2 le informatíon that reaches theublicisnever a fuIl recordof

sanddevelo ments in the world .•It 15, [ather, a hi hl seledivearid"

Information, predispositions, and opinion View of what has taken place. It could hardly be otherwise. But eveú i .(
flhe public wou ave lule desire to be kept c10sely informed abdl.1t"~¡:¡
'~orId beyond its personal experience. It requires news presentatiol1s ,.-
short, simple, and highly thematic - in a word, stereotyped. Thus,'"
er (1984), in a close study of how a sample of citizens monitored theri
d that her subjects "grumbled frequently about the oversimplified treátrrf
eIevision] news ... " Yet when
¡al news programs and neWspaper features presented a small opportunity for '
,lE '" 'ive exposure to issues, they were unwilling to seize it. For the most part, rc' .
i I ,very OPImOnI~_.~. man:!ª~5iJ:l~ infoE!11.ati~1::E~di~p.9sitioll;,jnformation to not read and study carefully the more extensive versions of election and othe
, form a mental nlcture of the glven issue d d····· " . spapers and news magazines. Masses of specific facts and statistics were unit..
1·-"·-- .---"''': . --~..- <...... _....~<lº_ R.r.~ ISposltlOn tº,..!!!~te Sorne
1 c2ncluslOn about. It. ~~~.f.~ntr~Iaill1 of. this book is to show.how terized as dull, confusing, and unduly detailed... , (p. 105)

~
:,!~
•
... . e._,._.:.t•.•an
mdlvIdual
".
..•-g
,.y.'.-'..'
...e.....'tu..
dIfferences '. ,' " .v',.a".,r'. ,.i,¡l', .t ~):~,O.:lÍ S.ín'" ~,.Io.:~.r.-,m~'ÚQ~'.c.ªir¡ .
·.t.hy.. '.ii.·,•.n fiqr.matlQn.,,,ªº-ºIndlvIdual
.:O.'•,.',,. f ..i.S. es. in attention to ~,'~._.
'.
.,this
. . , .... ' ".,,,,-,,,,,.,,,,.,-_...
•. . .e.,d'.' ~r.~~:c.,~~:~
..i.n,.-:"""'-";""-''''
,-','.,.,;.'•ft.t. dlfferences
pmann, who remains perhaps the most insightful analyst of the proce~
h thepublic comes to form an understanding of complex and distant ev~
mp?IItIc~1 .values and other predispositi~!!~jQi.utly."deter';ine.ih~_.~"oñiourso f ted a large section of Public Opinion to news stereotypes, or what tod .•. .
~u~I~:O~I~IO~. T~e bo.o~, lhus, is most cruciaIly about the relationship ';;;;;~g often calIed frames of reference. In one lucid passage, he described
mformatlOn, predlsposItlOns, and opinion.
I as it would probably have been perceived by a character in
.~he present chapter introduces and defines these key terms, examines sorne is's Main Street:
cntIcaI problems associated with their study, and shows in a preliminary way
how they r~late to one another. In so doing, it develops the intuitions behind the s Sherwin of Gopher Prairie is aware that a war is raging in France and tries to<.c
more techmcal core of the book, which begins in Chapter 3. c it. She has never been to France, and certainly she has never been along wha
. the battlefront. Pictures of France and German soldiers she has secn, but it is
ible for her to imagine three mili ion men. No one, in fact, can imagine them,
fessionals do not try. They think of them as, say, two hundred divisions. But
INFORMA nON AND ELITE DISCOURSE crwin has no access to the order of battle maps, and if she is to think about the
To an extent that few like but none can avoid, citizens in large societies are de- c fastens upon Joffre and the Kaiser as if they were engaged in a personal duel. Per.
you could see what she sees with her mind's eye, the image in its composition ITl
j)endent on unseen and usuaIly unknown others fue rnQst Df their. information not unlike an Eighteenth Century engraving of a great soldier. He stands there b6
a?out th~ larger worId in which they live. As Walter Lippmann wrote in his cl;- \ruffled and more than life size, with a shadowy army of tiny Httle figures winding
SIC treatlse, Publie Opinion (1922/1946), oto the landscape behind. (p. 8)
Each of us lives an~ works on a small part of the earth's surface, moves in a small circle As suggested by Miss Sherwin's reliance on an eighteenth-century e.n!l"rllvirl2'.
a~d of these acquamtances knows only a few intimately. Of any public event that ha~ mann doubted that individuals can ersonalIy create the stereoty es
wI~e effects we see at best only a p~ase and an aspect. , .. Inevitably our opinions cover bther symbolic representatio~''the pictures in our heads" - by which remote
a bIgger space, a longer reach of tIme, a greater number of things, than we can directly
observe. The~ hav~, therefore, to be pieced together out of what others have reported and Hnd even proximate events are understood. ~athe!:
what we can Imagme. (p. 59)
(¡In the great blooming, buzzing confusion of the outer world we Rick out what our cul-
~ "others" on whom we d.epend, directIy or indirectly, for information ture has already defined for us, and we tend to perceive that which we have picked out in
líclorm stereotyped tor usoyour culture. (p. 61)
about the world~, for the most part, persons who devote themselves fuIl time
t~me .asp~ct of politics QLQublic affairs - which is to say~cal elites: Ma~o v.:h.ic.h Lippmann refe~s ~~~yerm~~~tfeatm~s
These el~t~s mclude politicians, higher-Ievel government officiaiS;jüurnalists, 01' the culture - the corrupt pohtIClan, the la~or stnke, the eIec1lOn contest,
sorne actIvlsts, and many kinds of experts and policy specialists. Even when we lhe yeom~;r;rmer. Bu~because society is always churning up new issues and
Iearn from. friends or family members about sorne aspect of public affairs, often problems, many stereo!.Y.~.J!f§.~s.!:.t:ª-!.i2.I.!~: For example, research has
we may sttll be secondhand consumers of ideas that originated more distantly shown how:-in thedeb;te' over the Equal Rights A-~endment, stereotypes of ,uni-
among sorne type of elite. sex toilets and women combat troops carne into being as a reflection of the

NATURE AND ORIGINS OF MASS OPINION
organizational and ideo~o~ical needs of the contending activists (Mansbridge,
1~86). Luk~r has done sll~I1ar research on the origins ofthe Pro-Choice and1pro-
Llfe labels m the contrastmg world vi~ws of abortion activists (Luker, 1984). A
powerfu1 stereotype that .has emerged m recent years is that of "the homeless."
