Let me tell you a story…: a study about the effectiveness of narrative advertising on the creation of brand memory structures within consumers’ minds

Let me tell you a

story…

A study about the effectiveness of narrative advertising on the

creation of brand memory structures within consumers’ minds

University of Amsterdam Graduate school of Communication Master track: Persuasive Communication Author: Sanne Houterman Student number: 10615466 Master’ thesis supervisor: Ewa Maslowska 1st of February 2019

Table of Content

BRAND KNOWLEDGE REFLECTED BY BRAND MEMORY STRUCTURES ... 8

Categorizing the nodes in the brand memory structure ... 9

Benefits of strong brand memory structures... 10

BUILDING IMPLICIT AND EXPLICIT BRAND MEMORY STRUCTURES ... 11

Different ways to process information ... 11

Consumers’ attention towards advertising ... 11

Acquiring implicit and explicit memories ... 12

Advantages of implicit learning and implicit memory ... 13

NARRATIVE INFORMATION PROCESSING ... 14

Narrative transportation ... 15

RESEARCH METHOD ...19

RESEARCH DESIGN ... 19

SAMPLE & DATA COLLECTION ... 20

STIMULI MATERIAL ... 20 PROCEDURE ... 20 Pre-test ... 20 Experiment ... 21 MEASURES ...22 Mediator ... 22 Moderators ... 23 Dependent variables ... 26 Control variables ... 30 RESULTS ...31 HYPOTHESIS 1 & 2 ... 31 HYPOTHESIS 3 ... 33

New ‘Narrative transportation (total)’ measure ... 34

HYPOTHESIS 4 ... 36

HYPOTHESIS 5 ... 36

HYPOTHESIS 6 ... 39

DIRECT EFFECTS OF ADVERTISING STRATEGY ON BRAND KNOWLEDGE ... 40

DISCUSSION & LIMITATIONS ...41

APPENDIX 1 – DEVELOPMENT STIMULI MATERIAL & RESULTS PRE-TEST...44

APPENDIX 2 – TABLES & RESULTS MEASURES ...49

APPENDIX 3 – PROCEDURE & DESIGN IAT ...51

APPENDIX 4 – RESULTS H3, H4, H5 AND H6 ALTERNATIVE ANALYSIS ...56

APPENDIX 5 – TABLES & FIGURES RESULT SECTION ...67

APPENDIX 6 – SURVEY FLOW & SURVEY QUESTIONS ...70

Abstract

In order to build a strong brand and increase consumer-based brand equity, the

main goal of advertising should be to increase brand knowledge by building strong brand memory structures within consumers’ minds. But, little is known about the most effective advertising strategy to build brand memory structures and thus to

increase brand knowledge within consumers’ minds. In this article we investigated the effect of narrative advertising versus nonnarrative advertising on consumers’ brand

knowledge, with the role of narrative transportation as a mediator. Brand knowledge

was operationalized as two dimensions, namely brand awareness and brand image.

Brand awareness was measured with brand recall and brand recognition. Brand image

was measured with the quantity and quality of explicit brand associations

(operationalized as the number, strength, favourability and uniqueness of explicit

brand associations) and implicit brand associations (operationalized as the strength &

favourability of implicit brand associations).

Results showed a strong positive relation between a narrative advertising

strategy and narrative transportation. Subsequently, narrative transportation had a

positive relationship with brand recall, brand recognition, the number and

favourability of explicit brand associations, and the strength and favourability of

implicit brand associations. These results therefore strongly indicate that narrative

advertising is an effective tool to create or enhance both implicit and explicit brand

memory structures and thus brand knowledge within the consumers’ minds. These

finding can be of great importance to the fields of education, health care and

communication. Therefore, more research into the topic of ‘narrative communication as a tool to build memory structures’ is highly necessary.

Acknowledges

I would first like to thank my thesis supervisor dr. Ewa Maslowska of the

University of Amsterdam. Professor Maslowska was always there to brainstorm with

me, whenever I experienced trouble with my writings or data analyses. This help was

especially useful for the parts of the thesis in which I had little to no experience, such

as the implicit association test, the repeated-measure ANOVA analyses and the

logistic regression analyses. Thank you for keeping me motivated during the entire

process.

Furthermore, I would like to thank Gijs de Beus (Head of Strategy), Rogier

Heijning (Creative Director) and Felicia Spans (Junior Creative) of the creative

agency Lemon Scented Tea. Thanks to the expertise of Gijs and Rogier I was able to

create stimuli material for the narrative condition, which moved the hearts of my

participants and resulted in high levels of narrative transportation. Gijs also helped me

out by brainstorming with me, when I was stuck during the research process. I also

want to thank Felicia for helping me out with her editing skills, by creating the two

commercials of the experimental manipulations. The results of this thesis confirmed Lemon Scented Tea’s philosophy; you can indeed ‘create brands with the power of

Introduction

Consumer-based brand equity is widely used to measure the strength of a

brand (Agarwal & Rao, 1996; Keller, 1993; Romaniuk & Nenycz-Thiel, 2013). Keller

(1993, p 1) defined consumer-based brand equity as “the differential effect of brand

knowledge on consumers response to the marketing of the brand”. Brand knowledge

is reflected by the number and strength of the brand associations within the brand

memory network of consumers (Keller, 1993). Having a strong brand has numerous

benefits for companies. A strong brand improves consumers’ perceptions of product

performance and increases customer loyalty (Hoeffler & Keller, 2003). Furthermore,

a company with (a) strong brand(s) is less vulnerable to competitive marketing

actions, has larger margins and its customers respond more inelastic to product price

increases (Hoeffler & Keller, 2003).

In order to build a strong brand and increase consumer-based brand equity, the main

goal of advertising should thus be to increase brand knowledge by building strong

brand memory structures within consumers’ minds (Romaniuk, 2013). It is therefore

extremely important for marketers to know what kind of advertising strategy is most

effective in creating brand knowledge within consumers’ minds.

One advertising strategy that could be effective in increasing brand knowledge

is narrative advertising. According to Chang (2009), advertising takes either a

narrative or a nonnarrative form. A narrative advertisement persuades consumers by

telling a story and is therefore often called storytelling (Chang, 2009). The

advertisement has one or multiple characters that pursue a certain goal, and they have

specific means to reach that goal (Kim, Ratneshwar, & Thorson, 2017). A

nonnarrative advertisement provides product information in a factual manner. It often

contains only the product attributes and/or product benefits. Prior research found three

First of all, narratives are a natural way for people to process, understand and

store information (Adaval & Wyer, 1998; Bruner, 1991). Most of the social

information people acquire in their daily lives takes the form of narratives (Adaval &

Wyer, 1998). People create stories to organise their experiences, create order in

presented information, and explain unexpected events (Bruner, 1991; Escalas, 2004a;

Heider & Simmel, 1946; Mcadams & Mclean, 2013). This tendency of humans to use

narrative as a way to understand the world and learn new concepts is used in narrative

advertising.

