You are here

  • Home
  • Archive
  • Volume 23, Issue 3
  • The natural history of antismoking advertising recall: the influence of broadcasting parameters, emotional intensity and executional features

Article Text

Article menu

Download PDFPDF

Research paper
The natural history of antismoking advertising recall: the influence of broadcasting parameters, emotional intensity and executional features
  1. Sally M Dunlop1,2,
  2. Donna Perez2,
  3. Trish Cotter3
  1. 1Sydney School of Public Health, University of Sydney, Sydney, New South Wales, Australia
  2. 2Cancer Prevention Division, Cancer Institute NSW, Sydney, New South Wales, Australia
  3. 3Victorian Comprehensive Cancer Centre, Melbourne, Victoria, Australia
  1. Correspondence to Dr Sally Dunlop, Sydney School of Public Health, University of Sydney; Research Associate, Prevention Division, Cancer Institute NSW, Level 9, 8 Central Ave, Australian Technology Park, Eveleigh NSW 2015, Australia; sally.dunlop{at}cancerinstitute.org.au

Abstract

Background The necessary first steps for televised media campaign effects are population exposure and recall. To maximise the impact of campaign funding, it is critical to identify modifiable factors that increase the efficiency of an advertisement reaching the target audience and of their recalling that advertisement.

Methods Data come from a serial cross-sectional telephone survey with weekly interviews of adult smokers and recent quitters from the state of New South Wales, Australia, collected between April 2005 and December 2010 (total n=13 301). Survey data were merged with commercial TV ratings data (Gross Rating Points (GRPs)) to estimate individuals’ exposure to antismoking campaigns.

Results Multivariable logistic regression analyses indicated that GRPs and broadcasting recency were positively associated with advertisement recall, such that advertisements broadcast more at higher levels or in more recent weeks were more likely to be recalled. Advertisements were more likely to be recalled in their launch phase than in following periods. Controlling for broadcasting parameters, advertisements higher in emotional intensity were more likely to be recalled than those low in emotion; and emotionally intense advertisements required fewer GRPs to achieve high levels of recall than lower emotion advertisements. There was some evidence for a diminishing effect of increased GRPs on recall.

Conclusions In order to achieve sufficient levels of population recall of antismoking campaigns, advertisements need to be broadcast at adequate levels in relatively frequent cycles. Advertisements with highly emotional content may offer the most efficient means by which to increase population recall.

  • Media
  • Cessation
  • Social marketing

https://doi.org/10.1136/tobaccocontrol-2011-050256

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Recent reviews conclude that there is strong evidence that televised antismoking campaigns can contribute to reductions in adult smoking.1 ,2 Higher levels of exposure to public health media campaigns designed to encourage smoking cessation have been associated with increased quitting activity 3–5 and decreased smoking prevalence.6 While there is growing interest in identifying the types of messages4 ,7 ,8 and broadcasting parameters5 ,9 that might optimise campaign effects, there is a limited amount of population-level evidence showing how variations in advertisement features and media strategies affect campaign impact. In advertising research, the ‘hierarchy of effects’ model postulates that advertising works by first being perceived and retained;10 and health communication scholars recognise that exposure to and recall of a campaign are necessary first steps in the process of campaign effects.11 ,12 In order to maximise the impact of scarce public health funding, it is critical to identify modifiable factors that increase the efficiency of an advertisement reaching the campaign's target audience and of their recalling that advertisement. The primary objective of the current study was to explore modifiable factors related to recall of antismoking advertising, focusing on broadcasting parameters and advertisement characteristics.

    Determining the impact of broadcasting parameters on advertisement recall is a difficult task due to the need for population-level data relating to real-world campaigns. Previous research has demonstrated that recall of antismoking advertisements is related to the reach and frequency of an advertisement in a population (measured using commercial TV ratings data providing Gross Rating Points or GRPs), showing that more frequently broadcast advertisements are more likely to be recalled.11 ,13–16 A decay in recall has also been observed when advertisements cease to be broadcast.13 Additionally, it is posited that advertisements can ‘wear out’ over time and with repeat airings,17 with some evidence showing that newly created antismoking advertisements may drive more change than familiar advertisements.18

    The current study utilises a continuous tracking methodology to collect weekly population-level data on adult smokers’ and recent quitters’ recall of antismoking advertisements broadcast in the Australian state of New South Wales (NSW) over a period of 5 years. This methodology extends previous research by tracking recall of specific advertisements when they are on air as well as in the weeks and months following by assessing both ‘prompted’ and ‘unprompted’ recall, and by including a large and varied set of advertisements. We consider the impact of three broadcasting parameters—frequency (measured by GRPs), recency and novelty—on recall, expecting that advertisements achieving higher GRPs will be more likely to be recalled (frequency; H1), that the more time that has passed since the last broadcast the less likely an advertisement will be recalled (recency; H2) and that advertisements are most likely to be recalled when in their launch phase (novelty; H3).

