Tropes in Advertising: A Web-Based Empirical Study

Stage One: Operationalisation of experimental stimuli

The advertisement used as the experimental stimulus was a replica of one of the ads used in the Toncar and Munch (2003) study [see HREF3, HREF4 and HREF5]. The ad contained a graphic of a portable CD-player with no brand name, accompanied by a single product claim. Twelve versions of the ad were used, as there were four ads with claims structured as simple tropes; four with claims structured as complex tropes; and four structured as explicit versions of the claims. This differed from the ad Toncar and Munch (2003) used, as their ad contained four product claims, within the same claim structure, on the one ad. Arguably, an ad structured in this way, with more than one product claim per ad, would not be what one would expect to see in a print ad. Consequently, in this study, one single product claim was placed on each ad.

By restructuring the ads to look like what one would expect to see in a print ad, this study attempted to overcome the limitations of the Toncar and Munch (2003) study in this regard.

Pre-test: Rationalisation of advertisement claim structure

Because this study was conducted in a different cultural context, the validity of this study’s experiment relied on the assumption that the explicit claims, simple tropes, and complex tropes had similar meanings for the subjects who viewed them. In order to determine the equivalence of meaning, a pre-test needed to be conducted, where focus group were used to carry out this task.

The sample for this pre-test consisted of university undergraduate students. Participation was voluntary and did not affect students’ marks or assessment for the unit. To conduct the test, two focus groups of eight students were invited to participate in a moderated group discussion. To begin the focus group task, all subjects were shown all three versions, and all four sets of the product claims developed by Toncar & Munch (2003) (See Table 2). Subjects were only shown the claims used in the ads, and not the actual print ads, as the full ad may have distracted the participants, and thus, influenced their responses. After viewing the product claims, they were asked to express the perceived meaning of each type of claim. In particular, they were asked whether each trope and claim had the same meaning, if there were any differences, and if there were differences, how they differed. The focus groups were lead by the principal researcher.

The researcher examined and categorised the focus group responses. It was thereby ensured that the groups of claims to be used in the ads had similar meanings.

Table 2. Claims used in the experiment

Source: Toncar and Munch, 2003 p. 14.

Stage Two: Testing hypotheses

Sampling

Given the limited time, budget, and resources, it was decided to utilise a convenience sample of University undergraduate students in the present study. This was consistent with the convenience sample of University undergraduate students Toncar and Munch (2003) used. One university undergraduate Marketing unit (subject) was adopted as the sampling frame for this research. As a subject taught completely online, it had a count of 460 local and international students. Despite the large number of students in the unit, difficulty was encountered, as many participants were unwilling to participate in the study. Of these 460 students who were invited to participate, 123 questionnaires were completed, giving a response level of 26.7% overall.

Students were invited to participate in the experiment through a written invitation (Plain Language Statement) on the unit website, where participation was entirely voluntary. While the subjects from the Toncar and Munch (2003) study came from North America, the subjects from the current study were a mix of Australian and International students. Of these subjects, a total of 71 were from Australia and 52 from outside countries. Hence, this study was conducted in a different context, however, it was anticipated that it would yield similar results to those of Toncar and Munch (2003).

Procedure and Data Collection Method

Testing the hypotheses involved the use of an online experiment, where a self-administered online questionnaire was used to collect data. While Toncar and Munch (2003) chose to physically deliver the experiment to participants, the current study utilised the Internet for experimental research. The type of medium may have an impact on the results, and hence, this study explored if a different medium would yield the same results.

To begin the experiment, an active browser link to the online experiment with the particular ad was placed on the unit website on the intranet. Subjects were asked to view the ads and answer the questions that followed them. Of the twelve ads developed, four were placed on each web page, where these four contained copy (text) with the same claim structure [HREF3]. This between-subjects design meant that subjects were only allocated to one group.

Immediately after viewing the ads, subjects completed a questionnaire designed to measure the dependent variables [HREF3]. The final items on the questionnaire sought respondent information on demographics, including age; gender; and country of birth.