The second benefit of narrative advertising is narrative persuasion. Prior

research found that when people consume and process a story, they engage in

narrative transportation (Chang, 2009; Escalas, 2004a; Green & Clark, 2013; Hamby

& Brinberg, 2016; Lee & Leets, 2002; Phillips & McQuarrie, 2010; Shen, Sheer, &

Li, 2015; van Laer et al., 2012). Narrative transportation is “the phenomenon in which

consumers mentally enter a world that a story evokes” (van Laer et al., 2012; p. 797).

Narrative transportation leads to narrative persuasion, which are enduring persuasive

effects (van Laer et al., 2012). Prior research stated that narrative persuasion increased consumers’ affective responses (Chang, 2009; Escalas, 2004; Moore & Homer, 2008), increased consumers’ product evaluations (Polyorat, Alden, & Kim, 2007), decreased

advertising intrusiveness (Wang & Calder, 2006), increased advertisement and brand

attitude (Lundqvist, Liljander, Gummerus, & Van Riel, 2013; Wentzel, Tomczak, &

Herrmann, 2010), increased willingness to pay for the brand (Lundqvist et al., 2013),

increased donating money to charities (Barraza, Alexander, Beavin, Terris, & Zak,

2015), and led to story consistent beliefs (Green & Brock, 2000; Slater & Rouner,

2002). Narrative transportation and persuasion are thus interesting processes for

The third benefit of narrative advertising is that it can be processed effectively

with little attention of the consumer. Prior research states that consumers pay low

levels of attention to advertising (Heath, 2001). The advantage of narrative

advertising is that the brain processes stories even when the brain is in rest and is not

actively working (Dehghani et al., 2017; Spreng, Mar, Kim, 2009). So even if

consumers do not pay attention to narrative advertising, their brains will

unconsciously process the presented information and still create or strengthen brand

memory structures.

These three advantages of a narrative form give us reason to believe that

narrative advertising might be an effective strategy to increase consumers’ brand

knowledge. Still little is known about the influence of narrative advertising on

creating or enhancing brand knowledge. Prior research on narrative advertising

focussed on narrative information processing (Adaval & Wyer, 1998; Chiu, Hsieh, &

Kuo, 2012; Escalas, 2004; Kim, Ratneshwar, & Thorson, 2017; Lang, 1989; Lien &

Chen, 2013), narrative transportation and persuasion (Ching, Tong, Chen, & Chen,

2013; Green, 2004; Kim, Lloyd, & Cervellon, 2016; Lee & Leets, 2002; Phillips &

McQuarrie, 2010; Polyorat et al., 2007; Shen et al., 2015; van Laer et al., 2012; Wang

& Calder, 2006) and emotional effects of narrative advertising (Barraza et al., 2015;

Boller & Olson, 1991; Hamby & Brinberg, 2016; Mattila, 2000; Zambardino &

Goodfellow, 2007). However, little is known about the effects of narratives on

information storage & memory. The few studies that did look into the effects of

narratives on memory, found that chronological presentation of events (Adaval &

Wyer, 1998; Lang, 1989), the story gist (Delgadillo, 2004), and a narrative form had a

positive effect on brand and story recall (Laer et al., 2012). Furthermore, van Laer et

al. (2012) found that consuming narrative advertising leads to story consistent beliefs,

But, there still exists a large research gap about the effectiveness of narrative

advertising on creating and enhancing brand knowledge. Therefore, this study

investigated the effect of narrative advertising on creating brand knowledge by

creating brand memory structures. This study will answer the following research

question by conducting an experiment:

RQ: What is the effect of narrative advertising versus nonnarrative advertising on consumer’s brand knowledge?

Theoretical framework

Brand knowledge reflected by brand memory structures

As stated in the introduction, consumers’ brand knowledge, and thus their

brand memory structures, are extremely important factors to build successful and

profitable brands. The Associative Network Theory of Memory (ANTM) states that

memory consists of nodes that hold information and/or concepts. These nodes are

connected with each other through links that vary in strength. When two or more

concepts are linked with each other they make up a network of associative

information. When a consumer is exposed to a brand in a specific context, links in

memory can be created or reinforced between concept nodes (brand associations) and

the brand node (Keller, 2003). A brand is here defined as “a name, term, sign,

symbol, or design or combination of them which is intended to identify the goods and services of one seller or a group of sellers” (Keller, 1993; p. 2). The brand

associations carry the meaning of the brand for the consumers (Keller, 1993). This

network of information linked to the brand node in memory is called the brand memory structure and forms the consumers’ brand knowledge (Keller, 2003).

Categorizing the nodes in the brand memory structure

The brand memory structure refers to the network of physical nodes and links

in consumer’ brains. Brand knowledge is the construct that describes and categorizes

all the nodes in the brand memory structures. To better understand and measure the

brand memory structures, we use the Dimensions of Brand Knowledge (Keller, 1993)

in this study (see Figure 1).

Figure 1 Dimensions of Brand Knowledge by Keller (1993, p.7)

Brand knowledge consists of two dimensions, namely brand awareness and

brand image. Brand awareness is defined as the strength of the brand node in the consumer’s memory (Keller, 1993). Brand awareness is reflected as the consumer’s ability to identify the brand under different conditions (Lintas, Usa, & Rossiter,

1980). Brand awareness can be divided between brand recognition and brand recall.

Brand recognition states the ability of consumers to confirm prior exposure to a brand

when the brand is given as a cue (Keller, 1993). Brand recall on the other hand, states

category is given (Keller, 1993). Brand recall and brand recognition measure

therefore a different level of brand awareness within the consumers’ mind (Heath &

Hyder, 2005).

Brand image is defined as the perceptions about a brand, and are reflected by

the different types of brand associations linked to the brand node in the consumers’

memory (Keller, 1993). Each brand association node can differ in terms of

favourability, uniqueness and strength, which results in different strengths of the link

between the brand node and the brand association node (Keller, 1993). The different

brand associations are formed based on experience and exposure and can be different

across consumers (Keller, 2003).

Benefits of strong brand memory structures

When consumers are in buying situations internal and external cues activate nodes in consumers’ minds (Romaniuk & Sharp, 2003), which can in turn spread to nodes of the brand memory structure, thus activating the entire brand memory

network (Romaniuk & Sharp, 2004). The probability that a brand is retrieved from

memory and comes to mind during purchase situations is called brand salience and is

determined by the size and quality of the brand memory network within consumers’

minds. A brand with a strong brand memory network that consists of many brand

associations (nodes) that are strongly connected with each other has more

opportunities to be activated by internal and external cues, than a brand with a small

and weak brand memory structure. Consumers, who can retrieve more brand

knowledge from their brand memory structure, thus experience more brand salience

(see Figure 1). A higher brand salience leads to an increased likelihood that a brand is

considered during the purchase process (Baker, Hutchinson, Moore, & Nedungadi,

1985; Macdonald, Sharp, Optus, & Engel, 2000), an increased product consumption

2004), and an increased consumer-based brand equity (Chandon & Wansink, 2002).