~~e.es and frames hke these are important to the illocess by which the pub-
~IC keeps mformed because they determine what the public thinks it is becomin
I~~orm~d a out, w![ch In rn olten determines how ¡teople take sirles on po-
ht.lcal Issues (Ed~lman, 1964; Bennett, 1980; Gamson and Modigliani 1987'
Kmder and Sanders, 1990). ' ,
~Ithoug? cultur~lIy given and elite-supplied stereotypes may be most power-
fuI m sha~m~ pu~hc understanding of events that are "out of reach, out of sight,
out of ml~d . (Llppmann, 192211946, p. 21), they can be important even for
matters wlthm peopl~'s powers of direct observation. For example, Iyengar
(1991) has us~d e~penmental evidence to argue that whether television news fo-
cuses on "eplsodIC" cases of individual poverty, or the societywide conditions
that cau~e p~v~rty, affects the public's attribution of blame for poverty and
thereby ItS wdhngness to support programs aimed at alleviating it.
Perha s .the most fundamental uestion about news stereot es, or frames of
~ef~r~nce, IS .whet er t e public is given any choice about them whether, that
~s, It IS permltted to choose between alternative visions of what the issue is. For
IU.tbe absence of such choice. lhe public can do little more than follow thee'"i'ite"
~onsensus on.what sh?uld be done; Forexample, in the early phase of American
mvolvement m the VIetnam War, the public was offered only one way to think
abo~t th:, war, namely as ~ struggle to preserve freedom by "containing Com-
mumsm. Even news stones that criticized government policy did so within a
fra~ework that ~ssumed the paramount importance of winning the war and de-
f~atmg commumsm (~alb~rstam, 1979; Hallin, 1986). During this period, pub-
hc support for Ame~lcan mvolvement in the war was very strong, and those
??em.b~rs ~f t~e pubhc most heavily exposed to the mass media supported the
offlclal Ime most strongly.
. In the later phase of the war, however, journalists began to present information
In ways sug~esting that it was essentially a civil war among contending Viet-
namese factI~ns and hence both inessential to U.S. security interests and also
perhaps unwmnable. Coverage implicitly supportive of the war continued, but it
no long~r had .near-monopoly status. Owing, as I show in Chapter 9, to this
change m medIa coverage, public support for the war weakened greatly. Also,
heavy expo~ure to the .ma~s media was no longer associated with support for the
~ar, but wlth.~ polanzatlOn of opinion that reflected the division in political
dlscou~~. PohtIcally att~ntive libera~s within the general public tended to adopt
the POSltIO~ taken by el~tes ~onventlOnally recognized as liberal, while politi-
cally atte~tlve c?~servatlves In the general pubJic moved toward the position of
conservatlve opInlOn leaders.
S~n elites uphold a_,clear Ricture gf~whaL~ould. be done, the public
t;nds to .,~::~~~~ t~~~~Of~~~L~ith th~JrioSlP.Oíiiis~~.!!ent_!~~
Information, predispositions, and opinion
~~~~"J;i,'>""b!","n",c,""m""""o""st,-,=,Jik.clY to ad.opt the elite-I!ositi2P' ~en elites di\lid~
ers of the ublic tend to follow the elites sh~IiQg their general idéoI0gi~13.r
rtisan predisposition, .with the. most politic~ att!1!!lv~members ofth~.
IC mirroriñ'gmóstSh3.rply the ideological dixisions ~!!!.onB.. the elite. .. ; ,
ese c1aims about the e fects o~Jite discourse, which are an importa~t;·
t of what this book will attempt to demonstrate, are obviously quite Stro~?f'
s. By way of further preliminary examination, I would like to give an oVéf~i
W of the evolution of American racial attitudes in the twentieth century,
ngly emphasize that my purpose in reviewing this sensitive subject is not
vince anyone of the final correctness of my view, but only to iIlustraté
r1y as possible the general vision that underlies the more specific argumeri
ter chapters.
Elite discourse and racial attitudes
the turn of the century, the United States was a deeply racist society ~
y in the caste structure of the southern states and in the, widespread p~act~
iscrimination, but in the political ideas that informed ehte and mass thID
ut race. Although there was sorne mainstream elite disagreement on
of race, it was confined to a very narrow range. Virtually all white
epted sorne notion of the inferiority of other racial groups
71). It is both distasteful and unnecessary to recount these ideas, but one
important to the argument 1 wish to make. It is that racist ideas about blacks-
d, indeed, about most non-Anglo-Saxon groups, including Asians, southern
.'. d eastern Europeans, and Jews - had the support of the biological and psY"
ological science of that periodo Racist ideas, thus, were not confined to
tremist or backwater fringe; they were as common among the nation's ... ~úf~.·".
tellectual leaders as among other types of whites. Given this pattern of
titudes, any attempts to mobilize white support for black equality, wrletl1er PY"f
acks themse1ves or sympathetic whites, were bound to fail.
By 1930, however, the attitudes of political elites seemed to be changing.
year, President Hoover's nomination of John Parker of North \...-a.lVII11«
Supreme Court was rejected in large part because of a ten-year~old sp(~eC~I1·.··'.·
which Parker had said that "The Negro as a c1ass does not deslre to
""Ii',,,c" and that the "participation of the Negro in politics is a source of
¡lnd danger to both races" (cited in Kluger, 1975: p. 142). That ~ single
spcech, of a type that was entirely conventional throughout the mneteenth and
carly twentieth centuries, could become a basis for the rejection of a Supremc
Court nominee by the Senate was an indication that attitudes toward race were
undergoing a historic shift. .. .
Despite this, race was apparently not a major pubhc Is~ue.ID the . I93.0s
(Sheatsley 1966: p. 217). Moreover, Gunnar Myrdal (1944), ID hls. masslv~ !Il-
n
vcstigation of American race relations. found that neither the matenal.condltlo
(Ir blacks nor the amount of discrimination they faced were much dlfferent ¡n
 lnformation, predispositions, and opinion
.v NATURE AND ORIGINS Of MASS OPINION

..•.. dency of elite discourse has been .to blame .inequalit~:i~th'ffe,pr'''f.,'nt:'';'';: ;';:"Y'j
1940 than they had beeo in the immediate aftermath of the Civil War. To the.