    A growing body of literature suggests that specific advertisement characteristics can have implications for how an advertisement is processed by the viewer. Antismoking advertisements can be categorised along a number of dimensions, including message theme or topic, emotional intensity, and executional characteristics of the advertisement such as the use of a testimonial or narrative structure or the presence of graphic imagery.4 ,7 ,8 ,13 ,19–22 In both experimental and population-based survey studies, antismoking advertisements that use graphic imagery or personal stories to depict the negative consequences of smoking have been associated with greater recall and impact than other types of advertisements.4 ,8 ,9 ,13 ,20–24 This advantage is typically theorised to be due to the emotional content of such advertisements. Research conducted by Lang and colleagues suggests that television content that elicits negative emotion is more likely to be attended to and remembered than that without emotional content,25 and that discrete emotional responses increase the viewer's memory and the likelihood of their recalling the advertisement.26 One study to date has shown a difference in recall for highly emotive advertisements with different executional characteristics, demonstrating the highest recall rates for advertisements with testimonial formats, followed by advertisements with graphic imagery.7 The current study extends previous research on emotionality and other advertisement characteristics in two main ways. The first is methodological: the population-level dataset measures recall (both prompted and unprompted) of a large and varied set of advertisements as they are on air and in the months following, rather than in experimental exposures or one-off campaigns. Consistent with previous research, we expect that more emotionally arousing advertisements will be more likely to be recalled than less emotional advertisements (H4). Given the lack of evidence as to how executional characteristics are related to recall, we pose a research question as to whether emotionally-evocative advertisements with graphic imagery will be recalled more frequently than emotionally-evocative advertisements with a narrative format (RQ1). The second way we extend previous research is by exploring whether emotional intensity and executional characteristics moderate the impact of broadcasting parameters (RQ2).

    Method

    Survey data

    The Cancer Institute NSW's Tobacco Tracking Survey (CITTS) is a serial cross-sectional telephone survey monitoring smoking-related cognitions and behaviours, along with recall of and responses to antismoking mass media campaigns, in adult smokers and recent quitters (quit in the last 12 months) from NSW. Households are recruited using random digit dialling of landline telephone numbers and a random selection procedure is used to select participants (selecting the nth oldest eligible adult in the household). As a continuous tracking survey, 50 interviews per week are conducted across most weeks of the year; analyses were limited to respondents interviewed between April 2005 and December 2010 (n=13 301), in which an overall response rate of 40% (using the American Association for Public Opinion Research Response Rate #4) was achieved.27

    Measures

    Recall and recognition of antismoking advertisements

    A key feature of the CITTS is the ability to track recall of antismoking advertisements as they are currently on air and in the weeks following. Prompted or aided recall (henceforth labelled ‘recognition’) is often used as a self-reported measure of campaign exposure,7 ,9 ,11 ,15 while unprompted recall can also be used as a measure of advertisement ‘cut-through’ or the advertisement's ability to be readily recalled from memory.28

    All respondents were asked ‘Have you seen any television advertising recently about tobacco smoking?’ Respondents who answered ‘yes’ (n=9453, 71% of total sample) were asked to describe the advertisement, with descriptions recorded verbatim and matched to a list of television advertisements that appeared during the study period (unprompted recall). If the response did not match one of the listed advertisements (eg, an advertisement for NRT), it was excluded. Following this, interviewers probed for recall of any other antismoking advertisements, with responses again matched to the list of available advertisements (unprompted recall, n=7190, 54% of total sample).

    Recognition of specific advertisements currently or recently on air was then measured by describing advertisements and asking respondents whether they remembered seeing the advertisement on TV recently. The brief descriptions gave respondents enough information to recognise the advertisement, but not enough for them to falsely indicate awareness. On average, participants were asked about four different advertisements (range 2–7). The number of weeks that recognition was tracked for an advertisement after it had ceased being broadcast varied across the different advertisements (M=4, SD=6).

