|
?:linkContext
|
-
Die ZDMA-Arbeitsberichte (neue Folge) haben eine hausinterne Begutachtung durchlaufen und werden vom Geschäftsführenden Direktor zusam men mit den übrigen Wissenschaftlichen Lei tern herausgegeben. Die Berichte dieser Reihe sind zur allgemeinen Weitergabe nach außen bestimmt. A bstract Three experiments indicate that the numeric values provided as part of a rating scale may influence respondents' interpretation of the endpoint labels. In experiment I, a represen tative sample of German adults rated their success in life along an ll.-point rating scale, with the endpoints labeled " not at all successful" and " extremely successful." When the numeric val ues ranged from 0 (" not at all successful" ) to 10 (" extremely successful" ), 34 percent of the respondents endorsed values be tween 0 and 5. However, only 13 percent endorsed formally equivalent values between -5 and 0, when the scale ranged from -5 (" not at all successful" ) to + 5 (" extremely successful" ). Experiment 2 provided an extended conceptual replication of this finding, and experiment 3 demonstrates that recipients of a re spondent's report draw different inferences from formally equiva lent but numerically different values. In combination, the findings indicate that respondents use the numeric values to disambiguate the meaning of scale labels, resulting in different interpretations and, accordingly, different subjective scale anchors. Rating scales with labeled endpoints are probably the most widely used measurement instrument in social and psychological research. Leaving some concerns about their psychometric properties aside (see , the use of these scales does not seem to be very controversial (see for a careful discussion of their general properties and for empirical and psychological justifica tions for their use). In general, 7-point scales seem to be best in terms o f reliability, percentage of undecided respondents, and respondents' ability to discriminate between the scale values (e.g., . Thus, seven plus or minus two is the usual recommendation. Moreover, re spondents are able to use rating scales consistently, even in telephone interviews without visual aids (e.g., . In addition, verbal rating scales, which provide a label for each scale point, have been found to be more reliable than scales that provide labels for the endpoints only . Finally, re searchers have observed that the terms used to label the endpoints, or to designate the separate values of verbal rating scales, may affect the obtained distribution (e.g., . This suggests that respondents pay close attention to the meaning of the labels provided to them, much as one would hope. Whereas the number of scale points, the inclusion or omission of a neutral point, and the choice of scale labels have received considerable attention in the literature (see , the specific numeric values provided have, to our knowledge, not been the topic o f theoretical analysis and empirical investigation. Apparently, re searchers assume that, for example, a 7-point scale that ranges from " 1" to " 7" is equivalent to a 7-point scale that ranges from " -3 " to " + 3 ," as long as the same endpoint labels are provided. In the present article, we will question this assumption. Drawing on survey data from the Allensbach archive, we will first demonstrate that the specific nu meric values provided in a rating scale can have a dramatic impact on the obtained results. We will then discuss different underlying pro cesses and will test their viability in laboratory experiments. In July 1988, the Allensbach Institute conducted a split-ballot experi m ent as part o f a representative survey of the adult population of the Federal Republic of Germany, using a quota sample of 1,032 respon dents, based on an intersection o f region, sex, and age. In face-to-face interviews, all respondents were asked to report how successful they have been in life, along an 11-point rating scale with labeled endpoints. The rating scale was presented on a show card, in the form of a ladder, and ranged either from 0 = " not at all successful" to 10 = " extremely successful," or from -5 = " not at all successful" to + 5 = " ex tremely successful." Respondents were randomly assigned to one of the two numeric value conditions, and the question read: " How suc cessful have you been in life, so far? Please use this ladder to tell me. This is how it works: 0 [ -5 , respectively] means 'not successful at all,' and 10 [ + 5, respectively] means that you were 'extremely successful.' Which number do you choose?" As shown in table 1, the numeric values provided on the respective rating scales had a pronounced impact on the obtained reports, x2 (10) = 105.1, p < .001. For example, whereas 34 percent of the respon dents who were given the 0-10 scale endorsed a value between 0 and 5, only 13 percent of the respondents who were given the -5 to + 5 scale endorsed one of the presumably equivalent values between -5 and 0 (z = 8.11, p < .0001, for this contrast; cf. ). Coding both scales from 0 to 10, this pattern results in mean ratings of M = 6.4 on the 0-10, but M = 7.3 on the -5 to + 5 version o f the scale. In addition, an inspection of the distributions along both scales indicates that the responses are dislocated toward the high end of the -5 to + 5 scale, as compared to the 0-10 scale. This is also reflected in markedly different standard deviations, SD = 1.03 and .56 for the 0-10 and -5 to + 5 scale, respectively. These findings may reflect either that respondents hesitated to assign themselves a negative score with regard to their success in life or that the numeric values provided on the scale influenced respondents' interpretation of the endpoint labels. Specifically, " not at all success ful" may be interpreted as referring to the absence of success if com bined with a numeric value of 0, but as referring to the presence of explicit failure if combined with a numeric value of -5. Before we address these possibilities in more detail, however, it is informative to consider the findings of an extended replication. To test the reliability of the above finding, we conducted a conceptual replication with a sample of 101 students at a German university. In a self-administered questionnaire, half of the respondents were asked to report, along 11-point rating scales, how successful they have been in Note.-Percentages rounded. Data based on a quota sample of 1,032 German adults, randomly assigned to conditions. The question read, " How successful have you been in life, so far?" with scale endpoints labeled " not at all successful" (0 or -5) and " extremely successful" (10 or +5). See Appendix for full wording, x? (10) = 105.1, p < .0001. life and how happy a childhood they had. The other half of the respon dents rated the success and childhood happiness of their parents along the same scales, thus extending the present study to proxy reports. The key manipulations consisted in variations of the numeric values and the type of endpoint label provided. Specifically, the scale ranged either from 0 to 10 or from -5 to + 5 , replicating experiment 1. In addition, the endpoints were either labeled " unsuccessful" and " very successful" (or, " unhappy" and " very happy," respectively) or they were labeled " not so successful" and " very successful" (or, " not so happy" and " very happy," respectively). The latter wordings were introduced to explore the impact of differentially ambiguous scale la bels, and the exact wordings of the questions are given in the Appen dix. In sum, these manipulations resulted in a 2 (numeric values) x 2 (scale labels) x 2 (self-or proxy reports) factorial between subjects design, which was analyzed by analysis of variance. The obtained data provide a robust replication of the previously observed impact of numeric values on self-as well as proxy reports of success and childhood happiness. Coding both scales from 0 to 10, respondents reported higher success in life for themselves (M = 7.38), as