University of Cincinnati

U UNIVERSITY OF CINCINNATI

Date:

I, , hereby submit this original work as part of the requirements for the degree of:

It is entitled:

Student Signature:

This work and its defense approved by:

Committee Chair:

Approval of the electronic document:

I have reviewed the Thesis/Dissertation in its final electronic format and certify that it is an accurate copy of the document reviewed and approved by the committee.

Committee Chair signature:

Hearing, Remembering, and Branding: Guidelines for Creating Sonic Logos

A dissertation submitted to the Graduate School of the University of Cincinnati in partial fulfillment of the requirements for the degree of

Doctor of Philosophy (Ph.D.)

in the Department of Marketing of the College of Business Administration by

Vijaykumar Krishnan Palghat

PGDM, Indian Institute of Management, Calcutta B. Tech., Civil Engineering, Indian Institute of Technology, Delhi

Committee

Dr. James J. Kellaris (Chair) Dr. Frank R. Kardes Dr. Wei Pan Abstract

Sonic Branding is the strategic use of sound to create an authentic auditory identity for the brand. Conventional applications of sound in branding are tactical and lean on classical conditioning theory by repetitive pairing of sound and brand to create desired associations. In contrast, sonic branding leans on processing fluency theory leveraging sound as information in and of itself. Often such auditory information is nonverbal and nonlinguistic. Sonic logos are good examples illustrating this phenomenon.

A sonic logo, “sogo,” the auditory analog of a visual logo, is a typical sonic branding device. Sogos are short melodies not lasting more than six seconds. Some interesting examples are the 5-tone Intel sogo, windows vista’s 4-tone start-up chime and NBC’s 3-tone sogo. Sogos vary in their design characteristics. They may comprise different number of tones. They may have an ascending pattern (Windows Vista), descending pattern (windows XP) or a zigzagging contour (Intel). A sogo may be easier to remember because it comprises ‘chunks’ (Miller 1956) of similar tones. Thus, number of tones they comprise, their contour and their chunkability may characterize Sogos.

Per logo literature (Henderson and Cote 1998), good sogos should engender favorable consumer responses on recognition, affect, and familiarity dimensions. For instance, sogos with fewer tones should be easier to remember; thus obtain high true recognition on a subsequent encounter. On the other hand, because they are easier to process, they may engender illusions of familiarity (Whittlesea 1993) leading to high false recognition. Sogos with a zigzagging contour may be more difficult to recall but may be perceived novel and so liked more. In other words, consumers experience differing levels of ease in processing sogos based on the design characteristics.

iii

This subjective experience of ease of processing (Reber, Wurtz and Zimmerman 2004;

Whittlesea 1993; Janiszewski and Meyvis 2001; Winkielman et al 2003) incoming auditory

information is misattributed to the judgment at hand: Familiarity (Whittlesea 1993), Positive

affect (Reber, Winkielman and Schwarz 1998; Winkielman and Cacioppo 2001), judgments of

truth (Reber and Schwarz 1999) and brand Evaluation (Lee and Labroo 2004). This research explores the systematic influence of three design characteristics of sogos : number of tones, contour, and chunkability across five studies on response dimensions. Overall, these studies evidence processing fluency mediation of these influences.

Results show that several response dimensions vary systematically with the sogo design characteristics, thus providing for guidelines. Leader brands would want a high true recognition and a low false recognition; brands in a low involvement product segment (e.g., bread) could profit from high false recognition and illusions of familiarity. Huge investments are made to create and air auditory branding stimuli; rights for use of popular songs in commercial jingles

may top $500,000 such as for “stand by me,” deployed by Citibank, (Krasilovsky and Shemel

2007). In conclusion, given that sogos are branding devices, guidelines from this research should reduce the precarious dependence of marketers on musicians (Bruner 1990), and provide for greater precision over sonic branding.

Vijaykumar Krishnan Palghat

Acknowledgements

First, I thank Prasanna, my wife, for her continued love, patience, encouragement and

support for twenty five years now without which any and all of my achievements, including this dissertation would be impossible and quite meaningless. I thank my two daughters Aditi, and

Anaga for their beautiful presence in our lives that makes this exercise at all worthwhile.

Next, I thank my parents Smt. Akhila Krishnan and Shri P.N. Krishnan at whose feet I learnt music to the extent I could. I also thank violin vidwan Shri M.S. Sundaresan at whose feet

I learnt to play the violin to the extent I could.

I thank all my teachers and professors at DTEA Sr. Secondary School, Indian Institute of

Technology Delhi, Indian Institute of Management Calcutta and University of Cincinnati, for imparting knowledge and wisdom to me with sincerity and grace.

I thank Dr. James Kellaris, my committee chair for his excellent guidance throughout this program. I also thank Dr. Frank Kardes and Dr. Wei Pan, my committee members for their valuable suggestions and guidance in this dissertation research.

I thank Dr. Bruce Pfeiffer, University of New Hampshire, for his wonderful voiceovers in the stimuli used in study 1.

I thank Doina Chichernea, University of Cincinnati, for her valuable help with development of the bootstrap SAS routine.

Finally, as Saint Thyagaraja observes in his immortal composition in Raga Sri “EndarO mahAnubhavulu antarIki vandanamulu”

… To the innumerable great souls [known or unknown, of past present or future],

…[my] salutations.

Table of Contents

Chapter 1 Introduction 1

Chapter 2 Literature Review 9

1. Consumer Responses to Musical Stimuli 9

2. Processing Fluency 14

Chapter 3 Defining Sonic Branding 19

Chapter 4 Theory and Hypotheses 24

Chapter 5 Research Studies 39

1. Study 1 41

2. Study 2 45

3. Study 3 50

4. Study 4 58

5. Study 5 64

General Discussion 69

Beyond this Dissertation 74

Some Final Thoughts 79

References 82

Appendices 96

Appendix A 96

vii

Chapter 1

Introduction

Sound, in its many forms, plays a central role in branding. Distinctive audio logos such as the Windows chimes or Nokia ring tones are well known throughout the world. Older readers will recall the three-note NBC television network chimes and the ascending scale accompanying the spelling of J-E-L-L-O. Jingles such as “I'm Chiquita banana and I've come to say,” “Pepsi cola hits the spot,” and Coca-cola’s “I’d like to teach the world to sing” achieved the status of cultural icons. Wallace (1991) cited the Oscar Meyer song, noting that “Just saying the brand name is usually enough to start… the jingle playing through your mind” (p. 239).

Yet, despite its undeniably central role, sonic branding is yet to be unambiguously defined and consequently there is a paucity of systematic research on sonic branding. This leads to what Bruner (1990) terms a ‘precarious dependence’ of marketers on musicians for sonic designs. Although the identity of brands is often expressed both visually and sonically, branding research has been almost exclusively conducted in the visual domain. For example, expounding on the marketing strategy of planned visual communications, Sthal (1964) emphasizes visual synergy across brand advertisements, packaging, label tags, catalogs and brochures for visual brand identity. Krugman’s (1965) research on influence of TV advertisements on consumer visual behavior and consequent learning under low involvement is another early example focusing on visual domain.

This inadvertent preference for visual stimuli surfaces in several laboratory experiments.

Lichtenstein and Srull (1985) conducted experiments to disentangle on-line versus memory- based judgments. Respondents were presented an informationally heavy product print ad. They were either required to form an evaluation of the product or an evaluation of the advertisement in

terms of grammar and style. Subsequently, all respondents were asked for their product

evaluation at session 2. Results showed greater correlation among judgments in the product

evaluation condition because these respondents simply retrieved previously formed judgments

while those in the advertisement evaluation condition had to retrieve individual attributes at session 2 from memory and then form judgments. Perhaps this experiment could have been operationalized by exposing respondents to a radio spot; i.e., stimulus mode need not interact with this pattern of results. On the other hand, it is an empirical question if this distinction between on-line versus memory-based judgments holds across different stimulus modalities.

In a study investigating attitude formation toward unattended stimuli, Janiszewski (1988) manipulated the placement of the picture and demonstrated that brands are more positively evaluated when picture is on the left of the brand name and therefore in focus with respect to the compatible right hemisphere. Janiszewski (1988) notes the examples of right hemisphere compatible tasks to include processing of both of music and visuospatial information. That is, left hemisphere is good at recognizing individual pieces of information while the right hemisphere is good at integrating holistic information auditory or visual. However, the chosen experimental stimuli were visual in nature. Given that music is processed holistically by the right hemisphere, perhaps an alternative way to operationalize this experiment might have been to play (or not) music that is affectively equivalent to the picture and have respondents rate the brand name.

Thus, several studies – pre-attention (Krugman 1965), subliminal processing (Janizewski

1988), salience effects on judgment of stimuli (Fiske and Taylor 1991), camera angles and

salience effects (Meyers-Levy and Perrachio 1992), -- exemplify this preference for visual

stimuli, although branding stimuli are presented as often in auditory mode in practice.

Huge investments are made for creation and airing of auditory branding stimuli. Rights

for popular songs for use in commercial jingles may top $500,000 such as for “stand by me,”

deployed by Citibank, “Bye Bye Blackbird,” for Pontiac commercials (Krasilovsky and Shemel

2007, p261). Microsoft incurred a licensing cost of $12 million for the song by Madonna “Ray of

Light,” to launch Windows XP (Krasilovsky and Shemel 2007). Further, a 30-second spot may cost anywhere between $250,000 for a moderately popular program to $705,000 for a top program such as American Idol (Krasilovsky and Shemel 2007). These levels of investments underscore the need for intensified research in sonic branding.

Sound as information

Interestingly, other fields of enquiry e.g., information sciences are extensively exploring the ‘sound as information’ idea. According to the NSF report (Kramer et al 1997) by the

International Community for Auditory Display (ICAD), ‘sonification is the use of non-speech sound to convey information.’ This report considers sonification to be a multi-disciplinary research with a significant component of psychological research in perception and cognition.

ICAD explores the innately integrative properties of auditory information. Whereas visualization techniques of high-dimensional scientific data are reaching limits of comprehension

(Kramer et al 1997), the integrative features of auditory information open versatile possibilities.

For example, audification (Hermann 2002) is the mapping of ordered data such as seismic data

into sound streams that enable ‘hearing’ for unusual rumblings of a volcano rather than looking

for unusual data patterns. Earcons and auditory icons (Hermann 2002) are used for auditory

display of information at the man-machine interface. While earcons are simple beeps that alert,

auditory icons exploit our meaning for everyday sounds, e.g., the sound of a bottle filling up to

indicate the progressing download in the background. Hermann (2002) shows exploratory data

3 analysis in a multi-dimensional space by mapping data to sound parameters. Whereas, Geiger counters and Pulse-oxy meters are traditional examples of use of sound as information, these recent developments at the behest of ICAD indicate revival of interest in informational properties of sound.

Sound-ready audience

This renewed scientific interest in use of sound as information elsewhere, is matched by a more sound-ready audience. Consumers are becoming more sonically perceptive because of the explosive proliferation of electronic gadgets in their lives. These gadgets have multiplied consumers’ access to auditory interfaces. In the book Sonic Branding, Jackson (2003) refers to auditory interfaces as “sonic touch points.” Radio, TV, cell phone, music played during telephone hold, and customer interactions in an IVR (Interactive Voice Response) system are but a few examples of the fourteen touch points Jackson identified. This access has powered the ten- fold growth in music industry from 4 billion dollars in 1970 to 43 billion dollars in 2006

(Krasilovsky and Shemel 2007). In 2006, Americans spent over twenty hours per week on radio, downloaded over 50 million ring-tones and bought iPods for $12 billion (Krasilovsky and

Shemel 2007).

Consumers in the past often encountered simplistic sonic information, such as busy versus ring tones. Today consumers routinely process complex sonic information, such as arrival of an SMS or an email. They are reassured that their remote just locked their car even as they walk away form the just-parked car immersed in animated conversation. They routinely respond to “wear your seat belt” beep before turning on the ignition. They easily keep track of the customer associate ringing their purchases at the checkout, even while visually distracted. Sound card, a standard component in computers has contributed to the proliferation of sonic cues. For

4 instance, ‘sonification’ of a scroll bar that mimics upward (downward) scroll with increasing

(decreasing) pitch allows users to process the textual information visually and the scroll bar movement sonically (Brewster 1997) enhancing consumer experience. Surrounded and immersed thus in sound, consumer is all ears today – just a shout away.

Therefore, given these new insights on sound as information and easy access to sonically perceptive consumers, might sound provide identifying information for brands, qualify brand awareness, brand recognition, brand recall or contribute to the distinctiveness and brand personality? Sonic touch points are important for branding, because whereas an audience can close their eyes or turn their head away from a visual stimulus, they cannot shut their ears.

Auditory attention is largely involuntary (Eyenck 1982). Moreover, despite the fact that visual processing tends to dominate (Colavita 1974), accompanying sound can shape the perception of visual information (Boltz 2001, 2004). However, unlike other non-verbal elements of branding

(e.g., visual logos), despite its frequent use in practice, sound is relatively underrepresented in the branding literature.

Sonic Logo -- Sogo

Perhaps the most basic form of sonic branding is the sonic logo, a “sogo.” A sogo is a unifying, focal sonic branding device. Whereas brand knowledge has been conceptualized as an associative network memory involving brand recognition and brand image (Keller 1993), a logo may be thought of as a compact “zip file” of this brand knowledge network. Corporate identity literature views a logo as an entity’s signature on its materials (Snyder 1993). Analogously, I define a sogo to be a sonic branding device that plays the role of a short distinctive auditory signature lasting between three to six seconds. A sogo is the auditory analog of a visual logo that

works as an auditory zip file. The 5-tone (G↑-C↑ -F↑ -C↑- G↑) IntelTM is a sogo. Nokia’s 5-tone

(G –F- E -G -C) is another example.

Sogo duration: three to six seconds

I recommend the lower limit based on extant popular exemplars; the sogos for Intel and

Nokia, the Yahoo Yodel and the Jell-O audio signature are all three seconds long. Indeed, Intel’s

brief to the Austrian composer Werzova who created the Intel’s five-tone sogo was in fact for

three-second duration (Droney 2004). Research on thin slice judgments (Ambady and Rosenthal

1993) shows that respondents are able to make accurate judgments of personality of targets based on thin slices of silent videos as brief as only six seconds long. A sogo is a thin slice of non-

verbal brand information. Thus, insofar as the sogo is a short mnemonic and not the message

itself three to six seconds duration appears appropriate.

The sogo ‘activates’ the entire brand knowledge that includes brand attributes, brand

benefits and brand attitudes. It may also make accessible the brand’s jingle if any, and other communication attributes that already exist for the brand. Thus, the sogo is not the message.

Rather, it is a short (three to six seconds) potent cue that activates and renders the entire brand message accessible.

The Chiquita jingle is not a sogo, although it is sonic branding. An appropriate three to six seconds sonic sub-set carved off that jingle (i.e., a musical phrase, such as the cadence) may be a retrospective manifestation of a sogo. A multi product company may have different jingles for each product, tied together by the corporate sogo. For instance, an adaptation of the Chiquita jingle wrapped around the same sogo may convey a different meaning for a new brand extension, while retaining the overarching sonic identity. Thus, a brand extension of a smooth, thick, banana-flavored Chiquita yogurt may have a jingle that is congruently slow paced, but wrapped

around the same sogo. As another example, sonic branding for different models of the Nokia

phones may be different but related musically as a family, signed off by the same sogo. Finally,

while acknowledging that non-musical sounds like the Elsie’s moo or the MGM lion roar may be

effective sogos, given their greater prevalence and versatility I limit my exploration in this

dissertation to musically based sogos.

Whereas a logo is a summary brand expression visually, a sogo is the most parsimonious

brand expression in sound. Therefore, research on sogo will inform us of the design

characteristics, and responses to sonic branding at its most rudimentary atomistic level. In

investigating the properties of a good logo, Henderson and Cote (1998) identify memorability

captured as true or false recognition, likeability and subjective familiarity at first exposure to be

some of the defining response characteristics. They then identify, using experimental aesthetics procedures, the design characteristics of the logos manifesting in these desirable responses. In this dissertation research, I extend and contribute to this stream of literature in five ways.

First, I incorporate the dependent variables study in the auditory domain and find interesting similarities. Some sogos are easier to remember just as some visuals are easier to recall (true recognition). Some sogos engender illusions of familiarity just as some visuals are

(false recognition). Some sogos are more liked than others are.

Second, I conceptualize recognition as stimulus detectability and deploy d-prime analysis

(Singh and Churchill 1986; Singh, Rothschild and Churchill 1988) thus linking true and false recognition. Higher detectability (sensitivity) should simultaneously imply a higher true

recognition rate and a lower false recognition rate to stimulus. For instance, a given stimulus

might obtain both higher true and false recognitions simply because of participants’ affirmative

bias; i.e., both hit rates and false alarm rates are high although overall sensitivity is low. D-prime

analysis teases this apart. This is a refinement over the approach adopted by Henderson and Cote

(1998) who cast true and false recognitions as independent aspects of recognition of stimuli.

Third, while the Henderson and Cote (1998) study is a significant contribution and

managerially relevant in terms of understanding implications of design characteristics on

response dimensions of logos, this study demonstrates these effects without delving very deeply

into explaining why they occur. Later developments in processing fluency literature suggest

possibilities for further investigation and conceptual depth.

Fourth, although I use the same set of dependent variables (with the signal detection

refinement of recognition), I use design characteristics (independent variables) relevant for

sogos. Unlike the Henderson and Cote (1998) study where the design characteristics are measured, I manipulate the design characteristics through lab experiments thus providing for

greater conviction on causality.

Finally, I incorporate some methodological refinements. Because several stimuli are rated

by respondents, a multilevel-modeled regression is appropriate to account for repeated measures.

I control for order effect by randomizing stimulus presentation and I minimize carry-over effects

through distraction tasks between stimuli presentations.

Thus, my dissertation makes a significant contribution to this stream in a) exploring an

alternative (auditory) domain, b) refining dependent variables on recognition with sensitivity

analysis, c) enquiring processing fluency mediation, d) using an experimental manipulation of

the design characteristics and e) incorporating multilevel modeling for methodological

robustness.

Chapter 2

Literature Review

Consumer Responses to Musical Stimuli and Auditory Imagery

Research on listener responses to music enjoys a long and rich history. This is not an

exhaustive review of this vast body of literature. Reader is referred to Kellaris (2008), the most

recent review of literature on music and consumer responses – itself an update, built upon the

seminal review article titled “Music, Mood and Marketing by Bruner (1990) and the book The

Social Psychology of Music by Hargreaves and North (1997), especially the chapter on “Music and Consumer Behavior.” Likewise, the book Auditory Imagery by Reisberg (1992) is an excellent repository of an equally rich research stream exploring a somewhat different question.

