THE MEANING OF PARTY LABELS Ralph W. Bastedo and Milton Lodge
To determine the meaning(s) of the concepts "Republican," "Democrat," and "Independent," the most frequently cited attributes of each party label were scaled in terms of their semantic centrality. An analysis of the magnitude scale values demonstrates that the labels "Republican" and "Democrat" have unique cognitive properties which easily discriminate one label from another. The most characteristic and discriminating properties refer to (1) voting, (2) electioneering, and (3) other forms of electoral behavior. Although these two labels have many strong properties over which there is considerable agreement, such consensus is lacking for the fewer and weaker properties which characterize and discriminate the label "Independent." Whereas "Republican" and "Democrat" are sharply delineated, semantic inversions of one another, the concept "Independent" is ambiguously defined and only weakly distinguishable from other concepts.
For several decades political scientists have used a variety of categorical scales to measure subjective identification with a party. Within the domain of American politics, these scales have usually b e e n based on the two-part inquiry periodically made by SRC/CPS. The first part of this question asks, "Generally speaking, do you usually think of yourself as a Republican, a Democrat, an I n d e p e n d e n t , or what?" A follow-up probe asks those respondents who consider themselves Republicans or Democrats w h e t h e r they are "strong" or "not so strong'" partisans, or, if they are Independents, w h e t h e r they feel "closer to the Republican or Democratic party." To produce a single, ordinal scale, some political scientists (e.g., Campbell et al., 1960) use all seven categories created by the SRC/CPS format. Others (e.g., Converse, 1966; Burnham, 1970) collapse "'pure" Ralph W. Bastedoand Milton Lodge, Laboratoryfor BehavioralResearch, Department of Political Science, SUNY-StonyBrook. Political Behavior
© 1980 Agathon Press, Inc., New York
Vol. 2, No. 3, 1980 0190-9320/80/030287-22501.50 287
288
BASTEDO AND LODGE
Independents and the two types of "leaning" Independents into one category to produce a five-point scale. Still other political scientists (e.g., Nie, Verba & Petrocik, 1976) ignore the second part of the SRC/ CPS format, favoring the use of the initial three categories "Republican," "Democrat" and "Independent." And more recently, Keith et al. (1977) have suggested a quite different three-point category scale in which all respondents except "pure" Independents are labeled as Republican or Democratic partisans. Different classifications of the responses made to the SRC/CPS question make different assumptions about the meanings of the labels to the respondent: The seven-category scale assumes that each category evokes a separate, consensual meaning for all who choose it to describe themselves, while the five-category scale favored by Converse and Burnham assumes that "leaning" and "'pure" Independents are similar to one another, but different from those respondents who choose one of the partisan labels in the first part of the SRC/CPS format. Keith et al., on the other hand, challenge Nie et al. on the basis of an analysis of the seven-category scale showing that "leaners" behave as closet partisans, not as "pure" Independents; thereby raising the question why these respondents did not opt for a partisan label when it was offered. Such a multiplicity of categorizations on a single variable is especially problematic here, as each classification yields different interpretations of the correlates and consequences of partisanship. The p r o b l e m - - e n d e m i c to all research relying on the analysis of languagemis that social-psychological concepts are inevitably ambiguous: The words and phrases denoting them carry multiple meanings, connotations, and associations which affect what information p e o p l e extract, process, store in m e m o r y , a n d recall, a n d r e consequently--affect how they interpret and evaluate the concepts as stimuli. Where such variation exists in the meaning(s) of the stimuli, stimulus and response effects are confounded. THE MEANING OF CONCEPTS
Rather than assume that each of the party labels evokes a single, consensual meaning as a stimulus, we are asking: What do these concepts mean to respondents? What are the cognitive properties of each label? Which meanings are central? Which are peripheral? Are the meanings consensual among Republican identifiers, Democratic identifiers, and Independents? Similarly, what discriminates one label from another? For a category scale to be meaningful, the categories and labels must evoke distinct meanings. Given this emphasis on the
THE MEANING OF PARTY LABELS
289
