Other Possible Influences on Priming Tasks

6.3.2.1 Knowing the Purpose of This Main Experiment or

After target judgment in the questionnaire, subjects are requested whether they figure out the key purpose of this research. Once they say “yes”, they still have to write

down their own ideas about the purpose. Among all 124 valid samples, only eight persons’ answers are closed to the true goal of this study. Participants possibly spend a little more efforts and time so as to figure out and elaborate the answer after the target judgment task. Hence, it is hardly able to affect apparently their previous evaluating process.

Participants are asked such question because this study concerns whether subjects guess right on the real research objective will affect the average points of each brand and finally bias the dependent measures regarding the emergence of assimilation and contrast effects. This would be a kind of “demand artifact”. “Demand artifacts” refer to the bias itself, the “error of inference regarding the cause of an observed effect”

1975, P. 103). However, according to Shimp, Hyatt, and Snyder’s study about demand artifacts in consumer research in 1991, they propose that three conditions must exist for a subject to be demand biased. Firstly, the subject must “encode” a demand cue that would alert him or her to the research hypothesis. Secondly, s/he must “discern” the correct research hypothesis or guess another hypothesis that is incidentally correlated with the true research hypothesis. Thirdly, s/he must “act” on the conforming to a certain role that leads to biased responses on measures. In equation form (Shimp, et al., 1991),

Pr(Bi) = Pr(Ei) × Pr(D₁｜Ei) × Pr(Ai｜Di),

where Pr(Bi) = probability that the i^th subject will be demand biased; Pr(Ei) = probability of encoding a demand cue; Pr(D₁｜Ei) = conditional probability of discerning the true experimental hypothesis or a correlated hypothesis; and Pr(Ai ｜ Di) probability of acting on the

This multiplicative junction stresses that all three conditions (i.e, encoding a demand cue, discerning the hypothesis, and acting on the hypothesis) must apply for a subject to be demand biased.

However, among the eight persons whose answers are closed to the true goal of this study, they merely encode a demand cue that previous advertisements read might affect their following target judgment, but they all fail in discerning the hypothesis. That is, they can not tell which direction the evaluated target would shift after being primed.

Hereafter, the probability that these eight subjects would be demand biased (i.e., Pr(Bi)) indeed decreases a lot due to the least Pr(D1｜Ei). The consequence of subjects guessing right on the experimental purpose is not as serious as originally expected.

What’s more, removing hypothesis guessers may lead to systematic between comparison groups along the lines of the selection-bias threat validity noted by Cook and Campbell (1979, p.53). In addition to threatening internal validity, removing hypothesis guessers may also adversely affect an experiment’s external validity and in turn compromise its construct validity (cf. Lynch, 1982). As a result, since those eight purpose guessers do not successfully discern the correct research hypotheses (i.e., they do not indicate the accurate direction the average prestige point of targets would shift when being primed), they shall not lead to any remarkable demand bias. Thus, the eight samples are still kept in further analyses for dependent measures.

6.3.2.2 Have Ever Heard those Existing Brand Names or not

At the last of the questionnaire, participants are asked whether they have ever heard those presented brand names before. Based on the question, among those analyzed samples, there are some of them who have not heard at least one brand name. Brand names especially concerned here are those priming exemplars and familiar targets, such as Calvin Klein, Polo Ralph Lauren, Tommy Hilfiger, Lacoste, Sunday’s, and 陶板屋. Fictitious brand names are all excluded because they are new to participants.

It is essential to examine whether participants have ever heard those brands or not

will influence their rating of 1) average points in terms of target judgment and 2) prestige ranges; which may finally affect dependent measures, the assimilation and contrast effects.

First, as for point estimates for seven product dimensions of target judgment, participants assigned in the four manipulated conditions which the targets are existing brands all have ever heard of targets’ brand names. Thus, such doubts will not happen to affect those target point ratings.

As for the exemplars’ prestige ranges, there are six, twenty-three, ten, samples respectively that have “not” heard of Calvin Klein, Tommy Hilfiger, Polo Ralph Lauren, and Lacoste. In order to examine whether their prestige ranges are different from participants having heard of these brands and those having not heard of, one-ANOVA is used for proceeding checks.

Firstly, among the samples that have ever heard of those exemplars’ names, the same numbers as those having never heard of are randomly chosen. Thus, respectively, six, twenty-three, ten, and six random samples that have ever heard of Calvin Klein, Tommy Hilfiger, Polo Ralph Lauren, and Lacoste. Continuously, four individual one-way ANOVA analyses would be applied to test if there exists obvious diversity between these two kinds of groups. Through one-way ANOVA, no significant variations between the two groups (i.e., samples having and having not heard of four individual prime) in exemplar’s prestige range are shown (see Table 6-3~ 6-6). Thus, evaluations of four primes’ prestige ranges do not diverse apparently no matter participants have heard those brands or not. (see Table 6-7)

For this reason, here comes the conclusion that even samples have not hear at least one exemplar brands within the main test; they are still covered in further analyses for dependent measures. These samples will not lead to any critical effect and bias.

Table 6-3. One-Way (Samples Having Not Heard of v.s. Samples Having Heard of) ANOVA

Table 6-4. One-Way (Samples Having Not Heard of v.s. Samples Having Heard of) ANOVA Result of Tommy Hilfiger

Between 1 1.761 1.179 .283

Within 44 1.494

Total 45 Hilfiger High

bound

Between 1 2.174 1.823 .184

Within 44 1.193

Total 45

Table 6-5. One-Way (Samples Having Not Heard of v.s. Samples Having Heard of) ANOVA Result of Polo Ralph

Between 1 2.813 1.669 .213

Within 18 1.685

Table 6-6. One-Way (Samples Having Not Heard of v.s. Samples Having Heard of) ANOVA

Between 1 1.333 .672 .431

Within 10 1.983

Total 11

Between 1 .333 .519 .488

Within 10 .642

Total 11

Table 6-7. Treatment Means in Low and High Limit of Prestige Range Dimensional

Low 4.833 (1.861) 5.4167 (1.357)

High 6.917 (1.463) 8.333 (1.329)

NOTE: In the parentheses shows the standard variation of each brand’s lowest and highest bond on the prestige dimension.

6.4 Dependent Measurement