Advertising time expansion, compression, and cognitive processing influences on consumer acceptance of message and brand

Available online at www.sciencedirect.com

h 62 (2009) 420–431
Journal of Business Researc
Advertising time expansion, compression, and cognitive processinginfluences on consumer acceptance of message and brand

Carol M. Megehee ⁎

Nicholls State University, United States

Received 1 August 2007; received in revised form 1 November 2007; accepted 1 January 2008

Abstract

This article examines the nature of consumer process involvement and cognitive processing of advertising content as mediating variablesbetween commercial message executions (e.g., broadcast time compression and expansion and using broadcast versus print media) on attitude andbehavioral intentions. The article proposes a framework that builds on the prior work of Krugman, Wright, and MacInnis and colleagues; theframework includes hypotheses of an advertising execution and processing involvement interaction effect on cognitive processing of commercialmessages and a substantial direct effect of cognitive processing on attitude and behavioral intention. The article includes details of an experimenttesting hypotheses in the framework. The findings provide strong support of the hypotheses. Implications for advertising strategy include adoptinga conservative view on the use of time compression in advertising commercials and nurturing low consumer processing involvement of commercialmessages.© 2008 Elsevier Inc. All rights reserved.

Keywords: Advertising; Interaction; Consumer process involvement; Expansion; Compression

Advertising in some contexts (e.g., television or radiomedia) may work “just by changing perceptions [cognitiveprocessing] toward the product in the course of merely shift-ing the relative salience of attitudes, especially when the pur-chaser is not particularly involved in the message” (Krugman,1965, p. 349). While not relying explicitly on Krugman'sproposal in developing his own models, Wright (1973, 1980)concludes that (1) consumer cognitive processes mediatethe acceptance of advertising messages and (2) the use ofspontaneous free-response recording of thought processesappears to be an:

… extremely promising method for studying communica-tion effects. Such measures offer important advantages overresearcher-imposed measures. The information contained insuch protocols is extremely rich compared to sterile, fre-

⁎ Corresponding author. Nicholls State University, College of BusinessAdministration, Thibodaux, Louisiana 70310 United States.

E-mail address: [email protected].

0148-2963/$ - see front matter © 2008 Elsevier Inc. All rights reserved.doi:10.1016/j.jbusres.2008.01.019

quently uninvolving measures requiring nothing more than aquick checkmark on the part of the subject (Wright 1973,p. 61).

Friestad and Wright (1994) propose that in addition totemporal contingencies, cultural and individual differences arelikely to influence people's motivation to process and use per-suasion knowledge. Consumers are likely to vary substantiallyin their involvement in processing an advertising message; suchvariation is one example of temporal contingencies likely toaffect the direction and strength of ongoing cognitive activitywhen hearing, watching, or reading a commercial message—ageneral proposal that follows from Krugman's more specificproposal that television advertising is a low involvement mes-sage receiving context.

The present article builds on prior proposals by Krugman(1965) and the research by Wright (1973), Friestad and Wright(1994), and related literature (MacInnis et al., 1991; MacLachlan,1982; MacLachlan and Siegel, 1980; Montigny, 2007; Schlingeret al., 1983) to theoretically and empirically examine how con-sumer message involvement may mediate media and temporal

mailto:[email protected]

http://dx.doi.org/10.1016/j.jbusres.2008.01.019

421C.M. Megehee / Journal of Business Research 62 (2009) 420–431

advertising influences on the creation and use of cognitiveprocessing information by consumers and possible attitudeoutcome influences of such a process. The present report alsocontributes by proposing and testing how media and temporaladvertising manipulations affect consumers' cognitive responsesand how such responses influence consumers' advertisingacceptance and use from the perspective of an importantindividual difference (i.e., consumer process involvement withthe advertising message). The theoretical propositions andempirical evidence in this report offer a rich contingency per-spective of the unique contribution of message involvementinfluences on how advertising dynamics affect consumer cog-nitive responses and use of persuasion knowledge.

This report expands and empirically tests some of MacInniset al.'s (1991) propositions. MacInnis et al. (1991) contribute aframework proposing that variations in advertising strategies andcues affect consumer processing-related activities (e.g. counter-arguing and generating product support arguments) and thatsuch activities mediate the impact of strategies and cues oncommunication outcomes—attitude toward the message, pro-duct, spokesperson, intention to purchase, and recommending theproduct to friends and family members.

Despite the plethora of research on (1) the relationshipbetween executional cues and communication outcomes and(2) the impact of brand information processing on commu-nication outcomes, few investigators have studied the me-diational role of MOA [consumer motivation, opportunity,and ability] on the executional/brand-processing relationship.Investigating this mediational role is critical, however, as (1)MOA in the typical [advertising] exposure setting is oftenlow, (2) executional cues are controllable aspects of ad designthat can enhance MOA, and (3) enhancing (MOA) in ads canproduce enduring brand attitudes and memories. (MacInniset al., 1, pp. 45–46)

The present study is the first report in the literature on bothmedia (print versus audio) and temporal (time compression versusexpansion) affects on cognitive processes—consumers' inter-pretations that likely mediate the influences on communicationoutcomes. Wright's 1973 observations support the value of suchadditional research on message modality influences on mediatingcognitive influences. In his laboratory study, “Counterargumentsproved to be a significantly stronger mediator of acceptanceamong subjects receiving the Audio mode message than amongthose receiving the Print version” (Z=2.10, pb .04, Wright, 1973,p. 57). The following 1973 conclusion remains valid into the 21stcentury, “The influence of message modality on the process ofattitude modification has not received much attention amongbasic or applied researchers. Reasons for this apathy are notreadily apparent since the mode of information transmissionrepresents one of the most basic dimensions of any communica-tion setting” (Wright, 1973, p. 61).

Research on time-compressed versus normal speed influ-ences on communication outcomes produce conflicting find-ings. LaBarbera and MacLachlan's (1979, p. 30) conclude fromthe findings of their laboratory experiment that “…radio ad-vertisers might achieve heightened impact, and require less time

for their messages, if they use electronic speech compression.”They emphasize, “There is no ‘Donald Duck’ effect as the speechis speeded up. In fact, as long as the speech is not speeded up bymore than 50% the listener will be unaware that there has beenan electronic alteration of the original recording” (LaBarbera andMacLachlan, 1979, p. 30).

