal is high and excitement when arousal is low). In conclusion, the following is proposed:

H1: It is expected that the combination of a high (vs low) level of arousal and a positive (vs negative) mood, and the combination of a low (vs high) level of arousal and a negative (vs

positive) mood increases the chance of choosing an exciting (vs sophisticated) brand. 2. Researching the influence of arousal on brand personality perception

The next expectation is based on the excitation-transfer theory, by Schachter and Singer (1962). According to the excitation-transfer theory, the residual excitation from one stimulus will amplify the excitatory response to another stimulus, regardless of its valence. Following this theory, previous research by White et al. (1981) has established that emotionally aroused individuals show a polarization in their judgements; they appear to

(19)

18 misattribute the residual arousal that was induced by either exercise or emotional material as part of their affective response. Therefore, in line with these findings, it is expected that the individuals who are in the high-arousal priming condition will misattribute the residual arousal and polarise their evaluation of the presented brands. As such, it is for instance expected that they will evaluate a brand that is associated with the brand personality dimension “excitement” as being more exciting and the brand that is associated with the brand personality dimension “sophistication” as being more sophisticated, than the subjects who were not primed and those who were low-arousal primed. In conclusion, the following is proposed:

H2: Subjects in the high-arousal priming condition will show a polarisation in their evaluation of a brand’s personality traits, compared to those in the low-arousal and

no-priming condition. 3. Researching the influence of visual stimuli modality

Lastly, as the processing on word-stimuli and picture-stimuli rely on different cognitive systems, respectively the semantic- and lexical system, it is expected that the modality in which the prime is presented might influence its effectiveness. As the majority of previous research has either focused on picture modality, or word-modality, it is difficult to compare these results in terms of effectiveness. Even though neuroimaging techniques has shown that the processing of word-targets take up greater brain activity, meaning that it is more effortful processed (Zhang et al., 2006); it is believed that picture stimuli will evoke greater effects. This, based on research by Hermans, Houwer and Eelen (1994), which suggests that the semantic network (to which pictures have privileged access) primarily contains affective information. As our research aims to manipulate one’s affect, it is expected that the use of picture stimuli will show a greater effect compared to the use of word stimuli. Therefore, the final hypothesis is as follows:

H3: Subjects who have been subliminally primed with picture-stimuli, will be more affected by the priming condition compared to those who have been primed with word-stimuli.

(20)

19

3. Methods

3.1 Participants and design

248 individuals participated in the experiment; they were all recruited via Amazon Mechanical Turk (MT). Only “Mechanical Turk Masters” could participate in current research, as individuals who have been granted this title have shown a high degree of accuracy across a variety of requesters (Crump et al., 2013), thereby enhancing the reliability of the results.

From the initial sample, 9 participants were excluded from further analysis. Of these exclusions, 6 participants were excluded because the video did not play on their mobile device, and 3 participants were excluded because they recognized the subliminal message. Hence, 239 participants were reviewed. Of the participants there were 135 males, and 97 females. Ages varied from under 18 to 84, the vast majority (49.4 percent) of the participants were in the 25-to-34 age bracket. In table 1, a more detailed overview of the participants is provided. N % Sex Male 135 56.5 Age Under 18 18 7.5 18 – 24 26 10.9 25 – 34 118 49.4 35 – 44 40 16.7 45 – 54 19 7.9 55 – 64 15 6.3 65 – 74 2 0.8 75 – 84 1 0.4 85 or older - 0

Education Less than highschool 19 7.9

Highschool graduate 19 7.9 Some college 22 9.2 2 year degree 12 5 4 year degree 81 33.9 Professional degree 75 31.4 Doctorate 11 4.6 Table 1. Demographics

Current research is designed to investigate how subliminally priming arousal influences brand personality preference, and how the valence of one’s mood influences this relationship. Also, it will be studied what priming modality will evoke greater effects; picture

(21)

20 primes or word primes. Given that the valence of one’s mood is manipulated, and different priming conditions are administered, current research is experimental in nature.

