1. Number of Scale Points

2. 3 Points How would you say this magazine has impacted your overall opinion of actors? It has made my opinion more favorable It has had made no difference It has made my opinion less favorable

3. 4 Points Now I want to ask you about one of the articles in this magazine. You may or may not have already mentioned it. The article was titled “Riding Horses for Fun.” How much of this article did you read? All or most of it Some of it A little of it Did not read it

4. 5 Points Which of the following statements best describes your feelings about the magazine? I liked it very much I somewhat liked it I neither liked nor disliked it I somewhat disliked it I disliked it very much

5. 7 Points Overall, how satisfied were you with the meal you ate? Please enter a number from 1 to 7, where: 1 = Not at all satisfied, and 7 = Extremely satisfied

6. 10 Points Now I am going to read you some attributes that might describe this type of a magazine. Using a scale of 0 to 10, where the higher the number the better the rating, please tell me how you would rate this magazine on these attributes. The more you think the magazine has that attribute, the closer to 10 you should choose. Is easy to read Offers important information

7. 101-Points As you know, there were many people mentioned this past year as possible candidates for president by political parties. We would like to get your feelings towards some of these people. I have here a card on which there is something that looks like a thermometer. We call it a “feeling thermometer” because it measures your feelings towards these people. Here's how it works. If you don't feel particularly warm or cold toward a person, then you should place them in the middle, at the 50 degree mark. If you have a warm feeling toward a person or feel favorably toward him, you would give him a score somewhere between 50 degrees and 100 degrees depending on how warm your feeling is toward that person. On the other hand, if you don't feel very favorably toward a person -- if you don't care too much for him -- then you would place him somewhere between 0 degrees and 50 degrees. How would you rate …”

8. Theoretical Issues 1) Mapping: More is better

9. Mapping Attitudes To Response Scales Attitude Dimension Person A Person B  Response Dimension OpposeFavor ++ --

10. Mapping Attitudes To Response Scales Attitude Dimension Person A Person B  Response Dimension OpposeFavor  Attitude Dimension Person C  Response Dimension OpposeNeutralFavor ++ --

11. Mapping Attitudes To Response Scales Attitude Dimension Person A Person B  Response Dimension OpposeFavor  Attitude Dimension Person C  Response Dimension OpposeNeutralFavor  Attitude Dimension --++ --  Response Dimension Oppose a great deal Oppose somewhat Neutral Favor somewhat Favor a great deal

12. Theoretical Issues 1) Mapping: More is better 2) Information gain: More is better 3) Clarity of meaning: Too many become ambiguous 4) Satisficing: Middle alternatives on bipolar dimensions may be dangerous PREDICTION: Increasing precision up to a certain length, decreasing precision thereafter.

13. Data Quality Data Quality

14. Number of Rating Scale Points Completion Errors Longer better than shorter Time Longer greater than shorter

15. 5-7 points Completion Time Number of Scale Points

16. Number of Rating Scale Points Completion Errors Longer better than shorter Time Longer greater than shorter Reliability Bipolar: 7 pts. Unipolar: 5 pts.

18. Number of Rating Scale Points Completion Errors Longer better than shorter Time Longer greater than shorter Reliability Bipolar: 7 pts. Unipolar: 5 pts. Correlational Validity Bipolar: 7 pts. Inter-Rater Agreement Unipolar: 5 pts. Object Differentiation Context Effects Natural Discrimination Bipolar: 7 pts.

19. Include Middle Alternatives on Bipolar Dimensions?

20. Euro-Barometer 1992 13,000 Interviews 12 Countries BelgiumPortugal DenmarkGreat Britain East GermanyNorthern Ireland West GermanyThe Netherlands Greece Luxemburg SpainItaly FranceIreland

21. Science makes our way of life change too fast. StronglyAgreeNeitherDisagreeStrongly Agree AgreeDisagree nor Disagree Half the Respondents had the Middle Alternative Omitted

22. Substantive Middle Don’t SubstantiveDon’t Answer Alternative Know AnswerKnow 76.0% 14.6% 9.4% 88.6%11.4% Middle Alternative Offered Middle Alternative Omitted

23. MA Offered MA Omitted 48.6% 54.4% 27.4%34.2% Strongly Agree/Agree Strongly Disagree/Disagree Difference = 5.8%Difference = 6.8%

24. Dependent Variable Middle Alternative Omitted Middle Alternative Offered Predictor“Don’t Know”“Don’t Know” Middle Alternative Amount of Factually Correct -.34*** -.33*** -.01 Knowledge About Science Perceived Knowledgeability -.11 -.16*.02 About Science Understanding of Science -.14* -.21** -.24*** Interest in News About Science -.38*** -.25*** -.31*** Exposure to News About Science -.36*** -.40*** -.03 Number of Television Channels -.20** -.18*.17* Education -1.70* -1.92** -.25 Age.81***.93***-1.16*** Gender.34***.24**.00 _____________________ _____________________________________________________ R 2.25.25.02 N 6141 6152 6152 DK and MA Are Not The Same – Different Predictors

25. Random Error Variance Middle Alternative Offered Middle Alternative Omitted 1.18 1.23 Of the 19 questions, 15 showed statistically significant increases in random measurement error.