But, before companies can benefit from these effects of consumer’s brand knowledge, they have to build strong brand memory structures within consumers’ minds.

Building implicit and explicit brand memory structures

Different ways to process information

Consumers have different ways of processing information and therefore

different ways of creating memory structures. According to the Low-Involvement

Processing Model (Heath, 2001), consumers can either process information via

explicit processing, shallow processing or automatic processing (Heath, 2000; Heath,

2001) (see Figure 7). These forms of processing differ from each other by the level of

involvement and attention of the consumer with the presented information. Explicit

processing requires high levels of involvement and is in other models of information

processing often called active processing or central processing (Petty, Cacioppo, &

Schumann, 1983). Explicit processing is a conscious way of processing information.

Shallow processing requires a lower level of involvement and consumers are not

completely aware of this type of processing (Heath, 2001). The last form of

processing is automatic processing. When consumers engage in automatic processing,

they process information without awareness of the information (Heath, 2001).

Shallow processing and automatic processing are in other models of information

processing often referred to as peripheral processing (Petty et al., 1983). Each level of

attention and thus each type of processing leads to different memory types.

Consumers’ attention towards advertising

Consumers do not pay attention to everything in their environment. Before

they are even aware of it, their brain already selects from the environment what it will

Krugman, 1965). Krugman (1965) referred to this unconscious selection process as

pre-attentive processing.

According to Heath and Hyder (2005) all advertising is processed with low

attention by consumers, especially television commercials (Krugman, 1965). Petty

and Cacioppo (1996), stated that it is impossible for consumers to carefully think

about each message they receive and that they reject most of the messages because

they are not worth their time and energy. Therefore, consumers have low attention for

the processing of advertising information.

Acquiring implicit and explicit memories

Heath’s Low Involvement Processing Model can explain the consequences of consumers’ limited attention towards advertising (see Figure 2).

Figure 2 The Low Involvement Processing Model (Heath, 2001; p.79)

As can be seen in Figure 2, when consumers process advertising with low levels of

attention, they excluded explicit processing, and engage either in automatic

Automatic processing results in implicit learning and shallow processing

results in passive learning. Consumers who process information automatically, are not

consciously aware that they are learning, but implicit learning does create and/or

reinforces memory structures (Schacter, 1971). Implicit memories are able to store

concepts and meanings that are based on previous learning. Memories acquired by

automatic learning are implicit and we are not conscious aware of them (Graf, Peter,

Schacter, 1985; Heath, 2001).

When consumers decide to pay a little more attention to advertising, then

information is processed via the shallow processing route and this leads to passive

learning and implicit memories. According to Heath (2001), passive learning can

sometimes via an external cue also result in explicit memories, which are normally

only the result of explicit processing and active learning. Explicit memories are

capable of storing new ideas in memory and we are always conscious aware of them.

Shallow processing and learning can thus result in both implicit and explicit

memories.

Concluding, most advertising is thus automatically processed and this results

in implicit memory structures. In contrast to explicit memories, implicit memories

cannot store new concepts. Automatic processing and implicit learning may thus not

seem like an effective way to create (new) brand memory structures. But, implicit

memories turn out to have some advantages over explicit memories.

Advantages of implicit learning and implicit memory

First, unlike active processing, which is activated by volition, passive and

shallow processing are automatic processes. This means that consumers will engage

automatically in implicit processing at any given time, and they cannot switch this

form of processing off (Heath, 2001). This is an advantage because most consumers

information by violation in an active manner. Secondly, without active processing

consumers cannot analyse advertising messages and will store information just as

perceived. This will withhold them from forming critical thoughts about the

advertising message (Heath, 2001; van Laer, Ruyter, Visconti, & Wetzels, 2012).

Thirdly, implicit memories are more enduring than explicit memories (Shapiro &

Krishnan, 2001). People better remember implicit memories even in cases of delay

and divided attention, than explicit memories over time (Schacter, 1996; Shapiro &

Krishnan, 2001). Finally, implicit memory is inexhaustible in its capacity, whereas

explicit memory is limited in its capacity. Prior experiments showed that the capacity

of explicit memory was limited to a maximum number of 18 items that participants

could recall (Rose, 1992), while the capacity of implicit memory was not limited and

participants could recognize at least 10.000 items (Standing, 1973). These findings

suggest that implicit memory is inexhaustible, especially compared to the limited

explicit memory.

These benefits of implicit processing imply that automatic processing and thus

implicit memories may be not as ineffective in creating memory structures, as

researchers thought they would be.

Narrative information processing

In general, advertising is processed with low attention and thus low

involvement, but different advertising strategies might influence the consumers’ level

of involvement towards an ad, and therefore influence the way information is

processed and stored.

Nonnarrative advertising consists of only factual (product) information. When

information is presented in a factual way with clear arguments, consumers are

prior research available of the processing route of narrative versus nonnarrative

advertising, but due to the presentation of only factual information, we assume that

nonnarrative advertising is processed more explicitly than narrative advertising.

Because consumers are in general not highly involved in advertising, we think

nonnarrative advertising will be processed via shallow processing leading to both

implicit and explicit memory structures.

Narratives are process by the default network mode of the brain (Dehghani et

al., 2017; Spreng, Mar & Kim, 2008). The default network mode is the modus in

which the brain is in rest (Dehghani et al., 2017). Therefore, stories can be even

processed by the default network modus of the brain, when people do not pay

attention and are not involved. This leads us to think that narrative information is

processed in an implicit way. But, to fully understand narrative information

processing we should consider the mediating processes of narrative transportation.

Narrative transportation

When people consume and process a story they engage in narrative

transportation, which influences the way information is processed (van Laer et al.,

2012). Narrative transportation is “the phenomenon in which consumers mentally

enter a world that a story evokes” (van Laer et al., 2012; p. 797). These findings lead

to the first hypothesis:

H1: Narrative advertising, in contrary to nonnarrative advertising, has a positive effect on the level of narrative transportation.

Figure 3 shows the Extended Transportation-Imagery model (van Laer et al., 2012).

The Extended Transportation-Imagery Model provides insights into the factors of the

storyteller (brand) and story-receiver (consumer) that influence narrative

transportation (van Laer et al., 2012). It is important to consider the storyteller and

Figure 3 The Extended Tranportation-Imagery Model (van Laer et al., 2012; p.801)

characters, an imaginable plot and verisimilitude (believability of the story) within the

narrative message enhance narrative transportation of the story-receivers (van Laer et

al., 2012). Furthermore, van Laer et al. (2012) found that some story-receiver

characteristics mediated the relation between a narrative form and the level of

narrative transportation. Participants with more topic familiarity, higher levels of

transportability, more ad attention, a higher education and women, experienced higher

levels of narrative transportation (van Laer et al., 2012). This led to the second

hypothesis:

H2: Topic familiarity, transportability, ad attention, education and gender have a positive influence on the relationship between narrative advertising and the level of narrative transportation.