    Covariates

    Demographic items measuring age, gender, income and level of education were included in the survey. The income and education variables were combined into dummy variables indicating low, middle or high socioeconomic status (SES). Postcodes were used with the Socio-Economic Indices for Areas29 to indicate neighbourhood SES (quintiles 4–5=low SES, quintiles 1–3=moderate–high SES). A dummy variable was used to indicate media market (Metropolitan, Northern NSW or Southern NSW). For smokers, cigarette consumption was measured with a three-level categorical variable (<10, 11–20, >21 per day). Some of the advertisements were targeted at parents of young children (focusing on the potential impact of their smoking-related disease on their children). For this reason, parental status (whether there were any children younger than 17 years living in their household) was controlled for in all analyses. A linear time variable was included as a covariate to account for any secular trends in the data (numbered consecutively with 1 as the first survey week).

    Person-advertisement data structure

    Two person-advertisement-level datasets were created (Recall and Recognition), with the data structured so that each observation corresponded to an individual's responses to one advertisement. In the Recall dataset, each observation corresponded to an individual's recall score for a specific advertisement (1=recall, 0=no recall). Given that individuals could not recall advertisements that had not yet been broadcast, for each advertisement the dataset was limited to respondents interviewed after the date of the first broadcast. Inspection of the Recall dataset showed that recall for advertisements that had not been on air for more than 12 months was rare (3% of advertisements recalled). This fact, coupled with the fact that the recall measure referred to seeing advertisements recently, led to a restriction of the dataset to include advertisement-specific recall scores only for respondents interviewed within 12 months of the last broadcast. We note, however, that the pattern of results was the same using the full dataset. The resulting Recall dataset consisted of 106 733 person-advertisement observations corresponding to one respondent's recall of an advertisement broadcast within the previous 12 months. The Recognition dataset consisted of 38 026 person-advertisement observations corresponding to one respondent's recognition of a recently broadcast advertisement. In each dataset, because individuals reported on more than one advertisement, it was possible that an individual's responses might be correlated across advertisements. To adjust for possible inter-correlations, we adjusted all estimated SEs by using the individual as a cluster variable in all regression analyses. Sample characteristics of each dataset are shown in table 1.

    View this table:
    Table 1

    Sample characteristics of recall and recognition datasets

    Advertisement-level data

    A total of 32 cessation-focused antismoking advertisements were broadcast in NSW during the study period. The vast majority of antismoking messages in NSW are aimed at adult smokers and provide a strong health effects message. Some of these advertisements also present advice and information about cessation support services available to smokers. The advertisements were classified according to two dimensions: emotional intensity and executional characteristics. The advertisements were first rated by 16 independent adults on three items measuring emotional intensity:4 ,9 emotional, intense and powerful (each scale 1–10; α=0.91). The emotional intensity scale was an average of these three items (M=5.99, SD=2.15) and was used to classify advertisements into low and high emotion categories (low emotion=less than 5, high emotion=5–10). Following this, advertisements were then further classified by two trained coders according to two executional characteristics: the presence of graphic imagery or the use of a narrative format (see table 2). Discrepancies between coders were discussed until consensus was achieved. From this coding, it was apparent that none of the advertisements in the low emotion category had graphic imagery or narrative formats. Therefore, there were three categories of advertisements: low emotion, high emotion graphic imagery and high emotion narrative.

    View this table:
    Table 2

    Antismoking advertisements broadcast in New South Wales, 2005–2010

    We ascertained GRPs for each of the advertisements based on OzTAM (Australian Television Audience Measurement) for adults aged 18 years and older including both free-to-air and cable TV. GRPs are a product of the percentage of the target audience exposed to an advertisement (reach) and the average number of times a target audience member would be exposed (frequency). Hence, 200 GRPs might represent 100% of the target audience receiving the message an average of two times over a specified period or 50% reached four times. The advertisements differed in length from 15 to 60 s; we took these different lengths into account by calculating GRPs equivalent to a 30 s spot for each advertisement.30 Emerging evidence suggests that the effects of antismoking advertising occur in a relatively short time-frame of between 2 and 3 months.5 ,6 ,31 For this reason, and the fact that the recall and recognition measures in the CITTS referred to ‘recent’ advertising, we computed a cumulative 3-month GRPs variable as a measure of media weight. We chose not to include GRPs from the week of interview, given that respondents are interviewed throughout the week and may therefore be interviewed either before or after potential exposures for that week. To aid in interpretation, this variable was divided by 1200 so that a 1-unit increase represented an average of one additional potential exposure per week for 100% of the target audience. To test for diminishing returns at increasing levels of advertising exposure, we included a quadratic term (squared GRPs) in our analyses.

    A variable indicating the number of weeks since an advertisement had last been on air was used to indicate recency of broadcasting (coded 0 if the advertisement was presently on air). A dummy variable was also included that identified respondents interviewed during the launch phase of an advertisement's broadcasting life, the first 4 weeks that it was on air (1=launch phase, 0=any other time).