meaning of the key words and phrases used to refer to political concepts, what "causes" some respondents to call themselves "Republican," others to call themselves "Democrat," and still others to adopt the label "Independent?'" The problem is more basic than that of question wording (Bishop, Oldendick & Tuchfarber, 1978; Sullivan, Pierson & Marcus, 1978). Until we can determine the meaning(s) of the stimuli used in our research instruments, it is impossible to separate the variance in response resulting from differences in the cognitive interpretation of a stimulus from differences in the affective evaluation of the stimulus, regardless of how precisely we measure the direction and strength of response. Recent developments in cognitive psychology, particularly in its subfields of information processing and semantic memory, provide welcome theoretical guidance for determining the cognitive content and structure of concepts (Collins & Quillian, 1969; Loftus & Loftus, 1976; Rips, Shoben & Smith, 1973). Substantial evidence has been amassed in the past decade showing that much of human memory is organized semanticallymthat is, structured by concepts, each made up of various elements, instances, and features that characterize it and discriminate it from other concepts (Rosch, 1975). In every culture there are hundreds of semantic concepts that are so well-learned as to activate representations in memory which almost everyone in the society shares. The concepts "bird" and "animal" are examples of such well-learned, consensual concepts. Most Englishspeaking people can decide quickly and confidently--within 700 to 900 m i l l i s e c o n d s h t h e instances, attributes, and characteristics of Such well-learned concepts. Just as some exemplars of a concept are more representative than others (a hawk is a better exemplar of the concept "'bird" than a chicken is, but less so than a robin), some features of a concept are more characteristic, more typical, than others: Feathers, wings, and flying are strongly defining features of "bird." A welllearned concept such as "bird" activates in virtually all people a unitary representation in memory, and not just an assortment of discrete properties which must be assembled by complex compare-and-contrast memory processes. But other concepts, including many used to refer to such abstract things as government and politics, are neither so well-learned as to evoke a unitary representation in memory, nor so strongly referenced as to activate a consenual meaning among users of the language: The instances, attributes, and characteristics of such concepts are ambiguous, and the boundaries discriminating one category from another are
290
BASTEDOAND LODGE
fuzzy. Most social-psychological concepts appear to be of this type (Tanenhaus & Foley, 1979; Tanenhaus, 1979). Judging, for example, by the number, diversity, and disparity of responses to SRC/CPS openended questions asking respondents for their likes and dislikes of the major parties and candidates, the concepts "Republican," "Independent," and "Democrat" vary in meaning (that is, cognitive interpretation) as well as in affect. Conversely, judging from the low but positive correlations routinely found to relate each party label with some political attitudes and behaviors, the concepts are not idiosyncratic. Although the leading models of semantic memory differ in their description of exactly how information is stored in memory, there is fundamental agreement among them that with a less-well-learned concept, people must assemble its meaning from whatever exemplars, features, instances, or experiences are activated by the stimulus. Some of these attributes of a concept may be central to its meaning (and presumably processed instantly for comprehension); while other properties are semantically distant; only peripherally related (and consuited, if at all, to make fine distinctions or deal With new, disparate, or ambiguous information). To delineate the meanings of the three categorical labels employed as stimuli in the first part of the standard SRC/~PS intiuiry to ascertain party identification, a local area survey was administered which elicited magnitude estimates of the centrality of cognitive properties to each concept. RESEARCH METHOD Collection of Exemplars and Properties of Party Labels
The requisite first step for ascertaining the content and structure of the concepts "Republican, .... Independent," and"Derlaocrat'" involved the collection of all potential elements of each label A pool of 200+ words and phrases were culled from the responses to the open-ended questions in the 1964-1976 SRC/CPS national election studies asking for likes and dislikes regarding the Democratic and Republican parties and major presidential candidates. An additional 40 items were obtained from responses to queries included in several local area surveys asking for associates of the three labels. When redundant items, synonyms, bygone leaders, local officials, and obsolete issues were dropped from this population of items, other respondents in subsequent local-area surveys were asked to indicate which of these words and phrases were "characteristic of' or "part of' each of the three con-
THE MEANING OF PARTY LABELS
291
cepts. From these frequency data, the 54 most common references to each of the labels were selected for this study. The items, as worded on our questionnaire, are grouped here for convenience. Six of the items refer to ideology: 1. right-wing 2. conservatism 3. moderate 4. middle of the road 5. liberalism 6. left-wing Seven items express general affect: 7. corrupt 8. wastes money 9. good 10. responsible 11. trustworthy 12. even-handed 13. efficient Eleven items reflect preference for a party leader: 14. votes for George McGovern 15. votes for Edward Kennedy 16. votes for Jimmy Carter 17. votes for Hugh Carey (1978 Democratic governor of N.Y.) 18. votes for George Wallace 19. approves of President Carter 20. votes for Gerald Ford 21. votes for Ronald Reagan 22. votes for Nelson Rockefeller 23. votes for Perry Duryea (1978 Republican candidate for N.Y. governor) 24. votes for Riehard Nixon Five items refer to partisan-civic activities: 25. campaigns for Republicans 26. campaigns for Democrats
BASTEDOAND LODGE
292
27. criticizes Republicans 28. criticizes Democrats 29. listens to political debates Six items relate to partisanship of vote: 30. 31. 32. 33. 34. 35.