Since an online questionnaire was used, the data was automatically collected to a database on a Web server, thus eliminating the need for manual data entry and data entry errors. JavaScript was used to prevent any item-skipping by respondents on all inventories. This ensured there was no missing data. The data needed only to be imported into the statistical software package SPSS.

Measurement of Claim Believability & Importance (Attitude)

Consistent with Toncar and Munch (2003), scales measuring the beliefs and importance of claims were used to determine if tropes had an influence on attitude (where they use Fishbein’s theory of attitude formation and change to justify using these measures). The seven-point scale was chosen as it is more sensitive than a three or five-point scale, where it more accurately measures variability in responses (Zikmund, 1997 p. 357). The advantages of using such scaled response questions is that are relatively easy to use and they are useful where information is difficult to quantify (Frazer & Lawley, 2000). The main disadvantage of scaled response questions should also be highlighted, where response bias may occur (Frazer & Lawley, 2000). Nevertheless, the present study used these scales, where they came directly from the Toncar and Munch (2003) study.

Immediately after reading the tropes, subjects were asked to rate the believability and importance of the claims by using single item seven-point semantic differential scales [See HREF3]. The believability scales were anchored by “not very likely/very likely”. The importance scales were anchored by “not very important/very important”. Both of these scales came directly from the Toncar and Munch (2003) study. This meant that they had been pre-tested, and, therefore, adequately served their purpose in the research process. That is, these scales had already been tested for reliability and validity and had been found to measure the key variables involved in this study. Subject responses were analysed using one-way analysis of variance (ANOVA).

Measurement of Counterarguing

The effects of the use of tropes on counterarguing were measured by the summation of two semantic differential scales [See HREF3]. These scales were anchored with “I argued/agreed with the ad”, and “I rejected/accepted the ad’s point”. This self-reported degree of counterarguing was taken directly from McQuarrie and Mick (1992). They used these scales to measure the amount of counterarguing produced by ads, where it was modelled on that used by Deighton et al. (1989). Therefore, these previously developed items were already pre-tested for their validity and reliability to measure the key variable in question (counterarguing). The summed scale for counterarguing had satisfactory reliability, with alphas ranging from 0.89 to 0.94. Subject responses were analysed using one-way analysis of variance (ANOVA).

Research Findings and Discussion

Response Analysis

Of the 460 students invited to participate in the study, a total of 123 students chose to complete the questionnaire. When speaking with the instructors involved in the undergraduate units the sample came from, they initially predicted that subjects would be interested in the experiment as it was related to topics they were studying. However, as this was the first time the Marketing Management unit had been taught online, they could not predict the low interest levels. Late in the semester the lecturers were able to identify where the problem lay. They posited that subjects were more interested in completing online tasks, such as weekly online tests, that counted toward their assessment for the unit. That is, the study conflicted with their desire to participate in online discussions and other workshop tasks which counted towards their assessment, and would help their overall marks for the unit.

It is possible for this online questionnaire to be criticised for nonresponse bias, where there may have been a difference between respondents and nonrespondents (Armstrong & Overton, 1977). Armstrong and Overton (1977) suggest that late responders may have answered differently than early responders. Thus, late respondents are considered to provide a good measure of the characteristics of nonrespondents (Armstrong & Overton, 1977). These early and late responses are said to occur over ‘successive waves’, where ‘waves’ refers to the response generated by the stimulus (Armstrong & Overton, 1977). In order to test nonresponse bias, the responses from the second wave (or late participants) can be compared to the responses from the first wave (or early participants) to determine if they are significantly higher, significantly lower, or not significantly different. A two-tailed test (0.05 level) would be used to do this. However, because the response levels for the current study were relatively low, this test of nonresponse bias was not possible. That is, to conduct such t-tests would have required (at least) 30 early and (at least) 30 late respondents.

Data Integrity

As previously mentioned, JavaScript was used on the online questionnaire to prevent any item-skipping by respondents on any (and all) inventories. The result was an absence of missing data.