well as for their parents (M = 8.13), along the -5 to + 5 scale than along the 0-10 scale (M = 5.96 and 7.04 for self and parents, respectively), resulting in a pronounced main effect of numeric values (^[1,93] = 16.21, p < .001). Similarly, respondents reported higher childhood happiness along the -5 to + 5 scale than along the 0-10 scale, again both for themselves (M = 8.08 and 6.17 for the -5 to + 5 and the 0-10 scale, respectively), as well as for their parents (M = 7.04 and 5.38, respectively), with F(l,93) = 5.02, p < .03 for the main effect of numeric values. The remaining effects that reached significance were of little theoret ical interest and reflected that respondents perceived their parents as having been more successful in life, but less happy during their child hood. N either the self/proxy variable, nor the ambiguity of the scale labels, however, moderated the impact of the numeric values, all Fs < 1 for the interaction terms. As alluded to above, two processes, which are not mutually exclusive, may underlie the consistently observed impact of numeric values. As*, a first possibility, respondents may hesitate to assign themselves a negative score, reflecting self-presentation concerns. Although we can not rule out an impact of self-presentation concerns, some aspects of our findings suggest that self-presentation may not be the key factor that drives the observed phenomenon. First, respondents in experi-ment 2 provided their reports anonymously in a self-administered ques tionnaire, thus reducing the potential impact o f social desirability (cf. ). Moreover, self-presentation considerations suggest that the impact of numeric values should be more pronounced for selfthan for proxy reports, which was not the case. Finally, why should respondents hesitate to endorse a negative scale value, unless negative scale values communicate a different meaning than positive values? These considerations suggest that respondents may have used the numeric values provided as part of the rating scale to interpret the meaning of the labels. As note, " Even the m ost unambiguous words show a range of meaning, or a degree of 'semantic flexibility,' . . . that is constrained by the particular context in w hich these words occur." (See also and Bransford for a more general discussion.) F or example, respondents who are asked to rate their success in life need to determine what the researcher means by " unsuccessful," " not so successful," and the like. Does that term refer to the absence of remarkable successes, or does it refer to the presence of failure? Depending on how respondents interpret the term , respondents who have, for example, neither experi enced particular successes nor particular failures but have done " al right" in life may choose very different scale values. Thus, the present findings may reflect that the numeric values changed the meaning of the endpoint labels, resulting in different responses, much as has been observed in studies that explicitly varied the wording of the scale labels (e.g., . According to this account, the finding that the different labels used in experiment 2 did not result in a differential impact of num eric values presumably reflects that the terms " unsuccessful" or " unhappy" are as ambiguous as the terms " not so successful" or " not so happy," which we introduced on the intuitive assumption that they may be more ambiguous than the former. N ote that this interpretation does not require self-presentation con siderations as a necessary condition to account for the observed find ings. N evertheless, it implies that the impact of numeric values may increase with increasing self-presentation concerns, because these concerns are known to increase respondents' reluctance to endorse values with unfavorable implications . Accordingly, semantic interpretation and self-presentation processes are not mutu ally exclusive. Rather, the operation of the latter requires the operation o f the former in the first place. Experiment 3 was designed to provide a more direct test of the shift in meaning hypothesis offered here, under conditions that render self-presentation concerns very unlikely. To accomplish this, respon dents were given transcripts of another person's reports along rating scales with different numeric values and were asked to draw inferences about the target person. If the numeric values change the meaning of the scale labels, this should be reflected in different inferences about the target. Twenty-two students at a German university participated in a study that was purportedly concerned With the accuracy of the inferences that people can draw about others on the basis of minimal information. All participants received a short description of two target persons and a transcript of these persons' responses to a survey question (see Ap pendix for question wordings). The first target person reported his health satisfaction along an 11-point rating scale, ranging from " dis satisfied" (0 or -5) to " very satisfied" (10 or +5). Depending on the numeric values provided on the scale, this person had allegedly checked a value of -4, or a numerically equivalent value of 1. Simi larly, the second target person had allegedly rated his success on aca demic exams, again along 11-point rating scales, with the endpoints labeled " not so successful" (0 or -5 ) and " successful" (10 or +5), and had checked a minus three or a two. Subjects were randomly assigned to one o f the two numeric values conditions constituted by the scale along which the target persons had allegedly given their reports. As dependent variables, subjects were asked, in an open response format, to estimate how frequently the first target person had to see a doctor during the last month and how often the second target person had to repeat an exam due to failure. If the numeric values affect subjects' interpretation of the endpoint labels of the rating scales along which the target persons gave their reports, one should expect that they draw more extreme inferences if the numeric values range from -5 to + 5 rather than from 0 to 10. The findings support this assumption. Specifically, subjects esti mated that the first target person had to see a doctor twice as often when he checked a minus four on the -5 to + 5 scale (M -2.2), than when he checked one on the 0-10 scale (M = 1.0), F(l,20) = 4.86, p < .04. Similarly, they assumed that the second target person had failed on twice as many exams (M = 1.73) when he checked a minus three on the -5 to + 5 scale, than when he checked a 2 on the 0-10 scale (M = .91), F(l,20) = 5,63, p < .03. In combination, these findings indicate that respondents drew more extreme inferences from reports given along a -5 to + 5 scale than from formally identical reports given along a 0-10 scale. This is consistent with the assumption that the numeric values presented on the rating scales affected respondents' interpretation of the meaning of the report. We conclude from the reported findings that respondents may use the numeric values provided on a rating scale to disambiguate the meaning of scale labels. If the numeric values range from 0 to 10, as was the case in our studies, their very structure seems to suggest that the researcher is interested in the absence or presence of the attribute to which the scale pertains, that is, success or happiness in the above examples. If the numeric values range from -5 to + 5 , including a zero at the midpoint, their structure seems to suggest that the absence of the attribute corresponds to zero, whereas the negative values refer to the presence of its opposite, that is, failure or unhappiness in the above examples. In more general terms, scales that provide a contin uum from negative to positive values may indicate that the researcher has a bipolar conceptualization of the respective dimension in mind, whereas scales that present only positive values may indicate a unipo lar conceptualization. If so, the choice of numeric values may either