The goal in this section is to highlight the extent and scope of extant knowledge by reviewing illustrative examples that have a more direct bearing on this dissertation.

Many studies in consumer psychology have followed the approach established by early research in psychology (e.g., Heinlein, 1928; Gundlach, 1932, 1935; Hevner, 1935, 1936, 1937;

Rigg, 1940; Henkin, 1955, 1957), in which listener responses are examined as a function of

objective properties of music, such as tempo (Holbrook and Anand 1990), pitch, and their

combinations (Kellaris and Kent, 1991, 1992; Schubert, 2004). Numerous studies have examined

influences of music on consumers (Bruner 1990; Hargreaves and North, 1997). These include

the study of the roles of music in ads (Gorn, 1982; Kellaris and Cox, 1989; Kellaris, Cox and

Cox, 1993), in retail settings (Mattila and Wirtz 2001; North and Hargreaves 1998; Turley and

Milliman, 2000), the influence of music on time perceptions (Kellaris and Kent, 1992; Mantel

and Kellaris 2003), product perceptions (Zhu and Meyers-Levy 2005), co-varying personality

traits of the music and the listener (Rentfrow and Gosling 2003, 2006) and influence of in-store

music on consumer preferences (wine selections) by country of origin (North, Hargreaves and

McKendrick 1999). One way to organize this body of research is in terms of the response

dimensions. Consumer responses to musical stimuli have shown a wide variation along cognitive, affective and behavioral dimensions (Kellaris 2008).

It is well recognized that music evokes emotions and arguably, music from across cultures elicit uniform effects. The mean scores or ratings (higher score implied a rating closer to the emotion intended a priori by the music) across three genres of music viz. Japanese, Western and Hindustani for three emotions joy, anger and sadness showed strikingly similar results

(Balkwill, Thompson and Matsunaga 2004) among Japanese. Musical stimuli differing in major and minor mode may respectively evoke joyful or melancholic moods even among musically uninitiated (Hevner 1935). Specific structural constituents of music – tempo, tonality and texture

(Bruner 1990) may vary in the emotions they evoke.

Kellaris and Kent (1992) show that whereas tempo affects pleasure and arousal dimensions and tonality affects pleasure and surprise dimensions, texture moderates effects of tonality on pleasure. Recent research (Alpert, Alpert, and Maltz 2005) shows that equally liked music could nevertheless evoke divergent mood responses depending on its constituent structural properties of tonality, the consequent evoked mood coloring purchase intentions for products differentially. A happy mood induction led to increase in purchase intention for a birthday card – a happy occasion – while a sad mood induction led to increase in purchase intention for a get- well-soon card. These studies support the notion that music can be deconstructed into constituent elements, characterized along these dimensions and engineered to deliver specific responses.

That is, reduce marketers’ dependence on musicians at least in sonic strategy formulation for the brand if not its execution.

Experiments conducted in the visual perception and brand choice streams enjoy the advantage that visual stimuli are perceptually sustainable for any desired duration. For instance, in an experiment investigating truth effects, influence of ease of readability on believability is examined by presenting factually correct or incorrect statements presented in either bright or dull contrasts thereby varying readability (Reber and Schwarz 1999). Results demonstrated illusions of truth effects; easily readable statements were evaluated to be truer. This experiment allowed respondents to view the stimuli as long as they wished before making the judgment, i.e. the stimulus was available for continued perception.

Visual stimuli are spread in space while auditory stimuli are spread in time (Halpern

1992). Were we to conduct an analogous experiment for truth effects in the auditory domain, the question, how would the respondent consult the sonic stimulus heard in perceptual field, arises.

Assume that respondents across groups hear a statement with varyingly degraded acoustic clarity and evaluate the truthfulness of the statement. Once they hear the statement, the stimulus vanishes from the perceptual field. On what basis would they judge? A plausible answer is that the respondent would imagine what he heard and base his response on that experience.

Experiments in the auditory domain need to recruit auditory imagery.

Intons-Peterson (1992) defines auditory imagery to be “introspective persistence of an auditory experience, including one constructed from components drawn from long-term memory, in the absence of direct sensory instigation of that experience.” Typical experimental processes in the auditory domain involve comparisons between recently presented stimulus with another stimulus, and an assessment of similarity between them (Halpern 1992). Notice that the previous

11 stimulus is not available simultaneously. It has to be imagined. Importantly, even the most recent stimulus is not available and needs to be imagined as well. However, are processes in auditory imagery functionally equivalent to those in perception?

Halpern (1984) conducted several experiments by varying one element of tempo, contour, or tonality and evaluated levels of correct subsequent identification in each case.

Analysis of results showed that respondents discerned contours very well but were often confused between musical fragments pitched in different tonality. This suggests that musical contour is a more enduring characteristic.

In other experiments, Halpern (1988a) showed that temporal extent of musical imagery is equivalent to that in real-time. Respondents were given start and target words from popular songs (e.g., “The Star-Spangled Banner”). The target words were either original or similar words to that in the song. The start words were typically at the beginning of the song and the target words were several pre-assigned beats away i.e., the ‘step-size’ (Halpern 1988a) was varied.

Respondents saw the start-word and target-word appear on a computer screen and judged whether the target-word truly belonged be hitting ‘Yes’ or ‘No’ buttons. In addition, some respondents were told to start mentally playing the song beginning at the start-word. Results showed a response time delay that increased with the step-size. Interestingly, this pattern obtained with even respondents not asked to imagine. In a variation, Halpern (1988a) required respondents to identify if the target-word was pitched higher or lower in order to eliminate the possibility that respondents may have merely processed a word list rather than the music itself.

The response time pattern replicated. Clearly, respondents were imagining the music in real- time.

Farah and Smith (1983) asked participants to imagine a tone at 715 Hz. or at 1000 Hz.

Respondents subsequently judged a target pitch. When the target pitch was same as the imagined pitch it led to a priming effect such that lesser loudness was required to identify the target.

Weber and Brown (1986) demonstrated the equivalence of auditory perception and imagery for pitch. Participants were instructed to draw lines with heights indicative of higher/lower pitches as they either imagined or sang (perception). No significant differences were observed in the pattern of lines.

In an experiment investigating imagery of timbre (Crowder 1989), participants heard a sine tone and imagined an instrument it might be on (Clarinet, oboe). Subsequently they heard a tone corresponding in timbre to one of the two instruments, and identified it as previously heard or not. Response time was higher for same pitch responses when the instruments imagined and heard differed than when both the pitch and the instrument imagined were identical to that heard.

In another experiment investigating imagined rhythm, Halpern (1988b) had participants listen to music at a certain tempo. Later participants ‘corrected’ the tempo to that previously heard on a computer by speeding up an inordinately slowed version of the previously heard music. Perceived and imagined tempo were found to be highly correlated (r=.63).

All these studies involved participants that were unselected for musical competence i.e., they were not particularly musically competent. Taken together this body of research on auditory imagery shows that there is a close correspondence between auditory imagery and music perception across pitch, timbre, temporal extent, rhythm and contour such that one may reasonably stand for the other. Thus, although experiments in the auditory domain need to recruit auditory imagery, it appears to be a reasonable investigation procedure. For instance, we may study underlying ease of processing of auditory stimuli that vary on one or more of these design

characteristics such as contour or tonality. We may investigate musical ease of perception by

asking respondents’ assessment of ease of imagination of musical fragment heard, by seeking

descriptions of the contour (ascending/descending) of the musical fragment heard etc.

More recently, Krishnan, Machleit and Kellaris (2008) have developed a musical

intelligence scale to measure individual differences among respondents to a given musical

stimulus.

Processing Fluency

The most critical response expected of a branding stimulus is its easy and correct recognition upon exposure as having been previously encountered in the brand’s context. This recognition may be based either on recollection – a conscious, intentional, controlled process that is slow and realizing from effortful reconstruction of the events associated with and surrounding past encounter, or familiarity – a non-conscious, introspectively unavailable, automatic meta- cognitive, subjective experience of past encounter (Jacoby and Dallas 1981; Whittlesea 1993;

Greenwald and Banaji 1995; Toth 1996). This familiarity experience may be at the ‘fringes of consciousness’ evoking fuzzy feelings that are somewhat and somehow informative about the stimulus in focus (Reber, Wurtz and Zimmermann 2004).

Two-stage memory model (e.g., Mandler 1980) distinguishes between free recall and

recognition. Whereas recognition involves discrimination of presented information, free recall

entails independent reproduction of the previously encountered stimulus information, requiring

the two stages of retrieval and, subsequent subjugation of the retrieved information to a

recognition check (Lynch and Srull 1982). Free recall might be considered a higher hurdle than

recognition because proceedings may fail either at the retrieval stage or at the discrimination

stage. However, it is not the case that retrieval is more difficult than discrimination. In fact,

either retrieval or discrimination may be more difficult. For instance, high frequency words are easier to retrieve because they have low activation potential, whereas low frequency words are easier to detect because they perceptually stand out (Higgins 1996; Gregg 1976).

The relative ease of retrieval and detection combine in complex ways to engender an overall meta-cognitive feeling of the ease of processing termed ‘processing fluency’ and is typically experienced as a direct consequence of prior exposure to the stimulus (Bornstein and

D’Agostino 1992). However, this phenomenal experience need not necessarily flow from prior exposure; it may alternatively reside in present circumstances such as stimulus clarity, thus flowing from size, regularity, goodness of form, or symmetry (Reber, Winkielman, and Schwarz

1998). It may obtain from processing facilitation through priming or duration of current exposure (Reber, Winkielman, and Schwarz 1998). It may reside in greater semantic connectivity of the stimulus (Whittlesea 1993) in its associative mental representation. This processing fluency is a signal that is picked by a ‘meta-cognitive feedback mechanism’ and made available to other processing systems such as the affective processing system (Winkielman et al

2003) leading to its detection either automatically or consciously as a processing ease.

This phenomenal experience begs an attribution to the most plausible reason and the

(mis)attribution is dependent on context (Janiszewski and Meyvis 2001). In a context requiring assessment of stimulus familiarity Whittlesea (1993) shows that processing fluency is misattributed to greater familiarity. Similarly, where the context is likeability, processing fluency is misattributed to greater liking. Winkielman and Cacioppo (2001) monitored the facial electromyography (EMG) scans of the zygomaticus major – a region that shows increased activity indicative of an incipient smile; they report such increase in activity when respondents were exposed to relatively fluent stimuli. Reber and Schwarz (1999) show that statements

presented perceptually clearer were judged truer. Thus, processing fluency influences several

judgment contexts both affective and non-affective. Further, fluency is indicative of processes

and manipulations that occur at disparate levels (see Winkielman et al 2003). Implicit memory literature shows (Tulving and Schacter 1990) that processing fluency may be perceptual or conceptual with perceptual fluency occurring pre-semantically, relatively earlier.

Perceptual fluency is reflective of processes occurring at a low-level that is based on superficial features of the stimuli i.e. data-driven processes mainly governed by form, symmetry, contrast etc typically resulting in speed and accuracy of identification (Winkielman et al 2003).

In a study (Reber, Winkielman, and Schwarz 1998) show that respondents rated black dots on a white background to be prettier than dots in less contrasting gray background. Similarly, pictures visually primed by their encapsulating contours were rated prettier (Reber, Winkielman, and

Schwarz 1998).

Conceptual fluency, in contrast, obtains from phenomenal ease of experience in higher- level operations on semantic memory representations and thus is related to semantic priming, semantic predictability, and relatedness (Winkielman et al 2003). In a classic experiment,

Whittlesea (1993) presented sentence stems (e.g., “The stormy sea tossed the”) that either semantically predicted the subsequent target word (“boat”) or did not (“lamp”), and found faster response times for identification where the stem was predictive. In a study (McGlone and

Tofighbakhsh 2000), respondents evaluated rhyming aphorisms such as ‘what sobriety conceals alcohol reveals’ to be truer than non-rhyming phrases ‘what sobriety conceals alcohol unmasks.’

However, this effect disappeared when respondents were cautioned not to be swayed by the poetic quality of the phrases i.e., the misattribution disappeared. Thus, conceptual fluency may obtain from processing rhymes – another higher-order process.

As these studies show, both perceptual and conceptual fluencies have largely convergent evaluative outcomes. Further, a perceptually fluent stimulus is seen to be more meaningful and positively affect judgments of truth and fame and conversely a conceptually fluent stimulus positively affects perceptual judgments of clarity, thus mutually influencing the “other” domain

(Winkielman et al 2003). However, in a recent study, Lee and Labroo (2004) show, by independently manipulating both types of fluencies, that conceptual fluency of a stimulus may not necessarily be affectively positive and may be additionally influenced by the valence of the conceptual prime.

Measurement of processing fluency in literature is typically through response latencies. A word or a picture may be presented at various levels of clarity, visually primed or not, with a high contrast or not etc. and time taken to detect its presence is used as an objective measure of the subjective experience (e.g., Whittlesea 1993; Reber, Winkielman and Schwarz 1998).

According to Reber, Wurtz and Zimmerman (2004) stimulus detection is a lower hurdle

(example detecting that a letter is present ‘R’) in comparison to stimulus identification R (letter

‘R’ is present ‘R’ and rotated 30 degrees to vertical). Reber, Wurtz and Zimmerman (2004) evaluated subjective fluency ratings of readability of words presented by orthogonal manipulations of font type (identification stage) and clarity (detection stage). They found both main and interaction effects; therefore, different perceptual stages may contribute jointly to the phenomenal experience of perceptual fluency and only measuring the objective speed at detection (or at identification) could yield a confounded indication of perceptual fluency.

Conceptual Fluency is a function of the semantic characteristics of the stimulus

(Janizewski and Meyvis 2001). In studies investigating conceptual fluency (Whittlesea 1993; Lee and Labroo 2004), typically predictive context or is semantically priming has been manipulated.

Alternatively, conceptual fluency may be an inherent stimulus property. Shapiro (1999) exploits

this idea in the context of investigating conceptually fluent advertisements by either including a

consistent background or not. Similarly, Janiszewski and Meyvis (2001) deploy composite logos

comprising a picture, a brand name, and an industry descriptor either evoking a single meaning

or not, thus manipulating conceptual fluency. Therefore, to the extent a stimulus obtains

consensus meaning in a culture it is conceptually fluent. This idea has been operationalized

through ‘codability’ – the ability of a stimulus to generate consensus meanings (Henderson and

Cote 1998).

Finally, as noted here a variety of stimulus characteristics (Whittlesea 1993; Reber

Winkielman and Schwarz 1998; Lee and Labroo 2004) influence both processing fluency and

evaluative judgments of presented stimuli. Also, since identical stimuli are liked better when

preceded by visual priming (Reber, Winkielman and Schwarz 1998) it may be construed that it is

processing fluency that is predictive of evaluations rather than stimulus characteristics per se;

i.e., fluency mediates the effects of stimulus characteristics on evaluative response dimensions

(Reber, Schwarz and Winkielman 2004). Interestingly, studies on logo characteristics in

accordance with investigative traditions of experimental aesthetics (Henderson and Cote 1998,

2003) have taken an ‘objectivist’ view (Reber, Schwarz and Winkielman 2004) that the genesis

of response dimensions reside in the object (logo characteristics) producing the response, thus

missing the possible mediational role by processing fluency.

In summary, it appears that the effects ensuing from branding stimuli characteristics on

evaluative judgments – likeability, familiarity, recognition etc. are mediated by processing fluency. Thus, branding definitions must capture the role fluency plays in determining brand

stimulus effects.

Chapter 3

Defining Sonic Branding

According to Jackson’s (2003) account, the term marque sonique (“sonic brand”) was first used in France in the mid-1980s. Introduced by Jean Pierre Baçelon, the term was subsequently introduced into the English language by Jean Pierre and Diarmid Moncrieff, who crystallized the expression into “sonic branding.” Jackson describes sonic branding as “…the creation of brand expressions in sound and the consistent, strategic use of these properties across touch points” (p. 9). To the best of my knowledge, this is the only documented definition to date.

This definition comprises two facets – creation of “brand expression” and consistent deployment of that “expression”. The focus on consistent application is congruent with principles of integrated marketing communication (Stahl 1964). However, what is “brand expression?” Is it a designed configuration of components i.e. design characteristics, making up the branding stimulus? Alternatively, is it the perceiver’s holistic evaluative judgments of the stimulus? According to Reber, Schwarz and Winkielman (2004), aesthetic judgments such as beauty are anchored by processing fluency emergent at the interaction between properties resident in the stimulus and the beholder’s cognitive and affective processes.

Therefore, “brand expression” could reside in the brand stimulus, in the cognitive processes of the perceiver or emerge as an interaction. Insofar as the brand expression leads to an auditory identity for the brand, several questions regarding interrelations among the design characteristics, processing fluency and outcomes of the sonic branding emerge. What is auditory identity? What are the constituent components making up the auditory identity? On what design factors do these components depend? This deconstruction is required to understand the underlying psychological mechanisms by which sonic branding obtains. Without insights

revealed by such deconstruction tenuous dependence on musicians will continue. Further, such

auditory identity should not only be unique but also congruent to the product category and

interpretable with consensus across cultures. Thus, I propose (and subsequently elaborate upon)

the following definition for sonic branding.

Sonic branding is the creation and perpetuation of a consistent, distinctive, universal,

and appropriate non-verbal aural identity for a brand as a unique configuration of evaluative judgments of familiarity, liking, recognition and personality, through the considered arrangement of design characteristics using natural or synthesized sounds.

Design Characteristics refer to the properties inherent in the stimulus. Symmetry, clarity, goodness-of-form, proportion, balance, and harmony (Henderson and Cote 1998; Reber,

Schwarz and Winkielman 2004) are examples of design characteristics in a visual stimulus.

Number of tones, melodic contour, range and repetition of musical patterns are examples of auditory design characteristics. Although processing fluency – the phenomenal experience of ease in detecting and identifying an incoming stimulus (Reber, Wurtz and Zimmerman 2004; Lee and Labroo 2004) - is typically experienced as a direct consequence of prior exposure to the stimulus (Bornstein and D’Agostino 1992), it may alternatively reside in present circumstances such as stimulus clarity, size, regularity, goodness of form, or symmetry (Reber, Schwarz, and

Winkielman 2004b). It may obtain from processing facilitation through priming or duration of current exposure (Reber, Winkielman, and Schwarz 1998). It may reside in greater semantic connectivity of the stimulus (Whittlesea 1993) in its associative mental representation.