In this research, there are 10 different research conditions. Each participant is randomly assigned to one of these conditions. There are two mood valence conditions (positive / negative), and there are five subliminal priming conditions (high arousal images / high arousal words / low arousal images / low arousal words / no prime). Therefore, current research is a (2 x 5) between subjects design. In table 2, the distribution of the participants over the different experimental conditions is provided.

N %

Experimental Condition Negative Valence High Arousal Words 25 10.5 Negative Valence High Arousal Pictures 23 9.6 Negative Valence Low Arousal Words 23 9.6 Negative Valence Low Arousal Pictures 25 10.5 Negative Valence No Prime 22 9.2 Positive Valence High Arousal Words 26 10.9 Positive Valence High Arousal Pictures 20 8.4 Positive Valence Low Arousal Words 26 10.9 Positive Valence Low Arousal Pictures 27 11.3 Positive Valence No Prime 22 9.2 Total: 239 100%

Table 2. Descriptives: experimental conditions

3.2 General procedure

The experiment was conducted with the use of Qualtrics, a website which allows the results to be directly transferred into SPSS for further analysis. The link to the survey was distributed via Amazon Mechanical Turk.

The survey information read that the research investigates how certain emotions influence brand perception. It also informed the participants that their information is kept anonymously, and informed them about the amount of time it would take to complete the survey.

After filling in basic demographic information (age, sex, education), the participants were presented a short news story. This was either a positive or a negative story (see appendix 1 for a full version of the stories). They were then asked to rate the story on a mood scale that indicated how they would feel if they had read the story in their local newspaper. This served as a manipulation check.

Following the mood manipulation, the participants were instructed to watch a short video-clip. Within this video-clip, the priming conditions were embedded. In order to make

(22)

21 sure the participants kept their full attention while watching the clip, they were beforehand instructed that they had to answer a test questions about the shown clip. Moreover, they were instructed to not pause or replay the video-clip. The valence of the clip was neutral, so the initial mood manipulation would not get affected. Therefore, it was decided to use an instructional video that was pre-tested on its perceived valence. As such, the participant watched a short clip on how to correctly adjust their work chair to their desk. The test question that was asked about the instruction video was as follows: “What color was the desk in the video-clip?”

After the priming procedure, the participants were asked to indicate which of two presented brands they prefer, Haagen-Dazs or Ben & Jerry’s, as these brands are respectively considered to be regarded as a sophisticated and an exciting brand according to the pre-test. They made their choice by clicking on the brand of their choice, which were presented in randomized order.

In the last part of the survey, the participants were asked to think of a newly presented brand, Nespresso, as if it were a person, and were asked how well certain personality traits describe Nespresso. It was decided to use the Nespresso for this part of the survey, as the pre-test revealed that Nespresso is a well-known brand. The traits that were questioned in this part of the study were the following: imaginative, up-to-date, daring, spirited, charming, romantic, pretentious and glamorous.

Finally, the survey ended with a priming awareness check. As such, the subjects were asked if they noticed anything extraordinary about the video-clip they watched in an open-end response format.

On the last screen, the participants were thanked for their time and effort, and they received a personal code that they could fill in on Amazon Mechanical Turk to receive their monetarily compensation ($0,05).

3.3 Pre-test and Measures

In current research, the following variables were measured: mood valence, brand personality preference and perceived brand personality traits. Moreover, in order to establish the variables (video-clip, priming stimuli and brands) that were used in current research, a pre-test was conducted. The results of the pre-test and the measures of the variables will hereinafter be discussed in more details.

3.3.1 Pre-test

(23)

22 be neutral in terms of valence, so that the mood manipulation would not get affected. Therefore, the video-clip was tested on its perceived valence. The participants were instructed to watch a one-minute video clip of how to correctly adjust your work chair to your desk. The subjects were asked how they felt after watching the video, on a 5-point Likert scale where 1 = very negative and 5 = very positive. Results show that the majority of respondent (86%, N = 19) perceive the clip as somewhat negative, neutral, or somewhat positive (M = 3.5; SD = 0.85), and that ‘neutral’ was the most selected response (N = 10). Given these results, the video-clip is suited for current research purposes.