?/ividual effort or, in its common liberal va:la~t, ~n tdü¡coun;etcáll
extent that there had been any improvement at aH, it was ooly because sorné Inination against blacks. A more profound Shlft m elIte
blacks had migrated to the North, where conditions had always been somewhai
better. Nooetheless, Myrdal maintained that a period of great racial progress lay imagined. h n: t n
lhe lack of opinion data until the late 1930s, t e e ec s o d
just ahead. White Americans believed deeply in their creed of equality and ha this revolution in elite discourse cannot be fully docum~nte b'
come to realize that black demands for equality were justified. He ttlerefore s about public opinion are reasonably c1ear. First, t~er~ las een
thought the days of white resistance to racial equality were near their end. ift toward greater public support by whites for th~ pnnclp e
Thus, by Myrdal's account, which proved extraordinarily prescient, a change he shift has not extended as far as many woul~ lI~e .- ~ost notatllY,
in white attitudes preceded any change in the actual conditions of blacks. What, .sted many government efforts to combat dlscnmmatlOn
then, brought about the attitude change? ~~~:~ o osed to most efforts to make up for the effects .of ~ast
One can imagine many possibilities, but Myrdal found the explanation in . Nor i~~he sincerity of sorne people's professions of behef m "'Ul.a.'·
purely intellectual developments. Scientists, who as recently as 1920 had over- . uestion But evidence of great change is hard t~ de~y
whelmingly endorsed the notion that sorne racial groups were superior to others, ~obo 1985) For example, only 45 percent of whltes m a 1944
had by their subsequent research discredited it. The magnitude of the change in ks ':should' have as good a chance as white peopl~ t? get any
scientific thinking is captured by the following two passages from the work of I~c . 1972 this figure had risen to 97 percent. Slmllarly, the
Carl Brigham, who was for a time a leading authority on race. In 1923 Brigham \.w ~reas m " : and black students should go to the sa~e
concluded his Study al American lntelligence by c1aiming f1atly that "the in- saymg that whlte students f 32 percent in 1942 to 90 percent m
tellectual superiority of our Nordic group over the Alpine, Mediterranean, and 1rather than separate ones rose rom t the time of the first masS opin-
negro groups has been demonstrated" (p. 192). However, in a review of subse- ese changes may have begun tOhocc~r a es may have been already under
quent research that was published just seven years later, Brigham felt compeHed son race in the.early 1940s'horth~tc.:n;ass attitudes roughly coincides
that time. In elther case, t e SIl
to withdraw this conclusion. As he wrote in the final sentence of his paper,
e shift in elite attitudes. . d t the new elite discourse on race,
This review has summarized sorne of the more recent test findings which show that com- d h ople most heavIly expose o . h t
parative studies of various national and racial groups may not be made with existing n ,t e pe been most Iikely to support those Ideas t· a
the better educated, have Th the better educated are not
tests, and which show, in particular, that one of the most pretentious of these comparative
racial studies - the author's own - was without foundation. (Brigham, 1930: p. 165) ute the modern elite consensu~ on ra~e. UthS~ more controversial efforts
. I ort affirmatlve actlon or
ally IIke y to supp . h' h tend not to have consensual
Reviewing this and other research, Myrdal wrote that "A handful of social bat inequality, such as ~c~oo~ busm;r~i:n~~e support for the principIe of
and biological scientists over the past fifty years have gradually forced informed support; but they do exhlblt dISP~~~ discrimination, such as federal laws
people to give up sorne of the more blatant of our biological errors" (p. 92). As ity and for those efforts to com ortation systems that do enjoy main-
Degler (1991) has recently shown, changing scientific theories of race in the st segregated restaurants and transp 1 1985') Thus exactly as in
1920s were part of a much larger scientific movement away from biological ex- t (Allport 1954' Schuman et a ., . ,
m elite suppor . ' .' osure to elite discourse appears to pro-
planations of human behavior. I Vietnam case des~nbed ear~le~, i~xft (1 present further evidence on the ra-
With the inteHectual defeat of early theories of racial inferiority, psycholo- e support for the Ideas carne 'elite discourse in Chapter 8.)
gists shifted their research to the stigmatizing effects on blacks of what was now y liberalizing effects of exposure ~o rtisan elite cues on the subject of
taken to be white prejudice, and to the origins of racial prejudice in various o.' h br has been responslve to pa k f
l'¡ltlally, t e pu IC .' h f hich comes from the recent wor o
kinds of mental disorders and educational deficiencies (AlIport, 1954).2 ¡"ICC. The evidence on thls pomt, muc o w ., . sorne detail
In consequence of aH this, the stereotypes used to explain racial differences in .~. . (1989)' worth exammmg m .
{;urmines and Stlmson , IS 1960s elite Democrats and Republicans ex-
material conditions underwent a major change. Until about 1930 these stereo- Throughout the 1950s ~nd e~rly , n racial issues. The Democratic Party
types stressed racial inferiority as the reason for inequality. Since then the no consistent partlsan dlf~ere~ces 0 t notably Hubert H. Humphrey,
inent raCIal hbera1s, mQs h'
I Degler (1991) has also made the interesting argument that the new research was ideologically was home to many prom . H ' S Truman pressed to ac leve
h I dership of Presldent arry ,
motivated in the sense that its practitioners were nonrationally commitled to the defeat of racist Hnd had, un der t e ea . II 'n the military Yet racially conser-
11 rncasure of equality for blacks, ~specla y ~ithin the De~ocratic party. Mean-
theories. This may or may not be true. Bul Degler makes no argument tha! fraud or dishonesty was
present in the research, nor does he present any evidence that the large number of scientists who
s remained a maJor power d
decided to accept the research were motivated by anything other than their own judgments of the val ¡ve South erner . . . ht Eisenhower, though no crusa er on
facts of the matter. while, the Repubhcan ~reslde~t D;l: I Warren to be chief justice of the
2 The possibility Ihat white prejudice might have been due. at least in part, to the racist elite dis- race, appointed the raclally lIbera ar
course of previous decades was not, so far as 1 can telL given serious consideration.
 NATURE AND ORIGINS OF MASS OPINION
Information, predispositions, and opinion

a Voting trends on racial issues among U.S. Senators The effect of this change in party leadership cues is apparent in the 100..,er';p~jt."'
n of Figure 2.1. Rank-and-file Democrats and Republicans began in 19()4'tb..~;:.,;':
75
¡bit substantial amounts of party polarization on racial issues - the re~iull.'¡J.l\~)
65 luId seem, of the sudden change in the structure of party leadership
Democrats
Mean There is, however, an ambiguity in these results. Mass polarization a')'''Jtó,.''
liberalism rty lines could have come about from a reshuffling of party loyalties,
on racial ¡al liberals f10cking to the Democratic Party and racial conservatives l"\JV'l.l.~tó:
issues . r to the Republican side. This is the party conversion thesis. Or polariZ2lti~~ºr•. ·..·.·•••
uld have come about from opinion conversion, that is, existing D(~moc:ratstjeg
25-1--..----,--r--r--~__,~~ ming more racialIy liberal and existing Republicans becoming more
1945 50 55 60 65 70 75 80 nservative. This is the opinion leadership thesis. Although Carmines
Year son make no attempt to sort out these competing possibilities, it apP~ll~S >
both processes were at work. Petrocik (1989) has shown that, bej~inlnil'lg'!Q;/:;;
b. Racial attitudes among rank-and-file party identifiers , the Democratic Party lost Southern whites and gained blacks, wnmn'.IHi-'
tes a reshuffling of party loyalties along lines of preexisting racial opiniolj:í¡;V
60 Gerber and Jackson (1990) have shown that many existing Democrats
Democrats úblicans also changed their racial opinions to accord with the new
ership cues, which indicates a mass response to elite opinion le~lde:rsllipt·:
Meanscore
on racial 50 - :-,.;=.-....;.
r purposes of this book, the latter phenomenon is the more impo:rtanL Jf
Iiberalism cues can change racial opinions, which appear to be among the most
scale ()f mass opinions (Carmines and Stimson, 1982; Converse, 1964; Converse
dMarkus, 1979), they can probably affect most other types of opinions as well.