Stimulus exposure typically obtains two kinds of processing fluency – perceptual fluency and conceptual fluency (Janiszewski and Meyvis 2001). Whereas a feature-based memory representation of the stimulus leads to perceptual fluency, a meaning-based memory

representation leads to conceptual fluency. It is conceivable that a given sonic stimulus yields

leads to processing fluency that is a unique combination of perceptual and conceptual fluency.

Thus, design characteristics interact with the perceiver’s cognitive and affective systems

resulting in processing experiences. These phenomenal experiences are misattributed, depending

on context (Janiszewski and Meyvis 2001) influencing evaluative judgments such as familiarity

(Whittlesea 1993), likeability (Reber, Winkielman and Schwarz 1998), truth effects (Reber and

Schwarz 1999) or likeability for a brand (Lee and Labroo 2004). In other words, it is processing

fluency rather than design characteristic per se, that is predictive of the evaluative judgments on

brand i.e., the “brand expression”.

Response dimensions collectively refer to the typical evaluative judgments resulting from exposure to a branding device —the extent, to which it is recognizable, generates consensus interpretation in meaning or personality, evokes positive affect and appears familiar (Henderson and Cote 1998). Sonic branding should be designed to meet specific branding outcomes such as distinctiveness or elicitation of a targeted personality. Distinctiveness here means avoidance of a false feeling of recollection of the brand upon first exposure to the sonic stimulus. Some stimuli are easier to perceive, encode and retrieve than others in that they are inherently symmetric or otherwise possess goodness of form etc., (Reber, Schwarz, and Winkielman 2004) and are thereby processed more fluently. Processing fluency can also obtain from prior exposure

(Bornstein and D’Agostino 1992). Because a fluent stimulus processing experience is conveniently attributed to prior exposure as a practical heuristic, ‘illusions of familiarity’ obtain

(Whittlesea 1993). At the same time, a higher processing fluency engenders inclusion of the brand in the consideration set (Shapiro 1999). Thus, one of the desired outcomes for sonic branding is right sizing of processing fluency from a distinctive identity versus false familiarity

perspective. What factors or design characteristics locate a sonic branding device along

processing fluency continuum? How may we distill these factors? Answers to these questions

will help in right sizing processing fluency in turn meeting desired branding goals.

Processing fluency of the stimulus may designed to meet the role that sonic branding is

expected to play — namely, defining its objective in terms of brand recognition, brand recall, or

brand identity. Different answers will lead to different actions. For instance, brand recognition

role may encourage adoption of a low resistance path via piggybacking on a Top40 hit song – a

quick fix way to increase the low-level perceptual fluency. However, such licensed songs come

with identities of their own, the meaning of which can change over time. Thus, a sonic branding

design targeting brand recognition could undermine another key function – perpetuation of aural

identity.

In contrast, a design targeting distinctiveness will be unique such that it will require

enhanced repetitions to attain similar levels of processing fluency. An increase in the frequency

of stimulus exposure requires additional marketing investment. Specifically, the design

characteristics should be such that it maximizes potential for ready and distinctive recall in the

largest market segment of interest for a given level of repetition frequency. Thus, sonic branding

depends on the considered arrangement of the design characteristics and consequent processing fluency leading to the intended outcomes of recognizability, familiarity, meaning and affect for the associated brand.

Finally, sonic stimuli should work independently of confounding artifacts such as lyrics in vocal music. Indeed, many effects attributed to “music” may actually stem from the verbal content of lyrics (Kellaris and Kent 1993). Hence, I propose exclusion of human voice as an

element of sonic branding, when it is used to convey verbal messages. However, nonverbal

22 usage of the human voice, such as in yodeling, humming, whistling, and beat box, is included in sonic branding. Relatedly, sonic branding should strive to use universally familiar sounds that elicit consensus in the meanings they elicit.

In summary, sonic branding is the strategic use of sound to create an authentic auditory identity for the brand. Conventional applications of sound in branding are tactical and lean on classical conditioning theory by repetitive pairing of sound and brand to create desired associations. In contrast, sonic branding leans on processing fluency theory leveraging sound as information in and of itself. Often such auditory information is nonverbal and nonlinguistic.

Sonic logos are good examples illustrating this phenomenon.

Chapter 4

Theory and Hypotheses

Perhaps the most basic form of sonic branding is the sonic logo or “sogo.” A sogo is a unifying, focal sonic branding device. Whereas a logo is a summary brand expression visually, a

sogo is the most parsimonious brand expression in sound. Therefore, research on sogos will

inform us of the nature, function and outcomes of sonic branding at its most rudimentary

atomistic level. In investigating the nature of a good logo, Henderson and Cote (1998) identify

using experimental aesthetics procedures, the design characteristics of the logos manifesting in

desirable logo response dimensions – of memorability, likeability and subjective familiarity at

first exposure.

As noted earlier, this dissertation makes a significant contribution to this research stream

by a) exploring an alternative (auditory) domain, b) refining dependent variables on recognition

with sensitivity analysis, c) enquiring processing fluency mediation, d) using an experimental

manipulation of the design characteristics and e) incorporating multilevel modeling for

methodological robustness.

Particularly, whereas the Henderson and Cote (1998) study is a significant contribution

and managerially relevant in terms of understanding implications of design characteristics on

response dimensions of logos, their study demonstrates effects without delving very deeply into

explaining why they occur. Later developments in the processing fluency literature suggest

possibilities for further investigation and conceptual depth. Might design characteristics

influence processing fluency, thereby leading to different levels of response variables? For

instance, Henderson and Cote (1998) show that ‘proportional’ designs increase false recognition

of logos. Per processing fluency literature, stimulus clarity, regularity, goodness of form, or

24 symmetry (Reber, Schwarz, and Winkielman 1998) enhances fluency experience. Therefore, proportional designs could influence processing fluency leading to ‘illusions of familiarity’

(Whittlesea 1993) thereby false recognition. Thus, this dissertation extends the conceptual depth in exploring the meditational role of processing fluency.

Response Dimensions and Processing Fluency: What is a good sogo?

Intuitively good sogos should be likable and provide a targeted auditory identity for the brand. Therefore, what may be the response dimensions that identify a good sogo? Henderson and Cote (1998) delineate four defining characteristics for a logo—the extent to which it is recognizable, evokes positive affect, generates consensus interpretation in meaning and obtains subjective familiarity.

Willingness-to-Pay (WTP) is another frequently encountered dependent variable in the logo literature (Brooker and Eastwood 1989; Prelec and Simester 2001), though not considered by Henderson and Cote (1998). Brooker and Eastwood (1989) show systematic variations in

WTP for processed food presented with logos among different households. Feinberg (1986) shows that exposure to credit card logos inflates WTP. Analogously, it would be interesting to see if WTP responses differ between two variants of an advertisement differing only in the accompanying sogo.

True Recognition and False Recognition: Concerning visual logos, Henderson and Cote

(1998) observe that recognition occurs at two levels: a) recognition, which is memory for past encounter at the next encounter and b) recall, which is association of the logo with the brand.

Further, recognition may be either an acknowledgment of true past encounter i.e., true recognition or illusory, based on feelings of subjective familiarity even when there was no past exposure i.e., false recognition.

Analogously, consider an experiment wherein respondents are asked to identify either a

novel or a previously heard sogo. True recognition occurs when people correctly recognize a

previously heard sogo. Conversely, false recognition occurs when a new sogo is misconstrued as

having been previously heard. For instance, imagine that a consumer hears the three-tone NBC

sogo G3-E4-C4, and is subsequently presented the foil -- Airbus pre-announcement chime, C4-

G4-E4. Since the latter is perceptually as fluent, it is likely to be falsely recognized as having

been encountered before.

More recently, the “SCAPE – Selective Construction And Preservation of Experiences

(Kronlund, Whittlesea and Yoon 2008) framework for memory suggests that recognition is a

decision about past occurrence of an event rather than an explicit retrieval of content. According

to SCAPE (Kronlund et al 2008), this heuristic conclusion occurs by a constructive process in the

present comprising two facets – 1) a production of psychological events in interaction with the

situation and stimulus and 2) an evaluation of this production inducing the phenomenal

experience concerning the efficiency of this production. Therefore, recognition memory may be

construed as a heuristic inference based on the subjective ease of processing. Previous studies

converge with this idea. For instance, Jacoby and Whitehouse (1989) demonstrate that perceptual

fluency not only increases overall recognition but also proportion of false recognition by inducing ‘illusions of memory.’

Thus, both true and false recognitions are important response dimensions for sogo evaluation. For instance, leader brands may want sogos that are distinctive, i.e., prefer a high true

recognition score; in contrast follower brands would want to appear familiar to win consumer

franchise, i.e., prefer a high false recognition score.

Sensitivity and signal detection: Although both true and false recognitions appear to be independent parameters of signal evaluation, signal detection theory (Singh and Churchill 1986;

Singh, Rothschild and Churchill 1988) contests this because these evaluations are influenced by response bias. A liberal judgment criterion (tendency to say yes whether the signal is present or not) would increase both hit rates and false alarm rates i.e., both true and false recognitions would be high in this case. Conversely, both true and false alarm rates would be low if the judgment criterion is conservative. As such, a signal is highly discriminable when true recognition is high and false recognition is low. Signal detection theory (Singh and Churchill

1986) conceptualizes a sensitivity index based on the z-transforms of hit rates (HR) and False

Alarm Rates (FAR). A high HR indicates that true recognition is high; a high FAR indicates that false recognition is high and the z-transformed difference d’ (the sensitivity index d-prime) indicates that overall sensitivity is high. Mathematically, d’ = Z (HR) – Z (FAR). With a 100% affirmative bias, both HR and FAR would be 50% leading to a zero d’ i.e., low sensitivity Thus, d’ is an important response dimension for sogo evaluation although not considered in the

Henderson and Cote (1998) study.

Affective outcomes: It is well established that music can affect moods (Bruner, 1990).

Further, emotion seems the broadest, most basic outcome engendered by music (Meyer 1956), requiring the least musical competence for detection. Design characteristics of the stimulus may lead to affective outcomes by influencing processing fluency. In particular, the influence of perceptual fluency of the stimulus on the affective judgment is well researched and is always positive (Reber, Winkielman and Schwarz 1998). Reber and colleagues (1998) show that perceptual fluency may be manipulated in the visual domain by means other than mere-exposure, such as by a facilitating subliminal visual prime, by varying form-ground contrasts of a stimulus,

or by increasing the duration of exposure. Across these manipulations, they show that affective

judgment is always positively valanced rather than neutral, regardless of the focus of the

question. Participants were asked to rate either the prettiness or the ugliness of the stimulus i.e.,

the focus was varied. Perceptual fluency increased the prettiness scores and reduced the ugliness

scores suggesting that it always led to positive affective judgment.

Sogos may vary in their perceptual fluency too, depending on their design features. For

instance, sogos with fewer (many) notes, monotonic (zigzag) melodic contours, repeating (non

repeating) patterns and perfect (imperfect) intervals may be perceptually more (less) fluent

leading to a positive affective judgment. Perceptually fluent sonic stimuli would be easier to

hum, whistle or play in the head just as perceptually fluent visual stimuli such as the Oldsmobile

rectangle are easier to visualize or draw. Thus, perceptual fluency depends on stimulus features.

In contrast, conceptual fluency is based on stimulus semantics. Hevner (1935) demonstrated that music pitched in minor keys was characterized unhappy even by those with no musical training. Perhaps respondents effortlessly attribute affective meanings to the music modality (minor/major) – a design characteristic of the stimulus. That is, modality increases the conceptual fluency of the sonic stimulus by readily lighting up the associated meanings in memory. Conceptual fluency has been studied in the context of memory accessibility of brands and their consequent influence on a brand’s inclusion in consideration-set and choice (e.g.,

Nedungadi 1990). Recent research (Lee and Labroo 2004) shows that although conceptual fluency increases processing fluency, the affective outcome may be either positive or negative influenced by the valence of the conceptual fluency.

Ergo, sogo evaluations should depend on their ability to make a difference in, a) willingness-to-pay b) true recognition c) false recognition d) detection sensitivity and e) affect. I investigate these five response dimensions in this dissertation.

Design Characteristics

Sound can be characterized in terms of sonic properties, such as frequency, loudness etc.

Musical sound can be characterized by its three structural components pitch, time and texture

(Bruner 1990). The pitch component refers to variability ensuing from the nature and arrangement of tones in a sequence. Pitch can characterize a sogo in four different ways.

Number of tones: Just as a short brand name is easier to recall, sogos with fewer the number of tones, should be easier to recall. Further, the number of tones in the sogo can perform specific mnemonic functions. It can cue the brand name by mirroring the number of alphabets (J-

E-L-L-O) or N-O-K-I-A. It may parallel the number of syllables in the brand name such as in the kit kat commercial ‘give me a break… give me a …kit-kat bar’ where the last three tones C3-C3-

C3 parallel the syllables kit’kat’bar. Ideally, it may do both (N-B-C). Sogos should have at least three tones to gain any distinctiveness but may have utmost1 12 tones; after all, they are intended to be a short distinctive audio signatures and not expansive musical compositions. Thus, number of tones in a sogo is an important design characteristic that could influence response dimensions.

Contour: A second variation is the direction or the contour of the sogo. A rising pitch connotes increasing emotional intensity and music in higher pitches is considered happier; in contrast, a descending sequence is considered serene (Bruner 1990). However, both ascending and descending patterns are linear and do not provide any surprising auditory experience. Per experimental aesthetics literature (Berlyne 1971), hearing a zigzagging pattern may be a more

1 A comfortably pleasant tempo is about a tone a second (quartertone M.M = 60); four tones a second would be fast (quartertone M.M =240) or a maximum of 12 tones in three seconds. 29

novel experience, interesting and likeable. For instance, a sequence of notes can be ascending

tone-to-tone (J-E-L-L-O) descending (Windows XP shut down chime) or zigzagging (NOKIA).

Thus, the contour of a sogo may influence response dimensions.

Chunking: A third way pitch characterizes a sogo is by the repetition of tonal patterns.

Miller (1956) notes “the magical number seven, plus or minus two” as limits on human capacity to process information. A chunking (Miller 1956) strategy recodes information into chunks thus increasing processing efficiency. For instance, a telephone number like 5132930338 is processed more efficiently by grouping the information into three chunks 513-293-0338. Analogously, a

processing challenge posed by an increase in the number of tones in a sogo should invite a

chunking strategy. Repetition facilitates chunking. Consider two nine tone sogos, 1-2-3-4-5-6-7-

8-9 and 1-2-3-1-2-3-1-2-3 where the latter facilitates chunking by repeating a pattern three times over. Such repetition creates a musical emphasis and may lead to higher memorability and identity for the sogo i.e., repetition strives to make a musical point. It can take different forms;

for instance, a phrase can repeat an octave apart or on a different instrument. Therefore,

chunking is an important design characteristic that could influence the response dimensions

Musical Tonality, the fourth design characteristic, has been extensively studied (e.g.

Meyers 1956, Hevner 1935, 1936; Kellaris and Kent 1993); music in minor mode is perceived

less happy and pensive (Hevner 1935) while a sequence of descending tones seems serene

(Bruner 1990). While conceding that it is an important characteristic influencing affective judgments, since the effects of tonality are already well researched in literature, I exclude its study as a design characteristic in this dissertation.

A brief comment on the other two structural components of music – time and texture is in order. Time, simplistically, refers to the speed of the music heard. It also refers to particular

rhythms and metric structure. A firm rhythm may be interpreted as serious while a smooth

rhythm more relaxed (Bruner 1990). However, in the context of a sogo lasting only three to six seconds, it is almost impossible to discern a specific rhythm other than overall perception of speed. Therefore, I do not consider time as a design characteristic for sogos.

Texture refers to overall characteristics such as the volume and instrumentation. For instance, woodwind instruments have been found to be mournful while brass wind instruments found majestic (Bruner 1990). Texture appears to be more indicative of the sonic personality

rather than an influence on recognition memory. I exclude this characteristic from the study for

this dissertation research. However, I intend to explore its sonic branding role beyond this

dissertation discussed in chapter seven.

Finally, notwithstanding the fact that music may be deconstructed into its constituent

elements, it is more compatible with the right hemisphere and is processed holistically. Bruner

(1990) speculates and Kellaris and Kent (1993) confirm that constituents of music may not only obtain main effects but also combine to obtain interaction effects. For instance, perceptual fluency may decrease with the increase in number of tones only when there are no repeating elements. Similarly, number of tones and contour may interact to produce complex recognition effects. Therefore, interaction effects among design characteristics are anticipated and will be

explored in this dissertation.

Mediating Role of Processing Fluency

Some stimuli are easier to perceive, encode and retrieve than others in that they are inherently symmetric or otherwise possess goodness of form etc., (Reber, Schwarz, and

Winkielman 2004) and are thereby processed more fluently. Thus, sogos with fewer tones and

linear contour should be easier to process. Processing fluency can also obtain from prior

exposure or repetition (Bornstein and D’Agostino 1992). Further, sogos with repeating tonal

patterns should obtain easier processing through the chunking mechanism. Because a fluent

stimulus processing experience is conveniently attributed to prior exposure as a practical

heuristic, ‘illusions of familiarity’ obtain (Whittlesea 1993). Thus, fluency effects obtained from

design characteristics in turn influence recognition.

Per past literature, processing fluency (Lee and Labroo 2004) is affectively positive.

Thus, one may infer that easy-to-process sogos will be liked more. Henderson and Cote (1998)

report that harmonious and natural designs were liked more where harmony refers to visual

symmetry and balance and natural refers to the degree to which these objects are commonly encountered. Per processing fluency literature these properties should engender higher processing fluency, leading to greater liking. Thus, the proceeding fluency mediation is implicit in Henderson and Cote (1998) study though not explicitly articulated. More recently, it has been shown that the influence of temporal construal (Trope and Liberman 2003) on attitudes is mediated by the ease of retrieval of favorable or counter arguments (Herzog, Hansen and Wanke

2007).

As noted earlier, easy hummability is indicative of processing fluency. Humming requires temporary retention of the tones in the phonological store that retains short-term memory traces of encoded stimuli for only about two seconds before decay, unless kept activated through continuous rehearsal (Baddeley and Hitch 1974; Baddeley and Logie 1992, Baddeley

2000). As the number of tones in a sogo increase, the process of rehearsal must begin for the initial tones of the sogo already heard, even before all the tones have been completely presented.

Thus, processing fluency should be lower as the number of tones increase. Similarly, a

32 monotonically ascending or descending contour is more hummable than a zigzagging pattern, which should obtain lower fluency.

Finally, some of these effects may be countervailing. For instance, increasing the number of tones decreases fluency but not if such increase is due to pattern repetition. Thus, some interactive effects on processing fluency may be expected.