Selection of picture stimuli: In order to establish the images that are used for the

high-arousal and low-high-arousal picture priming conditions, participants were asked to indicate how certain images made them feel, on a 5-point Likert scale. In this sense, 1 = Calm, relaxed and 5 = Excited, tense. The picture that is most frequently selected as “calm, relaxed (1)” and “slightly calm, relaxed (2)” will be used for the low-arousal priming condition, whereas the picture that is most frequently selected as “excited, tense (5)” and “slightly excited, tense (4)” will be used for the high-arousal priming condition. It was decided to base the decision for the priming conditions on how frequently they were selected instead of using means, because it gives us a larger probability that the final sample will perceive the priming conditions as calming or exciting.

The images that were selected for the high arousal condition included a picture from a scary movie (a close-up of Jack Nicholson in The Shining), since fear is regarded as a high-arousing emotion (Lang, 1995). Following that same reasoning regarding fear, an image taken from a great height was also pre-tested on arousal as a fear of heights is one of the most common phobias (Fiset et al., 1989). Lastly, we pretested a picture of people in a rollercoaster, whose facial expression and widened eyes give them an excited look.

For the low-arousal priming condition, several scenery and nature photos were pre-tested on their perceived arousal. The decision for the low-arousal pictures is based on the assumption that people perceive scenery and nature photos as calming (Buchanan & Lovallo, 2001).

Results indicate that the picture of people riding a rollercoaster is most frequently selected as “excited, tense (5)” and “slightly excited, tense (4)” (N = 21). Therefore, this photo will be used for the high-arousal picture priming condition. The macro-photo of raindrops on a leaf is most frequently selected as “calm, relaxed (1)”, or “slightly calm, relaxed (2)” (N = 19). Therefore, this photo will be used for the low arousal picture condition.

(24)

23 See appendix 1 for an overview of the photos.

Selection of brands: Several paired brands that share a product category were pre-tested on

their perceived level of sophistication, and their perceived level of excitement. The pre-tested brands were as follows: G-Star Raw and Ralph Lauren (clothing), Ben & Jerry’s and Haagen-Dazs (ice-cream), Redken and Aussie (hair products), M&Ms and Dove chocolate (chocolates), and Nespresso and Starbucks (coffee). As a large portion of the sample was not familiar with Redken (45%; N = 10), Aussie (50%, N = 11) and G-Star Raw (37%, N = 8), these three brands were excluded from further analysis. As G-Star Raw was excluded from further analysis, it was decided to also exclude Ralph Lauren, as this brand could not be paired with another (clothing) brand. A graphic representation of the brands’ perceived excitement and sophistication (1 = not descriptive and 5 is very descriptive) can be found below.

As the graphs reveal, Ben & Jerry’s and Haagen-Dazs appear to be most suited for current research; as Ben & Jerry’s appears to be perceived primarily as an exciting brand, whereas Haagen-Dazs appears to be perceived as a sophisticated brand. In order to test this assumption, a paired-samples t-test was conducted to compare the means on perceived level of excitement and sophistication of the individual brands. Results indicate that there was a

1 2 3 4 5 Excitement Sophisticati on 1 2 3 4 5 Excitement Sophisticati on 1 2 3 4 5 Excitemen t Sophistica tion

(25)

24 significant difference (p = 0.001) in the outcomes for perceived sophistication (M = 2.92; SD = 0.99) and excitement (M = 3.82; SD 0.67) regarding Ben & Jerry’s, meaning that the brand is perceived significantly more exciting than sophisticated. Moreover, a paired sample t-test comparing the means on perceived sophistication (M = 3.15; SD = 0.74) and excitement (M = 2.63; SD = 0.74) of the brand Haagen-Dazs also revealed that these two means show a significant difference (p = 0.003). Henceforth, we can conclude that Ben & Jerry’s is regarded as primarily exciting, whereas Haagen-Dazs is considered a sophisticated brand. Therefore, these two brands were selected for current research.

3.3.2 Measures and Apparatus

Valence manipulation: The valence of the participants’ mood was manipulated following

research by Johnson and Tversky (1983). The participants were instructed to read a news story (positive versus negative). In line with the study by Johnson and Tversky (1983), the positive news story concerned an admission story to, in this case, Harvard University. The negative news story concerned a homicide. See appendix 2 for the full version of both stories. The participants were asked to rate the story on a mood scale that indicated how they would feel if they had read the story in their local newspaper. In this sense, the scale (1 = negative, depressed … 9 = positive, uplifted), served as a manipulation check.