Additional research results for audio and television commer-cials conflict with the LaBarbera andMacLachlan's findings andconclusions (see Hausknecht and Moore, 1985; Megehee et al.,2003; Moore et al., 1986; Murphy et al., 1986; Schlinger et al.,1983; Riter et al., 1983; Vann et al., 1987).). For example,Schlinger et al.'s (1983, p. 79) laboratory findings lead them toconclude “… time compression had only small effects on cog-nitive processing and postviewing attitudes. It appears that timecompression can result in somewhat fewer ideas being playedback in response to open-ended questions, inhibit both posi-tive and negative attitudes toward the advertised brand, anddepress positive emotional involvement with the execution.” InSchlinger et al.'s study time compression had no impact onconsumer buying intention.

The findings in the study in this article extend Schlinger et al.'s(1983) findings—time compression is likely to have a smallnegative influence on cognitive processing among consumershighly active in processing the advertising message but timecompression is likely to have a large negative influence on cog-nitive processing among consumers not very active in processingthe advertising message in comparison to consumers receivingadvertising messages at normal speeds. An update of Krugman's(1965) insights is telling—message involvement levels mediatemodal and temporal influences on consumer cognitive processing(i.e., learning without involvement is a valid interpretation ofadvertising's influence for both broadcast and print modes and forboth compression and expansion temporal contexts); Wright'sinsights also receive strong support—cognitive processes mediatethe influence of modal and involvement influences on messageacceptance (i.e., attitude measures).

1. Hypotheses

Fig. 1 represents the following hypotheses.

H1. Process involvement mediates modal (print and broadcast)influences on consumer cognitive processes, more specifically,(a) consumers relatively low versus high in processing theadvertising message counter-argue less with the advertisingmessage, derogate the source less, and generate more supportarguments; (b) this effect occurs to a greater extent for broadcastversus print ads because broadcast is a less involving mediumthan print medium. Krugman's (1965, p. 355) view offers arationale for H1b: “With high involvement [e.g., print versusbroadcast] one would look for the classic, more dramatic, andmore familiar conflict of ideas at the level of conscious opinionand attitude that precedes changes in overt behavior.”

H2. (a) The main effects of time compression and time ex-pansion on cognitive processing are not substantial because theinfluence of temporal variations depends on consumer processinvolvement. This hypothesis is an extension from Schlinger

Fig. 1. Advertising time compression, expansion, involvement, and cognitive processing influences on consumer outcome variables.

422 C.M. Megehee / Journal of Business Research 62 (2009) 420–431

et al.'s (1983) conclusion from their laboratory experiment,“Perhaps the most important conclusion to be drawn from thepresent data is that the impact of time compression on viewerresponses was not very great.” (b) The impact of timecompression is likely to be negative for consumers highlyinvolved and positive for consumers relatively uninvolved withthe advertising message. (c) When either medium generateshigh versus low involvement, net cognitive processing activityby consumers is negative; rationale: high versus low involve-ment generates substantial counter-arguing. “Net cognitiveprocessing activity” refers to Wright's (1973, Table 2, p. 58)Model B (∑i SAi−∑iCAi−∑iSDi), with SAi=support argu-ment i, CAi=counter argument i, and SDi=source derogation i.

H3. (a) Net cognitive processing activity has a positive influenceon all attitude outcomes such as attitude toward the message,speaker, and product, as well as purchase, use, and intention torecommend the product to friends. (b) This main effect of netcognitive processing activity on attitude outcomes is a strongerinfluence than process involvement influence on attitude out-comes; rationale: Krugman (1965) proposes that process involve-ment affects consumers perceptions of advertising messages andboth Krugman (1965) and Wright (1973) imply that perceptions(i.e., cognitions) are the principal influence on attitude outcomevariables. The expectation prior to running the experiment wasthat the findings would support this two-step process involvementinfluence on attitude outcome variables.

Note that Fig. 1 shows that all outcome variables relate toeach other positively (double headed arrows in the right side ofFig. 1). Prior research (Wright, 1973; Montigny, 2007) con-sistently reports such a pattern of relationships.

2. Methods

2.1. Informants

Two-hundred forty (240) students enrolled in two sectionsof principles of marketing at a large southwestern universityvolunteered for the study and received extra course credit for theirparticipation. The respondents were evenly divided by gender(49.6% male, 50.4% female), were predominantly businessmajors (63.9%), averaged 21.1 years of age (with 88% between19 and 22), and were mostly juniors (72%). Students were anappropriate population from which to draw for this study becausethey were expected to be the target market for the product pres-ented in the message.

2.2. Procedure

An experiment was conducted to test whether speech rate,tempo, and/or mode of delivery affect audience responses.Respondents were exposed to one of six message delivery treat-ments of the same script—the control, or “normal” speech; one ofthe four experimental voice treatments: pause-compressed, pause-


expanded, time-compressed, and time-expanded speech; and theread-to-oneself (not aloud)-only treatment—and then asked tocomplete a pencil and paper questionnaire. Table 1 shows thenumber of respondents assigned to each treatment condition.Responses measured in the experiment were the numbers, types,and proportions of cognitive and affective responses elicited;process involvement; attitudes toward the message itself, thespeaker presenting the message (for speech stimuli), andthe product in the message; behavioral intentions of purchasingand/or recommending the product to others, and the intendedfrequency of general and specific use of the product.

The experimental factors in this study were the speech rate asmeasured in words per minute (wpm), the tempo of messagedelivery as determined by method of speech rate change, andthe mode of the delivery of the message. Speech rates includedthe “normal” or base rate (achieved by the speaker talking in ananimated and persuasive manner, and used as the control in thisexperiment), and fast and slow speech rates achieved throughcompression and expansion of the base rate voice. The tempowas determined by the method of speech rate change (pausesampling and time sampling), and the two modes of deliverywere a printed text of the message (the read-only condition) andan orally presented/spoken delivery of the message (the listen-only condition). The control, or normal speech, was 154 secondslong and spoken at an average speech rate of 192 wpm. Fastspeech included pause-compressed and time-compressedspeech of equal duration (131 seconds, or approximately 85%of the 154 seconds base case). Expanded speech includedpause-expanded and time-expanded speech of equal duration(177 seconds, or approximately 115% of the base case).