    Statistical analyses

    Univariate logistic regression was used to explore associations between the dependent variables and the advertisement-level variables and covariates. From the univariate analyses, advertisement-level variables and covariates associated with the dependent variable at p<0.25 were selected to enter into multivariable logistic regression models predicting recall and recognition.32 The independent variables in these models included: GRPs and GRPs-squared (testing H1), broadcasting recency (H2), the launch phase of the advertisement (H3) and advertisement type (H4 and RQ1).

    In order to ascertain if the effects of broadcasting parameters on recall were moderated by advertisement type (RQ2), the multivariate models were also run with interaction terms between advertisement type and these broadcasting parameters included. Interactions were entered one by one and retained only if significant. Finally, we visually represented the relationship between advertisement type and GRPs by plotting the predicted probabilities of each outcome by GRPs and advertisement type, holding all other variables constant.

    Since cigarette consumption was only available for the smoker sample, the analyses were run in both the smoker sample and the full sample (of smokers and ex-smokers). In each analysis, the same pattern of results emerged, so we present only the results from the full sample.

    All analyses were conducted using Stata V.11. Population weights were applied to adjust for a slight over-representation of female subjects, older respondents and regional residents compared with the NSW population,33 and clustering on a unique individual identifier was used in all models to adjust for possible inter-correlations among observations.

    Results

    Unprompted recall

    Of the advertisements that were recalled and matched to NSW antismoking advertisements, 17% were low emotion, 34% were high emotion narrative advertisements and 48% were high emotion graphic advertisements. The first multivariable model in table 3 shows that, consistent with H1, there was a significant association between recall and GRPs, such that, with each additional 1200 exposures over the previous 3 months (equating to approximately one exposure per week for 100% of individuals), there was an increase of 4.46 in the odds of a respondent recalling an advertisement. Both types of high emotion advertisements were more likely to be recalled than the low emotion advertisements (consistent with H4), with a comparison of the point-estimates indicating that this difference was greater for recall of the graphic advertisements than the narrative advertisements (RQ1). Confirming H2, recency of broadcasting was also a significant predictor, with recall decreasing as the final broadcast became more distant. Finally, consistent with H3, respondents interviewed in the launch period of an advertisement were more than twice as likely to recall that advertisement than those interviewed at other times in the following 12 months.

    View this table:
    Table 3

    Unprompted recall of advertisements broadcast in last 12 months

    From the model with the interaction terms, it was apparent that broadcasting parameters had differential impact according to advertisement type (RQ2). GRPs were more strongly associated with recall for the high emotion graphic advertisements than for the low emotion advertisements. However, the significant quadratic term in which the OR is less than 1 indicates a diminishing effect of GRPs on recall for the graphic advertisements, as illustrated in figure 1. The launch phase of a campaign was a stronger predictor of recall for the low emotion than for the high emotion narrative advertisements, and the recency of broadcast was a stronger predictor for the high emotion graphic advertisements than for the low emotion advertisements.

    Figure 1
    Figure 1

    Predicted probability of unprompted recall. GRPs, Gross Rating Points.

    Recognition

    In all, 42% of respondents who were queried about a low emotion advertisement recognised the advertisement, 64% of those queried about high emotion narrative advertisements indicated recognition and 54% of those who were asked about high emotion graphic advertisements recognised them. Table 4 shows that there was a positive association between GRPs and recognition (consistent with H1), though the significant quadratic term (OR<1) indicated a diminishing effect of GRPs at higher levels. Recognition was positively associated with the launch phase of an advertisement (consistent with H3), and there was a negative association between recognition and weeks since last broadcast (consistent with H2). Both types of high emotion advertisements were more likely to be recognised than the low emotion advertisements (consistent with H4), and a comparison of the ORs indicates that this difference was greater for the narrative advertisements than the graphic advertisements (RQ1). From the model which included the interaction terms, it was apparent that GRPs were more strongly associated with recognition for low emotion advertisements than for the high emotion advertisements (shown in figure 2). Further, the positive association between recognition and the launch phase of an advertisement was stronger for low emotion than high emotion advertisements. Similarly, the association between broadcast recency and recognition was stronger for the low emotion advertisements than for the high emotion advertisements.

    View this table:
    Table 4

    Prompted recognition of antismoking advertisements

    Figure 2
    Figure 2

    Predicted probability of recognition. GRPs, Gross Rating Points.