Based on the foregoing discussion, I explore processing fluency mediation of the influences of design characteristics on response dimensions in this dissertation.

In summary, I explore the influence of three design characteristics of a sogo (number of tones, contour and chunking) on five response dimensions (Willingness to pay, true recognition, false recognition, signal sensitivity, and affect). In addition, I explore the processing fluency mediation of these influences.

Hypotheses

Willingness to pay: As the number of tones in a sogo increases, processing fluency should decrease because of the increase in auditory information. Thus, the 3-tone should be easiest to process and the 9-tone the most difficult to process with the 6-tone in between.

Therefore, illusions of familiarity should be highest for the 3-tone condition. Because familiarity also breeds contempt it is possible that WTP should be lowest for the 3-tone and increase with number of tones obtaining highest for the 9-tone with 6-tone falling in between. Recent research shows this disfluency effect. Galak and Nelson (2009) show that a non-fictional essay presented in a bad font (i.e., disfluent) is judged to be of greater quality than one presented in a good font.

Hence, a disfluent sogo should appear relatively less commonplace and more valuable. However, preferences vary in an inverted U shape with increasing complexity (Berlyne 1971) with greater

33 preference for moderate complexity in stimuli over either extreme. This would suggest that WTP should be maximal for the 6-tone sogo although the 9-tone sogo may be more disfluent. Thus,

H1: WTP should increase with number of tones in a sogo up to a point such that the 6- tone sogo should obtain highest willingness-to-pay in comparison to either the 3-tone or the 9- tone sogo.

H1a: The influence of number of tones on WTP should be mediated by the processing fluency of the sogos.

Recognition and fluency: Sogos with fewer tones should be easier to process and consequently engender greater processing fluency. Fluency should also be influenced by the musical pattern because some patterns facilitate processing (123-123-123) while others thwart

(1-2-3-4-5-6-7-8-9 processing. Therefore, a 3-tone sogo or a sogo with a facilitating pattern should obtain greater fluency experience.

H3: Number of tones should influence fluency such that fluency should increase with decrease in the number of tones in a sogo.

H4: Chunking should influence fluency such that fluency should increase when musical pattern in the sogo facilitates chunking rather than thwarts chunking.

When called upon to make a judgment on previous encounter, participants could lean on experienced processing fluency and recruit this experience as information to make the decision.

Alternatively, they could make use of content associated with that stimulus. Participants may note at test, that a given sogo has an ascending contour or a zigzag contour. They may note that a sogo has six tones or nine tones and attempt to reconstruct the past. However, content-based recognition could also depend upon salient recognition cues in the sogo. Sogos with many tones may be remembered because that characteristic sticks out. A zigzag sogo may be remembered

34 more vividly because attracts greater attention. Thus when an item (a sogo) is repeated from previous experience i.e., true recognition could be attributed to either route (Whittlesea 1993).

However, false recognition can only obtain from illusory familiarity consequent to fluent processing. Thus,

H5a: Number of tones should influence true recognition such that greater the number of tones in a sogo, the lesser the recognition.

H5b: Number of tones should influence false recognition such that greater the number of tones in a sogo, the lesser the recognition.

H5c: The influence of number of tones on false recognition should be mediated by fluency.

Although recognition measures are informative they may be confounded with affirmative

(“yea saying”) and negative biases (“nay saying”). Therefore true distinctiveness i.e., a high signal sensitivity for a stimulus should require recognition when present (high hit rate) and rejection when not (low false alarm rate). A 3-tone sogo has limited information. Therefore, it should be easy to remember leading to high hit rates. However, 3-tones are also easier to process and can lead to illusions of familiarity thereby leading to high false alarm rates. Because sensitivity captures the difference between these two rates, signal sensitivity is likely to be low for the 3-tone signal even though both hit rates and false alarm rates may be high. As the number of tones increases, information content increases. Consequently, hit rates should drop. However, false alarm rates should drop too owing to decreasing processing fluency.

Miller (1956) notes “the magical number seven, plus or minus two” as limits on human capacity to process information. Therefore, the hit rates between 3-tone and 6-tone sogos should not vary much although it is reasonable to expect some degradation. Consider a comparison

35 between two 3-tone sogos. If two out of the three tones are identical, there is 66% similarity between the two sogos. . Now consider a comparison between two 6-tones. If two out of six tones are identical, there is only 33% similarity between the two sogos. This means that the false alarm rates should fall significantly for 6-tone sogos. It is much harder to mistake a 6-tone sogo for another 6-tone sogo than it is to mistake a 3-tone sogo for another 3-tone sogo. In summary, sensitivity for the 6-tone sogo should be higher than it is for the 3-tone sogo.

Further increase would take the number beyond the magical limit so that hit rates should start to drop. However, it is harder to foresee how the false alarm rates would move. If we take the tone-to-tone comparison logic, we should expect a much lower false recognition. However, listeners might apply a length or number heuristic (I heard one with many tones before …this must be it). Although this is an empirical question, Wanke, Bless and Biller (1996) note that dependence on experience increases with erosion of confidence and this might encourage application of a heuristic, thereby increasing false alarm rates. If so, the 9-tone sogo should exhibit a lower sensitivity. Thus,

H6: Signal Sensitivity should exhibit a curvilinear, inverted U shaped relationship with increase in number of tones.

A linear contour (ascending or descending) is easier to remember in comparison to a zigzag contour because the directional information is simple. Therefore, one may argue that the hit rates should be higher for linear contours. Alternatively, because zigzag contour is novel and salient hit rates may be higher for zigzag contour because they leave stronger memory cues.

Further, false alarm rates should be higher because fluency experience should be higher for linear sogos. However, listeners might apply a content-based ‘kink’ heuristic (I heard one with many ‘kinks’ before …this must be it) rather than rely on fluency experience in which case false

alarm rates should be high. Thus, which contour will have greater sensitivity is an empirical question.

Positive Affect: All else equal, processing fluency has been shown to be “hedonically marked” (Reber, Schwarz and Winkielman 2004) and affectively positive. For instance, Reber,

Winkielman, and Schwarz (1998) show that respondents rated black dots on a white background to be prettier than dots in less contrasting gray background. Similarly, pictures visually primed by their encapsulating contours were rated prettier (Reber, Winkielman, and Schwarz 1998).

However, the key phrase is “all else equal”. In the two examples above, the stimulus was held

constant; only the context changed obtaining differential fluency. The same black dots were

evaluated against different backgrounds and the same picture was evaluated primed or not. The

overall likeability should depend both on objective properties (content) and on the fluency

experience. For instance, if a red dot and black dot are both presented against a white

background, red dot may be preferred because of color effect (content) and this preference may

be somewhat mitigated because of higher fluency effect the black dot obtains.

Analogously, between two 3-tone sogos one presented in a high fidelity music system

may be liked more. However, between a 3-tone and a 6-tone sogo, the 6-tone sogo may be liked

more because of increasing complexity and interestingness (content) somewhat mitigated by the

lower processing fluency (experience). Similarly, a zigzag contour may be preferred over linear

contour because of novelty (content) with that preference tempered by the lower processing

fluency (experience). Overall likeability should depend on the joint influence of the content and

the experience information on the judgment and, is therefore an empirical question. However,

fluency should certainly mediate overall likeability whatever the source. Thus,

H7: Fluency will mediate likeability assessments such that, all else equal, the more fluent a sogo, the more likeable it is.

In addition to these formal hypotheses, I have a general expectation that there may be other interactions among design characteristics and I treat this as an exploratory issue.

Chapter 5

Research Studies

The studies were divided by design characteristics. In all, five studies were conducted as described in this document. The first study explored the influence of number of tones on WTP in a between-groups test. The second study explored the influence of chunking on true recognition, false recognition, and signal sensitivity in a within-groups test. The third study explored the influence of number of tones on affect, true recognition, false recognition, and signal sensitivity in a within-groups test. The fourth study explored the influence of sogo contour on affect, true recognition, false recognition, and signal sensitivity in a within-groups test. The fifth study explored the joint influence of number of tones and contour on affect, true recognition, false recognition, signal sensitivity in a within-groups test. Additionally, two pilot studies were conducted – one exploring the influence of contour on WTP and the other replicating the third study.

In order to control for past familiarity no existing sogos were used. Instead, the stimuli were composed on PSR 275 YAMAHA keyboard to ensure objective unfamiliarity. Participants were undergraduate business majors from a Midwestern public university, USA who participated in these studies in exchange for extra academic credit. They were recruited by passing out sign- up sheets where they picked a suitable time from those available. Participants checked in at the appointed time, were taken through a warm-up session on the MediaLabTM software where they

were introduced to sonic logos. They heard several examples of these logos (Windows Vista,

Windows XP, Intel, and NOKIA) via headphones. The actual study began after this standard

warm-up session at which time participants were able to adjust their headphone and become

familiar with the MediaLabTM software. Typically, three levels of the design characteristic were

explored in every study. Participants were exposed to the treatment via the MediaLabTM software. The whole procedure took about 15-20 minutes.

In between-group studies, participants were exposed to only one of the three levels. In the within-group studies, participants were sequentially exposed to all the levels of the factors under consideration. However, in the within-group studies experimental procedure controlled for order effect by randomizing stimulus presentation and minimized carry-over effects through distraction tasks between stimuli presentations (Neter, Kutner, Nachtsheim and Wasserman 1996, p 1165).

Typically, these distraction tasks required the participant to recall a famous song, track a particular word several steps into the song and perform a task such as supply a rhyming word.

Past research (Halpern 1998, 1988a) shows that such tasks require auditory imagination of the song under question and therefore would wash away the previous treatment from the auditory buffer thus, minimizing carryover effects. Finally, appropriate analytical techniques such as repeated measures ANOVA or multilevel meditation modeling were deployed to account for participant response clustering and heterogeneity in within-group designs.

Study 1

The first study explored the influence of number of tones on WTP in a between-groups setting. Sample consisted of seventy-eight undergraduate business majors from a public

Midwestern university, USA, assigned to three groups. Participants checked in at the appointed time, were taken through a warm-up session on the MediaLabTM software where they were introduced to sonic logos. They heard several examples of these logos (Windows Vista,

Windows XP, Intel, and NOKIA) via headphones. The actual study began after this standard warm-up session enabling participants to adjust their headphone and become familiar with the

MediaLabTM software. After warm-up, they heard a 3-tone, a 6-tone or a 9-tone sogo per group assigned. Next, they rated fluency in response to the item “This sonic logo is easy for an average person to hum” (1— strongly disagree, 7— strongly agree). Next, they heard progressively fading radio music for 15 seconds followed by a commercial for a hypothetical brand of bread lasting 18 seconds. The commercials were rendered in a Male Voice-Over, and were identical across groups except for the accompanying sogo that was one of 3-tone, 6-tone or 9-tone per group assigned. All other objective properties of music were held constant across these sogos.

The commercial is reproduced below.

(MVO) “This program is co-sponsored by Zinto bread.”

(Sogo plays)

(MVO) “Zinto bread is wholesome and nutritious. Now, available at your local grocery store; Zinto bread”

(Sogo plays)

Next, participants rated their willingness-to-pay for this brand of bread given the benchmark price range between $1.50 and $2.50.

Finally, participants were thanked, debriefed and dismissed.

Results

Pre-Analysis revealed that all variables of interest were normally distributed and visual inspection did not reveal any significant skews. Two participants with a fluency rating of two on a seven-point scale for 3-tones were deemed outliers per Chauvenet’s criterion (Taylor 1997) and excluded from further analysis. Number of tones revealed a significant effect on WTP (F (2, 73)

= 3.53, p <.03). Post-hoc test were conducted using Bonferroni adjusted alpha levels of .017 per test (.05/3). Results (Figure 1) indicated that the WTP was significantly higher in the 6-tone condition (M = 1.80) than that in the 3-tone (M = 1.45). WTP in the 6-tone condition was also directionally higher than that in the 9-tone (M = 1.66) condition though not significant. Pairwise comparison of the WTP between the 3-tone and the 9-tone conditions was non-significant. Thus,

H1 is supported.

Number of tones revealed a significant effect on fluency (F (2, 73) = 6.35, p <.001).

Post-hoc tests were conducted using Bonferroni adjusted alpha levels of .0167 per test (.05/3).

Results indicated that the fluency rating was significantly lower for the 6-tone sogo (M = 5.08) than that for the 3-tone sogo (M = 6.27) or the 9-tone sogo (M = 6.08). Pairwise comparison of the fluency between the 3-tone and the 9-tone condition was non-significant.

A regression of WTP on fluency revealed that fluency significantly predicted WTP (β = -

.245, t = -2.18, p < .03). With fluency in the model, number of tones revealed a non-significant effect on WTP (F (2, 72) = 2.12, p >.12) providing evidence for fluency mediation of influence of number of tones on WTP. Thus, H1a is supported.

Figure 1: Influence of Number of Tones on Willingness-to-Pay

Discussion

The hypothesis that WTP should vary with number of tones in an inverted U shape is supported. As predicted, the commercial in the 6-tone sogo condition obtained higher WTP than

the commercials in the 3-tone or 6-tone conditions. Although the commercials were identical

except for the sogo included, WTP for the hypothetical brand of bread under 6-tone sogo

condition commanded a 16% price premium (.25 cents) over the average in the other two

conditions. As an aside, that the net income from operations2 for PaneraTM bread grew by a

comparable 17% between 2007 and 2008 puts the potential for sonic branding in perspective and

provides a practical motivation for this dissertation.

A surprising result is the comparable fluency rating for the 9-tone sogo and the 3-tone

sogo. A post hoc inspection of the sogo patterns is instructive. The sogos were respectively

2 Source: SEC filings for PNRA 43

musically patterned 1-2-3, 1-2-3-4-5-6 and 11-22-33-44-5 respectively. Although, the intention was to hold all other objective parameters constant across the three sogos, the 9-tone sogo pattern

appears to have been inadvertently processed as five groups of tones, thereby resulting in lower

fluency ratings. Perhaps this sogo was processed more like a 5-tone sogo and consequently

obtained higher fluency ratings than expected. Indeed, study 2 (presented next) evidences such

increase in processing fluency.

Study 2

The first study explored the influence of number of tones on WTP in a between-groups setting revealing a systematic variation, with the 6-tone condition commanding a price premium.

However, as discussed, the fluency results for the 9-tone sogo were not as predicted. Whereas, fluency for the 6-tone sogo was lower than that for the 3-tone sogo, i.e., fluency decreased with increase in number of tones, 9-tone sogo was more fluent than the 6-tone sogo. The sogos were musically patterned 1-2-3, 1-2-3-4-5-6 and 11-22-33-44-5 respectively. Perhaps the grouping effect for the 9-tone sogo made it seem like a 5-tone sogo.

People often recode information such as telephone numbers into chunks (Miller 1956) to increase processing efficiency. Further, one telephone number pattern (555-222-1111) may be more amenable to chunking than another (4726381905) and consequently easier to process.

Analogously, for a given number of tones, a changing musical pattern could change fluency ratings. In particular, a pattern such as 1-2-3-4-5-6-7-8-9 should thwart chunking and a pattern such as 123-123-123 should facilitate chunking. Consequently, the latter should be more fluent and, the fluency for the pattern used in study one (11-22-33-44-5) should fall between these two extremes. Thus, the second study explored the influence chunking on fluency in a within- subjects3 setting. Sample consisted of forty-seven undergraduate business majors from a public

3 Because within-subject designs involve exposing participants to multiple treatment conditions, concerns on demand effect confounds may be reasonably raised (Sawyer 1975). Equally, validity of between- subjects design has been questioned in situations involving subjective dependent variables that summon participants’ judgment (Birnbaum 1999). Birnbaum’s (1999) interesting experiment involves a judgment on two numbers 9 and 221, assigned to two different groups in a between-subjects design. Participants were asked to judge how large the number (9 or 221 per assignment) was on a ten- point scale. The number 9 was judged larger, thus dramatically demonstrating the misplaced overemphasis on between-subject designs in context dependent situations. In this experiment, non-availability of a benchmark (how large with respect to what?) presumably led participants to deploy trait or state based individual contexts. Other researchers advocate deployment of both designs (e.g., Hirt and Castellan 1988) especially in situations when clustering and unobserved heterogeneity could influence judgments. Randomization does not guarantee that the unobserved heterogeneity will be identical across cells in a between- subjects design. However, a within-subjects design can help in detecting clustering among participants as revealed by the intra class correlation (ICC) that can then be controlled for (Singer 1998). Further, whereas between-subject design minimizes demand effects, within-subject designs provide greater power. Finally, the studies in this 45

Midwestern university, USA. As in study 1, participants checked in at the appointed time, were

taken through a warm-up session on the MediaLabTM software where they were introduced to

sonic logos. They heard several examples of these logos (Windows Vista, Windows XP, Intel,

and NOKIA) via headphones. The actual study began after this standard warm-up session

enabling participants to adjust their headphone and become familiar with the MediaLabTM software. After warm-up, participants heard three target sogos all of which comprised 9-tones but with different musical patterns – one that facilitated chunking (123-123-123), another that thwarted chunking (1-2-3-4-5-6-7-8-9) and the one used in study 1 (11-22-33-44-5) as control.

Thus, they were exposed to all treatment conditions but in randomized order. Immediately after hearing a sogo, participants rated fluency in response to the item “This sonic logo is easy for an average person to hum” (1— strongly disagree, 7— strongly agree) for that sogo.

Next, they answered two other questions about the sogo they had just rated for fluency – a) whether the sogo was predominantly ascending, descending or zigzagging and b) the number of tones in the sogo. Nearly 80% identified the sogo contour correctly. The number of tones was correct to within two tones in 64.55% instances (thwarting 64.9%, facilitating 68.1% and control

60.65%). The average number of tones indicated was 7.89, 7.60 and 7.61 respectively for thwarting, facilitating and control sogos. This evidences that the participants possessed a vivid auditory imagery for the respective sogos when rating fluency. In order to minimize carry-over effects, participants performed a distraction task between treatments requiring them to recall a famous song, track a particular word several steps into the song and perform a task such as supplying a rhyming word.

dissertation involve recognition measures requiring an assessment of past encounter with a stimulus, therefore necessitating a within-subjects design. Thus, within-subject design seems appropriate. 46

Next, participants rated recognition for six sogos, the three targets they had heard and

rated for fluency before and three foils not heard before. The presentation order was

randomized. The foil sogos were similar on objective musical properties respectively to the

target sogos. Immediately after hearing a sogo, participants rated recognition in response to the

item “The sonic logo I just heard was played in the first set too” (1— strongly disagree, 6—

strongly agree) for that sogo.

Finally, participants rated the foils for fluency per procedure outlined above, were

thanked, debriefed and dismissed.