The product promoted in the message in this study, the“SmartCard,” was a form of debit and identification card. Threereasons for using this product in the message stimulus in thisresearch were: (1) prior knowledge of specific productarguments presented in the message was expected at the timeof data collection (1994) to be low so that results would not beconfounded by prior experience, expectations, or attitudetoward an existing product or brand; (2) understanding of,identification with, and facility with the product were expectedto be high from prior real or vicarious experiences (e.g., use ofcredit cards, IDs, library copying cards, and ATM cards); and(3) demand for the product was expected to be high for asubstantial portion of the respondent group by offering services

Table 1Number of respondents per treatment condition

Treatment Speech Sampling Method Total Number of respondents

Conditions: Time Pause None

Speech:Compression 30 30 60Expansion 30 30 60Control 60 60

Total speech 180Read 60 60Total 240

that respondents are already using, but in a more convenientform. The script for the message (see Appendix A) contained491 words based on aural rather than visual perception todetermine wpm (e.g., contractions and acronyms were countedas one word; compound words, hyphenated words, and othersymbols or words pronounced as two words were counted astwo words).

The normal (control) speech rate was operationally definedas the measured average word rate (wpm or total words per totalduration) of the actual reading or recitation of the message by aprofessional male communicator within the context of one-way,voice-only, marketing communication. In this research, duration(time in seconds) was also used to describe speech rate. Thenormal voice was analyzed for pauses to determine the extent towhich it could be compressed. The maximum amount ofpausing that could be removed without distorting the words inthe message (in this case, 15%) was the amount of compressionand expansion that would be undertaken to create the fourremaining voice treatment conditions. The number of words,491, was the same for all message treatments, and decreases(increases) in duration resulted in increases (decreases) in wpm.To achieve professional sound quality, the script was recordedin a soundproof room on digital audio tape (DAT) by a malegraduate communications student with experience as a profes-sional announcer and advertiser who was trying to speak in away that was natural for the persuasive context.

The four altered experimental voice treatments where createdusing two procedures: pause sampling and time sampling. Pausesampling is the process of removing or adding to the pauseportions of the speech spectrum between phrases and sentenceswithout altering the sound of words, syllables, or other seg-mental speech sounds. Time sampling is the process of sys-tematically selecting periodically spaced units of the speechsignal from the digitized speech spectrum for removal or ex-pansion. The speech spectrum is a graphic representation of thedistribution of digitized sound (or speech signal energy) inwhich intensity is plotted as a function of frequency over time.Silence, the absence of speech signal energy, occurs within aswell as between speech sounds.

In pause sampling, only inter-utterance silences were ma-nipulated such that only the pauses between phrases andsentences were lengthened or shortened. With pause sampling,the speech rate (wpm) and duration (time) of the message arechanged without altering the articulation rate (speed or durationof actual speech sounds).

Pause compression (pause removal) was performed on asystematic basis and was the initial step in manipulating thenormal voice to create additional voice treatments. Pause com-pression was limited to the extent that relevant inter-utterancepauses were present in the speech. Because of this limitationand the desire to create comparable speech rates among thevoice treatments, the maximum amount of pause compressionobtainable was used as the benchmark for creating the othervoice treatments. When 100% of the pause intervals that couldbe removed without altering the sound of vocalized speechwere removed, the threshold for compression (also used forexpansion) was found to be approximately 15%. The duration

Table 2Temporal speech characteristics of voice treatments

Treatment: Duration(seconds)

WPM Percentagespeech

ArticulationWPM

Time compression 131 a 226 a 75% 300Pause compression 131 a 226 a 86% 262 c

Normal 154 192 73% 262 c

Pause expansion 177 b 163 b 63% 262 c

Time expansion 177 b 163 b 75% 222

Note. All figures are rounded.a Equal duration, equal wpm for compression (fast rate).b Equal duration, equal wpm for expansion (slow rate).c Equal articulation wpm for pause sampling and normal speech.


of the pause-compressed speech was 131 seconds, or about 85%of the normal duration of 154 seconds; the duration of theexpanded speech, 177 seconds, was about 115% of the normalduration. The actual locations of pauses removed in compressionwere used to identify points in the speech spectrum for pauseexpansion.

The system used for the initial pause compression and lateranalysis of the five voice treatments was a 386 isx-PC upgradedto 8 MB of RAM with a math coprocessor and Kay ElemetricsCorporation's Model 4300 input–output box. The softwarerequirements were Kay Elemetrics' PC-based Computer SpeechLaboratory (CSL), Version 4, and DOS 6.22 (minimum ver-sion). The input settings used, given duration requirements andsystem limitations, were a sampling rate of 12.5 kHz with amaximum length of 200 seconds and an input level of 3 (on theModel 4300′s input–output box). DAT was used as the inputmedium in order to retain the maximum amount of the originalspeech information for analysis. PC-based software was used toload the entire 154 seconds message onto the computer, so thatone-millionth of a second duration intervals could be examinedthrough simultaneous viewing of and listening to markedsections of the speech spectrum, the relevant pause intervalscould be deleted, and the precise location and duration of eachdeletion could be marked for future reference. This system wasalso later used to analyze speech-to-silence ratios (speech as apercent of total duration) for all five voice treatment conditions.In the speech-to-silence ratio analysis all silences, includingthose within utterances, were counted.

A professional recording studio in Austin, Texas, wasemployed to create tapes representing equal durations of pauseand time compression, complementary expansions by way ofboth methods, and a base case tape of equivalent quality (fre-quency and amplitude). The five tapes were created using aMacintosh platform with CD quality and at least 5 MB of RAMper minute of input. The software, ProTools, was capable ofsampling the two-and-one-half minute tape at 48,000 sectionsper second with room to spare for expansion. Recording andcopying equipment was digitally based, providing high res-olution for manipulation and high quality output for taping. Thedigitally produced tapes were copied to analogue tapes so thatthe researcher could use available tape-playing equipment toprovide realism in the experimental setting.

In the cases of speech rate change through pause sampling,the amount of speech sound as a percentage of total durationchanged, but the normal articulation rate was maintained. Forthe time sampling cases, the change in duration without anaccompanying change in the speech-to-silence ratio altered thearticulation wpm. That is, while time sampling did not pro-portionately change the natural cadence or tempo of speech, thevocalized portion of the message was accelerated with com-pression and slowed down with expansion. Table 2 shows thetemporal effects of compressing and expanding the normalvoice by pause- and time-sampling.