votes votes votes votes votes votes
straight Republican ticket straight Democratic ticket split ticket for the best man, not the party on issues, not for a party in party primaries
Six items refer to group-constituency favoritism: 36. favors labor unions 37. favors large corporations 38. favors poor 39. favors rich 40. favors Jews 41. favors blacks And the final thirteen items reflect policy preferences: 42. favors busing for integration 43. more government jobs 44. big government 45. more inflation 46. more welfare spending 47. more taxes 48. combat unemployment 49. favors death penalty 50. tough on criminals 51. cut government spending 52. tough with Russia 53. balanced budget 54. more military spending The Magnitude Scaling of Centrality
Substantial experimental evidence confirms that some exemplars, features, and properties of a concept are more central or semantically
THE MEANINGOF PARTYLABELS
293
close to the concept's meaning than others (Collins & Loftus, 1975; Lakoff, 1972; Smith, Shoben & Rips, 1974). And as Rosch and her associates (Rosch, 1975; Rosch & Mervis, 1975; Rosch, Simpson & Miller, 1976) have demonstrated, people's ratings of "semantic typicality"--how typical or characteristic are the properties of a concept--can be used to ascertain the content and structure of a concept and predict how it will be interpreted. In this study we employed the magnitude scaling methods developed by Stevens (1966, 1975) to obtain a ratio measure of the semantic relatedness of all 54 properties to each of the three party labels. The questionnaire was administered during November 1978 to a nonprobability sample of 181 metropolitan New York adults. The median respondent was 35 to 39 years old, had two years of education beyond high school, had a total annual family income of $22,000 to $29,999 and, like the average American, did not participate in politics beyond voting. Each of the respondents was interviewed by one of 23 university students who received six hours of instruction in magnitude sealing, six hours of training and practice in interviewing, and satisfaction of a course requirement. Although this is not a representative sample of New York City, State, or environs, there is sufficient variance on basic demographic and political variables for the preliminary measurement of consensus within and discrimination between the three labels. Calibration of the Magnitude Estimation Measure
The interviewing followed explicitly the magnitude sealing procedures developed by Lodge & Tursky (1979, 1981) for survey research. First, to provide each respondent (R) with the necessary training and practice in using numbers to express magnitudes, R is instructed to assign numbers to nine line-length stimuli varying from 3 mm to 300 mm relative to a 30 mm reference line, which is given the reference number of 30. This "calibration" procedure provides the investigator with the information required to determine whether R's are properly using numbers to make proportional judgments, and, if so, how accurately. To verify the magnitude response measure, the magnitude estimates are logged, averaged by the geometric mean, regressed on loglog coordinates against the metric value of each line-length stimulus, and the empirical slope obtained from this sample compared to the characteristic slope of unity (1.0) relating stimulus to response. The empirically derived regression coefficient for the 174 R's who completed the interview is .962, a close match to the theoretical ratio of unitymthereby confirming that these R's are corrently using numeric
294
BASTEDOAND LODGE
estimation to m a k e ratio j u d g m e n t s (see L o d g e e t al., 1975, 1976a, 1976b). O n l y s e v e n o f the original R's w e r e u n a b l e to follow or u n d e r s t a n d t h e instructions g i v e n b y the i n t e r v i e w e r s a n d are t h e r e f o r e e x c l u d e d from t h e analysis.
Magnitude Scaling of Cognitive Properties I m m e d i a t e l y f o l l o w i n g the calibration task, R is r e a d instructions for m a g n i t u d e scaling t h e typicality o f p r o p e r t i e s to e a c h c o n c e p t . T h e instructions read: Just as you used numbers to say how much longer or shorter a line was compared to the reference line, we now want you to give numbers relative to your reference number 30 to indicate how characteristic, how typical, are different people, issues and ideas of a Democrat, Republican or Independent. (R shown packet of pages, each with one of the labels printed in large, bold-faced letters at the top, with 15 properties per page listed in a column, with each label counterbalanced, and each page presented in irregular order.) Let's again use the number 30 as your reference number to represent a neutral, in-between position for some person, issue or idea that is neither typical nor uncharacteristic of a Republican, Democrat, or Independent. So, if you feel that one of the items is typical of the word at the top of the sheet, write in a number bigger than 30, any number you think is appropriate to show how much more typical it is than the in-between position. If, on the other hand, one of the items is not typical, not characteristic of the word at the top of the page, give it a number smaller than 30. Give it any number smaller than 30 which indicates how uncharacteristic it is of an Independent, Republican or Democrat. And finally, if one of the items is in the middle, neither characteristic nor uncharacteristic, then write in the number 30 which stands for something neutral. You may use any numbers you wish. The key thing to keep in mind is this: the more characteristic some item is, the bigger the number compared to 30. The less characteristic an item is, the smaller your number compared to 30. So, for example, if you want to say some item is 10 times more characteristic than the middle, in-between position, you would give it the number 300, that's 10 times 30. On the other hand, if you feel another item is very uncharacteristic, about 10 times uncharacteristic, you could give it the number 3, that's l/a0th of 30. T h e calculation o f m a g n i t u d e scale (~) values for the estimates o f typicality is s t r a i g h t f o r w a r d - - t h e s i m p l e calculation o f the ratio o f e a c h R's r e s p o n s e n u m b e r to r e f e r e n c e n u m b e r . L e t t h r e e b r i e f illustrations suffice. Relative to t h e r e f e r e n c e n u m b e r 30 ( w h i c h stands for a m i d d l e , i n - b e t w e e n position n e i t h e r characteristic n o r u n c h a r a c t e r i s t i c o f a concept), s u p p o s e R assigns the n u m b e r 30 to indicate that " m o d e r a t e " is n e i t h e r typical n o r atypical o f the c o n c e p t " R e p u b l i c a n . " Since the
THE MEANING OF PARTYLABELS
295
response number in this instance is identical to the reference number, the resulting tk value is 1.0 (a ratio of 30:30 being equal to 1:1). Another itemwsay, "conservatism"--when judged as a property of the concept "Republican" may be assigned the number 150 to express R's judgment that it is five times more characteristic of"Republican" than the neutral position of 30. In this instance the ~ value is 5.0 (since the ratio is 150:30 or 5:1). And finally, for the sake of symmetry, were R to give the n u m b e r 6 to represent how atypical "liberalism" is of the concept "Republican," the t~value would be 0.2 (the ratio of 30:6 being one-fifth of 1.0). When, as here, we are interested in the average property-to-concept response, one simply takes the log of each R's response and reference number and calculates the geometric mean for each of the propertyconcept pairings.