facilitate or dilute the polarity implications of the endpoint labels that are provided to respondents. Accordingly, researchers may be well advised to match the numeric values that they provide to respondents with the intended conceptualization of the underlying dimension as uni-or bipolar. W hereas this point may seem obvious in the studies reported above-where the combination of an apparently unidirectional verbal scale (" not at all" to " extremely successful" ) with a bipolar numeric scale may strike some readers as awkward-the strong effects ob tained here suggest that the impact of numeric values should be even more pronounced when the polarity of the verbal labels themselves is more ambiguous. For example, political issue questions in the National Election Studies are intended to introduce two opposing positions, reflecting a liberal and a conservative viewpoint. Thus, respondents may be asked, " Some people believe that we should spend much less for defense. Others feel that spending should be greatly increased. W here would you place yourself on this scale? Greatly decrease de fense spending (1); greatly increase defense spending (7)." The current findings suggest that using values from -3 to + 3 , rather than from 1 to 7, may help to emphasize the intended liberal-conservative bipolar ity of these options. Similarly, our reasoning suggests that scales that follow the format of semantic differentials ) by providing polar opposites should make use of numeric values that range from negative to positive, as is frequently but not always the case. In con trast, scales that are intended to assess the intensity of a single attri bute should follow a zero-to-positive-values format to emphasize that the question pertains to the absence or presence of this specific attri bute, rather than the presence of its opposite. N ote, however, that the use of different scale formats may affect the obtained item variance. The conditions under which this is the case are not well understood, however, and further research is needed to explore this issue. Suppose that most people are more likely to experi ence success rather than failure in life. If so, using a -5 to + 5 format restricts the meaningful response alternatives for most respondents to the positive half of the scale, resulting in reduced item variance relative to a 0-10 format. Suppose, on the other hand, that many people would experience more failure than success in life. If so, a 0-10 format would provide fewer meaningful response alternatives for these respondents than a -5 to + 5 format, resulting in reduced item variance in the former case. This suggests that the choice of a scale format should be based on researchers' knowledge about the relevant distribution to avoid undesirable restrictions in item variance. Finally, the present reasoning bears on the comparability of data obtained under different administration modes (see for a general conceptualization of mode effects). The use of numeric values is most prevalent in face-to-face interviews and selfadministered surveys, whereas the increasing use of an unfolding for mat in telephone interviews avoids the presentation of numeric values. Based on the present findings, we hypothesize that data obtained in an unfolding format are more compatible with verbal rating scales that present labels for each scale point, or with scales that present unnum bered boxes or similar devices, than with scales that use numeric val ues. These considerations echo the general insight that the absolute values, or marginals, obtained in response to any survey question are difficult to interpret (see . To what extent changes in scale format affect the obtained relationships or changes over time, on the other hand, needs to be explored in future research. From a theoretical perspective, the present findings support the gen eral conclusions that we have drawn from related research into the impact of response alternatives on frequency reports of mundane behaviors and related judgments (see . Far from being " neutral measurement devices," the response alternatives that are provided to respondents do constitute a source o f information that respondents actively use in determining their task and in constructing a reasonable answer. While survey meth odologists have traditionally focused on the information that is pro vided by the wording of the question, we need to pay equal attention to the information that is conveyed by apparently formal features of the questionnaire. Respondents apply many of the rules that govern the conduct o f conversation in everyday life (cf. to the survey interview. In doing so, they assume that every contribution to the ongoing conversation is relevant and meaningful. In the survey interview, the researcher's contributions include the re sponse alternatives, the numeric values of rating scales, and the order ing of questions, as well as other features of questionnaire design. (See ], , ] for reviews and theoretical analyses.) The anal ysis o f the informational functions of apparently formal features of questionnaire design is therefore a key task in the collaboration of cognitive psychologists and survey researchers. been on your exams?" -" not so successful" (0 or -5) to " very successful" (10 or +5). Depending on type of scale, Peter K. allegedly checked " 2" or " -3 . " Scenario 2. Manfred D. was asked how satisfied he is with his health. He gave the following response: " How satisfied are you with your health?" -" not so satisfied" (0 or -5) to " very satisfied" (10 or +5). Depending on type of scale, Manfred D. allegedly checked " 1" or " -4 . " EXPERIMENT 1 " Wie erfolgreich waren Sie bisher in Ihrem Leben? Sagen Sie es bitte nach dieser L eiter hier. Es geht so: Null ( -5 ) bedeutet überhaupt nicht erfolgreich und 10 ( + 5) bedeutet, Sie waren bisher außerordentlich erfolgreich. Welche Zahl nehmen Sie?" EXPERIMENT 2. Self-reports. " Was meinen Sie: Wie glücklich war Ihre eigene Kindheit?" -" unglücklich" (0 or -5 ) to " glücklich" (10 or +5) " nicht so glücklich" (0 or -5 ) to " glücklich" (10 or +5). " Wie erfolgreich waren Sie im Leben?" -" erfolglos" (0 or -5 ) to " sehr erfolgreich" (10 or +5) " nicht so erfolgreich" (0 or -5 ) to " sehr erfolgreich" (10 or +5). Proxy reports. " Was meinen Sie: Wie glücklich war die Kindheit Ihrer Eltern?" -" unglücklich" (0 or -5 ) to " glücklich" (10 or +5) " nicht so glücklich" (0 or -5 ) to " glücklich" (10 or + 5). " Wie erfolgreich waren Ihre Eltern im Leben?" -" erfolglos" (0 or -5 ) to " sehr erfolgreich" (10 or +5) " nicht so erfolgreich" (0 or -5 ) to " sehr erfolgreich" (10 or +5). Scenario /. Der BWL-Student Peter K. wurde gefragt, wie erfolgreich er im BWL-Vordiplom war. Er machte die folgende Angabe: " Wie erfolgreich waren Sie in Ihrem Vordiplom?" -" nicht so erfolgreich" (0 or -5 ) to " sehr erfol greich" (10 or +5). Scenario 2. Manfred D. wurde gefragt, wie zufrieden er mit seiner Gesund heit ist. E r machte die folgende Angabe; " Wie zufrieden sind Sie mit Ihrer Gesundheit?"