When people engage in narrative transportation, they get engrossed with the

story and are more involved with the advertisement. According to Low-Involvement

Processing, a higher level of involvement will lead to more explicit message

processing. However, more involvement with a narrative does not lead to explicit

(Green et al., 2004; Green & Clark, 2013). Story-receivers experiencing narrative

transportation have more affective responses (emotions), less critical thoughts, more narrative thoughts (representations of the story’s structure) and more story consistent believes, attitudes and intentions (van Laer et al., 2012). Emotions are processed in an

implicit way (Batra & Ray, 1986; Heath, 2001). Less critical thoughts lead to a more

implicit way of processing information (Petty et al., 1983). These two consequences

of narrative transportation and the fact that the default network mode of the brain

processes stories lead us to believe that narrative advertising is processed in an

implicit way (Adaval & Wyer, 2015; Escalas, 2004; Heath, 2001; van Laer et al.,

2012). Therefore, we assume that narrative advertising is processed via implicit

processing and thus results in implicit memories.

For this reason, narrative advertising will be more effective in creating

implicit memory structures than explicit memory structures. Nonnarrative advertising

will be more effective in creating explicit memories than narrative advertising. Next,

we connect these information-processing and memory types to the creation of brand

knowledge. The first dimension of brand knowledge, brand awareness, can be

measured with brand recall and brand recognition. When consumers are asked to

recall a brand, they retrieve information from explicit memory (Heath, 2001). This led

to the following hypothesis:

H3: Higher narrative transportation leads to less brand recall than lower narrative transportation.

However, when consumers are asked to recognize a brand, they retrieve

information from both implicit and explicit memory (Heath, 2001). Narrative

transportation led to affective responses, which in turn enhanced the strength of the

narrative advertising can retrieve information from strong implicit memories during

the recognition task. This led to the following hypothesis:

H4: Higher narrative transportation leads to more brand recognition than lower narrative transportation.

The second dimension of brand knowledge, brand image, consists of explicit

brand associations and implicit brand associations. We assumed that narrative

advertising is processed with implicit processing, which results in implicit memories

and not in explicit memories. Therefore, narrative advertising is assumed to be a less

effective way to create explicit memory structures than nonnarrative advertising. This

led to the following hypothesis:

H5: Higher narrative transportation leads to a less effective creation of explicit brand associations within consumers’ minds than lower narrative transportation.

We assumed that narrative advertising creates strong implicit memories.

Therefore, we expect narrative advertising to be a more effective way to create

implicit brand associations, than nonnarrative advertising. This resulted in the last

hypothesis:

H6: Higher narrative transportation leads to a more effective creation of implicit brand associations within consumers’ minds than lower narrative transportation.

These six hypotheses are summarized in Figure 4. By analysing these six

hypotheses we hope to deepen our understanding of the effectiveness of the narrative

Figure 4 Conceptual model hypothesis 1-6

Research method

Research design

A one-factor (narrative versus nonnarrative) within-subjects online

experiment was conducted (see Table 1). We decided to conduct a within-subject

study, because of the nature of the survey-based implicit association test (IAT). For

further argumentation of the within-subject design see Appendix 3. The order in

which the participants watched the experimental (narrative) and control

(nonnarrative) advertisements was randomized. There were no observations of the

dependents variables before the manipulation, because this could have interfered

with the results.

Timings t1 t2

Group 1 R X1 O

R = Randomised

X = Experimental manipulation O = observation

Table 1 Research design: one-factor within-subjects online experiment

Sample & data collection

The population were people who were eighteen years or older and were living

in the Netherlands. In total 120 respondents were personally contacted via an online

message to participate in de survey in December 2018. The experiment was also

shared on several professional and social network profiles of the researcher. The

experiment was accessible for Dutch inhabitant on survey exchange websites.

Participants had a chance to win a small incentive in the form of a gift voucher of €30.00. In total 140 participants started the online experiment, of which only 49% of participants finished the experiment (N = 69). This high dropout rate was due to the

fact that the experiment was only accessible via a laptop/pc, and was disconnected

whenever the experiment was accessed by a mobile phone. Furthermore, it took on

average 35 minutes to finish the experiment, which was quite long and this led to a

higher dropout rate.

Stimuli material

We used the commercial ‘The Scarecrow’ of the brand Chipotle in the narrative condition (Chipotle, 2017), and the commercial ‘Fish Tacos’ of the brand

TacoTime in the nonnarrative condition (TacoTime, 2018). See appendix 1 for the

selection criteria, selection procedure of the stimuli material and the outcomes of the

expert interviews (N = 2) about the stimuli material.

Procedure

Before we conducted the online experiment, we conducted an online

both conditions, which could influence the nature of information processing, and

determine the associations for our implicit association test. See Appendix 1 for the

full procedure of the pre-test and the results. Based on the outcomes of our pre-test

we decided that our stimulus material had a high quality and could be used in our

experiment. Furthermore, we choose ‘healthy versus unhealthy’ as input for the

implicit association test.

Experiment

Participants were first presented with a brief explanation of the study, which stated that we studied ‘Watching TV series’. This was a topic diversion from the real research topic to decrease levels of attention towards the commercials, and thus

increase external validity. Next, participants signed an informed consent. Moderators,

such as topic familiarity and transportability, were asked before the manipulation,

because the manipulation could influence the results of certain moderators. Before the

manipulation we added a filler question about problematic TV series consumption to

enhance further research topic distraction and include a filler task after the moderator

questions. During the manipulation participants had to watch an unrelated filler video,

in this case three minutes of the TV show ‘How I met your mother’ (season 1, episode

1). The filler video was added to simulate a normal TV watching session, which

enhances the external validity. The TV show contained two commercial breaks,

during which the narrative and nonnarrative commercials were in random order

shown (see Table 3).

TV show

(one minute of ‘How I met your mother’)

Commercial break

(Experimental manipulation)

TV show

(one minute of ‘How I met your mother’)

Commercial break

(Experimental manipulation)

TV show

(one minute of ‘How I met your mother’)

After watching the filler video and the commercial breaks with the stimulus materials

the participants had to fill in a survey. The survey consisted of questions about the

narrative transportation, brand recall, brand recognition, explicit brand associations,

and implicit brand associations for both conditions. Finally, some control variables

were asked. The survey had 102questions in total and consisted of open and closed

question and an implicit association test (see Appendix 6 for full survey).

Measures

Mediator

Narrative transportation. To measure the mediator narrative transportation we

used the Narrative Transportation Scale form Green and Brock (2000). This scale

consists of twelve items measured with a seven-point Likert scale anchored from 1 = ‘strongly disagree‘ to 7 = ’strongly agree’. We replaced words like ‘reading’ with

‘watching the commercial’ to make the scale usable for TV commercials and deleted

one item from the original scale that was only relevant for written narratives, namely

‘While I was reading the narrative, I could easily picture the events in it taking place’. See Appendix 6 for all survey items. The items were in random order

presented in the survey.