RESULTS: SEMANTIC CENTRALITY Properties Characteristic of the Concept "Republican"
Table 1 lists the 21 of 54 items judged two or more times characteristic of the concept "Republican" than the neutral position by all 174 R's combined (column 1), by Republican identifiers (N = 58, column 2), by Independents (N = 61, column 3) and by Democrats (N = 55, column 4). For each item within each column, both the calculated ~ value and a parenthetical percentage are given. The latter number is the percent of R's who agree that the particular property is characteristic of the conceptwi.e., the percent of R's who gave numeric responses greater than the reference number. Since the magnitude scaling procedures employed in this study meet log-interval assumptions, the ~ values computed for each propertyconcept pairing represent proportions and as such reflect ratio relationships between items, both within and between groups. The most characteristic property of the concept "Republican" for all 174 R's combined is "campaigns for Republicans" (~ = 4.1), on average four times more characteristic than neutral, about 11/3 times more characteristic than "favors rich" (q~= 3.0), and twice as characteristic as "'responsible'" (~ = 2.0). It is also legitimate, given these magnitude data, to make item-byitem ratio comparisons between groups. For example, among Republican identifiers, the property "'favors rich" (~ = 1.5) is perceived as a weak characteristic of the concept "'Republican," whereas among Democratic identifiers (~ = 5.1) it is judged five times more characteristic of the concept "Republican" than the neutral reference. When
bo ~O
1. Campaigns for Republicans 2. Criticizes Democrats 3. Favors large corporations 4. Votes straight Republican ticket 5. Votes for Nelson Rockefeller 6. Conservatism 7. Votes for Ronald Reagan 8. Votes for Gerald Ford 9. Favors rich 10. More military spending 11. Tough on criminals 12. Tough with Russia 13, Votes for Perry Duryea 14. Favors death penalty 15. Votes in party primaries 16. Votes for Richard Nixon 17. Listens to political debates 18. Balanced budget 19. Right-wing 20. Responsible 21. Big government X of 21 most characteristic properties
Precise wording of item paired w/q~ value and % agree 4.1 (90%) 3.9 (85%) 3.9 (86%) 3.6 (87%) 3.3 (83%) 3.3 (79%) 3.2 (86%) 3.1 (82%) 3.0 (78%) 3.0 (79%) 2.9 (81%) 2.8 (80%) 2.8 (77%) 2.7 (72%) 2.6 (74%) 2.5 (75%) 2.4 (70%) 2.3 (69%) 2.2 (68%) 2.0 (62%) 2.0 (61%) 2.9 (77%)
Ratio of response nos. to reference no. for all R's (5/ = 174)
TABLE 1. Properties Most Characteristic of the Concept "Republican."
3.8 (86%) 3.0 (79%) 2.6 (72%) 3.3 (86%) 2.3 (76%) 3.4 (83%) 3.2 (86%) 3.3 (81%) 1.5 (64%) 2.4 (71%) 3.5 (90%) 3.0 (88%) 3.6 (88%) 2.9 (71%) 2,9 (78%) 1.8 (66%) 2,9 (78%) 3,0 (81%) 1.7 (55%) 3,1 (79%) 1.4 (55%) 2.8 (77%)
Republican (IV = 58)
(54%)
(62%) (69%) (67%) 1.9
(54%) 2.8 (77%)
1.8 2.5 2.1 1.8
2.2 (72%)
4.2 (95%) 3.6 (85%) 3.7 (90%) 3.0 (80%) 3.5 (87%) 2.7 (79%) 2.9 (9o%) 3.2 (87%) 3.5 (85%) 2.9 (84%) 2.9 (84%) 2.9 (8o%) 2.3 (70%) 2.8 (79%) 2.2 (72%)
Independent (N = 61)
(69%) (71%) (73%) (67%) (71%) (87%) (69%) (56%) (82%)
(82%)
(78%) (85%)
(8o%)
(89%) (91%) (96%) (91%) (85%) (93%)
(53%) 3.1 (76%) 3.5 (78%)
4.5 5.6 6.2 5.0 4.7 3.9 3.5 2.6 5.1 3.8 z.3 2.7 2.6 2.4 2.8 4.2 2.8 1.5 3.2 1.5
Democrat (N = 55)
Ratios of response to reference no. for party subgroups
THE MEANING OF PARTYLABELS
297
we examine separately the perceptions of Republican, Independent, and Democratic identifiers, it is clear that although the strength of association varies among groups, the magnitude estimates of the three subgroups are greater than unity for each of the 21 properties. Scanning the magnitude scores of typicality, one notes that 10 of the 21 most characteristic attributes of "Republican" relate to electioneering, voting behavior, and other forms of conventional electoral activity. Table 1 suggests that a Republican is usually perceived as a politically active citizen who votes for his party's slate, participates in party primaries, and votes for such Republican party leaders as Reagan, Rockefeller, Ford, and Nixon. As a critic of the Democrats, a Republican is characteristically conservative, favoring big business, rich people, and the military. The percent agree data given in Table 1 tallies the amount of consensus among all R's combined and within each party subgroup that a particular property is indeed typical of the concept "Republican." On average, 77 percent of the R's agree that the 21 items are characteristic. Among Republican identifiers, Democratic identifiers, and Independents, the percentage is similar. The items generating the greatest agreement among R's are "campaigns for Republicans" (90%), "votes straight Republican ticket" (87%), "'favors large corporations" (86%) and "votes for Ronald Reagan" (86%); while dissensus is greatest for such affect-laden properties as "big government" (61%), "responsible" (62%), and "right-wing" (68%). Properties Characteristic of the Concept "Democrat"
Table 2 lists the most characteristic properties of the concept "Democrat." All 174 R's combined, as well as each of the three subgroups, judge 18 of the 54 items to be characteristic of the concept "Democrat.'" On average, 78 percent of all R's agree that these items are typical. As with the concept"Republican," the most characteristic and consensual attributes of the concept"Democrat" relate to electoral behavior, while much disagreement shows up for affect-laden attributes. Properties Characteristic of the Concept "Independent"
Table 3 lists the most characteristic properties of the concept "'Independent.'" Only 8 of the 54 items are judged to be attributes of the label. Scanning the tabulation of features characteristic of the concept "'Independent," we observe that four of the eight most typical properties of the concept refer to voting behavior and other forms of political activity.