A final message treatment, read-only, was created tocompare purely verbal auditory stimuli with a purely words-only visual stimulus (i.e., the read-only respondents were onlyexposed to the words, or informational content, of the message).

In the read-only condition, respondents were asked to read thescript without listening to a tape. Respondents were instructedto read the script only once and then turn it over so that theycould not refer to it again.

2.3. Mediating and dependent attitude outcome measures

The dependent variables were attitude toward the message(Am), attitude toward the speaker (As), and attitude toward theproduct (Ap), purchase intention (PI), use intention (UI), andintention to recommend the product to a friend (RP). Variablesmediating the relationship between message treatment anddependent variables include process involvement (SI) and cog-nitive processes coded from open-ended responses. Productcounter arguments (CA), product support arguments (SA), andsource derogation (SD) comprise the cognitive processingvariables. Cognitive processes (CA, SA, and SD), PI, and RPwere single-item measures.

Each of the multi-item measures was factor analyzed usingprincipal factor analysis without rotation to reduce it to theitems that measured the construct of interest. Observations withmissing values were omitted from the factor computations.Items with factor loadings and/or item-total correlations lessthan .50 were eliminated from the scale. The scales for thesevariables, along with item-total correlations and coefficientalpha for multi-item scales, are shown in Appendix B. Thepsychometric properties of all scales were tested twice—oncefor data from a pilot study before the main study and for the datain the main study. The results in Appendix B are based on datafrom the main study.

2.3.1. Process involvementAt the end of the questionnaire, before the demographic

questions and debriefing statements, respondents were askedto respond to items related to their involvement in the study.The full scale was developed from Zaichkowsky (1990) andMcQuarrie and Munson (1991) and was reduced to the fouritems shown in Appendix B.

2.3.2. Open-ended responsesCognitive and affective responses were recorded to measure

thoughts and feelings about the product, the message, and thespeaker. The wording, placement, and coding of an open-endedquestion were adapted from prior research investigating the


processing of advertising-based feelings and cognitions (Andrewsand Shimp, 1990; Homer and Yoon, 1992; Vann et al., 1987). Thewording of the directionswas slightlymodified to include feelingsas well as thoughts. The question was placed at the beginning ofthe questionnaire so that respondents' cognitive-affective re-sponses would not be influenced by subsequent questions.

The coding of responses also followed precedence for codingof cognitive response items. Three independent judges, twopsychology graduate students and one marketing (consumerbehavior) graduate student, were hired to code the open-endedcognitive-affective response question. For each subject, judgescounted the total number of responses and then divided re-sponses into cognitive and affective categories and (positive,negative, or neutral) product-, message-, and speaker-relatedcategories. The judges were given copies of a coding form forguidance and were instructed to add new coding categories ifany were found. The judges were asked to code each subject'sresponses independently and then to come to an agreement onthe final tally (or score) to be assigned to each subject in eachresponse category.

The three judges returned their independently completedcoding sheets without reaching a consensus on scores, so all threesets of judges' codes were included in the data base and codesaveraged formeasurement scores. Average scoreswere used; first,because no consensus score was available, and second, becausethe objective in scoring cognitive-affective responses was to ex-amine the relative influence of product arguments, message, andspeaker, or alternatively, positive, negative, and neutral thoughtsand feelings. If judges disagreed as to the coding of a particularresponse, it was probably due to the fact that respondents' re-sponses were directed at multiple targets or had multiple mea-nings. For example, a comment about the message might havebeen attributed to the way product attributes were presented, theway the speaker presented the message, the words in the message,or some combination of all three factors.

All paired correlations (rsN .71) between judge's responseswere statistically significant at the pb .001 level except for“positive other responses” for judges 1 and 2 (pb .003) andNegative other responses for judges 1 and 3 (pb .017). However,in order to reduce the amount of error introduced by aggregatingresponse scores, response categories with any correlations lessthan .50 (positive and neutral message-related, neutral speaker-related, and all other responses) were not treated as individualdependent variables. In addition to being of little researchinterest, these categories appear to have been used as residual ordefault categories. That is, if the tallies across categories didnot add up to the total number of responses, then the “extra”response might have been placed into one of these categories.Because judge 3 did not count any neutral speaker-related re-sponses, there was no inter-judge correlation in that category ofresponse for this individual.

The three measures for cognitive processes came from thecoding of these open-ended responses. The number of negativeproduct-related responses became the single-item measurefor product counter argument (CA); the number of positiveproduct-related responses became the single-item measure forproduct support argument (SA); and the number of negative

execution-related responses, or the sum of negative message-and speaker-related responses, became the single-item measurefor source derogation (SD).

2.3.3. Attitude scalesAttitude was operationally defined as a person's predisposi-

tion (either positive or negative) to consistently think, feel, or acttowards a person, object, or concept (Sutherland, 1989;Wolman,1973). Attitude scales measured the strength and direction of aperson's attitude toward the product (the SmartCard), the mes-sage, and the speaker. That is, attitude scales were constructed tomeasure “how a person values something, i.e., how far he is foror against it” (Sutherland, 1989, p. 38). These scales were de-veloped specifically for the product, the message, and thespeaker in this research, although many of the individual itemswere adapted from published scales (Batra and Ray, 1986;Miniard et al., 1990; Moore et al., 1986; Olney et al., 1991;Scherer, 1986).

2.3.4. Behavioral intention scalesIn social psychology, a behavioral intention is “a predisposi-

tion to act in particular ways towards an object” (Sutherland,1989, p. 49). Behavior suggests action, and intent, “a presentresolve to perform some future action” (Wolman, 1973, p. 42). Inthis research, a distinction was made between attitudes andbehavioral intentions. Whereas an attitude is a predisposition tothink or feel a certain way about an object, a behavioral intentionis a predisposition to act a certain way towards an object.

Research examining the effect of speech rate on actual be-havior is very limited. Wheeless' (1970) dissertation and follow-up papers (Wheeless, 1970, 1971a,b) are an exception. Wheelesstested the effects of male and female speakers reading thesame persuasivemessage, the purpose of which was to sell a how-to-study booklet to students for $1.50. Wheeless found timecompression, which had the effect of removing up to 50% ofspeech spectrum information as well as doubling the rate ofspeech (e.g., twice the average word rate, half the duration), didnot adversely affect persuasibility of the message as measured byincidence of purchase. In fact, purchase frequency was slightlyhigher for the 40% compression rate group than the base speechrate group (33 versus 31%; not significant).