-"nicht so zufrieden" (0 or -5 ) to " sehr zufrieden" (10 or + 5). Comprehension and Semantic Flexibility Journal o f Verbal Learning and Verbal Behavior Human Cognition: Learning, Understanding, and Remembering Language Use and Language Users Handbook o f Social Psychology The Optimal Number of Response Alternatives for a Scale: A Review Journal o f Marketing Research Attitude and Opinion M easurement Handbook o f Social Psychology Logic and Conversation Syntax and Semantics: Speech Acts Telefoninterviews in Sozialforschung und Sozialpsychologie: Ausgewählte Probleme der Stichprobengewinnung, Kontaktierung und Versuchsplanung Kölner Zeitschrift fü r Soziologie und Sozialpsychologie The Impact of Verbal Labeling of Response Alternatives and Branching on Attitude Measurement Reliability in Surveys Psychometric Theory The Nature and Measurement of Meaning Psychological Bulletin Empirische Studien zur Entwicklung von Antwortskalen für die sozialwissenschaftliche Forschung Zeitschrift fü r Sozialpsychologie Contrast Analysis Impression Management Ordinary Questions, Survey Questions, and Policy Questions Public Opinion Quarterly Assessing Frequency Reports of Mundane Behaviors: Contributions of Cognitive Psychology to Questionnaire Construction Research M ethods in Personality and Social Psychology W hat Response Scales May Tell Your Respondents: Informative Functions of Response Alternatives Social Information Processing and Survey Methodology C ontext EfFects in Attitude Surveys: Applying Cognitive Theory to Social Research European Review o f Social Psychology Psychological Sources o f Response Effects in Surveys: The Impact of Administration Mode Special Issue: Cognitive Aspects of Survey Methodology Communicative Influences in Standardized Question Situations: The Case of Implicit Collaboration Language and Social Cognition The Salience o f Comparison Standards and the Activation of Social Norms: Consequences for Judgments of Happiness and Their Communication British Journal o f Social Psychology Die Wirkung von Antwortvorgaben bei Kalegorialskalen ZUMA-Nachrichten Determinants of Scale Responses: Labe] verus Position Journal o f Marketing Research Role of Sentence Context in the Encoding of Trait Descriptors Journal o f Personality and Social Psychology Schuessler Methoden der Analyse zeitbezogener Daten. Vortragsskripten der ZUMA-Arbeitstagung vom 25 ALLBUS Manfred Küchler Verallgemeinerte Lineare Modelle in der empirischen Sozialforschung Glenn R. Carroll Dynamic analysis of discrete dependent variables: A didactic essay Manfred Küchler Zur Messung der Stabilität von Wählerpotentialen Manfred Küchler Zur Konstanz der Recallfrage Rolf Porst " ALLBUS 1982" -Systematische Variablenübersicht und erste Ansätze zu einer Kritik des Fragenprogramms Peter Ph, Mohler SAR -Simple AND Retrieval mit dem Siemens-EDT-Textmanipulationsprogramm Cornelia Krauth Vergleichsstudien zum ALLBUS Bernd Wegener Tvo approaches to the analysis of judgments of prestige: Interindivi dual differences and the general scale Rolf Porst Synopse der ALLBUS-Variablen. Die Systematik des ALLBUS-Fragenprogramms und ihre inhaltliche Ausgestaltung Mohler Qualshop (ZUMA-Arbeitstagung zum Datenmanagement bei qualitativen Erhebungsverfahren") -Sammlung von Arbeitspapieren und -berichten Klaus Zeifang Die Test-Retest-Studie zum ALLBUS 1984 -Abschlußbericht Barbara Erbslöh, Michael Wiedenbeck Methodenbericht zum "ALLBUS 1986" Norbert Schwarz, Julia Bienias What Mediates the Impact of Response Alternatives on Behavioral Reports Günter Rothe Ein Ansatz zur Konstruktion inferenzstatistisch verwertbarer Indices Norbert Schwarz Assessing frequency reports of mundane behaviors: Contributions of cognitive psychology to questionnaire construction Norbert Schwarz, B. Scheuring (sub.) Judgments of relationship satisfaction: Inter-and intraindividual comparison strategies as a function of questionnaire strcuture Rolf Porst, Michael Schneid Ausfälle und Verweigerungen bei Panelbefragungen -Ein Beispiel -Cornelia Züll SPSS-X. Anmerkungen zur Siemens BS2000 Version Michael Schneid + Datenerhebung am PC -Vergleich der Interviewprogramme Bettina Scheuring Die Vergleichsrichtung bestimmt das Ergebnis von Vergleichsprozessen: Ist -Idealdiskrepanzen in der Partnerwahrnehmung Norbert Allgemeine Bevölkerungsumfrage der Sozialwissenschaften -ALLBUS 1988 -Norbert Schwarz Feelings as Information: Informational and Motivational Functions of Affective States Günter Rothe Jackknife and Bootstrap: Resampling-Verfahren zur Genauigkeitsschätzung von Parameterschätzungen Herbert Bless, Gerd Bohner, Norbert Schwarz und Fritz Strack Happy and Mindless? Moods and the Processing of Persuasive Communications Gerd Bohner Dominance and Cooperation Robert Huckfeldt Noncompliance and the Limits of Coercion: The Problematic Enforcement of Unpopular Laws Peter Ph. Mohler, Katja Frehsen und Ute Hauck CUI -Computerunterstützte Inhaltsanalyse Grundzüge und Auswahlbibliographie zu neueren Anwendungen Cornelia Züll, Peter Ph Rainer Mathes Modulsystem und Netzwerktechnik. Neuere inhaltsanalytische Verfahren zur Analyse von Kommunikationsinhalten Margrit Rexroth Der Interviewer im Pretest. Evaluation der Interviewerleistung und Entwurf eines neuen Pretestkonzepts Henrik Tham Crime as a Social Indicator Elisabeth Noelle-Neumann Response Order Effects in Dichotomous Questions: The Impact of Administration Mode Norbert Schwarz Elisabeth Noelle-Neumann, Thomas Münkel Response Order Effects in Long Lists: Primacy, Recency, and Asymmetric Contrast Effects Wolfgang Meyer Umweltberichterstattung in der Bundesrepublik Deutschland Michael Braun R a tin g S c a le s: Numeric Values May Change the Meaning o f Scale L ab e ls Norbert Schwarz Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany Bärbel Knäuper Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany Hans-J H Ip Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany Norbert Schw Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany Arz Bärbel Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany Knäuper Hans-J Hippler Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany Elisabeth Noelle-Neumann Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany Leslie Clark Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany George We Thank Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany Tory Bishop Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany Tracy Higgins Norbert Schwarz ZUMA P.O. Box 12 21 55 D-6800 Mannheim Germany R a tin g S c a le s: Numeric Values May Change the Meaning o f Scale L ab e ls GROBID - A machine learning software for extracting information from scholarly documents p ie r E lis a b e th Noelle-Neumann L e s lie C lark ZUMA-Arbeitsbericht Nr. 90/10 Zentrum für Umfragen Methoden und Analysen e.V. (ZUMA) Postfach 12 21 55 6800 Mannheim 1 The enclosed reprint Schw arz N . K nauper B . H ippier H . J . Noelle-Neumann E . & C lark F. R ating scales: Num eric values may change the m eaning of scale labels. Public O pinion Q uarterly. 1991 55 570-582 Seit Juli 1983 sind die ZüMA-Arbeitsberichte in zwei Reihen aufgeteilt: Die ZDMA-Arbeitsberichte (neue Folge) haben eine hausinterne Begutachtung durchlaufen und werden vom Geschäftsführenden Direktor zusam men mit den übrigen Wissenschaftlichen Lei tern herausgegeben. Die Berichte dieser Reihe sind zur allgemeinen Weitergabe nach außen bestimmt. Die Z U M A -T e c h n i s c h e n Berichte dienen zur hausinternen Kommunikation bzw. zur Unter richtung externer Kooperationspartner. Sie sind nicht zur allgemeinen Weitergabe b e stimmt. A bstract Three experiments indicate that the numeric values provided as part of a rating scale may influence respondents' interpretation of the endpoint labels. In experiment I, a represen tative sample of German adults rated their success in life along an ll.