We first conducted a principal component analysis on all eleven Chipotle

narrative transportation items (Q18_narr.trans.chip__1- Q18_narr.trans.chip__11)

with a Varimax rotation. Two components had eigenvalues bigger than one (5.31 and

1.36) and in combination explained 60.62% of the variance. The scree plot showed an

inflexion after the first component, so the first component was retained in the final

analysis. This component consisted of all eleven items, which had factor loadings

of the variance. The narrative transportation scale for Chipotle had a Cronbach’s

alpha of .89. Therefore the internal consistency of the scale was highly reliable.

Then we conducted the same analysis for the Tacotime narrative

transportation items (Q18_narr.trans.taco__1- Q18_narr.trans.taco__11). Three

components had eigenvalues bigger than one (4.77, 1.42 and 1.14) and in combination

explained 66.67% of the variance. The scree plot showed an inflexion after the first

component, so the first component was retained in the final analysis. The first

component consisted of nine items with factor loadings above .45 (see Table 5 in

Appendix 2), had an eigenvalue of 4.77 and explained 43.39% of the variance. There

were two items that had factor loading below .45 in this component. We wanted to

compare the Chipotle narrative transportation scale with the TacoTime narrative

transportation scale, therefore we found it important to keep the items within both

scales the same. All eleven items of the TacoTime narrative transportation scale still

had a Cronbach’s alpha of .89, and the scale was still highly reliable, despite the

inclusion of the two items of which the factor loadings were perceived to low for

component 1.

Moderators

Ad attention. We choose to use four attention questions extracted from a study

of De Graaf, Hoeken, Sanders, and Beentjes (2009) to measure the attention paid to

the commercials. De Graaf et al. studied the role of dimensions of narrative

engagement. One of the factors of narrative engagement was ‘Attention’ and

consisted of five items measuring the attentional focus to the story. We did not use the

fifth item in our survey because that focussed more on narrative transportation than on

attention for the commercials (see all attention scale items in appendix 6). Participants

‘strongly disagree’ to 7 = ‘strongly agree’. The items were in random order presented in the survey.

First, we took a look at the four TacoTime ad attention items

(Q22__attention_taco__1 - Q22__attention_taco__4) with a principal component

analysis. Results revealed one component with an eigenvalue over 1, this component

had an eigenvalue of 3.21 and explained 80.34% of the variance. The scree plot also

showed an inflexion after the first component. This component contained of all four

items, which all had a factor loading above .45 (see Table 6, Appendix 2). The

reliability analysis showed a Cronbach’s Alpha of .92, and therefore the scale was

highly reliable.

Then, we conducted a principal component analysis and a scree plot for the

four Chipotle ad attention items (Q19_attention_chipo__1- Q19_attention_chipo__4).

The first component, and only component with an eigenvalue above one, had an

eigenvalue of 3.32 and explained 82.89% of the variance. The four items within this

component had factor loadings above .45 (see Table 7, Appendix 2). The reliability

analysis showed a Cronbach’s Alpha of .93. The scale was therefore highly reliable.

Finally we formed two ad attention scales, one for each condition, containing

the mean of the outcomes of the four ad attention items. Then, we recoded the scales

in two categorical variables with the labels 1=‘High ad attention‘ and 0=‘Low ad attention’ to make them suitable for our final analysis. Because we wanted to compare the ‘TacoTime ad attention’ and the ‘Chipotle ad attention’ variable, we used the answer category ‘neither agree or disagree’ (4) to make the distinction between high and low ad attention and not the median of the scale.

Topic familiarity. In previous research familiarity with a topic was most often

measured with the Level of Contact Report scale (Corrigan, Green, Lundin, Kubiak,

which intimacy of contact with the mental illness topic is measured. The more

situations in which people were familiar with the topic, the more familiarity with the

topic people had. The scale was used in previous studies to measure familiarity with

mental illness (Corrigan et al., 2001), therefore we changed the ‘mental illness’ topic

into ‘healthy food’ to make the scale suitable for our research. Participants had to

confirm if a situation was familiar (1) or unfamiliar (0) to them. The items were in

random order presented in the survey.

We created a new scale by adding the outcomes of all the items together. None of the participants selected the negative formulated answer category “I have never seen a person that ate healthy food”, so we did not use this item in the total familiarity scale. Finally, we created a new categorical variable with the labels 1 = ‘High

familiarity with topic’ and 0 = ‘Low familiarity with topic”. The distinction between scores that were allocated to the high or low topic familiarity category was made

based on the median 5 of the scale (cumulative percentage of 68.10%).

Transportability scale. To measure the participants’ transportability we used

the widely used Transportability Scale (TS) (Dal Cin et al., 2004; Phillips &

McQuarrie, 2010; van Laer et al., 2012). This scale consists of 18 items, which could

be answered on a seven-point Likert scale, ranging from 1= ‘Strongly disagree’ till 7= ‘Strongly agree’. The items were in random order presented in the survey.

First, we recoded the negatively formulated items three, six and nine. Then,

we conducted a principal component analysis with Varimax rotation with all 18 items

(Q5__Transportability_1- Q5__Transportability_18). The analysis showed that there

were four components with eigenvalues higher than one (1.32, 1.50, 1.88 and 6.50),

which together explained 62.20% of variance. The scree plot showed an inflexion

after the first component. Therefore we selected the first component for the final

and consisted of 13 items with factor loadings higher than .45 (see table 8, Appendix

2). The new scale with had a Cronbach’s alpha of .91, which made it highly reliable.

Finally, we created a new categorical transportability variable with the labels 1

= ‘High Transportability’ and 0 = ‘Low Transportability’. The distinction between

high and low transportability was made based on answer category 4 = ‘neither agree or disagree’. All scores from 1-4 were coded as ‘Low Transportability’, and all scores 4-7 were coded as ‘High transportability’.

Education. We measured education with a multiple-choice item in the survey

ranging from no education to a Master’s degree. The choices were based on the

education levels used by the Centraal Bureau voor Statistiek ("CBS Statline", 2019).

Gender. We measured gender with a one-item multiple-choice question in the

survey. Participants could either choose ‘man’, ‘woman’, or ‘prefer not to

disclose/other’.

Dependent variables Brand awareness

Brand awareness was operationalized as brand recall and brand recognition.

Brand recall. The explicit brand awareness was measured with a cue-based

brand recall test (Rossiter, 2014; Srinivas & Roediger, 1990). With an open-ended

question we asked participants to name the two brands they saw during the

commercials breaks. We choose the answers ‘Taco Time’ and ‘TacoTime’, ‘Chipotel’

and ‘Chipotle’ as ‘adequate-for-choice’ answers. We coded these outcomes into two

categorical dependent variables, which respectively measured the recall of the

TacoTime brand and the recall of the Chipotle brand (1 = ‘recall’, 0 = ‘no recall’).