More recent studies have incorporated measures of beha-vioral intentions rather than actual behavior (Peterson et al.,1994; Schlinger et al., 1983; Vann et al., 1987). Speech rates arefound not to influence behavioral intentions in these studies.

Behavioral intentions measured in this research wereintended frequency of use of the product (in general, as a cashcard, for identification purposes, and to retrieve information),the likelihood of recommending the product to a friend, and thelikelihood of purchasing the product.

2.3.5. Debriefing questionsThe respondents were asked a series of debriefing questions

about their perceptions as to the objectives of the study and theirperceptions of the characteristics of the message (e.g., coded 1–7 for “was the message too short versus too long” and “was thespeaker slow or fast”). None of the respondents' perceived


objectives made reference to modal or temporal influences onmessage processing.

3. Findings

The analyses of the data include both univariant and multi-variant methods. The findings in this report focus on representa-tive samples of these analyses. Findings from applying alternativestatistical methods (e.g., multiple regression analyses) supportand extend the same conclusions that this section covers. Theseadditional findings are available from the author.

3.1. Processing involvement and cognitive processing

Fig. 2 summarizes a group level analysis (Bass et al., 1968)for process involvement. Rather than a median split, theanalysis includes three (equal number) groups of respondents bylow, moderate, and high process involvement levels to enablethe identification of segments of respondents varying substan-tially by process involvement (cf. McClelland, 1998). H1a

receives strong support: respondents relatively low versus highin processing the advertising message counter-argue less withthe advertising message, derogate the source less, and generate

Fig. 2. Audio time compression and expansion of pieces of sound, pauses, control, andfor informants by three process involvement levels.

more support arguments as the net average scores indicate inFig. 2 (∑i SAi−∑iCAi−∑iSDi). The findings are consistentacross all six treatment groups: average SA scores are higher forall six of lowest process involvement segments in comparisonwith the highest process involvement segments. The averageCA and SD scores are higher across all six process involvementsegments in comparison to the lowest process involvementsegments. This pattern of 18 for 18 comparisons is statisticallysignificant by a sign test (pb .01). The overall averages are forall respondents for the three process involvement levels.

The six treatment groups' average process involvementscores do not differ significantly. The findings do not supportthe hypothesis that the audio medium generates lower levels ofprocess involvement than the print medium.

H1b receives limited support. Shifting from low to highprocessing involvement, the decrease in net cognitive proces-sing is greater for the pause compression treatment group thanfor the print treatment group (e.g., 1–9− (−1.6)=3.4 versus2.1−(−0.1)=2.2). The decrease in net cognitive processing isalso significantly greater for the pause expansion treatment group(3.0) versus the print treatment group. However, a consistentpattern does not occur in comparing the five audio treatmentgroups versus the print treatment group. See Fig. 2 for details.

read only treatment influences on net cognitive processing (ΣSA−ΣCA−ΣSD)

Table 3Correlations of net cognitive processing with attitude and behavioral intensiondependent variables

Attitude/behavioralDependent variable

Correlation with netCognitive processing

Correlation with netcognitive processing

Controlling for processinginvolvement

Attitude toward message 47 42Attitude toward speaker 43 40Attitude toward product 47 46Recommend product 52 47Purchase intention 46 41Use intention 30 29

Note. Decimals omitted. All r values significant (pb .000); n 236 to 239, exceptfor Attitude toward Speaker, n=175.


3.2. Temporal and modal influences on processing involvement

The findings consistently support all parts of H2. (a) Themain effects of time compression and time expansion on cog-nitive processing are not substantial. (b) The impact of bothtime compression and expansion are negative for consumershighly involved and positive for consumers relatively unin-volved with the advertising message. (c) When both print andaudio media generate high versus low involvement, net cog-nitive processing activity by consumers is negative. The av-erage net cognitive processing scores (and respective standarderrors) for print equal 2.1 (.50), 1.9 (.71), and − .08 (.70); and forall 60 print treatment respondents the average equals 1.1 (.50).The average net cognitive processing scores (M) and respectivestandard error of the mean (se) for the five broadcast treatment

Fig. 3. Relationships (r) among processing involvement, ne

group equal 1.2 (.48), − .21 (.31), and − .80 (.28); and for all 179broadcast respondents the average equals − .14 (.20).

3.3. Processing involvement, cognitive processing, and attitudeoutcomes

Both parts of H3 receive strong support. (a) Net cognitiveprocessing activity has a positive influence on all attitudeoutcomes such as attitude toward the message, speaker, product,as well as purchase, use, and intention to recommend the pro-duct to friends. (b) This main effect of net cognitive processingactivity on attitude outcomes is a stronger influence than pro-cess involvement's negative influence on attitude outcomes.

Table 3 shows the correlations and partial correlations con-trolling for processing involvement of net cognitive processingand the dependent attitude and behavioral intention outcomevariables. Note that the partial correlations show a very mar-ginal decrease in correlations—processing involvement leveldoes not affect the main effect of net cognitive processing withthe dependent variables.

Fig. 3 shows that the absolute sizes of the correlations arehigher for net cognitive processing versus the processing in-volvement with the attitude and behavioral intention outcomevariables. Analyses for each of the six treatment groups and betacomparisons in multiple regression analyses consistently con-firm the same profile of relative absolute correlations—higherfor net cognitive processing in comparison to processing in-volvement and the net attitude/intention outcome variables.

Fig. 3 is an application of the “quick clustering” method(Kamen, 1970) of showing significant relationships relating toH3. Fig. 3 includes the simple method of showing correlations

t cognitive processing, and attitude outcome variables.


rather than findings from multiple regression analyses (MRA);Fig. 3 includes correlations rather than MRA findings since thedata were all collected in one setting. The use of MRAwould beappropriate for predicting attitude and behavioral intentionoutcome measures taken in a few days or weeks after the ad-ministering the advertising execution treatments—the study didnot include collecting data from informants from two differenttime periods.

However, MRAs were run and the results support the con-clusion that both net cognitive processing and processing in-volvement contribute substantially in predicting each of theattitude and behavioral intention dependent variables—with thebeta for net cognitive processing substantially larger thanthe beta for processing involvement for all treatment groups andthe total informants. The correlations among all attitude andbehavioral intention dependent variables are highly significant(pb .001).