26. Include Middle Alternatives on Bipolar Dimensions!

27. Mean Shifts More points = More moderate ratings Magnitude Scaling Infinite number of scale points No increase in reliability or validity

28. Branching "Generally speaking, do you consider yourself to be a Republican, a Democrat, an Independent, or what?" "Would you call yourself a strong (Democrat/Republican) or a not very strong (Democrat/Republican)?" "Do you think of yourself as closer to the Democratic Party or the Republican Party?" Strong Republican Not very strong Republican Republican leaning independent Neutral independent Democratic leaning independent Not very strong Democrat Strong Democrat Branching improves reliability and predictive validity

29. Percent of People Providing Identical Responses During Both Interviews Question Form Fully Attitude Labeled Branching Difference Significance Party Identification 77.894.3 16.5 .05 Ideological Orientation 52.868.6 15.8 .18 Defense Spending 58.3 60.0 1.7 ns Central America 50.045.7 -4.3 ns Social Programs 50.073.5 23.5 .05 Combined 57.868.4 10.6 .04 N 36 35

30. Like Neither Dislike Like a lot Like a little Dislike a lot Dislike a little Lean toward like Lean toward dislike Don’t lean

31. Scale Point Labels

32. End Point Labels Overall, how satisfied were you with this magazine? Please enter a number from 1 to 7, where: 1 = Not at all satisfied, and 7 = Extremely satisfied

33. Numbers Plus Words on End Points How important is this issue to you personally? Not at all Important 123456 Extremely Important 7

34. Words on End Points and Midpoint On a 1-7 scale where 1 indicates “not at all” and 7 indicates “extremely well”, please indicate how well aspirin stops your headaches: Not at all 123 Moderately Well 456 Extremely Well 7

35. Verbal Labels on All Points How do you rate your overall health? Excellent Very Good Good Fair Poor

36. Verbal Labels on All Points How do you rate your overall health? Excellent Above average Average Below average Poor

37. Goals of Scale Points Labels Respondents find it easy to interpret the meanings of the scale points. Respondents believe the meanings of each scale point to be clear. All respondents interpret the meanings of the scale points identically. The labels differentiate respondents from one another validly as much as possible. The resulting scale include points that correspond to all points on the underlying construct’s continuum.

38. Advantages and Costs of Labeling Numbers alone seem ambiguous; Longer scales = more ambiguous Partial labeling: Labels may attract people What if labels are vague? What if labels are overly ‑ specific? Effort: * Unlabeled scale requires interpretation * Labeled scale requires reading and interpreting labels  Administration: Difficult to administer numbered scales over the phone

39. Dimensions with No Natural Metric Liking Importance Certainty Friendliness Ambitiousness etc.

40. Dimensions with No Metric (e.g., Liking)  Data Quality - Respondent SatisfactionMore  Less

41. Dimensions with No Metric (e.g., Liking)  Data Quality - Respondent SatisfactionMore  Less - ReliabilityMore  Less

42. Percent of Respondents Providing Identical Responses During Both Interviews Partially Fully Respondents Labeled Labeled Difference Significance Low Education37.750.3 12.6 <.05 (N=32)(N=40) Medium Education40.958.117.2 <.001 (N=98)(N=79) High Education48.657.0 8.4 <.01 (N=114)(N=119)

43. Dimensions with No Metric (e.g., Liking)  Data Quality - Respondent SatisfactionMore  Less - ReliabilityMore  Less - Correlational ValidityMore  Less - Inter-Rater AgreementMore  Less - Discriminant ValidityMore  Less - Object DifferentiationMore  Less - Question Order EffectsMore > Less

44. Selecting Labels More widely-spread end points yields higher reliability. Respondents presume equal spacing is intended, so it is best to choose labels reinforcing this. Scaling studies provide bases for selecting labels.

45. Visual Representation of Vague Quantifiers

47. Excellent 94 92 79 93 89 96 98 96 92 90 95 91 99 100 99 Very Good 81 70 85 78 77 87 88 83 79 82 80 Good 70 65 72 67 69 76 72 72 71 70 66 58 73 75 Fair 51 56 43 54 57 51 51 48 45 Poor 21 28 38 21 10 25 12 31 21 18 16 23 10 Very Poor 11 21 12 07 15 11 10 11 01 X

48. Excellent94 Very Good81 Good70 Fair51 Poor21 13 11 19 30

49. Excellent941 Very Good813 Good705 Fair518 Poor2113 132 112 19 3 305

50. Term X Extremely 1.69 1.69 2.91 1.59 2.07 1.19 1.74 1.65 1.64 1.35 1.55 1.26 Very 1.46 1.40 1.53 1.30 1.32 1.66 1.50 1.41 1.48 1.78 1.71 1.19 1.20 Quite1.19 1.24 1.04 1.28 0.97 1.33 1.14 1.30 1.20 1.08 1.33 Somewhat0.62 0.69 0.83 0.49 0.45 Moderately0.66 0.80 0.61 0.65 0.84 0.40 Slightly0.44 0.44 0.56 0.74 0.42 0.41 0.40 0.43 0.31 0.23