40-l--..--.---r---.----.-_,.---,
1945 50 55 60 65 70 75 80
Year
Conceptualizing and measuring elite discourse
political information carried in elite discourse is, as we have seen, .ne~er
Figur~ 2. I. Racialliberalism scores of Democrats and Republicans, 1945- 1980. Source: te. It is, rather, an attempt by various types of elite actors to create a depICtJ0n
Carmmes and Stimson (1989).
eality that is sufficiently simple and vivid that ordinary people can grasp it.
is "information" is genuinely information in the sense that it consists of what
y be assumed to be sincere attempts to capture what is most impor~ant abo~t
Supr~me Co~~t, and ~sed federal troops to enforce its landmark school deseg- at is happening in the world and to convey it in its proper perspectJVe. But Jt
regatlOn deCJSlOns. FmalIy, despite the impression created by such prominent
never "just information," because it is unavoidably selective and unavoidably
Democratic liberals as Humphrey, Republican congressmen were more liberal
meshed in stereotypical frames of reference that highlight only a portion of
than Democrats on race, as shown in Figure 2.1. As a result of this lack of
al is going on.
c1early differentiated leadership cues, as Carmines and Stimson argue, Demo-
In consequence, the public opinion that exists on a given issue can rarely be
crats and Republicans in the general public did not differ on racial issues. 'consídere:d a straightforward response to "the facts" of a situation. Even topics
Beginning in late 1963, however, the Democratic Party, overcoming the re- are within the direct experience of sorne citizens, such as poverty, homo-
sistance of its Southern wing, stepped out as the party of racialliberalism, while ~exllality, and racial inequality, are susceptible to widely different understand-
the Republican party became the more racialIy conservative party. Thus, Pres- Ings, depending on how facts about them are framed or stereotyped, and on
ident Lyndon Johnson, as titular leader of his party, pressed for and won major
:1 AClually, Carmines and Stimson (1989) find the firsl evidellce of mass-level parlisan polarization
civil rights bilIs in Congress, while Senator Barry Goldwater, the Republican nI! race in a Harris poI! taken in November 1963. whlch was Just after Presldent John Kennedy
~res~dential nominee in 1964, became the most prominent opponent of this leg- declared his support for a major civil rights bil!. . .
IslatlOn. Also, congressional voting on racial issues began to follow c10sely ;\ Gcrber and Jackson (1990) reporl comparable evidence of opinion leadershlp on VIetnam. Fran-
klin and Kosaki (1989) show that a heightened mass polarization on abortion altitudes followed
Democratic-Republican party lines (see upper portion of Figure 2.1).
lhe 1973 Supreme Court decision in Roe v. Wade.
 l'lA.IUKE AND ORIGINS OF MASS OPINION lnformation, predispositions, and opinion

which partisan elites are assodated with which positions. In view of this, it' What the magazine bluntly characterizes as inadeqU~te;es~~~~:~tJJ~~~<'
difficult to disagree with Lippmann 's observation that while the 'by President Jimmy Carter an? Ronald Reagan, hls ep 1
e faIl eleetion, the story contmues:
orthodox theory holds that publie opinion constitutes a moral judgment of a group
facts ... [it is more reasonable to hold that) ... public opinion is primarily a morali . , d fi se posture is not what it could be - or
ittle quest~~n th~t Am~nca s : ena ed to the point of obsolescence - and
and codified version of the facts (1922/1946: p. 93). of the mlhtary s eqUlpment as g b need continuing and expenslve
.
'1 Mmutemen
ICBM's and B-'i~ 2 bom atin
ers costs have ravaged the ~"rvlc:esan(l
Thus, whén I refer in the course ofthis book to the "information carried in eli stay com.p~titive. Skyrocketmg °:a~es a~d increasingly long stretches orsea..·•. ··.·.. · ·','·1
discourse about politics," as 1 often will, 1 will be referring to the stereotype their trammg efforts. Low P~~, d airmen have prompted a mass
frames of reference, and elite leadership cues that enable citizens to form eo il?rs and overseas tours for ~~th~e~e:~t of any fighting force. These prclble:Il"l.S;"[/;>j{1
ceptions of and, more importantly, opinions about events that are beyond the enenced nonc~ms who ar\ bTt f the U.S. military to react to
cgitimate questlOns about t e a I I Y o
full personal understanding. The aim of the book is to show how variations i
in combat-
this elite discourse affect both the direction and organization of mass opinion
This conception of elite discourse, however, is more elaborate than can b es of this type were not unusual in the I~te 1970s and ~arlY ~9~~s7 By
fulIy measured and tested in this book. I have sketched it in order to indicate th h the 24 months prior to the 1980 eleetl~n, Ne~s~e;hi~~~~ollY or:ptC+;;
larger picturé into which my argument fits, and to acknowledge that elite dis e more or less directly on defense spendmg, 4 o
course is a more complex phenomenon than my simple measures will make it ou htly favored greater spending. t feature of NewS¡Veg¡
to be. For my measures realIy are quite simple and concrete. Often, I will mak d' sture was not however, a permanen
-spen mg p~ As the ~ew Reagan administration began to
only a dichotomous measurement - whether there is a monolithic elite point e of defense Issues. . . t nnlc<>hllv

/:~:~:::~~;~7:~;:~~!~~~;:;:I:i!:~~~~~:::~oel:;ur'¡-e~'I'::fl
view on what a given issue is and how it should be handled, or whether there ar
important elite disagreements over the issue (see especial1y Chapter 6). In a fe
other cases, 1 wil1 determine the relative intensity of opposing elite communi- ueezmg out social s~ g med a radieally different eharacter. NCIWI[llC\ +> :2i!
eations and how re1ative intensity ehanges over time. In these cases, 1 will be eek coverage of the Issue ~ss~ ¡: t' n about multimillion dollar
counting the number of media reports on a given issue, and the direction in fill d 't olumns wlth mlorma 10
jne le I s c . d other indieations of Pentagon
which eaeh report would tend to push opinion. ns, $600 air force, serewdnve~~e:~ it U S fighting force, the publie was
Yet, as much research has shown, even simple story counts are sufficient to gement. Instead of lmages of a f l P 'I't' y' Thus in the 24 months
show a c10se reJationship between elite discourse and mass opinion (Erbring, . f a bloated and waste u mi I ar . , "jh'ldh,,' ,:

Goldenberg, and Miller, 1980; MacKuen, 1984; Page, Shapiro, and Dempsey, ::~:u~~~; eleetion, there were 60 sto~~s ons:ef~~~; :~~:~~n:~r:d
1987; Fan, 1988; Page and Shapiro, in press; Brody, 1991). And, as the reader 1ed a posture opposed to defense spen mg., d ti spf:ndm!~',.lO
will see, they suffice equal1y well for the purposes of this book. rage swung from about four-to-on~ in favor of greater e ense
By way of iIIustration, let me briefly describe changes in news reports on the to-one in favor of redueed expe~dltures. d' tandem with these shifts in
issue of U.S. defense spending in the late 1970s and earIy 1980s, along with the lblic opinion on defense spendl,ng movWe mmost Amerieans wanted toeut
associated changes in public opinion that they appear to have produeed. At th end of the Vietnam ar, ' l
ia coverag~. e as 1975, only about 10 pereent feH too htt e
On the coverofthe Oetober 27, 1980, issue of Newsweek was the headline, "Is cnse spend~ng, and as late B t ' response to a steady stream of pro-
m
Ameriea strong enough?" The inside story began as follows: . . . ney was bemg spent on de~ensed' u "b d support for sueh spending rose
._ . f the type Just esen e , A .