3.4. Debriefing findings

The debriefing of informants included asking each if she orhe believed the message was “too short or too long”, “too slowor too fast”—both scored using 1 to 7 scales. In comparing theaverage responses for the three pause and time conditiontreatments, the findings are non-significant for too short/longand significant for too slow/fast; see Table 4. The findings on asignificant impact on compression versus control differ fromresults in previous studies. MacLachlan and Siegel (1980, p. 53)report, “A speedup of 25% which will shorten a [TV] com-mercial to 24 s[econds] is a subtle change that is not noticeablein either audio or video.” In a within-subject experiment,Schlinger et al. (1983, p. 81) used both open-ended and closed-ended questions to find out if respondents noticed differencesbetween TV time-compressed (by 20%) and regular timecommercials; for the closed-end question, the researchers asked“if the test commercial seemed faster, the same, or slower thatthe ‘other commercials you've seen.” Schlinger et al. (1983)report few and inconsistent comments and no top-of-mindawareness of time compression—and no consistent substantialdifferences in awareness of time compression from the closed-end responses.

Table 4Informants' perceptions that the message was “Too short/Too long” “Too slow/Too fast” (1–7 Scales): Means (M), Standard Errors (se) and Number ofInformants (n)

Treatment group Too short/Too long Too slow/Too fast

M se n M se n

Control 4.7 .15 60 4.8 .15 60Time compression 4.8 .23 30 6.2 .16 30Pause compression 5.0 .22 30 6.2 .17 30Pause expansion 4.6 .16 30 4.3 .17 30Time expansion 4.6 .27 30 4.6 .16 30Read 4.2 .16 59 NA NA NATotal 4.6 .08 239 5.2 .06 180

F = 2.01, d.f. = 5233pb .08, η2= .04

F=26.69, d.f. = 4175pb .00, η2= .38

Note. M=Mean, Se=Standard error, η2, NA=Not applicable.

Perceiving the commercial as too fast versus ordinary or tooslow does relate to net cognitive processing (∑i SAi−∑iCAi −∑iSDi); means (and standard errors) for perceptions of too fast(M=−53, se= .23, n=115), ordinary (M=+.67, se= .42, n=54),and too slow (M=− .20, se= .76, n=10), differ significantly(pb .03, η2= .04). Given that net cognitive processing relatespositively with attitude and behavior intention outcome variables,avoiding commercial messages generating the perception ofbeing too fast might be worthwhile.

4. Discussion

The findings in this article extend Krugman's (1965) pro-position that process involvement affects the occurrence andnature of cognitive processing by consumers in exposure con-texts of advertising messages. Process involvement mediatesthe influence of time and pause compression influences oncognitive processing of advertising messages within a broadcastmedium. However, the findings do not support the view that asubstantial process involvement main effect occurs for time andpause compression or expansion versus a normal speech rates;the findings support and extend Schlinger et al.'s (1983) generalconclusion that temporal influences are “not very great” onattitude and intention outcome variables.

Both pause compression and pause expansion may workwell in increasing net positive cognitive processing of broadcastadvertising in low but not high processing involvement contexts.Also, the findings support Wright's (1973) conclusions thatcognitive processing responses mediate attitude and intentionoutcomes of advertising messages. The findings in the presentarticle support Wright's proposition that Model B (∑i SAi−∑iCAi−∑iSDi) positively affects attitude and intention out-come variables; the level of processing involvement does nothave a substantial influence on these cognitive processing andattitude/intention outcomes.

The findings support the rather complex contingent model-ing propositions that MacInnis et al. (1991) offer and Friestadand Wright (1994) suggest. MacInnis et al. (1991, p. 46–47)stress that investigating “mediational roles” of consumercognitive processing variables is critical because a consumer's“(1) MOA [motivation, opportunity, and ability] in the typicalexposure setting is often low, (2) executional cues are con-trollable aspects of ad design, and (3) enhancing MOA in adscan produce enduring brand attitudes and memories.” Evidencehere supports all three points and serves to advance a contingenttheoretical view of how consumer cognitive processing me-diates variations in advertising execution cues in influencingattitudes and behavioral intentions.

As MacInnis et al. (1991, p. 48) stress, considerable workis necessary in developing MOA measures. Expanding the callfor necessary work needs to include developing multiplecomparisons of advertising executional cues such as examiningtemporally expanded as well as compressed messages, bothaudio and visual presentations, and storytelling (slice-of-life)versus information-oriented (lecture) formats (cf. Wells, 1989).Also, creative field experiments in unobtrusive contexts arenecessary to extend and confirm the findings from laboratory


experiments. The present study includes limitations relatingto choice of informants, the use of one brand in one productcategory, procedures used for message compression and ex-pansion, and collecting all data in a single setting. One aim of thepresent study is to stimulate additional research that addressesthese and other limitations inherent in findings from a singleexperiment.

4.1. Implications for advertising executive strategy

Neither time nor pause compression nor expansion is likely tohave a direct impact on consumer attitudes or behavioral in-tention. Schlinger et al.'s (1983) conclusion is likely to be a moreaccurate general assessment of the direct impact of time com-pression of broadcast commercials on behavioral intention thanMacLachlan's (1982) andMacLachlan and LaBarbera (1978)—time compression is unlikely to have a direct impact on con-sumer buying intentions.

Pause compression and pause expansion versus time com-pression and time expansion advertising commercials are likelyto increase net cognitive processing among consumers in lowprocess involvement contexts. Possibly, consumers may un-consciously or consciously perceive pause expansions as anordinary rate of speech—neither slow or fast in comparison toalternative speech rate adjustments. Pause expansion may be aneffective executional strategy in low process involvement con-texts for increasing net cognitive processing without stimulatingthe perception of “too fast” delivery. Pause expansion may bea useful tool for gaining increasing attention and interest—adramatic moment in storytelling; such a dramatic tool mayprovide a relevant explanation for the positive net cognitiveprocessing direct effect and consequent indirect positive impacton attitude and behavioral intention of pause expansions. In aworld of speed and noise, the pregnant pause can be highlydistinctive.

