«Assimilation and Contrast Effects in PartWhole Question Sequences: A Conversational Logic Analysis ...»
See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/31033013
Assimilation and Contrast Effects in PartWhole Question Sequences: A Conversational
Article in Public Opinion Quarterly · March 1991
Impact Factor: 2.25 · DOI: 10.1086/269239 · Source: OAI
3 authors, including:
Norbert Schwarz Fritz Strack University of Southern California University of Wuerzburg 358 PUBLICATIONS 27,400 CITATIONS 163 PUBLICATIONS 12,302 CITATIONS
SEE PROFILE SEE PROFILEAvailable from: Fritz Strack Retrieved on: 19 May 2016 Assimilation and Contrast Effects in Part‐ Whole Question Sequences: A Conversational Logic Analysis Norbert Schwarz, Fritz Strack, Hans‐Peter Mai ZUMA‐Arbeitsbericht Nr. 90/01 Zentrum für Umfragen, Methoden und Analysen e.V. (ZUMA) Postfach 12 21 55 D‐6800 Mannheim 1
Seit Juli 1983 sind die ZUMAArbeitsberichte in zwei Reihen aufgeteilt:
Die ZUMA-Arbeitsberichte (neue Folge) haben eine hausinterne Begutachtung durchlaufen und werden vom Geschäftsführenden Direktor zusammen mit den übrigen Wissenschaftlichen Leitern herausgegeben. Die Berichte dieser Reihe sind zur allgemeinen Weitergabe nach außen bestimmt.
Die ZÜMA-Technischen Berichte dienen zur hausinternen Kommunikation bzw. zur Unterrichtung externer Kooperationspartner. Sie sind nicht zur allgemeinen Weitergabe b e stimmt.
Assimilation and Contrast 1 Assimilation and Contrast Effects in Part-Whole
A Conversational Logic Analysis Norbert Schwarz Zentrum für Umfragen, Methoden und Analysen, ZUMA, Mannheim. FRG Fritz Strack Universität Mannheim,
The present paper is based in part on the third author's diploma thesis in psychology, conducted at the University of Heidelberg under direction of the first author. The reported research was supported by grant SWF0044-6 from the Bundesminster für Forschung und Technologie of the Federal Republic of Germany to N. Schwarz. Correspondence should be addressed to Dr. Norbert Schwarz, ZUMA, P.O. Box 122 155, DMannheim. W. Germany.
A theoretical model of the emergence of assimilation and contrast effects in part-whole question sequences is presented, and an experiment that tests its predictions is reported. Assimilation effects are predicted when one specific question precedes the general question and the two are not assigned to the same conversational context. If both questions are perceived as belonging together, however, conversational norms of non-redundancy prohibit the repeated use of information that has already been provided in response to the specific question when making the general judgment. Contrast effects may emerge in that case under specified conditions. If several specific questions precede the general question, however, the general one is always interpreted as a request for a summary judgment, resulting in assimilation effects even under conditions that foster contrast effects if only one specific question is asked. The model is supported by the reported experiment and is consistent with other findings reported in the literature.
Assimilation and Contrast Effects in Part-Whole Question Sequences: A Conversational Logic Analysis Survey researchers repeatedly observed that answering a specific question may influence the responses given to a subsequent general question (e.g.. McClendon & O'Brien. 1988 a, b; Schuman & Presser, 1981: Smith, 1982). However, the findings are inconsistent: In some studies, the responses to the general question are assimilated to the responses given to the specific question, whereas in others they are contrasted to the previous responses.
For example. Schuman and Presser (1981) found that respondents were less likely to report high general life-satisfaction when they had previously answered a similar question on marital satisfaction. Given that most respondents reported high marital satisfaction, this pattern reflects a part-whole contrast effect.
In contrast. Smith (1982; see also Smith, in press) obtained just the opposite result, although apparently using the same questions in the same order. Again, most respondents reported high marital satisfaction, but after having answered this specific question, they were subsequently more likely to report high general lifesatisfaction as well.
Thus, Smith's (1982) data reflect a part-whole assimilation effect. In the present paper, we describe a theoretical model that accounts for the emergence of these apparently inconsistent findings, and report an experiment that was designed to test our predictions.
Assimilation and Contrast 4 Cognitive Accessibility In a theoretical analysis of the above findings. Strack and Martin (1987) suggested that the emergence of assimilation effect?
on measure of general life-satisfaction reflects the increased accessibility of information about one's marriage that was used to answer the preceding marital satisfaction question. Specifically, individuals may use a variety of different aspects of their life to evaluate its overall quality, including their marriage, job.
income, housing, and so on (see Schwarz & Strack, 1989, in press, for a more detailed discussion). Which of these potentially relevant aspects they select in making a judgment depends on which is most likely to come to mind at that point in time (e.g..
