Comments on Methodological Issues in Advertising

Arch G. Woodside, University of South Carolina
[ to cite ]:
Arch G. Woodside (1983) ,"Comments on Methodological Issues in Advertising", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 288.

Advances in Consumer Research Volume 10, 1983      Page 288


Arch G. Woodside, University of South Carolina

Recall and recognition have not been shown to affect encoding, intention to buy, or sales measures in studies of advertising effectiveness. Evidence does exist that encoding does affect conations; encoding is the consumer's answer to Krugman's second question, "Is it for me?" "It" being the brand, benefit, or the message elaboration on an advertisement.

The question of recall versus recognition is not penetrating enough to answer the question of real concern in advertising copy testing: does the ad work to produce encoding?


To test the relative ability of recall versus recognition to measure some deciding activities, a between group design is most appropriate, e.g., recall and recognition should be measured separately in two groups of subjects exposed to the same stimuli. Using a within groups design, recognition tests will likely affect recall tests, and, as demonstrated by Singh and Rothschild, recall tests may affect recognition tests. Between-group designs control for such testing effects.

Precision is increased using a within- versus a between-group design but at the cost of confounding results. Using a within-group design to save the additional cost required by using larger sample sizes in between-group designs is a false economy.

Episodic Ecphory

Mental maps are needed of what occurs to consumers starting with being exposed to ads to thought verbalizations. Singh and Rothschild's introduction of Tulving's work to the consumer literature is helpful. Could Figure 1 be revised by Singh and Rothschild to (1) provide more information on how each event affects the next event and (2) allow for the occurrence of no further mental activity following each event?


Given that low-involvement ads were used and the subjects were tested for recall and recognition two weeks after viewing the commercials, accurate recognition b! any subject in Singh and Rothschild's study is surprising. Means and standard deviations for confidence only and accuracy only would have been useful additions to Tables 1 through 3.

Viewing of the news shows and commercials should have been done individually instead of by groups of 15 subjects. The use of groups for administering the treatments confounds the results.


Recall tests may produce confusion and frustration for subjects exposed to low-involvement ads. Confusion and frustration may be the cause of lower recognition scores o brand names following such recall tests.


Macklin demonstrates that task design affects memory responses in advertising testing of young children. Unfortunately, recall tests preceded recognition tests. The recall tests may have caused the nonverbal recognitions to be higher than otherwise would have been the case.

Within-group designs decrease external validity. A between-group design should be used with each subJect tested with one commercial.

The greater verbal recall ability of 5 versus 4 year olds is intuitively appealing. Macklin's thorough literature review and warnings on demand characteristics of recall tasks should be noted by researchers of advertising effects on children.


Given the low-involvement nature of much of consumer behavior, the questions asked by Mizerski et al. are appropriate. The provision for nonsubstantial responses (NSR's) is advisable in using recognition and choice tests in consumer research. "Don't know" or "could care less" may reflect reality for most low-involvement choice situations.

Mizerski et al. provide a useful literature review and framework in their Figure 1.


A total of 16 chi-square analyses would be more appropriate than the 8 tests run on the data in Table 1. The tests should be run on the control and treatment groups separately.

For both the control and treatment groups, the use of the "don't know" option provided lower preferences for "false" not higher as reported by Mizerski et al. for the last electric shaver recognition statement (3/21 v. 10/43 and 5/66 v. 11/80, respectively).

The Bottom Line

The main effect on the use of NSR's is on external validity. Reality is more closely approximated with NSR's. Mizerski et al. serve us well by stressing this point.