Subliminal primes: The presentation of the subliminally presented stimuli was controlled

through “Wondershare Filmora” for Mac. The word-stimuli (high arousal versus low arousal) for the experimental trials were modeled after the high arousal and low arousal primes used in research by Hinojosa et al. (2009). In the high-arousal word prime condition, the participants were primed with the word “victory”. In the low-arousal condition, they were primed with the word “bed”. The words were presented in Courier New font 18 point written in black on a white background, based on subliminal priming research by Aguado, Pierna and Saugar (2005). The high arousal and low arousal picture stimuli used in current research were based on the results of the pre-test, respectively a picture of people riding a roller coaster and a macro-image of a raindrop on a leaf (see appendix 1). The stimuli were presented one time, for the duration of one frame.

Brand personality preference: In current study, participants were able to select one out of

two brands that were presented; one brand perceived as exciting, one brand perceived as sophisticated. The choice of brands was derived from the pre-test, which researched the

(26)

25 perceived brand personality dimensions as discussed by Aaker (1997). In this sense, the pre-test revealed that Häagen-Dazs is regarded as a sophisticated brand, whereas Ben & Jerry’s is regarded as an exciting brand. Both brands belong to the same product category, in this case: ice cream. Henceforth, these brands will be used as the brand personality preference variables. The brands were presented on the left and right side of the screen, presented in randomized order so that order-bias was minimized.

Brand personality perception: In order to research how subliminally priming arousal

influences the way consumers perceive a brands’ personality, the participants were asked to evaluate a new brand on different brand personality traits. For this part of the study we asked the respondents to evaluate the character traits of the brand ‘Nespresso’, as the pre-test revealed that this brand was well known.

The questioned traits were descriptive traits for the sophistication- and excitement personality dimension, following research by Aaker (1997). The traits that were descriptive for the excitement dimension were “imaginative”, “up-to-date”, “daring”, and “spirited”. The traits that belong to the sophistication dimension were “charming”, “romantic”, “pretentious” and “glamorous”. Participants were asked how well descriptive these traits are for Nespresso, on a 5-point Liker scale where 1 = not at all descriptive, and 5 = extremely descriptive.

(27)

26

4. Analyses & Results

4.1 Manipulation check: Valence

First, we tested if our mood manipulation was effective. As such, an independent samples T-test was performed, in which we compared the means of the participants in the positive valence condition to the participants in the negative valence condition. Results indicate that individuals in the positive valence condition averagely reported to feel more positive and upbeat (M = 5.18, SD = 1.211) compared to those in the negative valence condition (M = 3.08, SD = 1.665). This result was significant (p < 0.001), which indicates that our valence manipulation was successful.

4.2 Researching hypothesis 1

“It is expected that the combination of a high level of arousal and a positive mood valence increases the chance of choosing an exciting brand over a sophisticated brand”.

Firstly, we conducted a crosstab to give us an idea of the distribution of brand choice over the different research conditions. Please regard table 3 for the frequency distribution table.

Mood condition Priming condition Brand Choice Count %

Negative Valence High arousal 1. Ben & Jerry’s 27 56

2. Häagen-Dazs 21 44

Total 48 100%

Low arousal 1. Ben & Jerry’s 33 69

Total 48 100%

No Prime 1. Ben & Jerry’s 14 64

Total 22 100%

Positive Valence High arousal 1. Ben & Jerry’s 39 74

Total 53 100%

Low arousal 1. Ben & Jerry’s 37 80

Total 46 100%

No Prime 1. Ben & Jerry’s 17 77

22 100%

Table 3: frequency distribution of brand choice over different research conditions.