The use of JavaScript to prevent missing data is not without its limitations. Because it forces subjects to answer all questions, it may have force subjects to give answers for things they wish not to. In addition, this imposed force can also result in premature exit from the questionnaire as a way for subjects to avoid answering questions they are reluctant to answer. In this instance, PHP tracking software designed to detect those participants who started the questionnaire but aborted before completing it, and used by researchers in previous studies (e.g., Zadeh et al., 2000), was not employed due to time, cost and university ethics committee considerations.

Respondent Characteristics

In total, 123 subjects participated in the experiment, and due to the use of Java Script in the HTML questionnaire, 123 useable questionnaires were analysed. Table 3 summarises the descriptive statistics of these subjects, illustrating that 46 subjects viewed ads with complex tropes, 41 with simple tropes, and 36 with explicit claims. The majority of subjects were under the age of 29, and many more females than males viewed the complex trope and explicit claim ads. The reason for this is not clear.

In terms of the region, most subjects, in all three ad conditions, indicated that they had lived most of the past 10 years in Australia. Other regions with large participant numbers were the Republic of China, South-East Asia, and ‘Other’ countries. Similar proportions of participants from each region viewed the different ads.

Table 3. Profile of experimental sample

Pre-test Findings

Subjects in both focus groups considered the meanings of each set of claims, in the first, second, and fourth set of claims to be the same, i.e., no differences could be identified. However, with regard to the third set of claims, subjects in both of the focus groups initially considered the claims to be different. When probed for further explanation as to how they differed, subjects explained that the simple and complex tropes had a much less explicit conclusion, and were open to much more interpretation than the explicit claims. Furthermore, they indicated that trope conclusions in this third set of claims were highly ambiguous and indeterminate, where the decoding difficulty level was not dissimilar.

Subjects pointed out that the simple and complex tropes in the first, second, and fourth sets of claims had equally simple interpretation difficulty levels, where the relative simplicity of them meant they could be decoded at similar rapid speeds. However, in saying this, they did consider the complex tropes to be more open in their conclusion than their simple counterparts.

Discussion of Pre-test Findings

The findings revealed that the claims were equivalent in their meaning, that the difficulty level in decoding the two types of tropes was equally simple, and that the complex tropes were slightly more open in their conclusion. The tropes were not changed to account for the decoding difficulty level, as they were still found to be equivalent in meaning; which was the aim of the pre-test. If the trope claims were changed to make the complex tropes more difficult to decode, then the results of this study could not be compared to Toncar and Munch’s (2003) results to see if their findings are robust enough to generalise beyond the sample they used.

With regard to the third set of claims, the results revealed a negative aspect of using tropes. That is, they can be too implicit in their conclusion to interpret. As Berlyne (1971) suggests, high levels of ambiguity can influence displeasure. Or, as McQuarrie and Mick (1992) suggest, the extreme levels of deviation in a rhetorical figure could make it possible that it may generate uncertainty instead of interest, and therefore, have a damaging effect.

In addition, the results show how tropes can be misinterpreted when they have such a hidden conclusion in the claim meaning. Ahearne, Gruen, and Saxton (2000), and Kardes, Kim, and Lim (1994) highlight this point. If the researcher was not there to moderate the discussion of these focus groups, then it is possible, as Sawyer and Howard (1991) suggest, that conclusions may have failed to be formed at all.

It is to be noted that for the same reasons, neither the degree of decoding difficulty in the adopted claims for the current study, nor the high level of ambiguity were changed in the current study.

Experimental Findings

Claim Believability and Importance (Attitude)

Tables 4 and 5 present the means for the believability and importance measures. Hypotheses one (H₁) and two (H₂) are not supported. Where each pair of the simple tropes and explicit claims was used, there was no significant difference in the believability of the claims, or perceived importance of the product features. This was also the case between each pair of the complex tropes and explicit claims, where there was also no significant difference in believability or perceived importance. Thus, hypotheses three (H₃) and four (H₄) are not supported.

Overall, it appears that claims structured as either simple or complex tropes, when compared with explicit claims, have no apparent effect on enhancing persuasion, or a positive attitude, for those who viewed them. In the case of the samples used in this study, structuring claims as tropes did not make them more effective, whether they were simple or complex.

Table 4. Means and standard deviation for Believability

Table 5. Means and standard deviation for Importance