Results

Pre-Analysis revealed that all variables of interest were normally distributed and visual inspection did not reveal any significant skews.

A multi-level unconditional means model revealed participant heterogeneity and clustering on recognition and fluency ratings with intra-class correlations exceeding 5% (ICC – true recognition = 9.49%; false recognition = 23.6%; fluency (targets) = 33.85%) thus justifying the within-group design.

Analysis of the influence of number of tones on fluency of targets via a repeated measures ANOVA revealed a non-significant sphericity (Mauchly’ W (2) = 3.11, p > .211).

Number of tones had a marginal effect on the targets’ fluency (F (2, 92) = 2.755, p <.069). Post- hoc tests were conducted using Bonferroni adjusted alpha levels of .017 per test (.05/3). Results indicated pairwise comparisons of the targets’ fluency between the facilitating (M = 4.85) and thwarting conditions (M = 4.26), was significant. Interestingly, the pairwise comparison of the targets’ fluency between the facilitating (M = 4.85) and control conditions (M = 4.79), was not significant i.e., the control sogo indeed facilitated chunking in study 1.

Analysis of the influence of number of tones on fluency of foils revealed a non-

significant sphericity (Mauchly’ W (2) = 4.05, p > .132). Number of tones had a significant effect on the foils’ fluency (F (2, 92) = 14.06, p <.000). Post-hoc tests were conducted using

Bonferroni adjusted alpha levels of .017 per test (.05/3). Pairwise comparison of the targets’

fluency between the facilitating (M = 5.17) and thwarting conditions (M = 3.75), was significant.

Interestingly, the pairwise comparison of the foils’ fluency between the facilitating (M = 5.00) and control conditions (M = 5.17), was not significant i.e., the control sogo indeed facilitated chunking in study 1 and H4 is supported.

Influence of chunking on true recognition was non-significant (F (2, 92) = 2.73, p >.070).

Influence of chunking on false recognition was non-significant (F (2, 92) = .57, p >.569).

Signal sensitivity was investigated using collapsed d-primes across participants

(Macmillan and Creelman 1991 p275) for the sogos under the two conditions (facilitating or thwarting) revealing non-zero sensitivity for both conditions (d’facilitating =1.19, z = 4.70, p < .000; d’thwarting = 1.06, z = 3.80, p < .000; d’9-tone =.88, z = 4.51, p < .000). However, comparison of the sensitivity index between the two conditions was non-significant (z = -.32, p > .32).

Discussion

The second study demonstrated the influence of chunking on fluency experience. Pattern

repetition undergirds chunking influence. A repeating pattern (Bornstein and D’Agostino 1992)

increases fluency by mere exposure. T-mobile’sTM ringtone is an interesting example. However, chunking appears to influence fluency in an incremental way qualifying the fluency effects resulting from number of tones in a sogo. In particular, chunking did not affect the recognition judgments nor were the signal sensitivities different across the conditions. Nevertheless, these results, taken together with the results from study 1, provide direction for stimulus correction.

Third study following next, explored recognition and affect response dimensions with the 9-tone stimulus corrected for chunking effects as informed by study 2.

Study 3

The third4 study explored the influence of number of tones on positive affect, true recognition, false recognition and signal sensitivity in a within-groups setting. Sample consisted of eighty-eight undergraduate business majors from a public Midwestern university, USA.

Participants checked in at the appointed time, were taken through a warm-up session on the

MediaLabTM software where they were introduced to sonic logos. They heard several examples of these logos (Windows Vista, Windows XP, Intel, and NOKIA) via headphones. The actual study began after this standard warm-up session enabling participants to adjust their headphone and become familiar with the MediaLabTM software. After warm-up, participants heard three target sogos – a 3-tone, a 6-tone and a 9-tone sogo. Thus, they were exposed to all treatment conditions but in randomized order. Immediately after hearing a sogo, they rated fluency in response to the item “This sonic logo is easy for an average person to hum” (1— strongly disagree, 7— strongly agree) for that sogo.

Next, they answered two other questions about the sogo they had just rated for fluency – a) whether the sogo was predominantly ascending, descending or zigzagging and b) the number of tones in the sogo. The three sogos were predominantly ascending in nature5. Sixty-three percent rated the sogos to be ascending and 87% rated the sogos to be either ascending or zigzagging. The number of tones was correct to within a tone in 84% instances (3-tone 98%, 6- tone 78% and 9-tone 77%). The average number of tones indicated was 3.07, 5.94 and 8.36 respectively for 3-tone, 6-tone and 9-tone sogos. This evidences that the participants possessed a vivid auditory imagery for the respective sogos when rating fluency. In order to minimize carry- over effects, participants performed a distraction task between treatments requiring them to recall

4 This study replicates and augments a pilot study not reported here that deployed the 9-tone sogo used in study 1 5 9-tone sogos had a marginal zigzag pattern 50

a famous song, track a particular word several steps into the song and perform a task such as

supplying a rhyming word.

Next, participants rated recognition for six sogos, the three targets they had heard and

rated for fluency before and three foils not heard before. The presentation order was

randomized. The 3-tone, 6-tone and the 9-tone foil sogos were similar on objective musical

properties respectively to the 3-tone, 6-tone and the 9-tone target sogos. Immediately after hearing a sogo, participants rated recognition in response to the item “The sonic logo I just heard was played in the first set too” (1— strongly disagree, 6— strongly agree) for that sogo.

Finally, participants rated the foils for fluency per procedure outlined above, were thanked, debriefed and dismissed.

Results

Pre-Analysis revealed that all variables of interest were normally distributed and visual inspection did not reveal any significant skews.

A multi-level unconditional means model revealed participant heterogeneity and clustering on recognition and fluency ratings with intra-class correlations exceeding 5% (ICC – true recognition = 8.6%; positive affect = 8.2%; fluency (foils) = 9.0%) thus justifying the within-group design.

Analysis of the influence of number of tones on fluency of targets via a repeated measures ANOVA revealed a significant sphericity (Mauchly’s W (2) = 13.89, p < .000).

Number of tones had a significant effect on the target fluency after applying Hyunh-Feldt corrected degrees of freedom for sphericity (F (1.78, 154.25) = 60.80, p <.000). Post-hoc tests were conducted using Bonferroni adjusted alpha levels of .017 per test (.05/3). Results (Figure 2)

indicated that all pairwise comparisons of the targets’ fluency among the three conditions, 3-tone

(M = 6.18), 6-tone (M = 4.81) or 9-tone (M = 3.74) were significant.

Analysis of the influence of the number of tones on fluency of foils revealed a significant

sphericity (Mauchly’s W (2) = 10.28, p < .006). Number of tones had a significant effect on the

foils’ fluency after applying Hyunh-Feldt corrected degrees of freedom for sphericity (F (1.83,

159.47) = 52.63, p <.000). Post-hoc tests were conducted using Bonferroni adjusted alpha levels of .017 per test (.05/3). Results (Figure 3) indicated that all pairwise comparisons of the foil fluency among the three conditions, 3-tone (M = 5.8), 6-tone (M = 4.81) or 9-tone (M = 3.84) were significant. Thus, H3 is supported.

Analysis of the influence of the number of tones on positive affect revealed a non- significant sphericity (Mauchly’s W (2) = 3.94, p > .14). Number of tones had a significant effect on positive affect (F (2, 174) = 4.60, p <.01). Post-hoc tests were conducted using Bonferroni adjusted alpha levels of .017 per test (.05/3). Results (Figure 2) indicated that the positive affect in the 9-tone (M = 3.34) condition was significantly lower than those in the 6-tone (M = 3.89) and the 3-tone (M = 3.89) conditions. Comparison of positive affect between the 3-tone and 6- tone conditions was non-significant.

Bootstrap estimates (Krull and Mackinnon 2001; Mackinnon, Lockwood and Williams

6 2004) of the indirect effects of fluency (β1β3 = -.381, β2β3 = -.68) on positive affect, via a multi-

6 Mediation is evidenced in the Baron and Kenney (1986) paradigm when the inclusion of the mediator dims or washes away the main effect. This procedure for mediation detection involves several OLS regressions that assume that errors are not correlated. This is not a valid assumption for clustered data used in within-subject designs. Multi- level mediation models (Krull and Mackinnon 2001; Mackinnon, Lockwood and Williams 2004) account for this clustering and heterogeneity and adopt an alternative equivalent procedure to establish mediation. Mediation is evidenced when the indirect path represented by the product β1β3 is significant where β1 represents the influence of independent variable on the mediator and β3 is the influence of the mediator on the dependent variable with the IV in the model. However, the distribution for β1β3 is not known or analytically derivable and therefore needs to be derived through bootstrapping techniques. Bootstrapping involves resampling of the data with replacement. Typically n cases are re-sampled several times where n is the actual sample size and the model estimated, thereby generating a synthetic distribution for β1β3. These estimates are ordered from smallest to the largest and the values at 52

level modeling with dummy coded column vectors X1 and X2 representing number of tones, was

significant (Figure 2) with corresponding 95% confidence intervals excluding zero (-.585 < β1β3

<- .21, -1.006 < β2β3 < -.41), providing evidence for fluency mediation of influence of number

of tones on positive affect. Further, a Sobel test revealed that the pathway from number of tones

to positive affect through fluency (β1β3 and β2β3 ) was significant (Z = -3.93, p < .000 and Z = -

3.62, p < .000 respectively). This pattern of results suggests that fluency mediates the influence

of number of tones on positive affect. Thus, H7 is supported.

Analysis of the influence of the number of tones on true recognition revealed a non-

significant sphericity (Mauchly’s W (2) = 3.90, p > .14). Number of tones had a significant effect on the true recognition (F (2, 174) = 8.27, p <.000). Post-hoc tests were conducted using

Bonferroni adjusted alpha levels of .0167 per test (.05/3). Results indicated that the true recognition in the 3-tone (M = 5.02) condition was significantly higher than those in the 6-tone

(M = 4.10) and the 9-tone (M = 4.25) conditions. Pairwise comparison of true recognition between the 6-tone and 9-tone conditions was non-significant. Thus, H5a is supported.

Analysis of the influence of the number of tones on false recognition revealed a significant sphericity (Mauchly’ W (2) = 35.16, p < .000). Number of tones had a significant effect on the false recognition after applying Hyunh-Feldt corrected degrees of freedom for sphericity (F (1.52, 132.04) = 23.20, p <.000). Post-hoc tests were conducted using Bonferroni

adjusted alpha levels of .017 per test (.05/3). Results (Figure 3) indicated that the false

recognition in the 6-tone (M = 1.44) condition was significantly lower than those in the 3-tone

(M = 2.83) and the 9-tone (M = 2.84) conditions. Pairwise comparison of true recognition

between the 3-tone and 9-tone conditions was non-significant. Thus, H5b is supported.

2.5% and 97.5% respectively serve as the confidence intervals. A significant mediation is indicated if this interval excludes zero value. In this dissertation, I deploy 1000 iterations. Appendix A contains the relevant SAS code. 53

Bootstrap estimates of the indirect effects of fluency (β1β3 = -.10, β2β3 = -.21) on false

recognition, via a multi-level modeling with dummy coded column vectors X1 and X2

representing number of tones, revealed marginal significance (Figure 3) with corresponding

97.5% upper confidence intervals falling just above zero (-.24 < β1β3 < .04, -.5 < β2β3 < .08),

providing evidence for fluency mediation of influence of number of tones on false recognition.

Further, a Sobel test revealed that the pathway from number of tones to false recognition

through fluency (β1β3 ) was marginally significant (Z = -1.45, p < .07). However, bootstrap

estimates of the indirect effects of fluency (β1β3 = -.05, β2β3 = -.08) on true recognition, was non-

significant with corresponding 95% confidence intervals including zero (-.284 < β1β3 < .15, -.47

< β2β3 < .30), Further, a Sobel test revealed that the pathway from number of tones to true

recognition through fluency (β1β3 and β2β3 ) was not significant (Z = -.46, p > .32 and Z = -.33, p

> .37 respectively), thus providing evidence for moderated fluency mediation of influence of number of tones on recognition. Thus, H5c is supported.

Collapsed d-primes across participants (Macmillan and Creelman 1991 p275) for the three conditions revealed non-zero sensitivity for all the three conditions (d’3-tone =1.55, z = 7.20,

p < .000; d’6-tone =1.93, z = 7.55, p < .000; d’9-tone =.88, z = 4.51, p < .000). Pairwise comparisons

of sensitivity across conditions revealed that the sensitivity to 9-tone condition was significantly

lower than that for 6-tone condition (z = -3.25, p < .000) and that for 3-tone condition (z = 2.20, p

< .011). The pairwise comparison of the sensitivity index between 3-tone and 6-tone condition

was non-significant (z = 1.13, p > .128). Thus, H6 is supported.

Figure 2: Influence of Number of Tones on Positive Affect

Figure 3: Influence of Number of Tones on False Recognition

Discussion

The hypothesis that fluency should increase with decrease in number of tones is supported. As predicted, the commercial in the 3-tone sogo condition obtained higher fluency than the 6-tone sogo condition, which in turn was more fluent than the 9-tone sogo. This also corroborates the chunking explanation for the anomaly seen in study 1 because we corrected the

9-tone stimulus in this study to eliminate the chunking effect.

The results evidence processing mediation fluency and extend the conceptual understanding of the mechanism by which positive affect and false recognition obtains. Fluency is expected per past literature to be affectively positive. The item “This sonic logo is easy for an average person to hum” (1— strongly disagree, 7— strongly agree) appears to capture the fluency construct and performs well. The results also increase nomological validity for the fluency measure. It is also interesting to see the joint action of content and experience on building positive affect. Positive affect in the 9-tone condition was lower than for those in the 6- tone and the 3-tone conditions, in line with our expectation that lowering fluency lowers likeability. However, the opposing effects of content and fluency render the 3-tone and 6-tone equally likeable.

Henderson and Cote (1998) show that natural, harmonious parallel or proportional design characteristics lead to false recognition. It is reasonable to expect such designs to have greater processing ease thus pointing in a fluency-based explanation for false recognition. Results evidence such mediation. More interestingly, this mediation occurs only for false recognition but not for true recognition. True recognition appears on to lean content but not on experience. In contrast, false recognition appears to recruit both types of information especially when an overarching heuristic is available. Because the 9-tone sogo was corrected for chunkability, the

fluency ratings fell in line with the 9-tone fluency rating significantly lower than the 3-tone

fluency. However, the false recognitions are still comparable. Give the mediation observed we

should expect 9-tone sogo to be less prone to illusions of familiarity. It appears that an

alternative content based representative heuristic (Tversky and Kahneman 1974) sets in. The 9-

tone sogo is longer (or rather denser because the durations were identical). Accordingly, a

previously unheard 9-tone is falsely recognized based on this content-based representative

heuristic (Tversky and Kahneman 1974) even though fluency experience dims. This content

effect also accounts for the marginal significance encountered for fluency mediation of false recognition. Because this heuristic does not set in between 3-tone and 6-tone (which is not dense enough), false recognition is in line with fluency experience

Distinctiveness for a branding device (logo or sogo) emanates not merely from true

recognition but from a combination of true recognition and correct rejection. It is interesting to

see that the sensitivity analysis with d-prime measures evidences such distinctiveness for the 6-

tone over the 9-tone and the 3-tone sogo. A guideline appears to emerge as a sogo design

characteristic– brands seeking distinctiveness would do well to target optimal number of tones in

the sogo. Conversely, brands in low involvement categories seeking to ride on illusory

familiarity should move to either extreme, preferably, have fewer tones in the sogo. Interestingly,

both Intel and NOKIA comprise five tones and appear to be in the optimal range and both have a

zigzag contour. Perhaps it is the contour and not the number of tones that makes the difference, the design characteristic I explore next in the fourth study.

Study 4

The fourth study explored the influence of contour on positive affect true recognition,

false recognition and signal sensitivity in a within-groups setting. Sample consisted of forty-two undergraduate business majors from a public Midwestern university, USA. Participants checked in at the appointed time, were taken through a warm-up session on the MediaLabTM software

where they were introduced to sonic logos. They heard several examples of these logos

(Windows Vista, Windows XP, Intel, and NOKIA) via headphones. The actual study began after this standard warm-up session enabling participants were to adjust their headphone and become familiar with the MediaLabTM software. After warm-up, participants heard three target sogos

each comprising five tones– one with an ascending contour, another with a descending contour

and a third with a zigzag contour, i.e., all treatment conditions but in randomized order.

Immediately after hearing a sogo, they rated fluency in response to the item “This sonic logo is

easy for an average person to hum” (1— strongly disagree, 7— strongly agree) for that sogo.

Next, they answered two other questions about the sogo they had just rated for fluency –

a) whether the sogo was predominantly ascending, descending or zigzagging and b) the number

of tones in the sogo. Nearly 73% rated the sogo contour correctly (88% ascending, 76%

descending, 54% zigzag). The number of tones was correct to within a tone in 96% instances

(ascending 94%, descending 96% and zigzag 96%). The average number of tones indicated was

4.79, 4.87 and 4.91 respectively for ascending, descending and zigzag sogos. This evidences that

the participants possessed a vivid auditory imagery for the respective sogos when rating fluency.

In order to minimize carry-over effects, participants performed a distraction task between

treatments requiring them to recall a famous song, track a particular word several steps into the

song and perform a task such as supplying a rhyming word.

Next, participants rated recognition for six sogos, the three targets they had heard and

rated for fluency before and three foils not heard before. The presentation order was

randomized. The ascending, descending and zigzag foil sogos were similar on other objective

musical properties respectively to the ascending, descending and zigzag target sogos.

Immediately after hearing a sogo, participants rated recognition in response to the item “The

sonic logo I just heard was played in the first set too” (1— strongly disagree, 6— strongly agree)

for that sogo.

Finally, participants rated the foils for fluency per procedure outlined above, were

thanked, debriefed and dismissed.

Results

Pre-Analysis revealed that all variables of interest were normally distributed and visual

inspection did not reveal any significant skews.

A multi-level unconditional means model revealed participant heterogeneity and

clustering on recognition and fluency ratings with intra-class correlations exceeding 5% (ICC –

true recognition = 11.96%; positive affect = 9.83%; fluency (targets) 45.89% fluency (foils) =

43.32%) thus justifying the within-group design.

Analysis of the influence of contour on fluency of targets via a repeated measures

ANOVA revealed a significant sphericity (Mauchly’s W (2) = 6.85, p < .033). Contour had a

significant effect on the target fluency after applying Hyunh-Feldt corrected degrees of freedom

for sphericity (F (1.80, 73.69) = 8.15, p <.001). Post-hoc tests were conducted using Bonferroni

adjusted alpha levels of .017 per test (.05/3). Results indicated that pairwise comparisons of the

targets’ fluency between the descending (M = 5.71) and zigzag (M = 4.83) was significant. The

fluency in ascending condition (M = 5.29) was not significantly different from the other

conditions.