Brand recognition. This type of measurement can recall information from

therefore used to measure the explicit ánd implicit brand awareness. We used iconic

recognition with a one-item multiple choice question to measure brand recognition

(Rossiter, 2014). We used the same logos as in the commercials. The Chipotle logo,

the TacoTime logo and 10 brand logos of other taco restaurants were presented in a

random order to the participants. They could select a maximum of two brand logos

that they recognized from the commercial breaks. There was also an “I don’t

recognize any” option they could select, to avoid guessing and increase the reliability

of the question.

Brand image

We made a distinction between the explicit brand associations and the implicit

brand associations, to measure the brand image both recalled from the explicit and

implicit memory network. The items in each question about the explicit and implicit

Chipotle and TacoTime brand associations were presented in a random order.

Explicit brand associations

Based on the Dimensions of Brand Knowledge (Keller, 1993) we measured

the number, the strength, the favourability and uniqueness of the explicit brand

associations to measure the quantity and quality of the explicit brand association

network, and thus the explicit brand image.

Number of explicit brand associations. In this study we used the unstrained

free elicitation measurement, which is often used to identify possible brand

associations (Danes, Hess, Story, & York, 2010; Danes, Story, & Vorst, 2012) We

used the open-ended question “When thinking about Chipotle/TacoTime, what

associations comes to mind?”. After writing down a brand association, three

follow-up questions appeared asking about the strength, favourability and uniqueness of that

commercial and were encouraged to mention as many associations as possible. Empty

brand associations lines and their follow up questions were scored as missing, and

then calculated the total number of brand associations per participants per condition. This created two variables namely, ‘Number of explicit brand associations (Chipotle)’ and ‘Number of explicit brand associations (TacoTime)’.

Strength of explicit brand associations. The strength of the mentioned brand

associations were measured on a seven-point Likert scale on which participants

selected how well they thought the association was linked to the brand, anchored by

7=“Extremely well” and 1=“Not well at all” (Schnittka, Sattler, & Zenker, 2012).

Next, we calculated the mean association strength per participant per commercial and

subsequently created two variables namely, ‘Strength of brand associations (Chipotle)’ and ‘Strength of brand associations (TacoTime)’.

Favourability of explicit brand associations. The question “Do you think this association is positive or negative?” was answered on a seven-point Likert scale,

anchored by 7=‘Extremely positive’ to 1=‘Extremely negative’ (Schnittka et al.,

2012). We calculated the average association favourability per participants per

condition, and thus created the two variables ‘Favourability of brand associations (Chipotle)’ and ‘Favourability of brand associations (TacoTime)’.

Uniqueness of explicit brand associations. Participants answered the question “How unique do you think this association is?” on a seven-point Likert scale,

anchored by 7=‘Extremely unique’ to 1=‘Extremely ordinary’ (Schnittka et al., 2012).

We calculated the average association uniqueness per participant per condition. By doing so we created the variables ‘Uniqueness of brand associations (Chipotle)’ and ‘Uniqueness of brand associations (TacoTime)’.

A commonly used implicit measure is the Implicit Association Test

(IAT)(Calvert, 2015; Greenwald, Mcghee, & Schwartz, 1998). Important benefits of

this test are that it can be integrated in Qualtrics, which is used to make and distribute

our online experiment, and that participants do not need equipment to use it (Calvert,

2015). This makes this measure highly scalable and cost effective. Therefore, we used

an IAT integrated in Qualtrics in this study.

IAT research design. We used a seven-block counterbalanced Implicit

Association Test (Greenwald et al., 1998). With the IATgen tool we generated the

needed code to implement the IAT into Qualtrics (Carpenter et al., 2009). The same

tool also calculated the D-scores and did the reliability analyses. Due to the nature and

the duration of the IATs, we only measured the strength and favourability of the

implicit associations. The D-scores could range from 1 (strong link between

Healthy/Good and Chipotle) till -1 (strong link between Healthy/Good and

TacoTime). A D-score of approximately zero indicated no difference in the

association strength for the Healthy/Good association between Chipotle and

TacoTime. See Appendix 3 for the complete procedure of the IAT, the IAT

parameters and the calculation of the D-scores.

Strength of an implicit association. As input for this IAT we used the explicit

association, Healthy, which was most often mentioned in our pre-test. Another benefit was that ‘healthy food’ was also the main topics in both commercials. By presenting Healthy and Unhealthy in an IAT we could test which condition created the strongest

link between the Healthy association and their brand name. We thus measured the strength between the categories ‘Healthy’ versus ‘Unhealthy’ and the brands ‘Chipotle’ versus ‘TacoTime’. See Figure 6 (Appendix 3) for the full IAT (Healthy/Unhealthy) research design.

The average error rate of the Strength (Healthy/Unhealthy) IAT was .08. One

participant was dropped due to excessive speed from the final analysis. The estimated

internal consistency of the Strength (Healthy/Unhealthy) IAT, based on split-half with

Spearman-Brown correction, was .78. This makes the scale reasonably reliable.

Favourability of implicit associations. To measure the favourability of an

implicit association prior research used the categories ‘Good’ and ‘Bad’ (Brunel,

Tietje, & Greenwald, 2004; Greenwald et al., 1998; Karpinski & Hilton, 2001; Lane,

Banaji, Nosek, & Greenwald, 2007). This IAT was focussed on the link between the

favourable association ‘Good’ and unfavourable association ‘Bad’ and the brands ‘Chipotle’ and ‘TacoTime’. See Figure 7 (Appendix 3) for an example of the Favourability (Good/Bad) IAT and the category dimensions.

Two participants of the Favourability (Good/Bad) IAT were dropped due to

excessive speed. The average error rate of this IAT was .07. The estimated internal

consistency of the Favourability (Good/Bad) IAT, based on split-half with

Spearman-Brown correction, was .74, which makes the scale reasonably reliable.

Control variables

Finally, we controlled for some factors that are known to influence attention to

the ad, and thus the form of processing by asking a few control questions.

Prior exposure to the brands. By using foreign stimuli materials we tried to

control for the prior exposure to the advertisements. To check is we succeeded we

asked the participants in two items if they knew the brand and if they had seen the

advertisement before the experiment. Prior brand knowledge was answered with a 1= ‘Did know brand before experiment’ and 0 =’Did not know brand before experiment’.

Usage of product category. We asked about the usage of the product category

tacos with one item on the survey. Participants answered the question ‘How often do

you eat Tacos?’ with answers ranching from: 1 = ‘never’ to 8 =‘Once every 2-3 days’.

Results

First, outliers and undefined categories (such as ‘I do not wish to disclose my gender’) were labelled as missing values across all variables. The participants (N = 69) were of twelve different nationalities. Most of them were Dutch (n = 43) followed

by British (n = 8) and German (n = 4). Seventy-five precent of the participants was

female (n = 52). The average age of the respondents was 27.84 years old (SD = 8.61). Most respondents had a Master’s degree (n = 33). Participants had low topic

familiarity (n = 47) and had a low transportability level (n = 57). During the narrative

commercial (Chipotle), most participants experienced a high level of ad attention (n =

46). In contrast, most participants (n = 57) experienced a low level of ad attention

during the nonnarrative commercial (TacoTime). Just over half (56.50%) of the

participants stated that they knew the Chipotle brand before the experiment (n = 39),

compare to only 7.20% of participants who stated they knew the TacoTime brand

before the experiment (n = 5). Most respondents that already knew Chipotle stated

that they did not know Chipotle ‘well at all’ (21.70%) or just ‘slightly well’ (18.87%).