    Discussion

    Consistent with prior literature4 ,8 ,9 ,13 ,20−24 and our hypothesis, emotionally arousing advertisements were more likely to be recalled (using both unprompted and prompted measures) than those which were less emotional. Emotional messages might be more likely to promote higher-order cognitive processing34 or to increase feelings of personal relevance,35 leading the audience to generate their own persuasive messages in response to the advertisement and increasing their memory for the content. It might also be that they generate greater discussion after exposure,36 ,37 reinforcing memory for the advertisement. It is possible, however, that some unmeasured factor related to emotional intensity, such as argument strength,38 is the real basis for improved recall, and this should be explored in future research. Nonetheless, for practical applications, emotional intensity is relatively easy to assess and might therefore be used by campaign developers as a way to select advertisements likely to generate high levels of population recall.9

    Among high emotion advertisements, advertisement execution was related to recall, though the results differed according to type of recall. High emotion advertisements with graphic imagery were most likely to be ‘top-of-mind’ using the unprompted recall measure. There are a number of potential explanations for this. The first is that graphic images increase message processing and therefore recall. An alternative explanation is that graphic advertisements are more readily recalled because they contain content most easily linked to the category cue (of antismoking or tobacco advertisement). Additionally, a number of the graphic images used in the advertisements were also used as graphic health warnings on cigarette packs,39 ,40 and repeated exposure to these images on packs might reinforce memory for the advertisements.41 In contrast, and consistent with another study using a recognition measure,7 high emotion narrative advertisements were more frequently recognised than graphic imagery advertisements, even when controlling for potentially different levels of exposure. Narrative effects are posited to arise from processes of identification and transportation into the story, so that perceptions of personal vulnerability are heightened, leading to greater impact.35 ,42–,44 This type of processing might lead to a deeper memory trace for these types of advertisements and should be a subject for further research. Future research might also investigate the relationship between different types of recall (prompted or unprompted) and advertisement effects in order to determine if one type of recall is more closely associated with advertisement-related cognitive and behavioural changes.

    Broadcasting parameters were closely related to both unprompted recall and prompted recognition. Consistent with our hypotheses, and with previous research,13 ,16 the likelihood of recall increased with increasing GRPs. There was, however, some evidence of a slowing of the effects of increasing advertising exposure on both recall and recognition. For the recognition measure, this effect was apparent for all advertisement types. For the unprompted recall measure, the diminishing effects were only apparent for the high emotion graphic advertisements. Since the graphic advertisements were most likely to be recalled in this measure, a plausible explanation might be that the other advertisement types would also show diminishing returns once they reached a level of recall similar to that of the graphic advertisements. These diminishing effects might be interpreted as meaning that advertisements will cease to reach new viewers with increasing exposures past a particular GRP level; nonetheless, increased frequency of exposure might still reinforce the message effects on those that have seen it.

    There was also evidence that, controlling for variations in level of broadcasting, recall for an advertisement was the greatest during its launch phase, possibly indicating that the novelty of an advertisement attracts greater attention, or that accompanying media releases and unpaid media exposure can boost recall. Once advertisements were removed from air, recall decreased with time. In order to maintain high levels of population recall, campaign planners should endeavour to ensure adequate exposure to advertisements over relatively frequent broadcasting cycles, and to include new or novel advertisements in their mix of campaigns.

    In general, broadcasting parameters were more closely related to recall for the advertisements arousing lower levels of emotion. Consistent with a series of studies focusing on youth recall of antismoking advertising,9 ,13 more emotionally intense advertisements required fewer exposures in order to achieve comparable levels of recall to the less emotional advertisements. This has an implication for campaign planners: if emotionally intense advertisements generate higher levels of recall at lower levels of broadcast, they may be best suited for making efficient use of scarce tobacco control funding. When lower emotion advertisements are included in the ‘media mix’ of antismoking campaigns, they need to be broadcast at sufficient levels to ensure adequate population recall.

    The strengths of this study are the length of the study period and resulting variation in antismoking advertising exposure, measurement of both unprompted recall and prompted recognition, the range of advertisements included, and the ability to track recall and recognition while advertisements were currently or recently on air. Limitations are the use of landline-only telephone numbers and the relatively low response rate, possibly leading to some bias in sample composition. We note that both these sampling issues were consistent across the study period, limiting their influence on the observed pattern of results, and that the inclusion of the linear time variable in all analyses helps to account for any gradual differences over time. The rate of mobile-only households in Australia was recently estimated at 14%, limiting concerns about excluding these individuals from the sample.45 The response rate is similar to that of other recent population telephone surveys in Australia 46 and the USA,47 and the overall rates of quitting in this sample are similar to that of larger population surveys of NSW smokers,48 suggesting that the smoker sample is somewhat representative.

