LECTURE 8.1


THE IMPLICATIONS OF RESEARCH FOR DOING SURVEY RESEARCH

I. The problems with asking "Why" questions

A. In Nisbett & Wilson's (1979) research, people didn't know why they made the "right" choice when selecting among identical stockings and TV sets

B. Zajonc's mere exposure research has shown that people often like things simply because they have seen them more. People don't know they do this and in some cases are even unaware that the pattern they like most is more familiar

C. Cognitive dissonance experiments have shown that people may rationalize behavior

D. Therefore, asking subjects why they did something may not get you the real reason. As a result, we use experiments, not surveys, to determine "why's" of behavior

II. The problems with asking "What will you do?" questions

A. Two (of many) reasons for wrong answers to this question

1. People grossly underestimate situational influences on behavior. ("If you had been in the Milgram experiment, would you have shocked the learner?")

2. When events are off in the future or hypothetical, the positive aspects of the choice are stronger relative to the negative aspects than they will be when the decision is to be made because the avoidance gradient is steeper than the approach gradient

B. Answers most likely to be accurate if:

1. To determine a specific behavior, you ask a specific question (Do you intend to attend this class tomorrow? vs. What's your attitude toward attending college classes?)

2. Behavior to be performed is simple and private

3. Person has performed behavior before

4. Short amount of time between when you ask question and when behavior is to be performed

III. Why you may get inaccurate answers to questions about past behavior and attitudes

A. Availability heuristic: Vivid or easy to remember events are remembered as occurring more frequently than less vivid events

B. We remember our own contributions as being greater than they are (remember doing more good deeds more frequently than we do)

1. Students and studying

2. Obese and eating

C. Fischoff's "knew it all along" phenomena

D. Loftus and reconstruction: The effect of leading questions

1. "How fast were cars going when they crashed (smashed) into each other?"

2. "Do you get headaches frequently (occasionally)? And, if so, how often?" (2.2/week in the "frequently" condition vs. .7/ week in the "occasionally" condition)

IV. Responses to questions about present attitudes and behaviors may be affected by

A. Response sets

1. Social desirability (e.g., the difference between what the exit polls say and the rate at which voters actually voted for black candidates)

2. "Yea-saying", "Nay-saying"

3. Choosing middle answer

B. Framing (Tversky and Kahneman, Science, 1981, 211, 453-458): It makes a difference whether the glass is half full or half empty; saving 200 of 600 people is not the same as having 400 of 600 people die

V. A dramatic example of problems with surveys: Nielson (pre-people meter)

1. Participants dropping out of sample

2. Participants waiting until the end of the week to fill out diaries

3. If participants like a particular TV series, they'll always say they watch it, even on weeks when they miss it.


Lecture 8.2

Back to Chapter 8 Main Menu