'O
1. Criticizes Republicans 2. Campaigns for Democrats 3, Votes for Hugh Carey 4. Votes tbr Jimmy Carter 5. Votes fbr Edward Kennedy 6. Favors labor unions 7. Votes straight Democratic ticket 8. Liberalism 9. Favors poor 10. Approves of President Carter 11. More welfare - s I ~ n d ~ 12. Wastes money 13. Combat unemployment 14. Favors Jews 15. Favors blacks 16. More government jobs 17. Favors busing for integration 18. Votes for :George McGovern X of 18 most characteristic properties
Precise wording of item paired w/~ value and % agree
Democrat (N = 65) 5.2 (85%) 4.4 (89%) 3.8 (80%) 4.6 (85%) 4.8 (82%) 4.1 (85%) 3.3 (76%) 4.2 (91%) 3.7 (78%) 3.7 (85%) 3.0 (80%) 2,3 (62%) 4.0 (89%) 3.2 (75%) 2.9 (80%) 3.1 (73%) 2.7 (71%) 2.9 (80%) 3.7 (80%)
3.1 (85%) 3.4 (92%) 2.7 (77%) 2.7 (84%) 2.5 (7.5%) 2.7 (75%) 2.1 (69%) 2.6 (79%) 2.2 (72%) 2.5 (82%) ,~ 1 ,(72%) 2.5 {a6%) 2.6 (74%) ,2.3 :(70%) 1.9 (62%) 2.0 (69%) 2.2 (79%) 1.7 (66%) 2.4 (75%) 3.5 (79%) 3.4 (83%) 4.0 (91%) 3.2 (86%) 3.0 (83%) 3.0 (83%) 3,7 (81%) 2.3 (76%) 3.0 (84%) 2.7 (72%) 3.5 (79%) 3.5 (79%) 1.6 (6o%) 2,2 (71%) 2.8 (76%) 2.4 (76%) 2,3 (81%) 1,8 (71%) 2.9 (78%)
3.8 (83%) 3.7 (88%) 3.4 (83%) 3.4 (85%) 3.3 (80%) 3.2 (81%) 2.9 (75%) 2.9 (80%) 2.9 (78%) 2.9 (80%) 2.8 (77%) 2.7 (69%) 2.6 (74%) 2.5 (72%) 2.5 (73%) 2.4 (73%) 2.4 (77%) 2.0 (72%) 2.9 (78%)
=
Independent (N 61)
Republican (N = 58)
Ratios of response to reference no. for party subgroups
Ratio of response nos. to reference no. for all R's (N = 174)
T A B L E 2. P r o p e r t i e s Most C h a r a c t e r i s t i c o f t h e Concept "Democrat."
tO "O ~O
1,8 (56%) 2.4 (67%)
1.8 (54%) 2.8 (72%) 2.3 (65%)
2.5 (68%)
of 8 most characteristic properties
3.6 (80%) 2.7 (75%) 2.2 (62%) 2.3 (72%) 2.0 (58%) 1.7 (53%)
2.2 (60%) 2.0 (6o%) 2.0 (56%)
2.3 (67%) 2.2 (67%)
2.7 (74%)
2.9 (76%) (87%) (87%) (70%) (75%) (66%) (66%) (72%)
4.1 3.6 3.0 2.4 2.4 2.6 2.4
Democrat (N = 55)
(78%) (72%) (78%) (62%) (62%) (55%) (53%) (59%)
3.1 2.6 2.5 22. 1.9 1.9 1.8 2.5
3.3 (80%) 3.3 (80%)
Independent (N = 61)
ratios of response to reference no. for party subgroups Republican (N = 58)
1. Votes for the best man, not the party 2. Votes on issues, not fbr a party 3. Votes split ticket 4. Listens to political debates 5. Combat unemployment 6. Moderate 7. Balanced budget 8. Middle of the road
Precise wording of item paired w/0 value and % agree
Ratio of response nos, to reference no. for all R's (N = 174)
T A B L E 3. P r o p e r t i e s Most C h a r a c t e r i s t i c o f the Concept "Independent."