-point rating scale, with the endpoints labeled " not at all successful" and " extremely successful." When the numeric val ues ranged from 0 (" not at all successful" ) to 10 (" extremely successful" ), 34 percent of the respondents endorsed values be tween 0 and 5. However, only 13 percent endorsed formally equivalent values between -5 and 0, when the scale ranged from -5 (" not at all successful" ) to + 5 (" extremely successful" ). Experiment 2 provided an extended conceptual replication of this finding, and experiment 3 demonstrates that recipients of a re spondent's report draw different inferences from formally equiva lent but numerically different values. In combination, the findings indicate that respondents use the numeric values to disambiguate the meaning of scale labels, resulting in different interpretations and, accordingly, different subjective scale anchors. Rating scales with labeled endpoints are probably the most widely used measurement instrument in social and psychological research. Leaving some concerns about their psychometric properties aside (see Nunnally 1978) , the use of these scales does not seem to be very controversial (see Dawes and Smith [1985] for a careful discussion of their general properties and for empirical and psychological justifica tions for their use). In general, 7-point scales seem to be best in terms o f reliability, percentage of undecided respondents, and respondents' ability to discriminate between the scale values (e.g., Cox 1980) . Thus, seven plus or minus two is the usual recommendation. Moreover, re spondents are able to use rating scales consistently, even in telephone interviews without visual aids (e.g., Hormuth and Bruckner 1985) . In addition, verbal rating scales, which provide a label for each scale point, have been found to be more reliable than scales that provide labels for the endpoints only (Krosnick and Berent 1990) . Finally, re searchers have observed that the terms used to label the endpoints, or to designate the separate values of verbal rating scales, may affect the obtained distribution (e.g., Rohrmann 1978; Wegner, Faulbaum, and Maag 1982; Wildt and Mazis 1978) . This suggests that respondents pay close attention to the meaning of the labels provided to them, much as one would hope. Whereas the number of scale points, the inclusion or omission of a neutral point, and the choice of scale labels have received considerable attention in the literature (see Dawes and Smith 1985) , the specific numeric values provided have, to our knowledge, not been the topic o f theoretical analysis and empirical investigation. Apparently, re searchers assume that, for example, a 7-point scale that ranges from " 1" to " 7" is equivalent to a 7-point scale that ranges from " -3 " to " + 3 ," as long as the same endpoint labels are provided. In the present article, we will question this assumption. Drawing on survey data from the Allensbach archive, we will first demonstrate that the specific nu meric values provided in a rating scale can have a dramatic impact on the obtained results. We will then discuss different underlying pro cesses and will test their viability in laboratory experiments. Experiments 1 and 2: The Impact of Numeric Values SURVEY DATA In July 1988, the Allensbach Institute conducted a split-ballot experi m ent as part o f a representative survey of the adult population of the Federal Republic of Germany, using a quota sample of 1,032 respon dents, based on an intersection o f region, sex, and age. In face-to-face interviews, all respondents were asked to report how successful they have been in life, along an 11-point rating scale with labeled endpoints. The rating scale was presented on a show card, in the form of a ladder, and ranged either from 0 = " not at all successful" to 10 = " extremely successful," or from -5 = " not at all successful" to + 5 = " ex tremely successful." Respondents were randomly assigned to one of the two numeric value conditions, and the question read: " How suc cessful have you been in life, so far? Please use this ladder to tell me. This is how it works: 0 [ -5 , respectively] means 'not successful at all,' and 10 [ + 5, respectively] means that you were 'extremely successful.' Which number do you choose?" As shown in table 1, the numeric values provided on the respective rating scales had a pronounced impact on the obtained reports, x2 (10) = 105.1, p < .001. For example, whereas 34 percent of the respon dents who were given the 0-10 scale endorsed a value between 0 and 5, only 13 percent of the respondents who were given the -5 to + 5 scale endorsed one of the presumably equivalent values between -5 and 0 (z = 8.11, p < .0001, for this contrast; cf. Rosenthal and Rosnow [1985] ). Coding both scales from 0 to 10, this pattern results in mean ratings of M = 6.4 on the 0-10, but M = 7.3 on the -5 to + 5 version o f the scale. In addition, an inspection of the distributions along both scales indicates that the responses are dislocated toward the high end of the -5 to + 5 scale, as compared to the 0-10 scale. This is also reflected in markedly different standard deviations, SD = 1.03 and .56 for the 0-10 and -5 to + 5 scale, respectively. These findings may reflect either that respondents hesitated to assign themselves a negative score with regard to their success in life or that the numeric values provided on the scale influenced respondents' interpretation of the endpoint labels. Specifically, " not at all success ful" may be interpreted as referring to the absence of success if com bined with a numeric value of 0, but as referring to the presence of explicit failure if combined with a numeric value of -5. Before we address these possibilities in more detail, however, it is informative to consider the findings of an extended replication. EXTENDED REPLICATION To test the reliability of the above finding, we conducted a conceptual replication with a sample of 101 students at a German university. In a self-administered questionnaire, half of the respondents were asked to report, along 11-point rating scales, how successful they have been in Note.-Percentages rounded. Data based on a quota sample of 1,032 German adults, randomly assigned to conditions. The question read, " How successful have you been in life, so far?" with scale endpoints labeled " not at all successful" (0 or -5) and " extremely successful" (10 or +5). See Appendix for full wording, x? (10) = 105.1, p < .0001. life and how happy a childhood they had. The other half of the respon dents rated the success and childhood happiness of their parents along the same scales, thus extending the present study to proxy reports. The key manipulations consisted in variations of the numeric values and the type of endpoint label provided. Specifically, the scale ranged either from 0 to 10 or from -5 to + 5 , replicating experiment 1. In addition, the endpoints were either labeled " unsuccessful" and " very successful" (or, " unhappy" and " very happy," respectively) or they were labeled " not so successful" and " very successful" (or, " not so happy" and " very happy," respectively). The latter wordings were introduced to explore the impact of differentially ambiguous scale la bels, and the exact wordings of the questions are given in the Appen dix. In sum, these manipulations resulted in a 2 (numeric values) x 2 (scale labels) x 2 (self-or proxy reports) factorial between subjects design, which was analyzed by analysis of variance. The obtained data provide a robust replication of the previously observed impact of numeric values on self-as well as proxy reports of success and childhood happiness. Coding both scales from 0 to 10, respondents reported higher success in life for themselves (M = 7.38), as well as for their parents (M = 8.13), along the -5 to + 5 scale than along the 0-10 scale (M = 5.96 and 7.04 for self and parents, respectively), resulting in a pronounced main effect of numeric values (^[1,93] = 16.21, p < .001). Similarly, respondents reported higher childhood happiness along the -5 to + 5 scale than along the 0-10 scale, again both for themselves (M = 8.08 and 6.17 for the -5 to + 5 and the 0-10 scale, respectively), as well as for their parents (M = 7.04 and 5.38, respectively), with F(l,93) = 5.02, p < .03 for the main effect of numeric values. The remaining effects that reached significance were of little theoret ical interest and reflected that respondents perceived their parents as having been more successful in life, but less happy during their child hood. N either the self/proxy variable, nor the ambiguity of the scale labels, however, moderated the impact of the numeric values, all Fs < 1 for the interaction terms. DISCUSSION As alluded to above, two processes, which are not mutually exclusive, may underlie the consistently observed impact of numeric values. As*, a first possibility, respondents may hesitate to assign themselves a negative score, reflecting self-presentation concerns. Although we can not rule out an impact of self-presentation concerns, some aspects of our findings suggest that self-presentation may not be the key factor