    This study demonstrates that both broadcasting parameters and advertising creative execution have significant impacts on advertising recall and recognition, with high emotion graphic and narrative advertisements being more likely to be recalled by smokers. Additionally, the results indicate that these types of advertisements are likely to require less GRPs (and hence less funds) to achieve higher levels of recall.

    What this paper adds

    • Previous research has shown that antismoking advertising can influence population-level smoking behaviours. Nonetheless, there is still a limited understanding of which types of advertisements are most likely to be successful at reaching and influencing smokers, and how broadcasting parameters can increase the likelihood of smokers’ recall of advertisements.

    • This study suggests that population recall of antismoking campaigns is related to the emotional intensity of the advertisements, with high emotion advertisements being more likely to be recalled than low emotion advertisements.

    • High emotion advertisements with graphic imagery were most likely to be ‘top-of-mind’ using an unprompted recall measure, whereas high emotion narrative advertisements were more frequently recognised than graphic imagery advertisements using a prompted recognition measure.

    • Recall was positively related to the frequency and recency of broadcasting, and also to the novelty of the campaign.

    • There was some evidence for a diminishing effect of increased Gross Rating Points on advertisement recall.

    • In general, advertisements which elicit high levels of emotion required fewer exposures to achieve similar levels of recall to low emotion advertisements, suggesting that they might present the most cost-efficient means of reaching and influencing smokers.