300
BASTEDOAND LODGE
The most characteristic properties are "votes for the best man, not the party" and "votes on issues, not for a party," both with a typicality score 3.3 times more typical of the concept "Independent" than the neutral position. Not only is the concept "Independent" described by considerably fewer features than are the concepts "Republican" (Table 1) and "Democrat" (Table 2), but the coneept is also more weakly defined. The mean of the magnitude estimates given for"Independent" is 2.5, while the means for the 21 properties of "Republican" and the 18 properties of "Demoerat" are both 2.9. When we divide the R's into Republiean, Independent, and Democratic identifiers, we see that all eight properties are consensual characteristics of the coneept "Independent"--in every instance the values of eaeh party subgroup are greater than unity. On average, however, only 68 pereent of the R's agree that these properties are indeed characteristic; this in contrast to a mean percent agreement of 77 percent for the label "Republican" and 78 percent for the label "Democrat." So while Table 3 illustrates a lack of intensity (i.e., of magnitude) in delineating the label "Independent," the data also reflect moderate disagreement among the R's as to what properties characterize the label. Although the labels "Republican" and "Democrat" have many strong properties over which there is minimal conflict, the label "Independent" has fewer and weaker properties over which disagreement is greater. RESULTS: DISCRIMINATION BETWEEN LABELS
While the magnitude scale (~) values presented in Tables 1, 2, and 3 reveal that some features are more eentral to a eoncept's meaning than others, for the three labels to be semantically distinct alternatives the attributes of each concept must discriminate one concept from another. The defining features of a party label should discriminate that label from the other two if all three labels are to serve as cognitively meaningful categories in a scaling task. If, on the other hand, a label is weakly or ambiguously differentiated, than the category scale is confounded. Properties Discriminating the Label "Republican"
Table 4 ranks the 21 most characteristic properties of the concept "Republican" according to how well each discriminates the concept "Republican" from the other two concepts. Column i tallies the average ~ value for all 174 R's relating each of the properties to the concept
ot°
4.1 3.6 3,2 3.3 3.1 3,9 3.3 2.8 2.5 3.9 3.0 2.2 3.0 2.7 2.8 2.3 2.9 2.4 2.6 2.0 2.0
for the concept "Republican . . . . .24 ..13 .31 .57 ,46 .51 .53 .39 .17 .72 .78 .58 1.3 1.0 1.5 1.1 1.6 1,7 1.9 1.6 1.8
.75 .40 .71 .88 1.0 1.6 .98 .85 .46 .74 .75 .78 .94 1.5 1.6 2.0 1,7 2.3 .89 1.9 .76
for the for the concept concept Democrat . . . . I n d e p e n d e n t " 18.6 16.9 14.9 13.3 12.6 12.4 12.3 12.2 11.9 10.6 8.0 7.9 6.4 6.2 5.8 4.7 4.6 4.4 3,3* 1.7" 0.4*
t value discriminating "'Republican . . . . from "Democrat" 13.0 16,2 10.9 8.9 9.2 8.8 10.4 9.8 9.5 10.5 9.7 7.5 8.8 4.7 6,5 1.1" 5.5 0.6* 7.4 1.0" 7.1
t value discriminating Republican" from "'Independent"
Two-tailed t test of significance
*Not significant at the .001 level. To attain a .001 level of significance w h e n the degrees of freedom vary from 158 to 173, at value must be greater than 3.3.
1. Campaigns for Republicans 2. Votes straight Republican ticket 3. Votes for Ronald Reagan 4. Conservatism 5. Votes for Gerald Ford 6. Criticizes Democrats 7. Votes for Nelson Rockefeller 8. Votes for Perry Duryea 9. Votes for Richard Nixon 10. Favors large corporations 11. Favors rich 12. Right-wing 13. More military spending 14. Favors death penalty 15. Tough with Russia 16. Balanced budget 17. Tough on criminals 18. Listens to political debates 19. Votes in party primaries 20. Responsible 21, Big government
Precise wording of item paired w/dg values and t statistics
Ratio of response nos. to reference no. for all R's (N = 174)
TABLE 4. Properties Discriminating the Concept "Republican" from the Other Party Labels.
302
BASTEDOAND LODGE
"Republican"; column 2 the ~ values for the same properties when judged as characteristics of the concept "Democrat"; and column 3 the t~ values relating how characteristic the properties are to the concept "Independent." A t~ value of 4.1 in column 1 for "'campaigns for Republicans," for example, shows that the average R gives a magnitude response number four times larger than the reference number, thereby indicating that this property is perceived as strongly characteristic of the concept "Republican." The qs value of .24 in column 2, however, reveals that this very same property is judged to be strongly uncharacteristic and atypical of the concept "Democrat"; here the average R gives a response one-fourth the size of the reference. Two-tailed t tests were computed to evaluate how well the 21 Republican properties discriminate the label "Republican" from the other labels. From these t statistics we see that the most discriminating properties of the concept "Republican" are "campaigns for Republicans," "votes straight Republican ticket," and "votes for Ronald Reagan." Eighteen of the 21 properties discriminate at the .001 level the concept "Republican" from "Democrat," and 18 of the 21 also discriminate it from "Independent." So in addition to characterizing the label "Republican," the listed properties also generally discriminate it from the other labels. For 15 of the 21 properties the t values discriminating "Republican" from "Democrat" are greater than those discriminating "Republican" from "Independent," indicating that the distinction between "Republican" and "Democrat" is more pronounced than that between it and the label "'Independent." The t statistics also indicate that the concept "Republican" is discriminated from the other concepts primarily on the basis of electionrelated behaviors. Most of the 10 properties above the median refer to voting behavior, electioneering, and other forms of conventional electoral activity: 8 of the 10 properties discriminating the label "'Republican" from "Democrat" refer to such electoral activites, as do 7 of the 10 discriminating "Republican" from "Independent." Properties Discriminating the Label "Democrat"
Table 5 ranks the 18 most characteristic Democratic properties according to how well each discriminates the concept "Democrat" from the other two concepts. The most discriminating properties are "votes straight Democratic ticket," "campaigns for Democrats," and "votes for Jimmy Carter." Seventeen of the 18 properties discriminate at the .001 level the concept "Democrat" from "Republican," and 17 of the 18 also discriminate it from "Independent," although here, too, the discrimi-
0
2.9 3.7 3.4 3.4 2.9 3.8 3.3 2.9 2.0 2.8 2.4 3.2 2.5 2,9 2.5 2.4 2.7 2.6
for the concept "'Democrat . . . . .15 .28 .39 .44 .45 .45 .37 .44 .27 .62 .57 .68 .71 .80 .92 1.3 1.5 1.7
.64 1.0 1.4 1.2 1.7 1.8 1.5 1.2 1.1 1.1 1.2 1.4 1.5 1.8 1.4 1.4 1.2 2.2
for the for the concept concept Republican . . . . I n d e p e n d e n t " 16;2 15.3 13.6 12.3 12.1 11,9 11.8 11.5 10.6 10.3 9.7 8.9 8.8 8.7 8.6 4.5 3.9 3.0*
t value discriminating "Democrat" from "'Republican . . . .