Analysis of the influence of the contour on fluency of foils revealed a non-significant

sphericity (Mauchly’s W (2) = 2.68, p > .262). Contour did not obtain a significant effect on the

foils’ fluency (F (2, 82) = 1.22, p >.336). Post-hoc tests were conducted using Bonferroni

adjusted alpha levels of .017 per test (.05/3). Results indicated that all pairwise comparisons of

the foil fluency among the three conditions, ascending (M = 5.8), descending (M = 4.81) or

zigzag (M = 3.84) were non-significant.

Analysis of the influence of the contour on positive affect revealed a non-significant

sphericity (Mauchly’ W (2) = .31, p > .857). Contour had a significant effect on positive affect

(F (2, 82) = 3.96, p <.027). Post-hoc tests were conducted using Bonferroni adjusted alpha levels of .017 per test (.05/3). Results (Figure 4) indicated that the positive affect in the zigzag (M =

4.45) condition was significantly higher than that in the ascending (M = 3.74) condition. The positive affect in the descending condition (M = 4.17) was not significantly different from those in the other two conditions.

Bootstrap estimates (Krull and Mackinnon 2001; Mackinnon, Lockwood and Williams

2004) of the indirect effects of fluency (β1β3 = .13, β2β3 = -.13) on positive affect, via a multi-

level modeling with dummy coded column vectors β1 and β2 representing contour, was significant (Figure 4) for β1β3 with 95% confidence intervals excluding zero (-.02 < β1β3 <- .28) and marginally significant for β2β3 (-.32 < β2β3 <.02) with the upper confidence limit just

falling outside zero. Further, a Sobel test revealed that the pathway from number of tones to

positive affect through fluency (β1β3 and β2β3) was significant (Z = 1.73, p < .04 and Z = -2.77,

60 p < .003 respectively), providing evidence for fluency mediation of influence of contour on positive affect. Thus, H7 is supported.

Figure 4: Influence of Contour on Positive Affect

Analysis of the influence of the contour on true recognition revealed a non-significant sphericity (Mauchly’ W (2) = .13, p > .935). Contour had a significant effect on the true recognition (F (2, 82) = 3.62, p <.031). Post-hoc tests were conducted using Bonferroni adjusted alpha levels of .0167 per test (.05/3). None of the pairwise comparisons of true recognition among the three conditions – ascending (M = 4.02), descending (M = 4.00) and zigzag (M =

4.85) was significant.

Analysis of the influence of the contour on false recognition revealed a non-significant sphericity (Mauchly’ W (2) = 1.18, p > .555). Contour had a significant effect on the false recognition (F (2, 82) = 11.72, p <.000). Post-hoc tests were conducted using Bonferroni adjusted alpha levels of .017 per test (.05/3). Results indicated that the false recognition in the

zigzag (M = 3.29) condition was significantly higher than that in the ascending (M = 1.79)

condition. The false recognition in the descending condition (M = 2.50) was not significantly

different from those in the other two conditions.

Collapsed d-primes across participants (Macmillan and Creelman 1991 p275) for the three conditions revealed non-zero sensitivity for all the three conditions (d’ascending =1.61, z =

5.01, p < .000; d’descending =.87, z = 3.06, p < .001; d’zigzag =.88, z = 2.97, p < .001). Pairwise

comparisons of sensitivity across conditions revealed that the sensitivity to ascending condition

was significantly higher than that for descending condition (z = 1.73, p < .046) and that for

zigzag condition (z = 1.68, p < .042). The pairwise comparison of the sensitivity index between

ascending and descending condition was non-significant.

Discussion

As expected, the results evidence processing mediation fluency of affect. Fluency is

expected per past literature to be affectively positive. Again the item “This sonic logo is easy for

an average person to hum” (1— strongly disagree, 7— strongly agree) appears to capture the

fluency construct and performs well. The results also increase nomological validity for the

fluency measure. It is also interesting to see the joint action of content and experience on

building positive affect. Purely based on content (i.e., the direct effect of contour with fluency in

the model) zigzag likeability outstrips descending and ascending contours (Mzd = 2.99, Mdd =

2.44; Mad = 2.14) because of its interestingness. Concurrently the contribution to likeability from

fluency kicks in on top but with a lesser increment for the zigzag contour than for others leading

to overall likeability scores. Thus, both content and fluency interact to contribute to overall

likeability in opposing ways.

False recognition appears to recruit content- based representative heuristic (the tones went up and down) for zigzag contours such that previously unheard zigzag tones induce illusory familiarity. Indeed the fluency ratings are not significantly different because a) the number of tones was held constant and b) zigzag patterns were almost linear with only one kink. Thus, fluency experience even if recruited is uninformative in this case. It would be interesting to see if the fluency experience trumps the content heuristic when number of tones increase such that it can afford a more complex zigzag pattern.

Distinctiveness for a branding device (logo or sogo) emanates not merely from true recognition but from a combination of true recognition and correct rejection. It is interesting to see that the sensitivity analysis with d-prime measures evidences such distinctiveness for the ascending contour over both descending and zigzag contours. Although zigzag contour is interesting, it also seems prone to high false recognition due to content-based representative heuristic. Another guideline appears to emerge as a sogo design characteristic– brands seeking distinctiveness would do well to target a linear contour preferably ascending contour.

Conversely, brands in low involvement categories seeking to ride on illusory familiarity should prefer a zigzag contour. On the other hand, brands in low involvement category or hedonic category may be driven by likeability where zigzag contour performs better. Thus, it is perhaps appropriate that Windows VistaTM sogo has an ascending contour being in a more utilitarian category and NOKIA sogo is zigzag. Interestingly, both Intel and NOKIA have a zigzag contour.

However, how might a zigzag contour with fewer tones perform against a linear contour but with many tones? I explore such interaction next, in the fifth study.

Study 5

The fifth study explored the joint influence of number of tones and contour on positive

affect true recognition, false recognition and signal sensitivity in a within-groups setting. Results

in the fourth study evidenced minimal differences between ascending and descending contours

although ascending contour evidenced more extreme results. Accordingly, ascending pattern was

chosen to represent linear contour, to be pitted against a zigzag pattern. Each level of contour

was represented by two sogos varying on the number of tones comprising either four or seven

tones. Thus, the stimulus consisted of four target sogos varying systematically on the two design

characteristics and four foil sogos respectively matched on other objective musical properties.

Sample consisted of sixty-eight undergraduate business majors from a public Midwestern

university, USA. Participants checked in at the appointed time, were taken through a warm-up

session on the MediaLabTM software where they were introduced to sonic logos. They heard

several examples of these logos (Windows Vista, Windows XP, Intel, and NOKIA) via

headphones. The actual study began after this standard warm-up session during which time

participants were able to adjust their headphone and become familiar with the MediaLabTM software. After warm-up, participants heard the four target sogos, i.e., all treatment conditions but in randomized order. As before, immediately after hearing a sogo, they rated fluency in response to the item “This sonic logo is easy for an average person to hum” (1— strongly disagree, 7— strongly agree) for that sogo.

Next, they answered two other questions about the sogo they had just rated for fluency – a) whether the sogo was predominantly ascending, descending or zigzagging and b) the number of tones in the sogo. Nearly 61% rated the sogo contour correctly. The number of tones was correct to within a tone in 92% instances. The average number of tones indicated was 4.1, 6.88

respectively for 4-tone and 7-tone sogos. This evidences that the participants possessed a vivid

auditory imagery for the respective sogos when rating fluency. In order to minimize carry-over

effects, participants performed a distraction task between treatments requiring them to recall a

famous song, track a particular word several steps into the song and perform a task such as

supplying a rhyming word.

Next, participants rated recognition for eight sogos, the four targets they had heard and

rated for fluency before and four foils not heard before. The presentation order was randomized.

The foil sogos were similar on other objective musical properties respectively to the target sogos.

Immediately after hearing a sogo, participants rated recognition in response to the item “The

sonic logo I just heard was played in the first set too” (1— strongly disagree, 6— strongly agree)

for that sogo.

Finally, participants rated the foils for fluency per procedure outlined above, were

thanked, debriefed and dismissed.

Results

Pre-Analysis revealed that all variables of interest were normally distributed and visual inspection did not reveal any significant skews.

Analysis of the influence of contour and number of tones on fluency of targets via a repeated measures ANOVA revealed a significant main effect of number of tones (F (1, 67) =

20.59, p <.000), a significant main effect of contour (F (1, 67) = 18.35, p <.000) and a significant interaction effect of number of tones X contour (F (1, 67) = 13.48, p <.000). Post-hoc tests were conducted using Bonferroni adjusted alpha levels of .0125 per test (.05/4). Results indicated that pairwise comparisons of the targets’ fluency between the linear and zigzag contours was

significant for the 7-tone sogos (Ms = 5.15 vs. 4.02 respectively) but not for the 4-tone (Ms =

5.68 vs. 5.54 respectively) sogos.

Analysis of the influence of contour and number of tones on fluency of foils revealed a

significant main effect of number of tones (F (1, 67) = 32.19, p <.000), a marginally significant

main effect of contour (F (1, 67) = 3.91, p <.052) and a significant interaction effect (Figure 5) of number of tones X contour (F (1, 67) = 7.21, p <.009). Post-hoc tests were conducted using

Bonferroni adjusted alpha levels of .0125 per test (.05/4). Results indicated that pairwise

comparisons of the foils’ fluency between the linear and zigzag contours was significant for the

7-tone sogos (Ms = 4.87 vs. 4.16 respectively) but not for the 4-tone (Ms = 5.63 vs. 5.71 respectively) sogos.

Analysis of the influence of contour and number of tones on false recognition revealed a non-significant main effect of number of tones (F (1, 67) = .04, p >.842), a non-significant main effect of contour (F (1, 67) = .49, p >.491) but a significant interaction effect (Figure 5) of number of tones X contour (F (1, 67) = 21.57, p <.000). Post-hoc tests were conducted using

Bonferroni adjusted alpha levels of .0125 per test (.05/4). Results indicated that pairwise comparisons of the false recognition between the linear and zigzag contours was significant for the 7-tone sogos (Ms = 2.59 vs. 3.41 respectively) and for the 4-tone (Ms = 3.60 vs. 2.49 respectively) sogos.

Collapsed d-primes across participants (Macmillan and Creelman 1991 p275) for the four conditions revealed non-zero sensitivity for all the four conditions (d’4-tone linear =.47, z = 2.13, p

< .020; d’4-tone zigzag =1.63, z = 5.11, p < .000; d’7-tone linear = 1.17, z = 6.61, p < .000; d’7-tone zigzag

=..45, z = 2.08, p < .0201). Further, pairwise comparisons of sensitivity across conditions

revealed that the sensitivity of zigzag contour was significantly higher than (z = 3.49, p < .000)

that for linear contour for 4-tones but, this pattern reversed for 7-tones with sensitivity of linear

contour being significantly higher than (z = -1.68, p < .042) that of zigzag contour.

Figure 5: Interaction Effects of Tone and Contour on Fluency and False Recognition

Discussion

Study 5 reveals interesting interaction between number of tones and contour on false recognition. First, there is no main effect on number of tones. This suggests that the content- based heuristic (I heard a denser sogo before) does not surface for seven tones as it presumably did for the 9-tone sogo in study 3. However, fluency experience should degrade as number of tones increase and consequently lead to lower false recognition for the 7-tone sogo in comparison to the 4-tone sogo. This pattern is seen for the linear contour. However, this pattern completely reverses for the zigzag contour. Although fluency goes down for zigzag contour (and this is expected because both contour and number of tones thwart fluency) false recognition goes up. Thus, the judgment appears to lean on content- based representative heuristic (the tones went up and down) for zigzag contours such that previously unheard zigzag tones induce illusory familiarity. It is interesting to see that the fluency experience is not able to trump the content heuristic when number of tones increase such that it can afford a more complex zigzag pattern.

Indeed, the more complex the contour pattern attained, the stronger the deployment of content

heuristic.

It is interesting to see that the sensitivity analysis with d-prime measures mirror the

influence of content heuristic on false recognition i.e., false alarm rates. Because fluency trumps

content heuristic for the linear contour, increasing number of tones reduces false alarm rates. In contrast, the content heuristic trumps fluency experience for the zigzag contour thereby increasing the false alarm rates as number of tones increase. Hit rates degrade but more gently with increase in information load leading to a dramatic reversal in sensitivity. Zigzag contour is more distinctive with less number of tones but linear contour is more distinctive with increase in number of tones.

General Discussion

The results across these five studies evidence the influence of several systematic variations of design characteristics on response dimensions and bring alive the proposed definition of sonic branding.

Sonic branding is the creation and perpetuation of a consistent, distinctive, universal, and appropriate non-verbal aural identity for a brand as a unique configuration of evaluative judgments of familiarity, liking, recognition and personality, through the considered arrangement of design characteristics using natural or synthesized sounds.

Sonic logos are good examples illustrating this phenomenon.

A sogo is a unifying, focal sonic branding device. Whereas a logo is a summary brand expression visually, a sogo is the most parsimonious brand expression in sound. Therefore, research on sogo will inform us of the nature, function and outcomes of sonic branding at its most rudimentary atomistic level.

As noted earlier, my dissertation makes a significant contribution to this stream in a) exploring an alternative (auditory) domain, b) refining dependent variables on recognition with sensitivity analysis, c) enquiring processing fluency mediation, d) using an experimental manipulation of the design characteristics and e) incorporating multilevel modeling for methodological robustness. It is also directly relevant to the marketing stream.

Managerial Contributions: Study 1 demonstrates a strategic outcome flowing from sonic branding in showing that willingness-to-pay can be systematically influenced through appropriate design characteristics for a sogo. That is, sonic branding is a strategic revenue enhancement tool. It could also increase value to the firm in optimizing branding costs. For instance, Microsoft incurred a licensing cost of $12 million for the song by Madonna “Ray of

Light,” to launch Windows XP (Krasilovsky and Shemel 2007). Further, a 30-second spot may cost anywhere between $250,000 for a moderately popular program to $705,000 for a top program such as American Idol (Krasilovsky and Shemel 2007). Although, awareness, familiarity and likeability can be achieved with sufficient repetition through classical conditioning, these levels of investments underscore the need for optimal designs. As these studies demonstrate, several differential outcomes obtained with just two exposures. For instance, WTP differed in study 1 between groups after just two exposures to objectively unfamiliar sogos purely based on design characteristics of the sogos.

Theoretical Contributions: Study 2 demonstrated the moderating influence of chunking.

Chunking increases fluency and could potentially drive downstream response dimensions like recognition, affect and distinctiveness – i.e., favorable brand outcomes. Study three through five help cobble together a theoretical framework in the nascent field of sonic branding in exploring the influence of number of tones and the contour of a sogo on the response dimensions and in evidencing the processing fluency mediation process in detail. Interesting insights obtain. False recognition and illusions of familiarity obtain through the joint interplay of the content-based representative heuristic and the processing fluency. Zigzag contour and denser sogos appear to recruit representative heuristic trumping the fluency influence. In contrast, linear contours and sogos with fewer tones rely on fluency experience to influence response dimensions. Why would

this happen? Tybout et al (2005) show for instance, that reliance on accessibility experience is greatest when the expectation of the fluency experience is moderate. When the content is easily accessible, fluency is blatant and therefore discounted. At the other extreme when knowledge is extremely difficult to access, it is attributed to task difficulty, i.e., disfluency is not diagnostic.

Perhaps similar mechanisms are at work here.

Sensitivity analysis is an interesting layer in this stream of research. Previous research

(e.g., Henderson and Cote 1998) explores the two recognition constructs independently.

Although, 3-tone sogo is better on both true and false recognition scores, 6-tone is truly distinctive. Although a zigzag sogo is more likeable, and scores high both on true and false recognition it is least distinctive. However, when number of tones is few, zigzag tones are more

distinctive.

Methodological Contributions: In isolation none of the methods deployed in this

dissertation are new; however, this dissertation brings together methods of analysis and

experimental design in an interesting way. Whereas experimental aesthetics research typically

favors correlational approach where IVs are rated and factor analyzed (e.g., Henderson and Cote

1998; Aaker 1997) this dissertation takes a causal approach in deploying experimental

manipulation of design characteristics. Within-subject design is not as often encountered in

consumer behavior research presumably because of demand effect concerns. This dissertation

leverages the power inherent in within-subject design while minimizing demand effects.

Although necessary and used per within-subject design – multilevel meditational analysis,

bootstrapping and repeated measures ANOVA – techniques deployed in this dissertation are

interesting techniques less commonly encountered in the consumer behavior research. Processing

fluency is either manipulated or measured through response latencies; neither approach is very

71 amenable to auditory research. The operationalization of the construct via hummability assessment is an interesting contribution. Taken together, this eclectic confluence of within- subject design, auditory stimuli and the accompanying analytical techniques bring forth some innovation.

Sogo Guidelines: The five studies illustrate general principles. I illustrate some specific examples where these principles may be applied as guidelines. Products may be classified along hedonic and utilitarian dimensions (Voss, Spangenberg and Grohman 2003). Products that are low on utilitarian dimension but high on hedonic dimension (e.g., beer video games) would benefit from a high false recognition and positive affect. Therefore, sogos for brands in this category should be designed with a zigzag contour (high on affect, study 4, high on false recognition), and six or fewer tones (high on affect, false recognition study 3). Products that are high on utilitarian dimension but low on hedonic dimension (e.g. alkaline batteries, shoelaces) would benefit from high true recognition and high affect (because the products in themselves are boring). These products would benefit from a sogo with fewer tones and an ascending contour

(study 3). Products high on both dimensions (e.g., automobiles, athletic shoes) need to be distinctive. Corresponding sogos should score high on distinctiveness and should deploy a 6-tone ascending sogo (high on distinctiveness study 3). High involvement product categories (e.g., laptop) would benefit from high distinctiveness and positive affect. They should benefit from a

4-tone sogo with a zigzag contour (high on distinctiveness study 5, high on affect study 4). Low involvement product categories would benefit from a high false recognition and positive affect

(5-tone zigzag contour, study 4).

Finally, whereas this dissertation illustrates the sonic branding potential it studied only two design characteristics in detail although several characteristics were suggested as

72 possibilities worthy of research (Figure 6). As this research stream advances, guidelines could be richer and fine-tuned to specific situations. For instance, zigzag sogos engender positive affect.