The findings in this article support and extend Krugman's(1965) propositions on “learning without involvement” forbroadcast commercial messages. Creative broadcast advertisingwork on presenting low process involvement contexts that nurturehigh net cognitive processing (analogous to Krugman's “chan-ging perceptions of the product”) may be more effective thancreative work generating high process involvement contexts.Examples of creating low process involvement commercialsmight include the Australian Tourism Commission advertise-ments starring Paul Hogan (see: http://en.wikipedia.org/wiki/Shrimp_on_the_barbie) and the series of professional and ama-teur Corona beer laid-back beach commercials (for one of dozensamateur productions of this low-processing context commercialsee http://www.youtube.com/watch?v=zv1IDSvpl_Y).

Wright's 1973 propositions on how cognitive processes me-diate the acceptance of advertising is a useful theoretical andpractical bridge from Krugman's (1965) propositions to thelater work of MacInnis et al. (1991) and Friestad and Wright(1994). Advertising research and theory on commercial mes-sage impact need to explicitly link executional cues to com-munication effectiveness via their impact on levels of processinvolvement and consumer cognitive process responses.

Appendix A. The script

You're a student.You know what it's like to stand in long lines.To be refused for an improper ID.To have to collect quarters for laundry, copying, and soft

drink machines.Or to go back through the whole bureaucratic process to prove

you passed step A to get to step D when steps B and C didn't tellyou that you needed step A in the first place—even though youalready had it (but didn't have “proper documentation”).

If you're a student at the University of Texas, you don't haveto put up with this anymore!

For just $15 you can own the SmartCard.What is the SmartCard? It's an all in one card that can be used

as an ID or access card, a cash or debit card, or as a storehouse ofinformation.

Because of its ability to store vast amounts of information,the SmartCard replaces the need to keep up with the cum-bersome paperwork. It replaces your student ID, and the longlines associated with the validation of IDs, purchase of parkingpermits or tickets to entertainment and sporting events, and theregistration, payment for, and access to courses, clubs, and otherstudent activities.

You can even keep track of your credits at the Coop.The SmartCard also makes your entire academic transcript,

fee-bill, and tuition status available—only to you through yourpersonal identification number or PIN—if and when you needit.

And if that isn't enough, you can also use the SmartCard inATMs or as a “storehouse of change.” You can charge up thecard for copying, telephone calls, washing, vending machinepurchases or other uses that require change and accept theSmartCard.

No exact change required with the SmartCard.One SmartCard copies your class notes, washes your clothes,

gets you into football games, dorms and computer labs—andstores all the information about these transactions. However, byusing your PIN, you limit “charging” or information access byothers.

The only information that others can read from your cardwithout your PIN is the information coded for access to thoseparticular events or activities.

And if you've paid your tuition and fee-bill and it's in the UTsystem but you haven't stood in a long line at the gym to get thesticker, try to check out a book at the library.

Don't hand over your old card—you don't need it any more.Hand them your SmartCard.When they run the scanner, not only will your ID be

accepted, but your SmartCard will be validated for the “sticker”that says you've paid your bills.

From now on, that validation is part of your SmartCardrecord—as well as your student record on the UT informationsystem.

The SmartCard remembers so that you don't have to.All you've got to remember is to carry this multipurpose

card—and remember your PIN.

http://en.wikipedia.org/wiki/Shrimp_on_the_barbie

http://en.wikipedia.org/wiki/Shrimp_on_the_barbie

http://www.youtube.com/watch?v=zv1IDSvpl_Y


Appendix B. Scale items

Process involvement (Coded 1–7: 1=Highly Negative, 7=The study was _______:

Item-total correlation

Boring/interesting
0.86 Unexciting/exciting 0.86 Mundane/fascinating 0.85 Not Fun/fun 0.82
(4 items, α=0.93)

Cognitive processes (Coded from open-ended responses)a. Product Counter Argument (CA)=number of negativeproduct-related responsesb. Product Support Argument (SA)=number of positiveproduct-related responsesc. Source Derogation (SD)=number of negative execution-related responses=sum of negative message- and speaker-related responses

Purchase Intention (PI) (Coded 1–7: 1=Not Very Likely,7=Very Likely)

How likely is it that you would purchase a SmartCard?

Use Intention (UI) (Coded 1–5: 1=Never, 2=Once per Monthor Less, 3=2–3 Times per Month, 4=Once per Week, 5=Daily)

If you had a SmartCard, how often do you think you woulduse it ________?


In general
0.81 For identification purposes 0.74 As a cash card 0.79 To retrieve information 0.68
(4 items, α=0.86)

Recommend Product to Friend (RP) (Coded 1–7: 1=NotVery Likely, 7=Very Likely)

How likely is it that you would recommend the SmartCard toa friend?

Attitude toward the Message (Am) (Coded 1–7: 1=HighlyNegative, 7=Highly Positive)

Was the message ______?


Dull/interesting
0.79 Unpleasant/pleasant 0.80 Not believable/believable 0.61 Not truthful/truthful 0.59 Inappropriate/appropriate 0.74 Bad/good 0.83 Unattractive/attractive 0.82 Not important/important 0.75 Boring/exciting 0.73 Not coherent/coherent 0.54 Irrelevant/relevant 0.72
Appendix B (continued )


Not informative/informative
0.61 Not useful/useful 0.79 Not persuasive/persuasive 0.78 Unfavorable/favorable 0.82 Weak/strong 0.77 Negative/positive 0.80 Ineffective/effective 0.83
(18 items, α=0.96)

Attitude toward the Speaker (As) (Coded 1–7: 1=HighlyNegative, 7=Highly Positive)

Was the Speaker ______?


Inappropriate/appropriate
0.81 Not believable/believable 0.84 Not trustworthy/trustworthy 0.82 Not attractive/attractive 0.82 Not knowledgeable/knowledgeable 0.69 Not credible/credible 0.82 Not irritating/irritating 0.78 Not effective/effective 0.89 Negative/positive 0.59 Unpleasant/pleasant 0.86 Untruthful/truthful 0.76 Dull/interesting 0.78 Not likeable/likeable 0.92 Not confident/confident 0.62 Bad/good 0.88 Not favorable/favorable 0.87 Not natural/natural 0.76 Not persuasive/persuasive 0.81 Boring/exciting 0.72 Not friendly/friendly 0.66 Not relaxed/relaxed 0.60 Cold/warm 0.56
(22 items, α=0.97)

Attitude toward the Product (Ap) (Coded 1–7: 1=HighlyNegative, 7=Highly Positive)

Was the SmartCard ______?