10.3 10.5 8.5 10.4 4.9 6.2 7.4 8.7 5.8 7.7 6.7 7.4 5.2 4.7 5.5 5.1 6.6 1.8"
t value discriminating "Democrat'" from Independent"
Two-tailed t test of significance
*Not significant at the .001 level. To attain a .001 level of significance w h e n the degrees of freedom vary from 156to 173, at value must be greater than 3.3.
1. Votes str~igh± Democratic ticket 2. Campaigns for Democrats 3. Votes for Jimmy Carter 4. Votes for Hugh Carey 5. Liberalism 6. Criticizes Republicans 7. Votes for Edward Kennedy 8. Approves of President Carter 9. Votes for George MeGovern 10. More welfare spending 11. Favors busing for integration 12. Favors labor unions 13. Favors blacks 14. Favors poor 15. Favors Jews 16. More government jobs 17. Wastes money 18. Combat unemployment
Precise wording of item paired w/~ values and t statistics
Ratio of response nos. to reference no. for all R's (N = 174)
T A B L E 5. P r o p e r t i e s D i s c r i m i n a t i n g t h e C o n c e p t " D e m o c r a t " f r o m t h e O t h e r P a r t y L a b e l s .
304
BASTEDO AND LODGE
nation between "Democrat" and "'Republican" is greater. Like the label "Republican," the label "Democrat" is discriminated from its foils primarily on the basis of conventional electoral activity: seven of the nine properties above the median which discriminate "Democrat" from "'Republican" refer to electoral behavior, while five of the nine discriminating "Democrat" from "Independent" do. In many respects, the properties most characteristic of the label "'Republican" are the properties least characteristic of the label "Democrat," while the properties most characteristic of the label "Democrat" are the properties least characteristic of the label "Republican." When the 21 properties judged most typical of "'Republican" and the 18 properties judged most typical of "Democrat" are regressed against each label, the Pearson product-moment correlation is strong and negative, -.904, indicating that the concepts "Republican" and "'Democrat" are sharply defined inversions of one another. Properties discriminating the label "Independent" Table 6 ranks the eight properties most characteristic of the concept "'Independent" according to how well each discriminates "Independent" from the concepts "Republican" and "Democrat.'" The t tests reveal that the most discriminating properties are "votes on issues, not for a party," "votes split ticket," and "votes for the best man, not the party." Like the labels "Republican" and "Democrat," the label "Independent" is discriminated primarily on the basis of electoral behavior: three of the four properties above the median which discriminate "Independent" from "Republican" refer to voting, as do three of the four discriminating it from "Democrat." What is particularly noteworthy about the results for the concept "Independent," when compared to those of the partisan concepts, is that all eight properties of"Independent" have smaller t values (column 4 and 5, Table 6) than do most properties discriminating " R e p u b l i c a n " and " D e m o c r a t . " This suggests that the distinction between the concept "Independent" and the partisan labels is much less sharply defined than is the difference between "'Republican" and "Democrat." CONCLUSION
Our analysis of the typicality ratings given for the three labels-"'Republican, .... Democrat," and "Independent"--demonstrates that the two partisan labels have unique cognitive properties which discriminate one label from another. The magnitude estimates, percent
ida
t~
3.3 2.7 3,3 2.0 2,2 2,2 2.0 2.3
.96 1.0 1.2 1.1 1,4 1,7 2.3 2.4
for the for the concept concept "'Independent" "Republican . . . . 1.7 1.2 1.0 1.0 1.6 2.6 1.1 1.7
for the concept Democrat"
9.3 8.1 8.0 4,6 4.1 2.2" LI* 0.6*
t value discriminating "Independent" from "Republican . . . .
5.0 7.0 8.4 5.5 3,5 1.8" 4,4 3,1"
t value discriminating "Independent" from Democrat"
Two-tailed t test of significance
*Not significant at the .001 level, To attain a .001 level of significance when the degrees of freedom vary from 168 to 172, at value must be greater than 3.3.
i. Votes on issues, not for a party 2, Votes split ticket 3. Votes for the best roan, not the party 4. Middle of the road 5. Moderate 6. Combat unemployment 7. Balanced budget 8. Listens to political debates
Precise wording of item paired w]~ values and t statistics
Ratio of response nos, to reference no. for all R's (N = 174)
T A B L E 6. P r o p e r t i e s D i s c r i m i n a t i n g the Concept " I n d e p e n d e n t " from the Other Party Labels.