Schwarz & Clore. 1983; Schwarz, Strack, Kommer, & Wagner. 1987;
Strack. Schwarz, & Gschneidinger, 1985). As a large body of literature in cognitive psychology indicates (see Bodenhausen & Wyer, 1937; Wyer & Srull, 1989 for reviews), individuals are unlikely to retrieve a l l information that may potentially bear on a judgment, but truncate the search process as soon as enough information has come to mind to form a judgment with sufficient.
subjective certainty. Accordingly, their judgments strongly reflect the impact of the information that is most accessible in memory at the time of judgment. This is usually the information that has been used most recently, e.g. for the purpose of answering a preceding question.
In line with this assumption, Strack, Martin and Schwarz (1988) observed in an experiment with American college students that the correlation between ratings of "happiness with dating" and
which both question were asked. I f the general happiness question preceded the dating question, both questions were essentially uncorrelated, r =.16. If the question order was reversed, however, this correlation increased to r =.55, z = 2.44, p.007, for the difference between both correlations.
These findings indicate that respondents were more likely to use information about their dating-life in evaluating the quality of their life-as-a-whole when this information was more accessible in memory, due to its use in answering the preceding question.
The Impact of Conversational Norms However, individuals do not always use the information that is easily accessible in memory. Under some conditions, they may intentionally disregard information that comes to mind, e.g., because it does not bear on the judgment at hand (Schwarz & Bless,
1990) or because other factors require that it should not be used.
As Strack and Martin (1987) pointed out, following related suggestions by Bradburn (1982) and Tourangeau (1984). a particularly relevant factor that may inhibit the use of easily accessible information in a survey context is provided by conversational norms. Specifically, one of the principles that govern the conduct of conversation in everyday life (Grice, 1975) requests speakers to make their contribution as informative as is required for the purpose of the conversation, but not more informative than is required. In particular, speakers are not supposed to be redundant and to provide information that the respondent already has. In psycholinguistics, this principle is known as the "given
contract", that emphasizes that speakers should provide "new" information rather than information that has already been "given" (Clark. 1985; Haviland & Clark. 1974).
Applied to question order effects in survey interviews, these considerations suggest that respondents may hesitate to reiterate information that they have already provided in response to a preceding question. Thus, respondents who have just reported their marital happiness may consider the subsequent question about their happiness with life-as-a-whole to be a request for new information. They may therefore interpret the general question to refer to other aspects of their life, much as if it were worded.
"Aside of your marriage, how happy do you feel about the other aspects of your life?". If so. these respondents may deliberately ignore information about their marriage in answering the general life-satisfaction question, despite its high accessibility in memory. Thus, conversational norms that discourage redundancy may provide the psychological rational that underlies what Schuman and Presser (1981) have called a "subtraction effect".
To provide a direct test of this assumption. Strack et al. (1988) explicitly manipulated the conversational context in which the specific and the general question were presented. This was accomplished by a joint lead-in to both questions that read. "Now, we would like to learn about two areas of life that may be important for people's overall well-being: a) happiness with dating, b) happiness with life in general." Subsequently, both happiness questions were asked in the specific-general order. Under this condition, answering the dating question prior to the general happiness question did not result in an Assimilation and Contrast 7 increased correlation, r =. 2 6 ; moreover, this correlation was significantly lower, z = 1.88. p. 0 3. than the correlation o f r =.55, obtained under the same order condition without a joint lead-in. Thus, respondents based their general happiness judgment on information other than their dating-life when both questions were explicitly assigned to the same conversational context — despite the high cognitive accessibility of the previously used marital information.
Although testing differences in correlations provides the strongest test of the theoretical assumptions, survey researchers are typically more interested in differences in means and margins. Accordingly, we will extend our analysis t o these differences in the present paper. Note in this regard, that the direction of differences in the means or margins depends on the valence of the information that is brought to mind by the specific question. For example, high dating happiness should result in reports of high general happiness if the specific information is included when making the general judgment, whereas low dating happiness should result in reports of decreased general happiness. While this prediction of part-whole assimilation effects is straightforward, the reverse does not necessarily follow. For example, disregarding one's happy dating life may not necessarily reduce judgments of general life-satisfaction. If respondents exclude information about one life-domain from consideration, they may turn to other life-domains as a basis of judgment. If so, their judgments may be determined by the evaluative implications of the new information they turn to.
If they happen to have wonderful jobs in addition to a great dating life, they may still report high Assimilation and Contrast 8 happiness when they use their job situation as a basis of judgment. Thus, while we can conclude that the impact o f dating happiness on general happiness will be reduced, and part-whole assimilation effects will not be obtained, it does not necessarily follow that a part-whole contrast effect will emerge. For this very reason, analyses of correlational differences rather than mean differences provide the theoretically more adequate test. The study reported in the present paper explores these possibilities, extending the analysis provided by Strack et al, (1988) from differences in correlations to differences in means.