Seldom in time of peace has the United States been so troubled by talk of war - and so .~íefcnse Images o 1 1981 a slight majority of meneans
much coneerned that the country is incapable of wagjng jI. The Army Chief of Staff, 1í.t(~adily in the late 1970s, S? that by ear Y ' Then as the news media began
1, 1" as bemg spent on d etiense, , ¡:
General Edward C. Meyer, complains publicly that he presides over a "hollow army," lblt that "too Itt e w . . a'nst defense spending, support lor
undermanned, undertrained, and underfunded. General Lew Allen, the Air Force Chief . deranee of mformatlon ag I , I
(;ttrrymg a prepon h 30 pereentage points within a smg e year,
of Staff, warns that his planes lack ¡he spare parts necessary to command the skies in any ¡m:ater spending feIl by more t an , . ased defense spending. Sueh
sustained fight. The Chief of Naval Operations, Admiral Thomas B. Hayward, protests " 1'" lopsidedly agamst mere . d '
leaving pub le oplmon, I h'ft in the information carne In
that he has a three-ocean mission and a "one-and-a-half ocean Navy," And for the first , r
changes m pub le oplmon,
.' hnked to e ear S I s
, l ' f this book, especially the latter
time since the missile-gap scare of the 1960 presidential campaign, a feeling is building
that American defenses have slipped - so badly that the nation may no longer be capable elite discourse, are a central tOPIC of ana YSIs o
of protecting its interests abroad, or containing Soviet expansionism. hall' of it.
16 NATURE AND ORIGINS OF MASS OPINION

This section has suggested why and how elite discourse affects mass opinion; 100
The next section wiII consider more carefully which members of the puBlic are
most susceptible to elite influence. 80
Percent
knowing 60
MASS A TTENTION TO ELITE DISCOURSE
Court decision
AIthough most Americans are, to use Downs's (1957) apt phrase, rationaIly ig- 40
norant about politics, they differ greatly in the degree of their ignorance. There
is a smaIl but important minority of the public that pays great attention to poli- 20
tics and is well informed about it. Members of this minority can recognize im-
portant U.S. senators on sight, accurately recount each day's leading news o,-!-------------¡
stories, and keep track of the major events in Washington and other world capi- Politica! awareness
tals. They are, thus, heavily exposed to elite discourse about politics.
Any attempt to gauge the absolute size of this highly informed minority is stimated diffusion of news about Supreme Court decision on ,WebS!er-
r respondents interviewed three days afte~ t~e SU'prem~ Court ~ d¡~CISI
essentially arbitrary (though see Bennett, 1989; Smith, 1989; Delli Carpini and case as calculated from model and coefftclents In P,nce and a e~"
Keeter, in press). Nonetheless, one indication of size is that when respondents to , , , measured by means of simple factual questlOns about pohtlc
reness IS , h' f +2
a National Election Study were asked to name as many members of the U.S. the Measures Appendix. The awareness measur~ In t IS Igure runs - . '.
Supreme Court as they could remember, about 1.9 percent of the public could NES pilot survey.
mention as many as half of the members, and a disproportionate number of
those who could do so were lawyers or educators. 5 Few Americans, it appears, iallegations of scandal, the first ti~e in Am~rican histor~ that
are deeply familiar with the operation of their government. (By way of compar- d The story was heavily covered In the medIa over a,penod of
ison, it is interesting to speculate what percentage of aduIts can name five or ne Yt t when about three weeks after Wright's resignatlOn, a naltiol~.ali<
more starters on their city's major league basebaIl team; almost certainly, the \s:ed abo~t his resignation, only 45 percent could supply any
figure is aboye 1.9 percent.) i'gnation - even so much as abare mention of scandal or wr"ons~~~;~~~~ .•.;;
At the other end of the attentiveness spectrum is a larger group of people who one other example: In the early summer of 1989, th~ U.S.
possess almost no current information about politics. In late 1986, for example, fiounced a major decision on women's right to. a~ortlOn, Wl'~bs,te"i;l
when George Bush was halfway into his second term as vice-president of the 'rive Ser vices. Because pro- and antiabortion actlVlsts he!d
United States, 24 percent of the general public either faiJed to recognize his .·ations in an attempt to lobby the Court, there was extenslve news
name or could not say what office he held. 6 People at this level of inattentive- the impending decision in the weeks before it was t,aken, and very .._..., . ,'....
ness can have only the haziest idea of the policy aIternatives about which poIl- when the decision was finally announced, Yet, m a survey don.e
sters regularly ask them to state opinions, and such ideas as they do have must d .. only about 50 percent of the public could say anythmg
often be relatively innocent of the effects of exposure to elite discourse. (lt~ h~~s:~~'court had ruled, and, as the survey continued over the nex!
Most citizens, of course, faIl between these extremes.? Probably from sorne Al weeks, this percentage feH gradually to about 35 percent (ZaIler and
combination of civic obligation and the entertainment value of politics, a ma-
jority pays enough attention to public affairs to learn something about it. But ·~.,1:9~~ did learo about the abortion decision were obviously not ~ ,random
even so, it is easy to underestimate how Jittle typical Americans know about ¡)se W " . h aid regular attention to pO]¡tICS were
pcrcent of the populatlOn: Cltlzens w o P " h d' d 't Fig-
even the most prominent political events - and also how quickly they forget ¡nore likely to learo about the abortion declSlon than thosde w o d,l nto' their
what for a time they do understand. For example, in the spring of 1989, the . , d t th survey were rate accor mg
2,2 makes this pomt. Respon ents o e " 'h 0-
speaker of the House of Representatives, James Wright, resigned the speaker- oral background levels of "poJitical awarene SS , Persons sco~e~ hl~e ~a:iual
5 No person among the 1,500 respondents to the 1966 National Election Study survey named all (¡'HI awareness if they were able to ,c~rrectly answ~r ~ vt~~e~o~s:I;~epresen-
nine justices,
m,lh¡'!l1;"IICln tests (such as whlch politlcal party con ro s f
6 1986 National Election Sludy,
7 Neuman (1986) attempts to be more precise than 1 am aboul the dislribulion of awareness in the whereas persons scored low on awareness if they could answer n~~eh~
general public, (See more generally Tichenor, Donohue, and ülien, 1970; Delli Carpini and qucstions. As can be seen in the figure, almost all 01' the most Ig Y
Keeler, in press; Bennett, 1989; Just, Neuman, and Crigler, 1989,)
 lnformation, predispositions, and opinion
LO NATURE AND ORIGINS OF MASS OPINION
etl dramatically in the past three decades, with t?e result thah~:o:~t>~-'
informed persons - upward of 95 percent of those interviewed within the fir "; . ¡; ¡; the I'ncumbent House members nave beenable,J)y
. appear sale lor . .