In order to test the first hypothesis, univariate associations between primed arousal and brand choice were analysed using a hierarchical logistic regression analysis. Unadjusted odds ratios with corresponding 95% confidence intervals (CI) were calculated for all predictors. Multiple logistic regression analyses were performed using enter procedure with an 0.05 alpha level of entry and 0.10 level of removal. The fit of the final model was assessed

(28)

27 using Nagelkerke R2 and Hosmer-Lemeshow goodness-of-fit test. The logistic regression analysis predicted the decision for the exciting brand, Ben & Jerry’s, over the sophisticated brand, Häagen-Dazs; as this was the most frequently chosen brand (70,7% of the participants chose Ben & Jerry’s over Häagen-Dazs). SPSS for Windows (version 22.0) was used to perform all statistical procedures. In total, four models were tested which will be discussed hereinafter.

The first model of our hierarchical logistic regression analysis studied the direct relationship of the control variables (age, sex and education) on our dependent variable; brand choice. This first model was statistically significant, indicating that the predictors as a set reliably distinguished between the choice for Ben and Jerry’s and Häagen-Dazs (χ2 = 17.421, p = .001 with df = 3). Nagelkerke’s R2 was .103, which suggests that the model roughly explains 10% of the variance. However, as Nagelkerke’s R2 is a pseudo R2, meaning that it gives us an approximation, it should not be overly emphasized (Field, 2013).

In the second model, we added the different priming conditions (high arousal, low arousal, no prime) to the model. Even though the model is still significant (p = .001, Nagelkerke’s R2

= .110), the results indicate that the arousal conditions did not make a significant contribution to the prediction (p = .256).

In the third model, the different valence conditions (positive versus negative) were added to the model. As it was the case for our second model, the model was still significant (p = .001, Nagelkerke’s R2 = .120). However, the outcome indicates that mood valence is not a significant contribution to prediction (p = .193).

In the last model, we added the interaction effect between mood valence and priming condition to the model. Results show that the model is still significant (p = .002, Nagelkerke’s R2

= .120), but the interaction effect between mood valence and priming condition did not significantly contribute to the prediction (p = .800). Lastly, the Hosmer & Lemeshow test of the goodness of fit suggests the model is a good fit to the data as p = 0.487 (>.05).

Given these results, the first hypothesis was not supported. Please regard table 4 for an overview of the results of the final model of the logistic regression analysis.

(29)

28

B se p

95% CI for Odds Ratio

Wald df Lower Exp(B) Upper

Included Constant -1.559 1.101 2.004 1 N.S. .210 Sex -.343 3.20 1.148 1 N.S.* .379 .709 1.329 Age .455 .133 11.759 1 0.01 1.215 1.576 2.043 Education .021 .110 .037 1 N.S.* .824 1.021 1.266 Arousal .076 .624 .015 1 N.S.* .317 1.079 3.668 Valence -.531 .601 .780 1 N.S.* .181 .588 1.910 Arousal*Valence .103 .408 .064 1 N.S.* .499 1.109 2.467

Note. R2= .084 (Cox & Snel), .120 (Nagelkerke). Hosmer & Lemeshow goodness of fit = 7.473, p = .487

* Statistically significant in the overall model, but neither sex, education, arousal, valence or the interaction of arousal and valence showed to significantly contribute to the prediction.

Table 4: Coefficients of the model predicting whether an individual would chose Ben and Jerry’s over Häagen-Dazs.

“Subjects in the high-arousal priming condition will polarise their evaluation of brands compared to those in the low-arousal and no-priming condition.”

In order to test this hypothesis, we analyzed whether the means of the evaluations between the arousal conditions (high arousal versus low arousal versus no priming condition) are different from each other. As such, we firstly transformed the answers of the personality trait questions into two new variables; the ‘sophistication dimension’ and the ‘excitement dimension’, as the traits from question 1 (charming), question 2 (glamourous), question 3 (pretentious) and question 4 (romantic) belong to the sophisticated dimension, and the traits from question 5 (up-to-date), question 6 (daring) question 7 (imaginative) and question 8 (spirited) belong to the excitement dimension (Aaker, 1997). Henceforth we conducted two analysis of variance’s with multiple factors (ANOVA’s), one for the ‘excitement dimension’ and one for ‘sophistication dimension’. The reason why ANOVA was chosen to analyze the data, is because the data can be distinguished on the basis of multiple independent categorical variables; in this case, the different primed arousal conditions. For a descriptive overview of the answers per questioned trait, categorized by research condition, see appendix 3.