306
BASTEDOAND LODGE
agreement scores, and t tests for the concepts "Republican" and "Democrat" show that the central cognitive properties of each, and what discriminates one concept from others, refer mostly to (1) voting behavior, (2) electioneering, and (3) other forms of conventional, electoral activity. The more peripheral cognitive properties reflect (1) general affect, (2) policy preferences, (3) ideology, and (4) groupconstituency favoritism. The magnitude scale values for these labels indicate that the concepts "Republican" and "Democrat" are sharply defined inversions of one anotherNthe most characteristic properties of one concept are the least characteristic properties of the other. The two partisan labels, then, appear to be well-defined and easily discriminated, primarily in terms of conventional electoral behavior. In contrast, the concept "Independent" is ambiguously defined and only weakly distinguishable from other concepts, possibly due to the lack of an appropriately Indep e n d e n t alternative on the ballot. For categorical labels to be meaningful alternatives, they must be semantically distinct. Because the label " I n d e p e n d e n t " is ambiguously defined and weakly discriminated, from "Republican" and "Democrat," its presence on scales of party identification may well undermine the validity and reliability of such scales. The overall pattern of cognitive association implies that party identification is based largely on past or expected electoral behavior, perhaps influenced by the respondent's "a priori causal theories" (Nisbett & Wilson, 1977). The mass public apparently uses a behavioral definition of party identification, rather than a definition based on perceptions of political issues, party platforms, or dogma. ACKNOWLEDGMENTS
Research supported by grant SOC 77-25539fromthe Political Science Division,National Science Foundation. REFERENCES
Bishop, G. F., Oldendiek R. W., and Tuchfarber, A. J, (1978). "Effects of question wording on political attitude consistency." Public Opinion Quarterly 42: 81-91. Burnham, W. D, (1970). Critical Elections and the Mainsprings of American Politics. New York: W. W. Norton. Campbell, A., Converse, P. E., Miller, W., and Stokes, D, (1960). The American Voter, New York: John Wiley & Sons. Collins, A. M., and Loftus, E. F. (1975). "A spreading-activation theory of semantic processing." Psychological Review 82: 407-428.
THE MEANING OF PARTY LABELS
307
and Quillian, M. R. (1969). "Retrieval time from semantic memory.'" Journal of Verbal Learning and Verbal Behavior 8: 240-247. Converse, P. E. (1966). "The concept of a normal vote." In A. Campbell et al. (eds.), Elections and the Political Order. New York: John Wiley & Sons. Keith, B., Magleby, D., Nelson, C., Orr, E., Westlye, M., and Wolfinger, R. E. (1977). "The myth of the independent voter." Paper delivered at the annual meeting of the American Political Science Association, Washington, D.C., September 1977. Lakoff, G. (1972). "Hedges: A study in meaning criteria and the logic of fuzzy concepts," Paper delivered at the eighth regional meeting of the Chicago Linguistics Society. Linguistics Department, University of Chicago. Lodge, M., and Tursky, B. (1979). "Comparisons between category and magnitude scaling of political opinion employing SRC/CPS items.'" American Political Science Review 73: 50-66. __, and Tursky, B. (1981). "'Workshop on the magnitude scaling of political opinion in survey research." American Journal of Political Science, fourthcoming. Lodge, M., Cross, D., Tursky, B., & Tanenhaus, J. (1975). "The psyehophysieal scaling and validation of a political support scale." American Journal of Political Science 19: 611-649. , Tanenhaus, J., Cross, D., Tursky, B., Foley, M. A., & Foley, H. (1976a). "The calibration and cross-modal validation of ratio scales of political opinion in survey research." Social Science Research 5: 325-347. --, Cross, D., Tursky, B., Tanenhaus, J., & Reeder, R. (1976b). "The psychophysical scaling of political support in the 'real world.' " Political Methodology 2: 159-182. Loftus, G. R., and Loftus, E. F. (1976). Human Memory: The Processing of Information. New York: John Wiley & Sons. Nie, N. H., Verba, S., and Petrocik, J. R. (1976). The Changing American Voter. Cambridge: Harvard University Press. Nisbett, R. E., and Wilson, T. D. (1977). "Telling more than we know: Verbal reports on mental processes." Psychological Review 84: 231-259. Rips, L. J., Shoben, E. J., and Smith, E. E. (1973). "Semantic distance and the verification of semantic relations." Journal of Verbal Learning and Verbal Behavior 13: 1-20. Rosch, E. (1975). "Cognitive representations of semantic categories."Journal of Experimental Psychology: General 104: 192-233. _,and Mervis, C. G. (1975). "Family resemblances: Studies in the internal structure of categories.'" Cognitive Psychology 7: 573-605. _, Simpson, C., and Miller, R. S. (1976). "Structural basis of typicality effect." Journal of Experimental Psychology 2: 491-502. Smith, E. E., Shoben, E. J., and Rips, L. J. (1974). "Comparison processes in semantic memory." Psychological Review 81: 214-241. Stevens, S. S. (1966). "A metric for the social consensus." Science 151: 530-541. (1975). Psychophysics: Introduction to its Perceptual, Neural, and Social Prospects. New York: John Wiley & Sons. Sullivan, J. L., Pierson, J. E., and Marcus, G. E. (1978). "Ideological constraint _
_
_
_
308
BASTEDO AND LODGE
in the mass public: A methodological critique and some new findings." American Journal of Political Science 22: 233-249. Tanenhaus, J. (1979). "Separating objects of specific and diffuse support: Experiments on presidents and the presidency.'" Laboratory for Behavioral Research, Department of Political science, SUNY-Stony Brook. Report No. 15. __, and Foley, M. A. (1979). '" 'The words of things entangle and confuse': Response effects and the ambiguous concept." Manuscript of the Departments of Psychology and Political Science, SUNY-Stony Brook.