three days of the Court's decision - could, when asked, supply the rudiments rown efforts, to build a "personal vo~e" that IS loyal to
the Court's ruling in the abortion case; but almost none of the persons at the lo t)artisan considerations. Thus it sometImes happens thatha seat
end of the awareness spectrum had absorbed any information about the decisio á particular incumbent for a decade or more, but that w en
Data such as these on differential attentiveness to political news have i tires the seat will quickly become safe for a person of the dO.IPP(!SitICi.';
mense implications for the impact of elite discourse on mass opinion, and takin de~:lopment has given House members an independent stan lOg
systematic account of them will be a central task of this book. . e among legislators in Western democracies and that ~eems
Figure 2.2 also shows that the people most strongly committed to women' ;l;~ected the performance of the American Congress (CalO, Ferc,
right to abortion - in particular, the minority who said women should have a "¡orina, 1987; Jacobson, 1991). . h
absolute right to decide for themselves whether to get abortions - were mo on for the rise of the personal vote, however, remalOs somew.at
likely to find out about the court decision than other persons. Yet their info than half of the eligible electorate can recal! t~e name ~f t~elr
mational advantage was rather modest. There was, moreover, no difference at a r~presentative, and this figure has not changed m the penod m
between men and women in awareness of the decision. Even women of chil shave become safer. But although mosto peop~e c~nnot
bearing age did not differ from the rest of the public in their awareness of thi . bout 80 percent can recognize 11. Thls dlscovery has
Í:l name, a in the votiflg 2
issue (Price and ZalIer, 1990). 1 mention this in order to give pause to reader for a ciaim that much of an incumbent's advantage occur~1 h .
who may suspect that, although citizens are often poorIy informed about politic "ere voters are asked only to recognize rat~er than to reca w o IS
in general, they stilI manage to learn about matters that are especialIy importan " ember of Congress (Mann and WolflOg~r, 1?80). . '
to them. Although there is sorne tendency for this to occur, as emphasized i )r m . h takes place m a 10w-lOforl113.
pical congressional electlOn, tus, f h' bent ana
Converse (1964), Iyengar (1990), DeIli Carpini and Keeter (1990), McGraw an nt in which a few people know the name o t e m.cum
t
Pinney (1990), and sorne other studies, the tendency appears not to be very grea lething about his or her record; many others can, wlth a pr~mtP ,
or very widespread (Price and ZalIer, 1990). e incumbent's name and perhaps hazily recall one or ~wo ac s
The two main points about political awareness, then, are (1) that people var , d or background' and still others know nothmg at al!
on s recor '
greatIy in their general attentiveness to politics, regardless of particular issues;
mben1. I ff, t th capacity of
and (2) that average overalI levels of information are quite low. More succinctly, differences in political awareness gre~t y a e.c e d t
there is high variance in political awareness around a generaIly low mean. develop a "personal vote" among thelr const!tuents - an ye
These points are widely familiar to professional students of public opinion ignored in research on the subject. In consequence, the
(Converse, 1975; Kinder and Sears, 1985; Luskin, 1987; Bennett, 1989). Yet fa-o . d murky To preview arguments that are mote
1vote have remame· ., . l!elleral'¡rre /;;':.
miliarity is often as far as it goes. Most of the time, when scholars attempt to ed in Cha ter 10, the people who know most about ~ohtlcs m
explain public opinion and voting behavior, they build models that implicitly as- . p d to the incumbent's self-promotlOnal efforts.
sume alI citizens to be adequately and about equaIly informed about politics, and :t::;:~tlrca~:~~~~ey are also better ~bl~ eval~at~i;~1yc~~~:eIlY
toh
hence to differ mainly in their preferences and interests. In other words, they ¡he new information they encounter. 0.10 t e en ~ share the party
build models that ignore the effects of political awareness. 8 One aim of this book o be little affected by incumbent camp~lgns. ~~~~ :hether he or she cam-
s of the incumbent, they wIII support~he m~u~ cumbent's values they will
is to provide a corrective for this dominant research practice.
It may be useful to give an example of why a corrective is needed. The ex-
ns vigorously or not; if they do noths areht de ~: campaigns. Me;nwhile, at
ample concerns the effects of campaigns on voting behavior in congressional t h' or her no matter ow ar ..
elections, but ¡he issues it raises paralIel those concerning the effect of elite dis- "se to suppor 1m m those who pay Iittle attention to pohtlcs
low end of the awa~eness spectru, ressional politics. Hence they are
course on public opinion generaIly.
lo get little or no mformatlOn about fc~ng. mbent to build a personal fol-
One of the most heavily researched problems in the congressional elections relatively unaffected by the efforts o t e mcuu ce tibie to influence: They
< {r)wing. This leav~s the moderatel~ ~w~~: :~~~~s~m~ts of the incumbent but
literature in recent years has been the advantage enjoyed by incumbents in the
House of Representatives in their reelection bids. The average winning margin cnough attention to be expose o
Ih(:k.' lhe resources to resist. . F' 2 3 which depicts the relationship
8 Among the exceptions are Stimson (1975), Nie, Yerba, and Petrocík (1976), and theír assorted . th' point IS shown m Igure . , .
critics. Studies done by psychologists and psychologically oriented polítical scientists pay some- EVldence on IS h that people will desert thelr own
what more atlention to polítical awareness. often referring to ít as "polítical expertise" (see Sni-
r' I reness and the e ances
b(:lwcen po Itica awa. . b
.
f Congress rather than thelr own
derman, Brody, and Tetlock. 1991; Kinder and Sanders. 1990; and the 1990 voJume of essays pnrly to vote for an mcumbent mem er o
edited by Krosnick).
 NATURE AND ORIGINS OF MASS OPINION
.75
Proportion of
outpartisans .50
defecúng to
incumbent
.25
.00 -r--'---r--r---.
Political awareness
N= 24 39
48 80 82
Figure 2.3. ~efections to .H~use incumbent among partisans of outparty. See the Mea-
Sures Appendlx for a descnptlOn of the measure of political awareness . Source'. 1978 NES
survey.
party's candidate. The ~ata involve contested seats in the 1978 general election.