However, 72.50% of respondent (n = 50) stated that they had never seen the Chipotle

commercial before the experiment. Finally, the average participant ate tacos once

every half-year (M = 2.96, SD = 1.59).

Hypothesis 1 & 2

In hypothesis 1 we stated that narrative advertising, compared to nonnarrative

advertising, would have a positive effect on the level of narrative transportation. The

variables; one for the narrative transportation during the narrative condition

(Chipotle) and one for the narrative transportation during the nonnarrative condition

(TacoTime). On average participants (N = 69) experienced a higher level of narrative

transportation during the narrative condition (M = 4.07, SD = 1.11) than during the

nonnarrative condition (M = 2.47, SD = .86).

In hypothesis 2 we stated that the categorical moderators ‘Topic familiarity’, ‘Transportability’ ‘Ad attention Chipotle’, ‘Ad attention TacoTime, ‘Education’ and ‘Gender’ had a positive influence on the relation between the condition and narrative transportation. We conducted a factorial repeated-measures ANOVA analysis to test

hypothesis 1 and 2. The ‘Narrative transportation’ (narrative & nonnarrative) and ‘Ad

attention’ (narrative & nonnarrative) variables were selected as within-subjects variables. ‘Topic familiarity’, ‘Transportability’, ‘Education’ and ‘Sex’ were selected as between-subjects variables and the continuous variables ‘Prior exposure

(Chipotle)’, ‘Prior exposure (TacoTime)’ and ‘Product usage’ were selected as control

variables. Sphericity is only an issue if there are more than two levels. Our

within-subjects variables only had two levels. That is why sphericity was met in our study and we looked at the ‘sphericity assumed’ row of the within-subject effects table (Field, 2017, p 475).

Narrative transportation. The test of within-subject effects revealed that the

participants experienced significantly higher levels of narrative transportation in the

narrative condition, than in the nonnarrative condition, F (1, 43)= 217.13, p < .000,

partial η2 = .84. The partial eta squared values showed a strong positive effect of the

advertisings strategy on narrative transportation and therefore hypothesis 1 was

confirmed.

Moderators. As for hypothesis 2, we found a significant, but weak positive

between the advertising strategy and the level of narrative transportation, F (1, 43) =

6.43, p = .015, partial η2 = .13. Participants with high levels of transportability scored

higher on narrative transportation in the narrative condition than participants with a

low level of transportability. This moderator effect of Transportability partially

confirmed Hypothesis 2.

The test of within-subjects effects also revealed ‘Ad attention’ as a second

moderator to have a weak positive effect on the relation between ‘Condition’ and “Narrative transportation’, F (1, 43) = 4.42, p = .042, partial η2

= .09. Participants

with a high level of ad attention therefore scored a higher on narrative transportation

in the narrative condition than participants with a low level of ad attention. This also

partially confirmed hypothesis 2.

Control variables. Concerning the control variables, we found a significant

interaction effect concerning ‘Prior exposure (Chipotle)’, F (1, 43) = 5.02, p = .030,

partial η2 = .11. This means that participants who knew Chipotle, showed a

significantly higher level of narrative transportation during the narrative condition

than participants who did not know Chipotle.

Additional main effect. Finally, we also found a main effect of condition on ad

attention in the test of within-subjects effects. Contrasts revealed a mediocre positive

relation between the condition and the level of ad attention, F (1, 43) = 10.71, p =

.002, partial η2 = .20. This means that the narrative condition evoked a significantly

higher level of ad attention than the nonnarrative condition.

Hypothesis 3

In hypothesis 3 we stated that higher narrative transportation would lead to

less brand recall than lower narrative transportation. More participants (n = 44) could

nonnarrative condition (TacoTime) (n = 13). On average participants experienced a

higher level of narrative transportation during the narrative condition (M = 4.07, SD =

1.11) than during the nonnarrative condition (M = 2.47, SD = .86). In the nonnarrative

condition 94% of the participants had a score below 4 (which indicates a low level of

narrative transportation). The responses in the narrative condition had a broader

variance, as 45% of the scores were below 4 and the rest of the scores were above 4

(which indicates a high level of narrative transportation). When we conducted the

binary logistic regression analyses and the linear regression analyses for both

conditions (to answer hypothesis 3, 4, 5 and 6) only the models based on the data of

the narrative conditions found significant results. See Appendix 4 for the full analysis

of hypothesis 3, 4, 5, and 6 separately for the two narrative transportation conditions.

The lack of significant results of the models based on the data of the nonnarrative

condition was probably due to the low variance of the nonnarrative transportation

(TacoTime) data. The ‘Narrative transportation (Chipotle)’ had a variance of 1.24 (M

= 4.07, SD = 1.11) compared to a low variance of .74 of the variable ‘Narrative transportation (TacoTime)’ (M = 2.47, SD = .86). These numbers indicated that there were simply no high narrative transportation scores in the nonnarrative condition,

which could be analysed and this probably led to non-significant results.

New ‘Narrative transportation (total)’ measure

To tackle this variance problem, we created eight new variables, in which we

combined the data of both conditions and thus treated the repeated scores as

independent scores. We made a ‘Narrative transportation (total)’ variable, which first the listed narrative transportation scores of participants during the narrative condition,

and then the narrative transportation scores of participants during the nonnarrative

‘brand recognition’, ‘number of brand associations’, ‘strength of brand associations’, ‘favourability of brand associations’, ‘uniqueness of brand associations’ and ‘prior brand exposure’. We made sure that one participant’s narrative transportation score

during the Chipotle commercial was linked to the recall score of Chipotle, and vice

versa for TacoTime (see figure 8, Appendix 4). The new ‘Narrative transportation (total)’ variable ranged from 1 till 6.55 (M = 3.27, SD = 1.27) on a scale from 1 = ‘Low narrative transportation’ till 7 = ‘High narrative transportation’. Within the new Narrative transportation variable 51% of participants had a score of 3 and below. The

data had a better variance of 1.63 (M = 3.27, SD = 1.27) than the data in the old

variables. We used these new variables to test hypothesis 3-6. First, we conducted two

binary logistic regressions to predict recall and recognition in order to analyse brand

awareness.