that drives the observed phenomenon. First, respondents in experi-ment 2 provided their reports anonymously in a self-administered ques tionnaire, thus reducing the potential impact o f social desirability (cf. Strack et al. 1990 ). Moreover, self-presentation considerations suggest that the impact of numeric values should be more pronounced for selfthan for proxy reports, which was not the case. Finally, why should respondents hesitate to endorse a negative scale value, unless negative scale values communicate a different meaning than positive values? These considerations suggest that respondents may have used the numeric values provided as part of the rating scale to interpret the meaning of the labels. As Woll et al. (1980, p. 60) note, " Even the m ost unambiguous words show a range of meaning, or a degree of 'semantic flexibility,' . . . that is constrained by the particular context in w hich these words occur." (See also Barclay et al. [1974] and Bransford [1979] for a more general discussion.) F or example, respondents who are asked to rate their success in life need to determine what the researcher means by " unsuccessful," " not so successful," and the like. Does that term refer to the absence of remarkable successes, or does it refer to the presence of failure? Depending on how respondents interpret the term , respondents who have, for example, neither experi enced particular successes nor particular failures but have done " al right" in life may choose very different scale values. Thus, the present findings may reflect that the numeric values changed the meaning of the endpoint labels, resulting in different responses, much as has been observed in studies that explicitly varied the wording of the scale labels (e.g., Rohrmann 1978; Wegner, Faulbaum, and Maag 1982; Wildt and Mazis 1978) . According to this account, the finding that the different labels used in experiment 2 did not result in a differential impact of num eric values presumably reflects that the terms " unsuccessful" or " unhappy" are as ambiguous as the terms " not so successful" or " not so happy," which we introduced on the intuitive assumption that they may be more ambiguous than the former. N ote that this interpretation does not require self-presentation con siderations as a necessary condition to account for the observed find ings. N evertheless, it implies that the impact of numeric values may increase with increasing self-presentation concerns, because these concerns are known to increase respondents' reluctance to endorse values with unfavorable implications (Schlenker 1980) . Accordingly, semantic interpretation and self-presentation processes are not mutu ally exclusive. Rather, the operation of the latter requires the operation o f the former in the first place. Experiment 3 was designed to provide a more direct test of the shift in meaning hypothesis offered here, under conditions that render self-presentation concerns very unlikely. To accomplish this, respon dents were given transcripts of another person's reports along rating scales with different numeric values and were asked to draw inferences about the target person. If the numeric values change the meaning of the scale labels, this should be reflected in different inferences about the target. Experiment 3: Inferences Based on Numeric Values Twenty-two students at a German university participated in a study that was purportedly concerned With the accuracy of the inferences that people can draw about others on the basis of minimal information. All participants received a short description of two target persons and a transcript of these persons' responses to a survey question (see Ap pendix for question wordings). The first target person reported his health satisfaction along an 11-point rating scale, ranging from " dis satisfied" (0 or -5) to " very satisfied" (10 or +5). Depending on the numeric values provided on the scale, this person had allegedly checked a value of -4, or a numerically equivalent value of 1. Simi larly, the second target person had allegedly rated his success on aca demic exams, again along 11-point rating scales, with the endpoints labeled " not so successful" (0 or -5 ) and " successful" (10 or +5), and had checked a minus three or a two. Subjects were randomly assigned to one o f the two numeric values conditions constituted by the scale along which the target persons had allegedly given their reports. As dependent variables, subjects were asked, in an open response format, to estimate how frequently the first target person had to see a doctor during the last month and how often the second target person had to repeat an exam due to failure. If the numeric values affect subjects' interpretation of the endpoint labels of the rating scales along which the target persons gave their reports, one should expect that they draw more extreme inferences if the numeric values range from -5 to + 5 rather than from 0 to 10. The findings support this assumption. Specifically, subjects esti mated that the first target person had to see a doctor twice as often when he checked a minus four on the -5 to + 5 scale (M -2.2), than when he checked one on the 0-10 scale (M = 1.0), F(l,20) = 4.86, p < .04. Similarly, they assumed that the second target person had failed on twice as many exams (M = 1.73) when he checked a minus three on the -5 to + 5 scale, than when he checked a 2 on the 0-10 scale (M = .91), F(l,20) = 5,63, p < .03. In combination, these findings indicate that respondents drew more extreme inferences from reports given along a -5 to + 5 scale than from formally identical reports given along a 0-10 scale. This is consistent with the assumption that the numeric values presented on the rating scales affected respondents' interpretation of the meaning of the report. Conclusions We conclude from the reported findings that respondents may use the numeric values provided on a rating scale to disambiguate the meaning of scale labels. If the numeric values range from 0 to 10, as was the case in our studies, their very structure seems to suggest that the researcher is interested in the absence or presence of the attribute to which the scale pertains, that is, success or happiness in the above examples. If the numeric values range from -5 to + 5 , including a zero at the midpoint, their structure seems to suggest that the absence of the attribute corresponds to zero, whereas the negative values refer to the presence of its opposite, that is, failure or unhappiness in the above examples. In more general terms, scales that provide a contin uum from negative to positive values may indicate that the researcher has a bipolar conceptualization of the respective dimension in mind, whereas scales that present only positive values may indicate a unipo lar conceptualization. If so, the choice of numeric values may either facilitate or dilute the polarity implications of the endpoint labels that are provided to respondents. Accordingly, researchers may be well advised to match the numeric values that they provide to respondents with the intended conceptualization of the underlying dimension as uni-or bipolar. W hereas this point may seem obvious in the studies reported above-where the combination of an apparently unidirectional verbal scale (" not at all" to " extremely successful" ) with a bipolar numeric scale may strike some readers as awkward-the strong effects ob tained here suggest that the impact of numeric values should be even more pronounced when the polarity of the verbal labels themselves is more ambiguous. For example, political issue questions in the National Election Studies are intended to introduce two opposing positions, reflecting a liberal and a conservative viewpoint. Thus, respondents may be asked, " Some people believe that we should spend much less for defense. Others feel that spending should be greatly increased. W here would you place yourself on this scale? Greatly decrease de fense spending (1); greatly increase defense spending (7)." The current findings suggest that using values from -3 to + 3 , rather than from 1 to 7, may help to emphasize the intended liberal-conservative bipolar ity of these options. Similarly, our reasoning suggests that scales that follow the format of semantic differentials (Osgood 1952 ) by providing polar opposites should make use of numeric values that range from negative to positive, as is frequently but not always the case. In con trast, scales that are intended to assess the intensity of a single attri bute should follow a zero-to-positive-values format to emphasize that the question pertains to the absence or presence of this specific attri bute, rather than the presence of its opposite. N ote, however, that the use of different scale formats may affect the obtained item variance. The conditions under which this is the case are not well understood, however, and further research is needed to explore this issue. Suppose that most people are more likely to experi ence success rather than failure in life. If so, using a -5 to + 5 format restricts the meaningful response alternatives for most respondents to the positive half of the scale, resulting in reduced item variance relative to a 0-10 format. Suppose, on the other hand, that many people would experience more failure than success in life. If so, a 0-10 format would provide fewer meaningful response alternatives for these respondents than a -5 to + 5 format, resulting in reduced item variance in the former case. This suggests that the choice of a scale format should be based on researchers' knowledge about the relevant distribution to avoid undesirable restrictions in item variance. Finally, the present reasoning bears on the comparability of data obtained under different administration modes (see for a general conceptualization of mode effects). The use of numeric values is most prevalent in face-to-face interviews and selfadministered surveys, whereas the increasing use of an unfolding for mat in telephone interviews avoids the presentation of numeric values. Based on the present findings, we hypothesize that data obtained in an unfolding format are more compatible with verbal rating scales that present labels for each scale point, or with scales that present unnum bered boxes or similar devices, than with scales that use numeric val ues. These considerations echo the general insight that the absolute values, or marginals, obtained in response to any survey question are difficult to interpret (see Schuman 1986) . To what extent changes in scale format affect the obtained relationships or changes over time, on the other hand, needs to be explored in future research. From a theoretical perspective, the present findings support the gen eral conclusions that we have drawn from related research into the impact of response alternatives on frequency reports of mundane behaviors and related judgments (see Schwarz 1990; Schwarz and Hippier 1987) . Far from being " neutral measurement devices," the response alternatives that are provided to respondents do constitute a source o f information that respondents actively use in determining their task and in constructing a reasonable answer. While survey meth odologists have traditionally focused on the information that is pro vided by the wording of the question, we need to pay equal attention to the information that is conveyed by apparently formal features of the questionnaire. Respondents apply many of the rules that govern the conduct o f conversation in everyday life (cf. Clark 1985; Grice 1975) to the survey interview. In doing so, they assume that every contribution to the ongoing conversation is relevant and meaningful. In the survey interview, the researcher's contributions include the re sponse alternatives, the numeric values of rating scales, and the order ing of questions, as well as other features of questionnaire design. (See Schwarz and Hippier [in press ], , and Strack and Schwarz [in press ] for reviews and theoretical analyses.) The anal ysis o f the informational functions of apparently formal features of questionnaire design is therefore a key task in the collaboration of cognitive psychologists and survey researchers. been on your exams?" -" not so successful" (0 or -5) to " very successful" (10 or +5). Depending on type of scale, Peter K. allegedly checked " 2" or " -3 . " Scenario 2. Manfred D. was asked how satisfied he is with his health. He gave the following response: " How satisfied are you with your health?" -" not so satisfied" (0 or -5) to " very satisfied" (10 or +5). Depending on type of scale, Manfred D. allegedly checked " 1" or " -4 . " German Question Wording EXPERIMENT 1 " Wie erfolgreich waren Sie bisher in Ihrem Leben? Sagen Sie es bitte nach dieser L eiter hier. Es geht so: Null ( -5 ) bedeutet überhaupt nicht erfolgreich und 10 ( + 5) bedeutet, Sie waren bisher außerordentlich erfolgreich. Welche Zahl nehmen Sie?" EXPERIMENT 2. Self-reports. " Was meinen Sie: Wie glücklich war Ihre eigene Kindheit?" -" unglücklich" (0 or -5 ) to " glücklich" (10 or +5) " nicht so glücklich" (0 or -5 ) to " glücklich" (10 or +5). " Wie erfolgreich waren Sie im Leben?" -" erfolglos" (0 or -5 ) to " sehr erfolgreich" (10 or +5) " nicht so erfolgreich" (0 or -5 ) to " sehr erfolgreich" (10 or +5). Proxy reports. " Was meinen Sie: Wie glücklich war die Kindheit Ihrer Eltern?" -" unglücklich" (0 or -5 ) to " glücklich" (10 or +5) " nicht so glücklich" (0 or -5 ) to " glücklich" (10 or + 5). " Wie erfolgreich waren Ihre Eltern im Leben?" -" erfolglos" (0 or -5 ) to " sehr erfolgreich" (10 or +5) " nicht so erfolgreich" (0 or -5 ) to " sehr erfolgreich" (10 or +5). EXPERIMENT 3 Scenario /. Der BWL-Student Peter K. wurde gefragt, wie erfolgreich er im BWL-Vordiplom war. Er machte die folgende Angabe: " Wie erfolgreich waren Sie in Ihrem Vordiplom?" -" nicht so erfolgreich" (0 or -5 ) to " sehr erfol greich" (10 or +5). Scenario 2. Manfred D. wurde gefragt, wie zufrieden er mit seiner Gesund heit ist. E r machte die folgende Angabe; " Wie zufrieden sind Sie mit Ihrer Gesundheit?"-"nicht so zufrieden" (0 or -5 ) to " sehr zufrieden" (10 or + 5). T able I able • The Impact of Numeric Scale Values on Reports along Rating Scales 0-10 Scale -5 to + 5 Scale Scale Scale Value Percentage Cumulative Value Percentage Cumulative 0 -5 1 1 1 • * » , , , -4 * . * 1 2 2 2 -3 1 2 3 5 7 -2 1 3 4 7 14 -1 1 4 5 20 34 0 13 6 14 48 + 1 22 7 20 68 + 2 23 45 8 20 88 + 3 35 80 9 6 94 + 4 94 10 3 97 + 5 98 Undecided 3 100 Undecided 100 N 480 N 552 S o u r c e . -IfD 5007, Juli 1988. Appendix English Translations of Question Wordings EXPERIMENT 1 See text for question wording. EXPERIMENT 2 Self-reports. " What is your opinion: How happy was your own childhood?" -" unhappy" (0 or -5 ) to " happy" (10 or +5); or " not so happy" (0 or -5 ) to " very happy" (.10 or +5), respectively. " How successful have you been in life?"-"unsuccessful" (0 or -5 ) to " very successful" (10 or +5); or " not so successful" (0 or -5) to " very successful" (10 or +5). Proxy reports. " What is your opinion: How happy was the childhood of your parents?" -" unhappy" (0 or -5 ) to " happy" (10 or +5); or " not so happy" (0 or -5 ) to " very happy" (10 or +5), respectively. " How successful have your parents been in life?"