As can be seen, defectlOn rates to the incumbent are markedly higher among
moderately aware persons than among persons at the extremes of either high or
low awareness. 9
.N~te that if, as sorne researchers have done, one checks for a linear relation-
shlp m th~se data: one .will.discover little of interest. Awareness has strong ef-
~ects-, but ItS relatlOnshIp .w1th vote defection is nonmonotonic. (A relationship
IS saId to be nonmonotomc w~en the association between variables is positive
over par~ of .the range of the mdependent variable and negative over the other
part, as m FIgure 2.3.)
..What is tme of congressional elections is tme of numerous other cases' Po-
htt~al aware~ess has important effects on many aspects of public opinion' and
V?tI.ng behavlOr, but these effects are often strongly nonlinear. This makes them
dlfflcult to detect and mode!.
Nonlinea~i~y is, however, far from the only complication in detecting the ef-
fe~ts of pohttcal awareness. In the example just discussed, we were concerned
wIth the e~fects of ~ s.ingle.political campaign, that of the congressional incum-
bent. Ob~lOusly, th~s IS a hIghly simplified account of electoral politics. In most
con~resslOnal electlOns,. ~here are two main campaigns, each having sorne ca-
pacIty to reach and mobIlize sympathetic persons (though that of the incumbent
nearly always has more).
Th~ same holds true in other mass persuasion siluations. Public opinion is
sometlmes f?r~ed by streams o~ a monolithically one-sided elite discourse, but,
more often, .1t IS shaped b~ mullIple and typicalIy conflicting information flows,
sorne of WhICh are more mtense, or easier to learn aboul, than olhers. Under-
9 There are three reasons for the small number of cases in Figure 2.3: First. the voting rate in off.
year congressJOnal electlOns IS relallvely low. even among NES respondenls; second, the figure
mvolves only members of lhe party not holding the given congressional seat; and third uncon.
tesled and open seat races have been eliminated. • hilo pcople's minds, which, in turn, is critical for intellectual engagement wlth
Informatíon, predispositions, and opinion
¡he effects of elite discourse on preference formation requires m()d(:lirlg
'ts of awareness in mediating exposure to each of the major cal:npaign
. in the environment, a task that presents serious complications.
lcomplication, as has already been suggested, is that opinion for·mllti()I\.
is¡age process, and awareness may affect different parts of
ly: Political awareness is associated with increased exposure to l:UI:rC;;ll!l
¡calions that might change one's opinion, but it is also associated with
ed capacity to react criticalIy to new information. These two effects
ross cutting, as in the case of congressional elections, where the most
rsons are most heavily exposed tO the incumbent's campaign but
istant. But this needn't be the case. There are, as we shall see, ca:ses In'•. (
e most aware persons are the easiest segment of the public to
, and other cases in which very inattentive persons are most sm,cept.iIDI~ •
sion. SystematicalIy explaining these and other ways in which politiqal,';;
s affects public opinion and voting behavior will be the most impolétal~t
ntribution of this book.
of the central importance of political awareness, ii is worthwhile to di~
iefly to consider how best to conceptualize and measure it.
'ss, as used in this study, refers to the extent to which an individual
to politics and understands what he or she has encountered. Attention
not enough, since people who, for example, watch the TV news while
the couch after dinner and a couple of glasses of wine wilI typicalIy fail
1ce their political awareness.
key to political awareness, then, is the absorption of political communi-
'. Political awareness denotes intellectual or cognitive engagement with
affairs as against emotional or affective engagement or no engagement at
olars have used a wide variety of concepts and measures to capture what
being called political awareness. The concepts in the research literature
le polítical expertise, cognitive complexity, polítical involvement, atten-
css, sophistication, and political acuity. Although choice of labels is per-
tnainly a matter of personal or disciplinary taste, my reason for preferring
Cal awareness is that this term, better than the others, seems to capture the
processes in the model to be introduced here, namely an individual's recep-
'and comprehension of communications from the political environment.
Scholars have also used several different types of questions to operationalize
1 am calling political awareness. These inelude media exposure, political
ij)'lillc'I)atj0l1, education, and self-described interest in politics. As 1 argue in
Mcasurement Appendix, political awareness is, for both theoretical and em-
reasons, best measured by simple tests of neutral factual information
politics. The reason, in brief, is that tests of political information, more
Wrtctly than any of the alternative measures, capture what has actually gott.en
pülitics. Typical information questions, as suggested earlier, ask which political
 lnformation, predispositions, and opinion
NATURE AND ORIGINS OF MASS OPINION
.. " l'Intel' vening variable
party controls the House of Representatives, 01' whether Mainland China is }. WI
'11 'f) predispoSltlons
SpeCI y ,
are the cntlca
d'
.... ; ..... J.'
"sIde'
t,'; communications people encounter in the mass me la, on one '. r .•••,t'i
member of the United Nations. 1
tátements of political preferences, on the othe~. .. b .....(iC ;"i'i~!1
Thus the information tests used to assess political awareness in this bQok a
' of variability in individuals' political predlsposl.tlOns. ~re e.YO~t·.",,"; .
strictIy neutral 01' factual. This point is stressed because, as indicated earlié ces . h t d Sposltlonsare·a .
much of the information carried in elite discourse is neither neutral nor strictl )f this book. My assumptio~, h?wever, IS. t a pr~ :Iudin childhoOº
. distillation of a person's hfetlme expenence~, m g . . ..' '.\!
factual. A news report implying that the Pentagon is awash in scandal an(J mi
~aand direct involvement with the raw .in~red.lents 0df pol~CYt~Ss~~:~;
management, 01' a presidential remark to the effect that most unemployed pe . ..' 'al discnmmatlon, an so 101' ' .....•..
n.. in g a' hVlln , pa~I~~~asX::i~~:~ld economic location and, probaob.: ly at.. • •. •. • .• . • •. • .•
sons could get jobs if they tried hard enough, constitute factual information i g
that they may contain sorne simple facts, and that they convey sincerely he s also part y depe d t stes I ....•...•.••
, . ed 01' ac uired personality factors an a .... •.•.•. . •. .••..
beliefs about factual states of affairs. Yet they are not neutral, since they ha 'ongly, on In~en.t the r~le of elite-supplied information in ShaplIlg::f
been framed for partisan purposes and can be reasonably disputed by fair IS book emp aSlzes ed t have an importá
, 1 wish to stress that elites are not assum o . '. 'd:'
minded people. Sorne kinds of assertions, such as the claim that the spread O ons, . .. 1 my argument pre·1
. . d' iduals' political predlspoSltlons. n ' .• :.
abortion signifies a degradation of American morals, are not even fulIy susce P~:fa:: 1;0 le's responses to elite information in the man~er J~st I
tibIe to empirical verification; yet the broadcast of this c1aim in the media woul . P ~. e not in the short run influenced by ehtes.'.?,
constitute a broadcast of information, since it would involve an assertion about ut predlspOSltlOnS al' the elite ideas that one internalizes hay~;.
the actual state of the nation.
