From the editor.

In my short tenure as Editor of the Journal of Advertising, I have noticed that certain issues arise as papers proceed through the review process. In this vein, I think it is appropriate to comment on my views of empirical research methods and related issues to which potential authors need to

The Journal has a history of being very open to all types of research paradigms and methods, and I intend to continue that tradition while I am Editor. To date, I have received over 70 manuscripts for review at the Journal, and it is safe to say that many different types of research methods have been employed in the studies that have been reported in these manuscripts. An unscientific sample of these methods indicates that studies that have been accepted and/or requested for revision have included the following research methods: experiments, surveys, content analyses, phenomenological methods, focus groups, and others. I hope that this is an indication of the eclectic nature of the Journal and research in the field of advertising.

First, I think it is important for me to go on record and say that I am also very open to publishing manuscripts that use empirical data to develop theory. Thus I view the broad category of research methods that have been labeled as "qualitative" as being very appropriate for the Journal. My concern about such studies (as is the case with most of the reviewers) is no different than my concern regarding so-called quantitative research--studies need to be rigorous, such that they follow the guidelines passed on by those who have standards for this type of research.

Second, some submitted manuscripts feature what are often referred to as "exploratory studies." Is there room in the Journal for these sorts of studies to be published ? My answer, to a limited degree, is yes! Keep in mind that I view the Journal's mission as disseminating information with respect to advertising theory, and developing theory often involves the use of some type of exploratory research method. To develop theory, researchers are likely to review literature, analyze cases, obtain the opinions of experts, and so forth. Thus, if studies that feature exploratory methods contribute to theory development, they are acceptable for the Journal. However, the analyses of data in such studies should be appropriate for the type of data that has been gathered, and it should be performed in a rigorous manner and reported in such a way that the rigor is apparent. Moreover, all manuscripts submitted to the Journal, including those that are exploratory, should make a contribution to theory.

While dealing with the domain of exploratory research, I believe it is important to elaborate on one point just a bit more (since several of the reviewers have indicated that this issue is sometimes a problem with submissions to the Journal). This point is best described via a fictitious example. Suppose manuscript A is submitted and includes a statement something like this:

   Given the exploratory nature of the study, a p < .10 significance
   level will be used in this research.

And, in response, a reviewer writes something like:

   The manuscript provides a theory-based model from which it
   derives a cogent set of testable hypotheses. It is not an
   exploratory study.

In such a situation, it appears that I am put in a position where I need to make a decision about whether a particular research study is exploratory and/or whether a p < . 10 significance level is appropriate (in fact, some papers claim to have provided "directional support" for a hypothesis when statistical tests are nowhere near significant but means are in the appropriate direction).

According to Churchill (2001), exploratory research deals with research whose major emphasis is on obtaining ideas and insights about a problem at hand. In fact, Churchill views exploratory research as a method to be used to discover ideas and insights so that a broad problem may be broken down into smaller subproblems. He clearly distinguishes this type of research from descriptive and causal research, which are concerned with describing the frequency with which something occurs and/or the extent to which two variables are related (either coincidentally or causally). Thus, it appears to me that whereas exploratory research is broad-based and non-technical, descriptive and causal research require a certain amount of technical work. As a result, I am reluctant to view, for example, a well-defined experiment that features moderate to large cell sizes, well pretested manipulations and manipulation checks, and resulting data that may be analyzed neatly with analysis of variance as being an exploratory study.

Recently, I have been made aware of a phenomenon called "exploratory experiments" (which is fairly prominent in the agricultural sciences). In these types of studies, researchers provide many treatments (sometimes 10 or 20) to subjects in hopes of uncovering some effects that may be studied with greater detail in future experiments. The idea is to simply see if potential effects might occur in given types of experimental treatments. In such studies, one might be tempted to argue that a p < . 10 significance level should be used. However, statisticians in these areas still warn against researchers using inflated alpha levels in such studies. For example, Snedecor and Cochran (1989) contend that the p-level should be divided by the number of treatments in exploratory experiments to make statistical tests even more conservative (and to protect against the occurrence of so-called type I errors). The logic is to reduce the possibility that researchers "think" they have an effect when it may simply be due to chance. Thus, in my opinion, the fact that an experiment is exploratory in nature still does nor seem to be reason enough to use a p < .10 significance level.

Having said this, does that mean that all statistical tests that are reported in the Journal should use a p < .05 significance level? My answer is an emphatic Yes! I am aware that using a p < .05 (as opposed to a p < .10) significance level increases the probability of type II errors, but I also believe that as scientists (albeit social scientists) we need to adhere to the rigor that is demanded in the scientific community. My main point here is that if a research project has used statistical methods (perhaps other than some method appropriate for exploratory research, such as metanalysis), it is unlikely that it should be classified as exploratory. Furthermore, even if a researcher is purporting to be performing an exploratory experiment, many statisticians contend that a more rigorous (as opposed to a less rigorous) p-level standard be used. As I stated above, methodological rigor is something that I want to see in every empirical article that is published in the Journal. Thus, if researchers choose to use a p < . 10 and they want to publish in the Journal, a very strong argument will need to be included in the manuscript suggesting why this is the case. However, if researchers are using true exploratory methods, it is unlikely that they should be testing hypotheses with statistical methods anyway.

REFERENCES

Churchill, Gilbert A. (2001), Basic Marketing Research, 4th ed., Fort Worth, TX: Harcourt.

Snedecor, G. W., and W. G. Cochran (1989), Statistical Methods, 8th ed., Ames: Iowa State University Press.

Russell N. Laczniak

Iowa State University