    References

    1. ↵
      National Cancer Institute. The role of the media in promoting and reducing tobacco use. Tobacco Control Monograph No. 19. USDHHS, NIH, NCI. NIH Pub. No. 07-6242. Bethesda, MD 2008.
    2. ↵
      1. Wakefield M,
      2. Loken B,
      3. Hornik R
      . Use of mass media campaigns to change health behaviour. Lancet 2010;376:1261–71.
      OpenUrlCrossRefPubMedWeb of Science
    3. ↵
      1. Hyland A,
      2. Wakefield M,
      3. Higbee C,
      4. et al
      . State-sponsored anti-tobacco television advertising and indicators of smoking cessation in adults: a cohort study. Health Educ Res 2006;21:296–302.
      OpenUrlAbstract/FREE Full Text
    4. ↵
      1. Durkin S,
      2. Biener L,
      3. Wakefield M
      . Effects of different types of antismoking ads on reducing disparities in smoking cessation among socioeconomic subgroups. Am J Public Health 2009;99:1–7.
      OpenUrlCrossRefWeb of Science
    5. ↵
      1. Wakefield M,
      2. Spittal MJ,
      3. Yong HH,
      4. et al
      . Effects of mass media campaign exposure intensity and durability on quit attempts in a population-based cohort study. Health Educ Res 2011;26:988–97.
      OpenUrlAbstract/FREE Full Text
    6. ↵
      1. Wakefield M,
      2. Durkin S,
      3. Spittal M,
      4. et al
      . Impact of tobacco control policies and mass media campaigns on monthly adult smoking prevalence. Am J Public Health 2008;98:1443–50.
      OpenUrlCrossRefPubMedWeb of Science
    7. ↵
      1. Niederdeppe J,
      2. Farrelly M,
      3. Nonnemaker J,
      4. et al
      . Socioeconomic variation in recall and perceived effectiveness of campaign advertisements to promote smoking cessation. Soc Sci Med 2011;72:773–80.
      OpenUrlCrossRefPubMed
    8. ↵
      1. Davis KC,
      2. Nonnemaker J,
      3. Farrelly MC,
      4. et al
      . Exploring differences in smokers’ perceptions of the effectiveness of cessation media messages. Tob Control 2010;20:26–33.
      OpenUrlPubMedWeb of Science
    9. ↵
      1. Biener L,
      2. Wakefield M,
      3. Shiner CM,
      4. et al
      . How broadcast volume and emotional content affect youth recall of anti-tobacco advertising. Am J Prev Med 2008;35:14–19.
      OpenUrlCrossRefPubMedWeb of Science
    10. ↵
      1. Lavidge R,
      2. Steiner G
      . A model for predictive measurements of advertising effectiveness. J Mark 1961;25:59–62.
      OpenUrlCrossRefWeb of Science
    11. ↵
      1. Southwell BG,
      2. Barmada CH,
      3. Hornik R,
      4. et al
      . Can we measure encoded exposure? Validation evidence from a national campaign. J Health Commun 2002;7:445–53.
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      1. Richardson A,
      2. Cullen J,
      3. Mowery P,
      4. et al
      . The path to quit: how awareness of a large-scale mass-media smoking cessation campaign promotes quit attempts. Nicotine Tob Res 2011;13:1098–105.
      OpenUrlAbstract/FREE Full Text
    13. ↵
      1. Biener L,
      2. Ji M,
      3. Gilpin EA,
      4. et al
      . The impact of emotional tone, message, and broadcast parameters in youth anti-smoking advertisements. J Health Commun 2004;9:259–74.
      OpenUrlCrossRefPubMedWeb of Science
      1. Emery S,
      2. Wakefield M,
      3. Terry-McElrath Y,
      4. et al
      . Televised state-sponsored anti-tobacco advertising and youth smoking beliefs and behavior in the United States 1999–2000. Arch Pediatr Adolecent Med 2005;159:639–45.
      OpenUrlCrossRef
    15. ↵
      1. Niederdeppe J
      . Assessing the validity of confirmed ad recall measures of public health communication campaign evaluation. J Health Commun 2005;10:635–50.
      OpenUrlCrossRefPubMedWeb of Science
    16. ↵
      1. Cowling DW,
      2. Modayil MV,
      3. Stevens C
      . Assessing the relationship between ad volume and awareness of a tobacco education media campaign. Tob Control 2010;19(Suppl 1):i37–42.
      OpenUrlAbstract/FREE Full Text
    17. ↵
      1. Tellis GJ
      . Effective advertising: understanding when, how, and why advertising works. Thousand Oaks, CA: Sage Publications, 2004.
    18. ↵
      1. Carroll T,
      2. Rock B
      . Generating Quitline calls during Australia's National Tobacco Campaign: effects of television advertisement execution and programme placement. Tob Control 2003;12(Suppl II0):ii40–4.
      OpenUrlAbstract/FREE Full Text
    19. ↵
      1. Farrelly M,
      2. Davis K,
      3. Nonnemaker J,
      4. et al
      . Promoting calls to a quitline: quantifying the influence of message theme, strong negative emotions and graphic images in television advertisements. Tob Control 2011;20:279–84.
      OpenUrlAbstract/FREE Full Text
    20. ↵
      1. Durkin S,
      2. Wakefield M,
      3. Spittal M
      . Which types of televised anti-tobacco campaigns prompt more quitline calls from disadvantaged groups? Health Educ Res 2011;26:998–1009.
      OpenUrlAbstract/FREE Full Text
      1. Biener L,
      2. McCallum-Keeler G,
      3. Nyman AL
      . Adults’ response to Massachusetts anti-tobacco television advertisements: impact of viewer and advertisement characteristics. Tob Control 2000;9:401–7.
      OpenUrlAbstract/FREE Full Text
    22. ↵
      1. Wakefield M,
      2. Durrant R,
      3. Terry-McElrath Y,
      4. et al
      . Appraisal of anti-smoking advertising by youth at risk for regular smoking: A comparative study in the United States, Australia, and Britain. Tob Control 2003;12(Suppl II):ii82–6.
      OpenUrlAbstract/FREE Full Text
      1. Terry-McElrath Y,
      2. Wakefield M,
      3. Ruel E,
      4. et al
      . The effects of antismoking advertisement executional characteristics on youth comprehension, appraisal, recall, and engagement. J Health Commun 2005;10:127–43.
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      1. Biener L