Minor tones create a somber mood. What might be the outcome of an interaction between these two characteristics?

Going beyond musical sounds, how might product related sounds (fizzing can, keyboard clicks, camera clicks) contribute to sonic branding? Might sonic branding convey a brand personality? These are areas for future research beyond this dissertation that I briefly discuss next.

Beyond this Dissertation

In this dissertation research, I assessed the relationship between design characteristics and response dimensions relating to recognition and affect. Taken together, I consider these dimensions to influence the brand awareness and I refer to this aspect of sonic branding as sonic awareness. However, I consider this dissertation as the beginning of a research stream with many questions needing investigation. I briefly describe three specific questions to illustrate possibilities in this research stream beyond this dissertation.

Conceptual Fluency and Sonic Branding: Lee and Labroo (2004) show that when a product is conceptually primed visually, preferences alter. Thus, participants that view mayonnaise may subsequently evaluate a brand of ketchup more positively because the former conceptually primes the latter. However, if an unrelated product is presented before evaluation, conceptual priming does not obtain. Sounds are uniquely linked to product/brand contexts and have the potential to provide authentic auditory identity. For instance, the shutter sound is innately associated with a camera and can semantically prime the product category and render the product conceptually fluent. Because fluency mediates several response dimensions it is possible that deployment of sounds inextricably linked to a product context could lead to positive evaluations for the product. Indeed many cell phones that have the capability to photograph but do not use a shutter mechanism nevertheless deploy the shutter sound. How might auditory semantic priming lead to conceptual fluency?

Sonic Personality: What is a good Sogo? The ‘Short Scale of Musical Preferences or

STOMP’ (Rentfrow and Gosling 2003) is a 14-item scale that identifies personality correlates based on listener preferences. For instance, listeners preferring Dance/Electronica, Rap/Hip-hop, and Soul/Funk are found to have an energetic and rhythmic personality. This study thus implies

74 that the music of a particular genre perhaps possesses the correlated personality. Two challenges arise in deploying the musical genre taxonomy to derive the personality type. First, the musical genres are ill defined (Aucouturier and Pachet 2000). Second, a short musical clip that lasts no more than 6 seconds virtually precludes such identification of genre. Yet, even a short clip reveals many things— the instrumentation, mode, whether or not there was a leap, the direction of the musical contour etc., that may conceivably reveal unique personalities. For instance, drums and trombone have been found to connote masculinity while violin and flute have been found to be feminine (Abeles and Porter 1978) and brass wind instruments found majestic

(Bruner 1990).

As discussed earlier, music can represent different motions, in turn painting different personalities. Design characteristics can determine the personality of a sogo. For example, the baby care products from Johnson and Johnson are positioned on the mildness, tenderness platform appropriate for the delicate treatment for a baby (The shampoo cleans gently – the no tears shampoo). This personality can perhaps be brought alive with a sogo that has just three, distinct and delicate tones in the upper octave played out on a xylophone. In contrast, the Ivory shampoo that ‘cleans and conditions gently’ (with a baseline -- not fancy just fabulous) is gentle, rich and flowing but not baby soft. A few legato bows on the violin around the middle C could bring alive this mellifluous personality steering away from delicate tenderness yet retaining the gentleness component. The sogo for a freshness shampoo may convey energy with bounding leaps, comprising three tri tones covering a wide range on the key board, and played out on an electric piano. Thus, the sonic personality may vary based on pitch, timing and instrumentation.

However, there is a need to convert these ad hoc speculations along a formal personality scale.

Aaker’s brand scale (1997) merits relevant discussion here. According to Aaker, (1997) consumers impute human personality traits or ‘animism’ to brands and thus have a symbolic relationship with the brands. Essentially, animism seeks responses to questions of the type “If this brand was a human being, will he/she be ______?” For instance, brands may be seen as down-to-earth, daring, intelligent or feminine. Based on exploratory/confirmatory factorial reductions, Aaker’s (1997) 42-item Brand Personality Scale (BPS) with five dimensions and fifteen facets has been found to be reliable, generalizable and valid (Sweeney and Brandon

2006). It is interesting to note that several items in the BPS scale, readily lend themselves to adjectival descriptions for auditory stimuli. For example, it is conceivable that a sogo may be described outdoorsy, sophisticated, exciting or sincere, paralleling dimensions and facets of the brand personality scale (Aaker 1997). I believe that animism is equally applicable to music as it is to a brand. It is equally feasible to ask the question, “If this sogo was a human being, will he/she be ______?” Given the related paradigm, i.e., measuring personality in a branding context, I propose a personality classification for sogos based on Aaker’s (1997) BPS scale.

Arguments may be raised against using a brand personality scale to glean the sonic personality. Indeed even the adequacy of the traits included in the brand personality scale has been questioned. For example, Sweeney and Brandon (2006) contest that brands positioned along negative traits— arrogant, cocky or controlling, may appeal to certain subcultures but

Aaker’s scale includes only positively valanced items. However, the essence of lexical approach is that representative items capture the underlying factors and a complete enumeration is both impossible and unnecessary given the factor analytic approach. Second, this scale is built off lexically based five factor personality models that have a long tradition (see Goldberg 1990 for a review) and a wide acceptance among the stakeholders in the personality literature. Finally, the

overarching principle in scale development is that new scales should be developed only if every

possibility of using existing validated scales is exhausted (Churchill 1979; Bruner 2003); Aaker’s

(1997) scale is a relevant and viable possibility with strong measurement properties (Sweeney

and Brandon 2006). Thus, sonic personality construct offers a rich extension to this research

stream.

When is Sonic branding appropriate? According to Construal Level Theory (see Trope

and Lieberman 2003 for a review), people create different mental representations of a judgment

object; higher-level construal is more abstract, captures the core, but omits the details while

lower level construal is more concrete and specific. When the object of judgment is at a distance,

spatially or temporally abstract judgments are formed. Conversely, concrete judgments are

formed when the judgment object is closer in time and space. Brands are judgment objects and

can be construed at different levels of abstraction too. A brand corresponding to an FMCG good

can be looked at, touched and its quality is relatively invariant from sample to sample. Often

more information may be available from prior consumption of FMCG brands. Thus, it is easy to

construe the brand promises of an FMCG brand. However, the brand promises of service brands

are more abstract given that services are a) intangible b) inseparable from producer c)

heterogeneous and d) perishable (Parasuraman, Zeithaml, and Berry 1985). Service brands promise a positive intangible experience, for instance, requiring respondents to imagine an efficient banking process or reassurance in seeking legal consultancy. Sonic branding is more relevant where an enduring identity is desired in a relatively abstract context.

Corporate identity is another area where the positioning is typically abstract. Corporate identity articulates ‘ethos, aims and values’ (Riel and Balmer 1997). For example, organizations may prefer to be identified as innovative (‘leap ahead,’ IntelTM) or easygoing but serious

(GoogleTM); and because music can evoke meaning relating to movement and in turn emotions, a

sogo is more versatile than a static logo in facilitating such abstract construal. When brand

extensions move a brand to newer product categories, the span of interpretation for the brand

expands to a more abstract umbrella position. In extreme cases, the brand is the corporate entity

(e.g., DellTM). For such brands, sonic branding is more relevant.

Finally, sonic branding is relevant for ingredient branding where the ingredient brand cannot be seen but only experienced (e.g., ‘IntelTM inside,’ ‘the LycraTM stretch’). For instance, the elastic properties of ‘the LycraTM stretch’ would come alive with a ‘Glissando’ from C4 to

G4 and back (or even to C5) on a violin. Thus, sonic branding is more appropriate for a)

corporate branding b) brands with wide span of extensions c) services branding and d) ingredient

branding. Thus, situational appropriateness is a third interesting extension to this dissertation.

Some Final Thoughts

Sonic branding is on ascendancy. A Google search on “sonic branding” tops twenty thousand results. Dozens of sonic branding websites advertise impressionistic solutions. Ford motors believes that consumers make conclusions about product quality based on gadget sounds and now designs Taurus car doors to make a “vault-like” sound (Kiley 2007). Mercedes has recently launched a three-tone sonic logo. To increase brand recognition, Lufthansa is introducing an ascending four-tone corporate sonic logo that is expected to invoke the idea of

“taking-off” and “well-being” (Walther 2006).

Despite this strategic appreciation for sonic branding, dependence on musicians’ subjective opinions continues. According to Steve Ball, Microsoft Corp., musician Robert Fripp was recruited to design the four-tone Windows Vista start-up sound and the effort took eighteen months (Linn 2006). Although industry captains, notably technology and auto sectors, have begun to incorporate sonic branding in their branding strategies because they see the strategic value in sonic branding, there is implementation ambivalence due to lack of research-based design guidelines in this area. It appears that there is a general reluctance to commit investments relying on anecdotal solutions provided by the mushrooming sonic branding industry.

This dissertation addresses this managerial need by offering a comprehensive definition and outlining a conceptual framework and five empirical studies to jump start research on sonic branding. Whereas this is a relatively new area of investigation, we expect these to be refined further as the propositions developed here meet investigation. Over time, findings in this stream could enable the marketers to make better-informed sonic branding decisions and to lean on musicians for execution rather than branding strategy. I also come back to the refrain that the auditory domain is relatively underrepresented in consumer judgment contexts. Research in this

79 area would extend the context of consumer judgments. After all, branding is not an activity that takes place in silence. Thus, sonic branding is an essential and somewhat ignored piece, to the development of the branding literature.

Ascending Descending

Symmetric valley Symmetric Peak

D. Direction of Motion

fol gers in a cup E. Leaps

Major Minor F. Modes

Figure 6. Illustrative Examples of Pitch Related Independent Variables for a Sogo

References

Aaker, Jennifer L. (1997), “Dimensions of Brand Personality,” Journal of Marketing Research,

34(3), 347-356.

Abeles, Harold F. and Susan Yank Porter (1978), “The Sex-Stereotyping of Musical

Instruments,” Journal of Research Music Education, 26(2), 65-75.

Alpert, Mark I., Judy I. Alpert, and Elliot N. Maltz (2005), “Purchase Occasion Influence on the

Role of Music in Advertising,” Journal of Business Research, 58, 369-376.

Aucouturier, Jean-Julien and François Pachet (2000), “Representing Musical Genre: A State of

the Art,” Journal of New Music Research, 32(1), 83-93.

Balkwill, Laura-Lee, William Forde Thompson and Rie Matsunaga (2004), “Recognition of

Emotion in Japanese, Western and Hindustani Music by Japanese Listeners,” Japanese

Psychological Research, 46(4), 337-349.

Baddeley, Alan D. (2000), “The Episodic Buffer: A New Component of Working Memory,”

Trends in Cognitive Sciences, 4 (11), 417-23.

Baddeley, Alan D. and Graham J. Hitch, (1974), “Working Memory,” in The Psychology of

Learning and Motivation, Bower, G.A. ed., New York: Academic Press, 47–90.

Baddeley, Alan D. and Robert Logie, (1992), “Auditory Imagery and Working Memory,” in

Auditory Imagery, Daniel Reisberg ed. Hillsdale NJ: Lawrence Erlbaum Associates, 179–

195.

Baron, Reuben M. and David A. Kenney (1986), “The Moderator-Mediator Variable Distinction

in Social Psychological Research: Conceptual, Strategic and Statistical Considerations,”

Journal of Personality and Social Psychology, 51(6), 1173-1182.

Berlyne, Daniel E. (1971), Aesthetics and Psychobiology. New York: Meredith Corporation.

Birnbaum, Michael H. (1999), “How to Show That 9 >221: Collect Judgments in a Between-

Subjects Design,” Psychological Methods, Vol. 4(3), 243-249.

Boltz, Marilyn G. (2001), “Musical Soundtracks as a Schematic Influence on the Cognitive

Processing of Filmed Events,” Music Perception, 18, 427-454.

Boltz, Marilyn G. (2004), “The Cognitive Processing of Film and Musical Soundtracks,”

Memory and Cognition, 32(7), 1194-205.

Bornstein, Robert F., Paul R. D’Agostino (1992), “Stimulus Recognition and the Mere Exposure

Effect,” Journal of Personality and Social Psychology, Vol. 63 (4), 545-552.

Brewster, Stephen A. (1997), “Using Non-Speech Sound to Overcome Information Overload,”

Displays, 17, 179-189.

Brooker, John R. and David B. Eastwood (1989), “Using State Logos to Increase Purchases of

Selected Food Products,” Journal of Food Distribution Research, February 175-183.

Bruner, Gordon C., II (1990), “Music, Mood, and Marketing,” Journal of Marketing, 54(4), 94-

104.

Bruner, Gordon C., II (2003), “Combating Scale Proliferation,” Journal of Targeting

Measurement and Analysis for Marketing, 11(4), 362-372

Churchill, Gilbert A. Jr. (1979), “A paradigm for developing better measures for marketing

constructs,” Journal of Marketing Research,16, 64-73.

Colavita, Frank. B. (1974), “Human Sensory Dominance: Auditory Versus Visual Stimulus,”

Perception and Psychophysics, 16, 409-412.

Crowder, Robert G. (1989), “Imagery for Musical Timbre,” Journal of Experimental

Psychology: Human Perception and Performance, 15, 472-478.

Droney, Maureen (2004), “L.A. Grapevine,” in Mixonline.com, (accessed April 28, 2008),

[available at http://mixonline.com/mag/audio_la_grapevine_66/]

Eyenck, Michael W. (1982), Attention and Arousal. New York: Springer-Verlag.

Farah, Martha J. and Albert F. Smith (1983), “Perceptual Interference and Facilitation with

Auditory Imagery,” Perception and Psychophysics, 33, 475-478.

Feinberg, Richard A. (1986), “Credit cards as spending facilitating stimuli: A Conditioning

interpretation,” Journal of Consumer Research, 13 (December), 348-356.

Fiske, Susan T. and Shelley E. Taylor (1991), Social cognition (2nd Ed.). New York: McGraw-

Hill.

Galak, Jeff and Leif D. Nelson (2009), “Complexity is Good: When Disfluent Communication

Signals Author Erudition,” in Proceedings of the Society for Consumer Psychology

Winter 2009 Conference, Alexander Chernev, Michael Herzenstein and Shilendra Pratap

Jain, eds. San Diego, CA, 70.

Goldberg, Lewis R. (1990), “An Alternative “Description of Personality”: The Big-Five Factor

Structure,” Journal of Personality and Social Psychology, 59(6), 1216-1229.

Gorn, Gerald. J. (1982), “The effect of music in advertising on choice behavior: A classical

conditioning approach.,” Journal of Marketing, 46, 94-101.

Greenwald, Anthony G and Banaji R Mahzarin (1995), “Implicit Social Cognition: Attitudes,

Self-Esteem, and Stereotypes Psychological Review,” Psychological Review, 102, 4-27.

Gregg, Vernon H. (1976), “Word Frequency, Recognition and Recall,” in Recall and

Recognition ed., John Brown, London: Wiley

Gundlach, Ralph H. (1932), “A quantitative analysis of Indian music,” American Journal of

Psychology, 44, 133-145.

Gundlach, Ralph H. (1935), “Factors determining the characterization of musical phrases,”

American Journal of Psychology, 47, 624-643.

Halpern, Andrea R. (1992), “Musical Aspects of Auditory Imagery,” in Auditory Imagery,

Daniel Reisberg ed. Hillsdale NJ: Lawrence Erlbaum Associates, 1-27.

Halpern, Andrea R. (1984), “Perception of Structure in Novel Music,” Memory and Cognition,

10, 496-512.

Halpern, Andrea R. (1988a), “Mental Scanning in Auditory Imagery for Songs,” Journal of

Experimental Psychology: Learning, Memory and Cognition, 14, 434-443.

Halpern, Andrea R. (1988b), “Perceived and Imagined Tempos of Familiar Songs,” Music

Perception, 6, 193-202.

Hargreaves, David J. and Adrian C. North (1997). The social psychology of music. New York:

Oxford University Press.

Heinlein, Christian P. (1928), “The affective characters of the major and minor modes in music,”

Comparative Psychology, 8, 101-142.

Henkin, Robert I. (1955), “A factorial study of the components of music,” Journal of

Psychology, 39, 161-181.

Henkin, Robert I. (1957), “A reevaluation of factorial study of the components of music,”

Journal of Psychology, 43, 301-306.

Henderson, Pamela W. and Joseph A. Cote (1998), “Guidelines for Selecting or Modifying

Logos,” Journal of Marketing, 62 (2), 14-30.

Herzog, Stefan M., Jochim Hansen, and Michaela Wanke (2007), “Temporal Distance and Ease

of Retrieval,” Journal of Experimental Social Psychology, Vol. 43, 483-488.

Hevner, Kate (1935), “The affective character of the major and minor modes in music,”

American Journal of Psychology, 47, 103-118.

Hevner, Kate (1936), “Experimental studies of the elements of expression in music,” American

Journal of Psychology, 48, 246-268.

Hevner, Kate (1937), “The affective value of pitch and tempo in music,” American Journal of

Psychology, 49, 621-630.

Hermann, Thomas (2002), “Sonification for Exploratory Data Analysis,” Dissertation Thesis,

(accessed April 27, 2008), [available at

http://sonification.de/publications/media/Hermann2002-SFE.pdf]

Higgins, Tory E. (1996), “Knowledge Activation: Accessibility, Applicability, and Salience,” in

Social Psychology: Handbook of Basic Principles, eds. E. Tory Higgins and Arie W.

Kruglanski, New York: Guilford, 133-168.

Hirt, Edward R. and N. John Castellan Jr. (1988), “Probability and Category Redefinition in the

Fault Tree Paradigm,” Journal of Experimental Psychology, Vol. 14(1), 122-131.

Holbrook, Morris B. and Punam Anand (1990), “Effects of tempo and situational arousal on the

listener’s perceptual and affective responses to music,” Psychology of Music, 18, 150-

162.

Intons-Peterson, Margaret Jean (1992), “Components of Auditory Imagery,” in Auditory

Imagery, Daniel Reisberg ed. Hillsdale NJ: Lawrence Erlbaum Associates, 45-72.

Jackson, Daniel M. (2003), Sonic Branding an introduction. Edited by Paul Fulberg, Palgrave

Macmillan.

Jacoby, Larry L. and Mark Dallas (1981), “On the Relationship between Autobiographical

Memory and Perceptual Learning,” Journal of Experimental Psychology: General,

110(3), 306-340.

Jacoby, Larry L. and Kevin Whitehouse (1989), “An Illusion of Memory: False Recognition

Induced by Unconscious Perception,” Journal of Experimental Psychology: General,

118(2), 126-135.

Janiszewski, Chris (1988), “Preconscious Processing Effects: The Independence of Attitude

Formation and Conscious Thought,” The Journal of Consumer Research, 15 (2), 199-

209.