Excitement: Concerning the excitement dimension, an analysis of variance showed

that the arousal condition of the participant did not have a significant effect on how exciting the brand is perceived [F(2, 236) = .235, p = .791].

(30)

29

Sophistication: In regards to the sophisticated dimension, an analysis of variance

showed that the experimental condition of the participant did not have a significant effect on how sophisticated the brand is perceived [F(2, 236) = .019, p = .981].

Therefore, we can conclude that the respondents’ evaluation of brand personality traits were independent of the arousal condition they were in; which rejects our second hypothesis.

“Subjects who have been subliminally primed with picture-stimuli, will be more affected by the priming condition compared to those who have been primed with word-stimuli.”

In order to test the last hypothesis, we first researched if the modality in which the prime was presented had a significant influence on brand choice. In order to obtain information about the distribution of our sample, we first conducted a three-way crosstab analysis, in which we added ‘brand choice’ and ‘arousal condition’ as rows and columns, to which we added ‘prime modality’ as a layer. Please regard table 5 for an overview of the frequency distribution across the different modalities.

Arousal condition Prime modality Brand Choice Count %

High Arousal Pictures 1. Ben & Jerry’s 30 58

Total 52 100%

Words 1. Ben & Jerry’s 36 74

Total 49 (101) 100%

Low Arousal Pictures 1. Ben & Jerry’s 33 77

Total 43 100%

Words 1. Ben & Jerry’s 37 73

Total 51 (94) 100

Table 5: frequency distribution of brand choice over different research conditions.

To research if the modality in which the primes were presented significantly predicted brand choice, a hierarchical logistic regression analysis was conducted, similar to the model used to test hypothesis 1. In our first model, we added the control variables (age, sex and education). In the second model, we added ‘priming condition’ (high versus low). In the third model we added ‘prime modality’, and in the fourth model we added the interaction between prime modality and priming condition. As the first two models are similar to what we

(31)

30 discussed in hypothesis 1, the results of the third and fourth model will be discussed hereinafter.

In the third model, the different priming modalities (pictures versus words) were added to the model. The results show that the model was still significant (p = .002, Nagelkerke’s R2

= .112). However, the outcome indicates that priming modality is not a significant contribution to prediction (p = .225).

In the fourth model, we added the interaction effect between arousal condition (high versus low) and prime modality (pictures versus words) to the model. Results show that the model is still significant (p = .002, Nagelkerke’s R2= .120), but the interaction effect between arousal condition and prime modality did not significantly contribute to the prediction (p = .220). Lastly, the Hosmer & Lemeshow test of the goodness of fit suggests the model is a good fit to the data as p = 0.792 (>.05).

Therefore, we can conclude that the modality in which the priming conditions were presented did not influence brand personality preference.

Please regard table 5 for an overview of the final results of the logistic regression analysis.

B se p

95% CI for Odds Ratio

Wald df Lower Exp(B) Upper

Included Constant -2.272 .792 8.228 1 .04 .103 Sex -.373 .318 1.374 1 N.S.* .369 .689 1.285 Age .473 .132 12.762 1 <0.01 1.238 1.605 2.080 Education .015 .108 .019 1 N.S.* .821 1.015 1.256 Arousal .123 .417 .086 1 N.S.* .499 1.130 2.559 Modality .657 .382 2.953 1 N.S.* .912 1.928 4.078 Arousal*Modality -.412 .341 1.463 1 N.S.* .340 .662 1.291

Note. R2= .085 (Cox & Snel), .120 (Nagelkerke). Hosmer & Lemeshow goodness of fit = 4.675, p = .792

* Statistically significant in the overall model, but neither sex, education, arousal, modality or the interaction of arousal and modality showed to significantly contribute to the prediction.

Table 6. Coefficients of the model predicting whether an individual would chose Ben and Jerry’s over Häagen-Dazs.

Lastly, we tested if prime modality influenced the evaluations of the brand personality traits within the different arousal conditions. As such, we analyzed whether the priming

Subliminally priming arousal : the effect on consumers’ brand personality preference and trait evaluation