      . Anti-tobacco advertisements by Massachusetts and Philip Morris: what teenagers think. Tob Control 2002;11(Suppl II):ii43–6.
      OpenUrlAbstract/FREE Full Text
    25. ↵
      1. Lang A,
      2. Dhillon K,
      3. Dong Q
      . The effects of emotional arousal and violence on television viewers’ cognitive capacity and memory. J Broadcasting Electronic Media 1995;39:313–27.
      OpenUrlCrossRefWeb of Science
    26. ↵
      1. Lang A
      . The limited capacity model of mediated message processing. J Commun 2000;50:46–70.
      OpenUrlCrossRefWeb of Science
    27. ↵
      American Association for Public Opinion Research. Standard definitions: final dispositions of case codes and outcome rates for surveys. Lenexa, Kansas: AAPOR, 2008.
    28. ↵
      1. Donovan R,
      2. Boulter J,
      3. Borland R,
      4. et al
      . Continuous tracking of the Australian National Tobacco Campaign: advertising effects on recall, recognition, cognitions, and behaviour. Tob Control 2003;12(Suppl II):ii30–39.
      OpenUrlAbstract/FREE Full Text
    29. ↵
      Australian Bureau of Statistics. 2006 census of population and housing: socio-economic indexes for area's (SEIFA). Canberra, Australia: Commonwealth of Australia, 2008.
    30. ↵
      1. Broadbent S
      . Modelling with adstock. J Mark Res Soc 1984;26:295–312.
      OpenUrl
    31. ↵
      1. Sly DF,
      2. Arheart K,
      3. Dietz N,
      4. et al
      . Effect of ending an antitobacco youth campaign on adolescent susceptibility to cigarette smoking- Minnesota, 2002-2003. Morb Mortal Wkly Rep 2004;53:301–4.
      OpenUrlPubMed
    32. ↵
      1. Hosmer DW,
      2. Lemeshow S
      . Applied logistic regression. 2 edn. New York: Wiley & Sons, 2000.
    33. ↵
      Australian Bureau of Statistics. Population by age and sex, Australian states and territories, June 2006. Canberra: Australian Bureau of Statistics, 2007.
    34. ↵
      1. Keller PA,
      2. Block LG
      . Increasing the persuasiveness of fear appeals: the effect of arousal and elaboration. J Consum Res 1996;22:448–59.
      OpenUrlAbstract/FREE Full Text
    35. ↵
      1. Dunlop SM,
      2. Wakefield M,
      3. Kashima Y
      . Can you feel it? Negative emotion, risk, and narrative in health communication. Media Psychol 2008;11:52–75.
      OpenUrlCrossRefWeb of Science
    36. ↵
      1. Dunlop SM,
      2. Kashima Y,
      3. Wakefield M
      . Predictors and consequences of conversations about health promoting media messages. ComM 2010; 77:518–39.
      OpenUrl
    37. ↵
      1. Dunlop SM,
      2. Wakefield M,
      3. Kashima Y
      . Something to talk about: affective responses to public health mass media campaigns and behaviour change. J Health Mass Commun 2009;1(3/4).
    38. ↵
      1. Strasser AA,
      2. Cappella JN,
      3. Jepson C,
      4. et al
      . Experimental evaluation of antitobacco PSAs: Effects of message content and format on physiological and behavioral outcomes. Nicotine Tob Res 2009;11:293–302.
      OpenUrlAbstract/FREE Full Text
    39. ↵
      Department of Health and Ageing. Tobacco Warnings Cigarette Packs Set A.
    40. ↵
      Department of Health and Ageing. Tobacco Warnings Cigarette Packs Set B.
    41. ↵
      1. Brennan E,
      2. Durkin SJ,
      3. Cotter T,
      4. et al
      . Mass media campaigns designed to support new pictorial health warnings on cigarette packets: evidence of a complementary relationship. Tob Control 2011;20:412–8.
      OpenUrlAbstract/FREE Full Text
    42. ↵
      1. Dunlop SM,
      2. Kashima Y,
      3. Wakefield M
      . Pathways to persuasion: cognitive and experiential responses to health promoting mass media messages. Commun Res 2010;37:133–64.
      OpenUrlCrossRefWeb of Science
      1. Hinyard L,
      2. Kreuter MW
      . Using narrative communication as a tool for heath behavior change: A conceptual, theoretical, and empirical overview. Health Educ Behav 2007;34:777–92.
      OpenUrlAbstract/FREE Full Text
    44. ↵
      1. Kreuter MW,
      2. Green MC,
      3. Cappella J,
      4. et al
      . Narrative communication in cancer prevention and control: A framework to guide research and application. nn Behav Med 2007;33:221–35.
      OpenUrlCrossRef
    45. ↵
      1. Pennay D,
      2. Bishop N
      . Profling the ‘mobile only’ population: a study of Australians with a mobile phone and no landline telephone. Melbourne: Social Research Centre, 2009.
      1. Thompson ME,
      2. Fong GT,
      3. Hammond D,
      4. et al
      . Methods of the International Tobacco Control (ITC) Four Country Survey. Tob Control 2006;15(Suppl 3):iii12–18.
      OpenUrlAbstract/FREE Full Text
    47. ↵
      1. Cantor D,
      2. Coa K,
      3. Crystal-Mansour S,
      4. et al
      . Health information national trends survey (HINTS) 2007: Final report. 2009.
    48. ↵
      Cancer Institute NSW. New South Wales smoking and health survey 2009. Sydney, Australia: Cancer Institute NSW, 2009.

    Footnotes

    • Contributors SD analysed the data and contributed writing for all drafts. TC contributed to the conception of the study and all drafts. DP contributed to the interpretation of the data and all drafts. All authors gave final approval of the paper.

    • Funding Sally Dunlop is employed by the University of Sydney. Donna Perez is employed by the Cancer Institute NSW, a state government organisation. Trish Cotter is employed by the Victorian Comprehensive Cancer Centre. She is also a technical advisor on mass media to the World Lung Foundation.

    • Competing interests None.

    • Ethics approval The NSW Population Health Services Research Ethics Committee.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data sharing statement Readers are encouraged to the contact the authors for further information regarding the data.