Janiszewski, Chris and Tom Meyvis (2001), “Effects of Brand Logo Complexity, Repetition, and

spacing on Processing Fluency and Judgment,” The Journal of Consumer Research, 28

(June), 18-32.

Kellaris, James J. and Anthony D. Cox (1989), “The effects of background music in advertising:

A reassessment,” Journal of Consumer Research, 16, 113-118.

Kellaris, James J., Anthony D. Cox, and Dena Cox (1993), “The Effect of Background Music on

Ad Processing - A Contingency Explanation,” Journal of Marketing, 57(4), 114-125.

Kellaris, James J. and Robert J. Kent (1993), “An Exploratory Investigation of Responses

Elicited by Music Varying in Tempo, Tonality, and Texture,” Journal of Consumer

Psychology, 2(4), 381- 401.

Kellaris, James J. and Robert J. Kent (1991), “Exploring tempo and modality effects, on

consumer responses to music,” Advances in Consumer Research, 18, 243-248.

Kellaris, James J. and Robert J. Kent (1992), “The influence of music on consumers' temporal

perceptions: Does time fly when you're having fun?” Journal of Consumer Psychology,

1, 365-376.

Kellaris, James J. (2008), “Music and Consumers,” in Handbook of Consumer Psychology,

Curtis Haugtvedt, Paul Herr, and Frank R. Kardes, eds., Mahwah, NJ: Erlbaum, 837-856.

Keller, Kevin L (1993), “Conceptualizing, measuring and managing customer based brand

equity,” Journal of Marketing, 57(1), 1-23.

Kiley, David (2007), “Fine-tuning a Brand’s Signature sound,” www.Businessweek.com,

(accessed May 31, 2008), [available at

http://www.businessweek.com/magazine/content/07_33/b4046054.htm?chan=search]

(December 6) 24-28.

Kramer, Gregory, Bruce Walker, Terri Bonebright, Perry Cook, John Flowers, Nadine Miner,

and John Neuhoff (1997), “Sonification report: Status of the field and research agenda,”

Technical report, International Community for Auditory Display (accessed April 27,

2008), [available at http://www.icad.org/node/400]

Krasilovsky, William M. and Sidney Shemel (2007), This Business of Music. New York: Watson

Guptill.

Krishnan, Vijaykumar, Karen A. Machleit and James J. Kellaris (2008), “Musical Intelligence:

Explication, Measurement, and Implications for Consumer Psychology,” in Proceedings

of the Society for Consumer Psychology Conference, Winter 2008, ed. Maria Cronley and

Dhanajay Nayakankuppam, New Orleans, LA.

Kronlund, Antonia, Bruce W. A. Whittlesea and Carolyn Yoon (2008), “Consumer, Memory,

Fluency and Familiarity,” in Handbook of Consumer Psychology, Curtis Haugtvedt, Paul

Herr, and Frank R. Kardes, eds., Mahwah, NJ: Erlbaum, 77-102.

Krull, Jennifer L. and David P. MacKinnon (2001), “Multilevel Modeling of Individual and

Group Level Mediated Effects,” Multivariate Behavioral Research, 36(2), 249-277.

Krugman, Herbert E. (1965), “The impact of Television Advertising: Learning without

Involvement,” The Public Opinion Quarterly, 29(3), 349-356.

Lee Angela Y. and Aparna A. Labroo (2004), “The Effect of Conceptual and Perceptual Fluency

on Brand Evaluation,” Journal of Marketing Research, 41(May), 151-165.

Lichtenstein, Meryl and Thomas K. Srull (1985), “Conceptual and Methodological Issues in

Examining the Relationship between Consumer Memory and Judgment,” in

Psychological Processes and Advertising Effects: Theory, Research and Application, L.

F. Alwitt and A.A. Mitchell eds., Hillsdale, NJ: Lawrence Erlbaum Associates, 113-118.

Linn, Allison (2006), “Start Me Up: The Sound of Vista,” www.msnbc.com (accessed on May 31

2008), [available at http://www.msnbc.msn.com/id/15656246/]

Lynch, John G. Jr. and Thomas K. Srull (1982), “Memory and Attentional Factors in Consumer

Choice: Concepts and Research Methods,” Journal of Consumer Research, 9(1), 18-37.

Macmillan, Neil A. and C. Douglas Creelman (1991), Detection Theory: A User’s Guide.

Cambridge: Cambridge University Press.

Mantel, Susan Powell, and James J. Kellaris, (2003), “Exploring determinants of psychological

time: The impact of cognitive resources required and available on the estimation of

lapsed time,” Journal of Consumer Research, 29, 531-538.

Mandler, George (1980), “Recognizing: The Judgment of Previous Occurrence,” Psychological

Review, 87(3), 252-171.

Mattila, Anna S and Jochen Wirtz (2001), “Congruency of scent and music as a driver of in-store

evaluations and behavior,” Journal of Retailing, 77, 273-289.

MacKinnon, David P., Chondra M. Lockwood and Jason Williams (2004), “ Confidence Limits

for the Indirect Effect: Distribution of the Product and Resampling Methods,”

Multivariaate Behavioral Research, 39(1), 99-128.

McGlone, Mathew S. and Jessica Tofighbakhsh (2000), “Birds in a Feather Flock Conjointly (?):

Rhyme as Reason in Aphorisms,” Psychological Science, 2(5), 424-428.

Meyer, Leonard B. (1956), Emotion and Meaning in Music. Chicago: The University of Chicago.

Meyers-Levy, Joan and Laura A. Peracchio (1992), “Getting an Angle in Advertising: The Effect

of Camera Angle on Product Evaluations,” Journal of Marketing Research, 29(April),

454-461.

Miller, George A. (1956), “The Magical Number Seven, Plus or Minus Two: Some Limits on

Our Capacity for Processing Information,” Psychological Review, 63(2), 81-97.

Nedungadi, Prakash (1990), “Recall and Consumer Consideration Sets: Influencing Choice

without Altering Brand Evaluations,” The Journal of Consumer Research, 17

(December), 263-276.

Neter, John, Michael H. Kutner, Christopher J. Nachtsheim and William Wasserman (1996),

Applied Linear Statistical Models, 4th edition, 1165-1167.

North, Adrian C. and David J. Hargreaves (1998), “The effect of music on atmosphere and

purchase intentions in a cafeteria,” Journal of Applied Social Psychology, 28, 2254-2273.

North, Adrian C., Hargreaves, David. J., and Jennifer McKendrick (1999), “The Influence of In-

store Music on Wine Selections,” Journal of Applied Psychology, Vol. 84 (2), 271-76.

Parasuraman, A., Valerie A. Zeithaml, and Leonard L. Berry (1985), “A Conceptual Model of

Service Quality and Its Implications for Future Research,” Journal of Marketing, 49(4)

41-50.

Prelec, Drazen and Duncan Simester (2001), “Always Leave Home without It: A further

Investigation of the Credit-Card Effect on willingness to Pay,” Marketing Letters,

12(10), 5-12.

Reber, Rolf and Norbert Schwarz (1999), “Effects of Perceptual Fluency on Judgments of

Truth,” Consciousness and Cognition, 8, 338-342.

Reber, Rolf, Piotr Winkielman and Norbert Schwarz (1998), “Effects of Perceptual Fluency on

Affective Judgments,” Psychological Science, 9(1), 45-48.

Reber, Rolf, Pascal Wurtz and Thomas D. Zimmermann (2004), “Exploring “Fringe”

Consciousness: The Subjective Experience of Perceptual Fluency and its Objective

Bases,” Consciousness and Cognition, 13, 364-382.

Reber, Rolf, Norbert, Schwarz and Piotr Winkielman (2004), “Processing Fluency and Aesthetic

Pleasure: Is Beauty in the Perceiver’s Processing Experience?” Personality and Social

Psychology Review, 8(4), 364-382.

Rentfrow, Peter J. and Samuel D. Gosling (2003),“The do re mi’s of everyday life: Examining

the structures and personality correlates of music preferences,” Journal of Personality

and Social Psychology 84, 1236-1256.

Rentfrow, Peter J. and Samuel D. Gosling (2006), “Message in a ballad – The role of music

preferences in interpersonal perception,” Psychological Science, 17(3), 236-242.

Reisberg, Daniel (1992), Auditory Imagery. ed. Hillsdale NJ: Lawrence Erlbaum Associates..

Riel, Cees B.M. van and John M.T. Balmer (1997), “European Journal of Marketing,” 31(5/6),

340-355.

Rigg, Melvin G (1940), “Speed as a determiner of musical mood,” Journal of Experimental

Psychology, 27, 566-571.

Sawyer, Alan G. (1975), “Demand Artifacts in Laboratory Experiments in Consumer Research,”

Journal of Consumer Research, 1, 20-30.

Schubert, Emery (2004), “Modeling perceived emotion with continuous musical features,”

Music Perception, 21, 561-585.

Singer, Judith D. (1998), “Using SAS PROC MIXED to Fit Multilevel Models, Hierarchical

Models, and Individual Growth Models,” Journal of Educational and Behavioral

Statistics, Vol. 24(4) 323-355.

Singh, Surendra N. and Gilbert A. Churchill, Jr. (1986), “Using the Theory of Signal Detection

to Improve Ad Recognition Testing,” Journal of Marketing Research, Vol. XXIII, 327-

36.

Singh, Surendra N. Michael l. Rothschild, and Gilbert A. Churchill Jr. (1988), “Recognition

Versus Recall as Measures of Television Commercial Forgetting,” Journal of Marketing

Research, Vol. XXV, 72-80.

Shapiro, Stewart (1999), “When an Ad’s Influence Is beyond Our Conscious Control: Perceptual

and Conceptual Fluency Effects Caused by Incidental Ad Exposure,” The Journal of

Consumer Research, 26 (June), 16-36.

Snyder, Adam (1993), “Branding: Coming Up for More Air,” Brandweek, 34 (December 6) 24-

28.

Stahl, Gerald (1964), “The Marketing Strategy of Planned Visual Communications,” Journal of

Marketing, 28(1), 7-11.

Sweeney, Jillian C. and, Carol Brandon (2006), “Brand Personality: Exploring the Potential to

Move from Factor Analytical to Circumplex Models,” Psychology and Marketing, 23(8),

639-663.

Taylor, John R. An Introduction to Error Analysis. 2nd edition. Sausolito, California: University

Science Books, 1997. pp 166-8.

Toth, Jeffrey P (1996), “Conceptual automaticity in recognition memory: Levels-of-processing

effects on familiarity,” Canadian Journal of Experimental Psychology, 50(1), 123-138.

(accessed on April 28, 2008, [available at Research Library database (Document

ID: 9665762)]

Trope, Yaacov and Nira Liberman (2003), “Temporal Construal,” Psychological Review, 110(3),

403-421.

Tulving, Endel and Daniel L. Schacter (1990), “Priming and Human Memory Systems,” Science,

New Series, Vol. 247, No. 4940. (Jan), pp. 301-306.

Turley, Lou W. and Ronald E. Milliman (2000), “Atmospheric effects on shopping behavior: A

review of the experimental evidence,” Journal of Business Research, 49, 193-211.

Tversky, Amos and Daniel Kahneman (1974), “Judgment Under Uncertainty: Heuristics and

Biases,” Science, Volume 185(4157), 1124-1131.

Tybout, Alice M., Brian Sternthal, Prashant Malaviya, Giorgios A. Bakamitsos and Se-Bum Park

(2005), “Content Versus Retrieval Ease in Judgments,” Journal of Consumer Research,

Vol. (32) 76-85.

Voss, Kevin E., Eric R. Spangenberg and Bianca Grohman (2003), “Measuring the Hedonic and

Utilitarian dimensions of Consumer Attitude,” Journal of Marketing Research, Vol. XL

(August), 31-320.

Wanke, Michaela, Herbert Bless and Barbara Biller (1996), “Subjective Experience versus

Content of Information in the Construction of Attitude Judgments,” Personality and

Social Psychology Bulletin, Vol. 22 (11), 1105-1113.

Wallace, Wanda T. (1991), “Jingles in advertising: Can they improve recall?” in Advances in

Consumer Research, Vol. 18, R.H. Holman and M.R. Solomon eds. Provo, UT:

Association for Consumer Research, 239-242.

Walther, Klaus (2006), “Lufthansa’s World of Sound: Corporate Sound increases brand

recognition,” www.Lufthansa.com (accessed on May 31, 2008), [available at

http://konzern.lufthansa.com/en/html/presse/pressemeldungen/?c=nachrichten/app/show/

en/2006/08/596/HOMands=0]

Weber, Robert J. and Brown Suellen (1986), “Musical Imagery,” Music Perception 13, 411-426.

Whittlesea, Bruce W.A. (1993), “Illusions of Familiarity,” Journal of Experimental Psychology,

19(6), 1235-1253.

Winkielman, Piotr and John T. Cacioppo (2001), “Mind at Ease Puts a Smile on the Face:

Psychological Evidence that Processing Facilitation Elicits Positive Affect,” Journal of

Personality and Social Psychology, 81(6), 989-1000.

Winkielman, Piotr, Norbert Schwarz, Tedra A. Fazendeiro and Rolf Reber (2003), “The Hedonic

Marking of Processing Fluency: Implications for Evaluative Judgment,” in The

Psychology of Evaluation: Affective Processes in Cognition and Emotion, J. Musch and

K.C. Klauer eds. Mahwah, NJ: Lawrence Erlbaum.

Zhu, Rui and Joan Meyers-Levy (2005), “Distinguishing between the meanings of music: When

background music affects product perceptions,” Journal of Marketing Research, 42

(August), 333-345.

Appendix A: SAS Code for Generating Synthetic Distribution by Case Resampling

************************************* *BOOTSTRAPPING WITH REPLACING*; ************************************; /*data Fluency; input id gender MI Index1 trueflu truerec truelike falseflu falserec falselike x1orth x2orth cenmi; datalines;*/

* or you ca n import from excel using; PROC IMPORT out=fluency REPLACE DATAFILE='C:\pkvk\Dissertation\Experiment set up\Data Analysis all studies\study 2 revised 4-16\restruc416.xls';*put your path here; SHEET="restruc416";*put the tab name here; GETNAMES=YES; run;

%macro randomsel(inputdata=,n=,r=,no=); ***************************************************************************** **** PURPOSE - macro randomly selects a certain number of observations from the input dataset with replacement, where the input dataset has repeated observations for one person (for example three obs for person 1, three obs for person 2 etc) INPUT DATA - inputdata is the original dataset which should have a variable called id that identifies each person and a variable called index1 that identifies the observation number per person - n is the total number of persons of the original dataset (in your case 42) - r is the number of persons you randomly want to select each time - no is the number of observations for one person OUTPUT - dataset fluencyran will contain the r randomly selected persons with all their lines - since this is with replacement one person can be drawn twice - to deal with this, we index the persons by a new variable called j that is unique for each draw (the number of observations in this final dataset is equal to the number of observations per person * the number of persons randomly selected) ***************************************************************************** ******; data random; do j=1 to andn; ran=ranuni(0);output; end; run; data random; set random; id=int(ran*(andn-0))+1;run; data random; set random; if _N_<=andr; run; data randomselected;set random; do i=1 to andno; index1=i; output;end; run; data randomselected;set randomselected; drop i ran; run; proc sort data=randomselected; by id index1; run; proc sort data=andINPUTDATA; by id index1; run;

96 data fluencyran; merge randomselected (in=in1) andinputdata (in=in2); by id index1; if in1; run; %mend;

%macro runmodels(inputdata=); ODS LISTING EXCLUDE ALL; ***************************************************************************** **** PURPOSE - macro runs the three models (three proc mixed) on the dataset called inputdata OUTPUT - final dataset estimates will contain one line with all the 10 observations of interest ***************************************************************************** ******;

*first model; proc mixed data=andinputdata noclprint covtest; class j; model like = X1 X2 /solution; random intercept /sub=j; ODS output solutionf=estimates1;*this outputs the required table in dataset called estimates1; run; * the rest of the code manipulates dataset estimates1 to obtain only the required info in the desired format (keep only the level of the estimates and put them on one line); data estimates1; set estimates1; keep effect estimate; run; proc transpose data=estimates1 out=estimates1t; id effect; run; data estimates1t; set estimates1t; drop _NAME_; rename intercept=interceptm1 x1=x1m1 x2=x2m1;*rename variables so you can identify what model they come from - m1 in the end refers to model 1; run;

*second model - it repeats the basic steps from the first model; proc mixed data=andinputdata noclprint covtest; class j; model trueflu = X1 X2/solution; random intercept /sub=j; ODS output solutionf=estimates2; run; data estimates2; set estimates2; keep effect estimate; run; proc transpose data=estimates2 out=estimates2t; id effect; run; data estimates2t; set estimates2t; drop _NAME_;

rename intercept=interceptm2 x1=b1 x2=b2; run;

*third model - it repeats the basic steps from the first model; proc mixed data=andinputdata noclprint covtest; class j; model like = trueflu X1 X2/solution; random intercept /sub=j; ODS output solutionf=estimates3; run; data estimates3; set estimates3; keep effect estimate; run; proc transpose data=estimates3 out=estimates3t; id effect; run; data estimates3t; set estimates3t; drop _NAME_; rename intercept=interceptm3 trueflu=b3 x1=x1m3 x2=x2m3; run;

*put all the estimates together in one row; data estimates; merge estimates1t estimates2t estimates3t; run; *this is the final dataset which will contain a line with

all estimates from one iteration; ODS LISTING; %mend;

%macro bootstrap(h=); ***************************************************************************** **** PURPOSE - macro repeats the procedure for h iterations (this is how many observations your final distribution will have) INTERMEDIARY STEPS - the do loop identifies the iterations - each iteration does the following: (1) randomly selects a number of persons from your dataset (2) runs the three proc mixed on the randomly selected sample (3) outputs one line with all estimates and adds it to a final dataset called allestimates OUTPUT - dataset allestimates will contain the final distribution of the estimates ***************************************************************************** ******; data allestimates; set _NULL_; %do i=1 %to andh; %randomsel(inputdata=fluency,n=88,r=88,no=3);*change r if you want to select more or less than 30 people at a time;

%runmodels(inputdata=fluencyran); data estimates; set estimates; iteration=andi;run; data allestimates; set allestimates estimates; %end; %mend;

%bootstrap (h=1000);run; *this does it for 5 iterations - change n if you want more; *this creates an excel file with all your estimates, but make sure you change the p ath fo r where you want it saved; proc export data=allestimates replace outfile='C:\Documents and Settings\krishnvu\Desktop\st2false.xls'; run;