Dull/interesting
0.76 Unpleasant/pleasant 0.73 Not beneficial/beneficial 0.84 Worthless/valuable 0.83 Not convenient/convenient 0.68 Bad/good 0.81 Useless/useful 0.85 Negative/positive 0.79 Not appealing/appealing 0.87 Awful/nice 0.71 Inefficient/efficient 0.68 Boring/exciting 0.77 Unimportant/important 0.78 Irrelevant/relevant 0.79 Unattractive/attractive 0.74 Ineffective/effective 0.70 Inappropriate/appropriate 0.77
(17 items, α=0.96)


References

Andrews J Craig, Terrence A Shimp. Effects of involvement, argument strength,and source characteristics on central and peripheral processing of ad-vertising. Psychol Mark 1990;7:195–214 (Fall).

Bass Frank M, Douglas J Tigert, Lonsdale RT. Market segmentation: groupversus individual behavior. J Mark Res 1968;4:264–70 (August).

Batra Rajeev, Michael L Ray. Affective responses mediating acceptance ofadvertising. J Consum Res 1986;13:234–49 (September).

Friestad Marian, Peter Wright. The persuasion knowledge model: how peoplecope with persuasion attempts. J Consum Res 1994;21:1–31 (June).

Hausknecht Douglas R, Moore Danny L. The effects of time compressedadvertising on brand attitude judgments. Adv Consum Res 1985;13:105–10.

Homer Pamela M, Yoon Sun-Gil. Message framing and the interrelationshipsamong ad-based feelings, affect, and cognition. J Advert 1992;21:19–33(March).

Kamen Joseph M. Quick clustering. J Mark Res 1970;7:199–204 (May).Krugman Herbert E. The impact of television advertising: learning without

involvement. Public Opin Q 1965;29:349–56 (Fall).LaBarbera Pricilla, MacLachlan James. Time-compressed speech in radio

advertising. J Mark 1979;43:30–6 (January).MacInnis Deborah J, Moorman Christine, Jaworski Bernard J. Enhancing and

measuring consumers' motivation, opportunity, and ability to process brandinformation from ads. J Mark 1991;55:32–53 (October).

MacLachlan James. Listener perception of time-compressed spokespersons.J Advert Res 1982;22:47–51 (April–May).

MacLachlan James, LaBarbera Pricilla. Time-compressed TV commercials.J Advert Res 1978;18:11–5 [August].

MacLachlan James, Siegel Michael H. Reducing the costs of TV commercialsby use of time compression. J Mark Res 1980;17:52–7 (February).

McClelland DC. Identifying competencies with behavioral-event interviews.Psychol. Sci 1998;9:331–9 (September).

McQuarrie Edward F, Munson Michael J. A revised product involvementinventory: improved usability and validity. In: John Sherry F, SternthalBrian, editors. Advances in Consumer Research, vol. 19. Provo, UT: TheAssociation for Consumer Research; 1991.

Megehee Carol M, Dobie Kay, Grant Jim. Time vs. pause manipulation incommunications directed at the young adult population: does it matter.J Advert Res 2003;43:281–92 September.

Montigny Michelle. Quantifying the isolated and synergistic effects of exposurefrequency for TV, print, and Internet advertising. J Advert Res 2007;47(3):215–21.

Miniard Paul W, Bhatla Sunil, Rose Randall L. On the formation and rela-tionship of ad and brand attitudes: an experimental and causal analysis.J Mark Res 1990;27:290–303 (August).

Moore Danny L, Hausknecht Douglas, Thamodaran Kanchana. Time compres-sion, response opportunity, and persuasion. J Consum Res 1986;13:85–99(June).

Murphy John H, Wilcox Gary B, Hardy Andrew P. Time-compression:additional evidence regarding its effects on audience response to televisioncommercials. In: Larkin Ernest F, editor. Proceedings of the 1986Convention of the American Academy of Advertising; 1986. p. R77–81.

Olney Thomas J, Holbrook Morris B, Batra Rajeev. Consumer responses toadvertising: the effects of ad content, emotions, and attitude toward the ad onviewing time. J Consum Res 1991;17:440–53 [March].

Peterson Robert A, Cannito Michael P, Brown Steven P. An exploratoryinvestigation of voice characteristics and sales effectiveness. J Pers SellSales Manage 1995;15:1–15 (Winter).

Riter Charles B, Balducci Phillip J, McCollum Donald. Time compression: newevidence. J Advert Res 1983;22:39–43 (December/January).

Scherer Klaus R. Vocal affect expression: a review and a model for futureresearch. Psychol Bull 1986;99:143–65 (March).

Schlinger Mary Jane Rawlins, Alwitt Linda F, McCarthy Kathleen E, GreenLeila. Effects of time compression on attitudes and information processing.J Mark 1983;47:79–85 (Winter).

Sutherland Stuart. MacMillan dictionary of psychology. London: MacMillanReference Books; 1989.

Vann John W, Rogers Robert D, Penrod John P. The cognitive effects of time-compressed advertising. J Advert 1987;16(2):10–9.

Wells WD. Lectures versus dramas. In: Cafferata P, Tybout AM, editors. Cog-nitive and affective responses to advertising. Lexington, MA: LexingtonBooks; 1989. p. 13–20.

Wheeless, Lawrence R. (1970), An experimental investigation of the persuasiveeffects of time-compressed speech, dissertation, Detroit, MI: WayneState University.

Wheeless Lawrence R. Effects of compression loss on persuasion. SpeechMonogr 1971a;38:327–30 (November).

Wheeless Lawrence R. Some effects of time-compressed speech on persuasion.J broadcast 1971b;15:415–20 (Fall).

Wolman Benjamin B. Dictionary of behavioral science. New York: VanNostrand Reinhold Company; 1973.

Wright Peter. The cognitive processes mediating acceptance of advertising.J Mark Res 1973;10:53–62 (February).

Wright Peter. Message-evoked thoughts: persuasion research using thoughtverbalizations. J Consum Res 1980;7:151–75 (September).

Zaichkowsky, Judith Lynne (1990), The personal involvement inventory: re-duction and application to advertising, Unpublishedworking paper, Burnaby,British Columbia: Simon Fraser University.

Documents

Advertising time expansion, compression, and cognitive processing influences on consumer acceptance of message and brand