Brand recall (total). A test of the model with our predictors versus the model

with only the intercept was significant, χ2

(8, N = 138) = 25.41, p < .001. The data

fitted the model well according to the Hosmer and Lemeshow test, which was not significant, χ2

(8, N = 138) = 1.38, p = .995. The total of the independent variables

explained 42% of the variance in the dependent variable (Nagelkerke R2 = .42). As

can be seen in table 18 (Appendix 5), the variable ‘Narrative transportation (total)’ and ‘Prior brand exposure (total)’ had significant partial effects. The odds ratio showed that when a participant knew the brand before the exposure, they were 8.26

times more likely to recall the brand and when a participant’s ‘Narrative

transportation (total)’ score increased with one, then the participant was 2.18 times

Hypothesis 4

In hypothesis 4 we stated that higher narrative transportation would lead to

more brand recognition than lower level narrative transportation. More participants

recognized the brand in the narrative condition (Chipotle) (n = 63) than the brand in

the nonnarrative condition (TacoTime) (n = 57). To test this hypothesis, we analysed

if the variables ‘Narrative transportation (total)’, ‘Prior brand exposure (total)’, and ‘Product usage’ had a predictive value on the probability of participants’ ability to recognize a brand. Our model with all the above predictors had no higher predictive

value than a model with just the intercept, χ2 (8, N = 138) = 14.74, p = .064. But, a

model with only ‘Narrative transportation (total)‘ as predictor did increase the

predictive value, compared to the model with only the intercept, χ2 (8, N = 138) =

5.03, p = .025. The data fitted the model well, as the Hosmers and Lemeshow’s test was not significant, χ2

(8, N = 138) = 14.44, p = .071. The total of the independent

variables explained 15% of the variance in the dependent variable (Nagelkerke R2 =

.15). Narrative transportation (total) had a significant strong positive effect (see Table

19, Appendix 5). The odds ratio showed that when a participant’s narrative

transportation score increased with one, they were 2.28 times more likely to recognise

the brand. These results showed that higher narrative transportation lead indeed to

more brand recognition than lower narrative transportation, and confirmed hypothesis

4.

Hypothesis 5

The fifth hypothesis stated that higher narrative transportation leads to a less

effective creation of explicit brand associations within consumers’ minds than lower

levels narrative transportation. The effectiveness of creating brand associations was

brand associations. On average participants could recall 3.11 (SD = 1.66) Chipotle

brand associations, compared to 2.65 TacoTime brand associations (SD = 1.74). The

average strength per association was almost equal in the nonnarrative condition (M =

5.52, SD = 1.14), as in the narrative condition (M = 5.50, SD = 1.24). The

favourability (M = 5.19, SD = 1.33) and uniqueness (M = 3.86, SD = 1.44) of explicit

brand associations was higher in the narrative condition than in the nonnarrative

condition (M = 4.20, SD = 1.42; M = 2.84, SD = 1.45). To analyse if these findings

were significant and caused by the level of narrative transportation we conducted four

linear regression analyses. In all four analyses we entered the ‘Prior exposure (total)’ and ‘Product usage’ as control variables. There were no outliers. The histograms of all the variables in the four analyses showed a normal distribution. The scatterplots with

the variables showed a linear line and the scatterplots with the residues of the

different variables showed homoscedasticity for all four analyses.

Number of explicit brand associations (total). The independent variable

‘Narrative transportation (total)’ and control variables predicted the dependent variable ‘Number of brand associations (total)’, F (3, 65) = 11.69, p < .000. These variables together explained 35 % of the variance in de dependent variable (R2 = .35). The coefficients table revealed that ‘Narrative transportation (total)’ had a significant strong positive effect, b* = .57, t = 5.71, p < .000, 95% CI [.65, 1.16] (see Table 20,

Appendix 5). When a participant scored one value higher on the narrative

transportation scale, then the amount of mentioned brand associations increased with

.86.

Strength of explicit brand associations (total). The model consisting of the

independent variable ‘Narrative transportation (total)’ and control variables did not predict the dependent variable ‘Strength of explicit brand associations (total)’ as the ANOVA was not significant, F (3, 65) = 2.62, p = .059. Therefore, we found no

significant effect of narrative transportation on the strength of explicit brand

associations.

Favourability of explicit brand associations (total). Our model with the

independent variable ‘Narrative transportation (total)’ and the control variables had

predictive value over the dependent variable ‘favourability of explicit brand associations’, F (3, 65) = 3.86, p = .014. In total these independent variables

explained 16% of the variance in the dependent variable (R2 = .16). Both ‘Narrative transportation (total)’ (b* = .27, t = 2.29, p = .026, 95% CI [.05, .67]) and ‘Prior brand exposure (total)’ had a significant weak effect on the favourability of brand

associations, b* = -.27, t = -2.23, p = .029, 95% CI [-1.36, -.08] (see table 21,

Appendix 5). When a participant scored one value higher on the narrative

transportation variable, their average favourability increased with .36. But, if the

participants knew the brand before the experiment, then the favourability decreased

with .72.

Uniqueness of explicit brand associations (total). Finally, the model with

‘Narrative transportation (total)’ and the control variables showed predictive value over the ‘Uniqueness of explicit brand associations (total)’, F (3, 65) = 3.50, p = .021. The total of our independent variable and control variables explained 15% of the

variance in the uniqueness of the explicit brand associations (R2 = .15). We found

only a mediocre significant negative effect of the ‘Prior brand association (total)’

variable, b* = -.36, t = -3.02, p = .004, 95% CI [-1.76, -.36] (see table 22, Appendix

5).

In contrary to hypothesis 4, higher narrative transportation led to a higher

number of and more favourable explicit brand associations than lower narrative

Hypothesis 6

In hypothesis 6 we stated that higher narrative transportation leads to a more

effective creation of implicit brand associations within consumers’ minds than lower

narrative transportation. On average participants showed a stronger link between the

term Healthy and Chipotle than between Healthy and TacoTime (M = .10, SD = .39).

On average participants showed a stronger link (M = .19, SD = .39) between Good

and Chipotle, than between Good and TacoTime. We conducted two linear

regression analyses to explore if these results were significant and included the control variables ‘Prior brand exposure (total)’ and ‘Product usage’. The scatterplot for both analyses showed a linear line, and the scatter plot with the residues had

homoscedasticity. The histograms revealed a normal distribution of the data in both

analyses.

Strength of implicit brand associations. The model with ‘Narrative

transportation (total)’ and the control variables was not significant, F (3, 64) = 2.52, p = .066. But a model with only ‘Narrative transportation (total)’ as a predictor was significant, F (3, 66) = 5.18, p = .026. This model explained 7% of variance in the

strength of the implicit association (R2 = .07). We found a significant, but weak effect of ‘Narrative transportation (total)’, b* = .27, t = 2.28, p = .026, 95% CI [.01, .19] (see table 23, Appendix 5). When participants scored one value higher on the

narrative transportation scale, then the strength between Chipotle and the Healthy

association increased with .10.

Favourability of implicit brand associations. The model with ‘Narrative

transportation (total)’ and control variables was not significant, F (3, 63) = 2.67, p =

.055. However, a model with only ‘Narrative transportation (total)’ as a predictor was

significant, F (1, 65) = 7.30, p = .008. This model explained 10% of the variance in