-" unsuccessful" (0 or -5) to " very successful" (10 or +5); or " not so successful" (0 or -5 ) to " very successful" (10 or +5). EXPERIMENT 3 Scenario 1. Peter K., an MBA student, was asked how successful he was on his exams. He gave the following response: " How successful have you ALLBUS 1984 ALLBUS , 1986 ALLBUS , 1988 Comprehension and Semantic Flexibility J R Barclay J D Bransford J J Franks N S Mccarrell K E Nitsch Journal o f Verbal Learning and Verbal Behavior 13 Human Cognition: Learning, Understanding, and Remembering J D Bransford Wadsworth Belmont, CA Language Use and Language Users H H Clark Handbook o f Social Psychology G. Lindzey and E. Aronson New York 2 Random House The Optimal Number of Response Alternatives for a Scale: A Review E P Cox Journal o f Marketing Research 17 Attitude and Opinion M easurement R M Dawes T Smith Handbook o f Social Psychology G. Lindzey and E. Aronson New York 2 Random House Logic and Conversation H P Grice Syntax and Semantics: Speech Acts P. Cole and J. L. Morgan New York Academic Press 3 Telefoninterviews in Sozialforschung und Sozialpsychologie: Ausgewählte Probleme der Stichprobengewinnung, Kontaktierung und Versuchsplanung S E Hormuth E Brückner Kölner Zeitschrift fü r Soziologie und Sozialpsychologie 37 The Impact of Verbal Labeling of Response Alternatives and Branching on Attitude Measurement Reliability in Surveys J A Krosnick K M Berent American Association for Public Opinion Research Lancaster, PA Paper presented at the annual meeting of the Psychometric Theory J C Nunnally McGraw-Hill New York The Nature and Measurement of Meaning C Osgood Psychological Bulletin 49 Empirische Studien zur Entwicklung von Antwortskalen für die sozialwissenschaftliche Forschung B Rohrmann Zeitschrift fü r Sozialpsychologie 9 R Rosenthal R L Rosnow Contrast Analysis Cambridge Cambridge University Press Impression Management B R Schlenker Brooks/Cole Belmont. CA Ordinary Questions, Survey Questions, and Policy Questions H Schuman Public Opinion Quarterly 50 Assessing Frequency Reports of Mundane Behaviors: Contributions of Cognitive Psychology to Questionnaire Construction N Schwarz Research M ethods in Personality and Social Psychology C. Hendrick and M. S. Clark CA Sage W hat Response Scales May Tell Your Respondents: Informative Functions of Response Alternatives N Schwarz H.-J Hippier ; P Biemer Social Information Processing and Survey Methodology H.-J. Hippier, N. Schwarz, and S. Sudman New York Wiley ---------. In press M easurement Error in Surveys C ontext EfFects in Attitude Surveys: Applying Cognitive Theory to Social Research N Schwarz F Strack European Review o f Social Psychology M. Hewstone and W. Stroebe Chichester Wiley 2 Psychological Sources o f Response Effects in Surveys: The Impact of Administration Mode N Schwarz F Strack H.-J Hippier G-Bishop Special Issue: Cognitive Aspects of Survey Methodology J. Jobe and E. Loftus 5 Communicative Influences in Standardized Question Situations: The Case of Implicit Collaboration F Strack N Schwarz Language and Social Cognition Semin. Beverly Hills, CA Sage In press The Salience o f Comparison Standards and the Activation of Social Norms: Consequences for Judgments of Happiness and Their Communication F Strack N Schwarz B Chassein D Kern D Wagner British Journal o f Social Psychology 29 Die Wirkung von Antwortvorgaben bei Kalegorialskalen B Wegner F Faulbaum G Maag ZUMA-Nachrichten 10 Determinants of Scale Responses: Labe] verus Position A R Wildt M B Mazis Journal o f Marketing Research 15 Role of Sentence Context in the Encoding of Trait Descriptors S B Woll D G Weeks C L Fraps J Pendergrass M A Vanderplas Journal o f Personality and Social Psychology 39 Schuessler Methoden der Analyse zeitbezogener Daten. Vortragsskripten der ZUMA-Arbeitstagung vom 25 Gerhard Arminger Willibald Nagl F Karl 09. -05.10.79 Erika Brückner Hans-Peter Kirschner Rolf Porst Peter Prüfer ALLBUS Peter Schmidt Methodenbericht zum Manfred Küchler Verallgemeinerte Lineare Modelle in der empirischen Sozialforschung Glenn R. Carroll Dynamic analysis of discrete dependent variables: A didactic essay Manfred Küchler Zur Messung der Stabilität von Wählerpotentialen Manfred Küchler Zur Konstanz der Recallfrage Rolf Porst "ALLBUS 1982" -Systematische Variablenübersicht und erste Ansätze zu einer Kritik des Fragenprogramms Peter Ph, Mohler SAR -Simple AND Retrieval mit dem Siemens-EDT-Textmanipulationsprogramm Cornelia Krauth Vergleichsstudien zum Manfred Küchler Thomas P Wilson Don H Zimmerman Integration Von Qualitativen Und Quantitativen Forschungsansätzen Gerhard Arminger Horst Busse ALLBUS Bernd Wegener Tvo approaches to the analysis of judgments of prestige: Interindivi dual differences and the general scale Rolf Porst Synopse der ALLBUS-Variablen. Die Systematik des ALLBUS-Fragenprogramms und ihre inhaltliche Ausgestaltung Werner Hagstotz Hans-Peter Kirschner Rolf Porst im ALLBUS 1980 und ALLBUS 1982 Peter Prüfer Methodenbericht zum "ALLBUS Mohler Qualshop (ZUMA-Arbeitstagung zum Manfred Küchler Peter Ph Datenmanagement bei qualitativen Erhebungsverfahren") -Sammlung von Arbeitspapieren und -berichten I + Ii Bernd Teil Wegener Gibt es Sozialprestige? Konstruktion und Validität der Magnitude-Prestige-Skala Klaus Zeifang Die Test-Retest-Studie zum ALLBUS 1984 -Abschlußbericht Barbara Erbslöh, Michael Wiedenbeck Methodenbericht zum "ALLBUS 1986" Norbert Schwarz, Julia Bienias What Mediates the Impact of Response Alternatives on Behavioral Reports Günter Rothe Ein Ansatz zur Konstruktion inferenzstatistisch verwertbarer Indices Ute Hauck Norbert Schwarz Assessing frequency reports of mundane behaviors: Contributions of cognitive psychology to questionnaire construction Norbert Schwarz, B. Scheuring (sub.) Judgments of relationship satisfaction: Inter-and intraindividual comparison strategies as a function of questionnaire strcuture Rolf Porst, Michael Schneid Ausfälle und Verweigerungen bei Panelbefragungen -Ein Beispiel -Cornelia Züll SPSS-X. Anmerkungen zur Siemens BS2000 Version Michael Schneid + Datenerhebung am PC -Vergleich der Interviewprogramme interv " und "THIS" Norbert Schwarz ; Schwarz George F Bishop -J Hans Hippler Bettina Scheuring Die Vergleichsrichtung bestimmt das Ergebnis von Vergleichsprozessen: Ist -Idealdiskrepanzen in der Partnerwahrnehmung Norbert Fritz Strack Psychological Sources Of Response Effects in Self-Administered And Telephone Surveys Allgemeine Bevölkerungsumfrage der Sozialwissenschaften -ALLBUS 1988 -Norbert Schwarz Feelings as Information: Informational and Motivational Functions of Affective States Günter Rothe Jackknife and Bootstrap: Resampling-Verfahren zur Genauigkeitsschätzung von Parameterschätzungen Herbert Bless, Gerd Bohner, Norbert Schwarz und Fritz Strack Happy and Mindless? Moods and the Processing of Persuasive Communications Gerd Bohner Michael Braun Reiner Trometer Michael Viedenbeck Methodenbericht Norbert Schwarz und Stefan E. Hormuth Die Stimmungs-Skala: Eine deutsche Version des "Mood Survey Dominance and Cooperation Robert Huckfeldt Noncompliance and the Limits of Coercion: The Problematic Enforcement of Unpopular Laws Peter Ph. Mohler, Katja Frehsen und Ute Hauck CUI -Computerunterstützte Inhaltsanalyse Grundzüge und Auswahlbibliographie zu neueren Anwendungen Cornelia Züll, Peter Ph Ulrich Mueller Evolutionary Fundamentals of Social Inequality Norbert Schwarz, Brigitte Chassein, Dieter Kern Rolf Porst Methodische Probleme bei der Durchführung telefonischer Befragungen: Stichprobenziehung und Ermittlung von Zielpersonen, Ausschöpfung und Nonresponse Dirk Vagner The Salience of Comparison Standards and the Activation of Social Norms: Consequences for Judgments of Happiness and their Communication Jutta Kreiselmaier. Qualität der Daten Rainer Mathes Modulsystem und Netzwerktechnik. Neuere inhaltsanalytische Verfahren zur Analyse von Kommunikationsinhalten Margrit Rexroth Der Interviewer im Pretest. Evaluation der Interviewerleistung und Entwurf eines neuen Pretestkonzepts Jutta Kreiselmaier Peter Prüfer Henrik Tham Crime as a Social Indicator Elisabeth Noelle-Neumann Response Order Effects in Dichotomous Questions: The Impact of Administration Mode Norbert Schwarz -J Hans Hippier Elisabeth Noelle-Neumann, Thomas Münkel Response Order Effects in Long Lists: Primacy, Recency, and Asymmetric Contrast Effects Wolfgang Meyer Umweltberichterstattung in der Bundesrepublik Deutschland Michael Braun Ulrich Mueller Expanding the Theoretical and Methodological Framework of Social Dilemma Research Hans-J. Hippier, Norbert Schwarz Reiner Trometer ALLBUS Bibliographie (8. Fassung, Stand: 30.6. 1989
(xsd:string)
|
![[This page as RDF]](https://data.gesis.org/gesiskg/static/rdf-icon.gif){kind=icon}