Nonneutral and not necessarily factual information, thus, is indistinguishable
::~~h:~~,~v~~I:~ ~~~go;~;; predispositions. B~t how~velr thi~m:~ ~Y;i;!
.'
fi rmation and change In partlcu al' s 01' ••.
.
from political argumentation. Neutral factual information, such as which part)' i:: s;~:~h~: ;~;:~~:, ~he long-term influence of ~lites on prediSpO",:,
d
th~ exte~t f ;r~i~~~s~;~~~s~e:~~~~~:i willtecei;\j~.
controls Congress, is important in this book insofar as it measures a person's
likely level of exposure to this other, nonneutral and not exclusively factual , that it exists values
information. vanous dlfferen~ ty~es o. ook This is because they seem to have<l,'
To avoid confusion between information of the neutral and nonneutral types, sustained attentlon m thls b . . ' s than any of the other pr~~
I will from this point onward use information exclusively in its nonneutral and more pel~vasive eff;c~fot~;:t~e;~~~~~~s, especially race and party
sense, as in "information about the deterioration of American moral s." Instead ional factors. But. som t and wiIl receive significant attention.
of referring to tests of neutral factual information, as used in measuring po- lent, are also very Importan. d ds" that hold a "more,.cene
litical awareness, 1 will simply refer to tests of political awareness 01' political es refer to "~enera~, a~d . e~?~~ngl ~t~:I:f systems (Kinder and Sears';.
ua
knowledge. sition than attltu~~s m m IVI k Sarticular positions on social issues);,
In order to remind the reader that information is normalIy used in its non- p. 674) and that lead us lO ta e Pie a erson strongly attached to th~
neutral sense, I will occasionalIy place the term in quotes. AIso, for aesthetic . ch, 1973:. p.. 13.)" Thu~, for e~~m~lI 'els:equal, be more likely to reje.B~
reasons, I will sometimes substitute cognates of attentiveness for awareness, as iof economlC mdlvlduahsm wo f.' . 1 e1fare spending than wouId
gument for higher taxes to pay 01' socia w
in "politicalIy inattentive" for "politicalIy unaware."
one less attached to this v,alue: have recently become objects of se"
litical values, understood m.thls w:tKinder and Sears, 1985). Although
POLITICAL PREDISPOSITIONS S scholarly study (for a revle:v" s. t'ng a previously moribund debate
Citizens differ greatIy in their levels of exposure to elite discourse, but these research has been quite. u~eful ~n mv~~~:~~ have two important weaknesses.
exposure differences can, by themselves, explain only a part of the variance in (he structure of mass opmlO ns , It app "bl fc individual differ-
. nahlY and oplmon IS pro ema 1 " ' 1
individual opinions. For citizens are more than passive receivers of whatever Although Ihe academlc hleralure on perso Ih 'usl differences in economlc and socia
, ' cal 'd pear 10 reflecl more an J ' f nalily
l~nces in pohll allllU es ap 'b h se underlying differences m terms o perso "
media communications they encounter. They possess a variety of interests, val- !ocalion, and it seems reasonable lO desc~1 :n~ ~hite 1956; McClosky, 1958; Altemeyer, 1981,
ues, and experiences that may greatly affect their willingness to accept - 01' al- (SCC Adorno et al." 1950; Smlt~, Brune " , "
ternatively, their resolve to resist - persuasive influences. Coslantini and Cralk, 1980; Wllson, 1983~ ve importanl effects on suscepllblhty 10 sub~equent
"lile ideas having been internahzed, can a. f "inertial resistance" to persuaslon In later
In this book, 1 refer to alI of these factors as political predispositions, by . , b
elite influence, ul, as s
1 how In my dlSCUSSlo n o r' I alues '
h t having any intervening effect on po ItJC~ v,
which I mean stable, individual-Ievel traits that regulate the acceptance 01' non- c1npters they are able to do so Wlt ou , ace _ that is the populatlOn IS loo one-
" B 1 there IS 100 hllle vaflance m r 'h' d
acceptance of the political communications the person receives. Because the to- 1,' bcept perhaps race, u , tral predispositional faclOr m t IS stu y,
sloedly white - to perroll makmg race a cen
tality of the communications that one accepts determines one's opinions (by
 bzformation, predispositions, and opinion
NATURE AND ORIGINS OF MASS OPINION

Send U.S. troops

One of these weaknesses wilI be centrally addressed in this book, but the seco A id to Contras 80
must be provisionally resolved by assumption.
The first limitation is that, like most public opinion research, the current r 60
erature on values largely fails to take systematic account of the vast differen
in political awareness that exist among citizens. This failure is unfortunate b Hawks
40
cause a frequent claim of the values literature has been that citizens wl10 a
as most scholars agree, too unsophisticated to possess "ideologies" nonethel
20 Doves
possess sufficient awareness to make reliable use of "values" to structure th
policy preferences. Thus, in a leading example of this research, Hurwitz and Pe Doves o..1-_ _--r-----O
fley (1987) propose a hierarchical model of foreign policy opinions in whi Political awareness
"core values" determine individuals' "general postures," which in turo dete
S ame N
mine opinions on particular foreign policy issues. The fact that many America 34 32
are quite ignorant of foreign affairs is, according to Hurwitz and Peffley, p 44 41
cisely the reason that individuals must often fall back on core values and gene Would you strongly favor, not so
postures to instruct their policy preferences: .you had a say in making up the strongly favor, not so strongly
deral budget, would you like to.s ee oppose, or strOngly oppose .
~nding for aid to the Contras In U. S. troops to Central Ame:nca? to
we see individuals as attempting to cope with an extraordinarily confusing world ... icaragua increased, decreased, or
structuring views about specific foreign policies according to their more general and a stop the spread of cornmUOlsm.
kept about the same?
stract beliefs. (p. 1114) ("Strongly favor" and "not so
. \Percent favoring increased strongly favor" have becn
Although this point is an excellent one, citizens must still possess sorne min spending.l collapsed. ]
imal degree of information in order to recognize the relevance of their values fo
a given issue, and, as 1 have been arguing, it is quite easy to underestimate ho questions about Central America policy. Source: 1987 NES pilot
often even minimal political information may be absent for sorne citizens.
By way of iIIustration, we may examine opinions toward the U.S. policy of
aid to the Contra rebels in Nicaragua. Figure 2.4 shows how citizens who dif- " t the spread of communism" in
fered in both their political awareness and in their predisposition toward use te.s should send troopS to s op b hawks and doves at al\
ere we find sharp differences etween
of military force responded in 1987 to a question on this topic. The persons
cIassified as "hawks" in the figure are ones who said, over a set of general S;ifference in response pattern to the.two. items, espeCtial~y:a'lmf'oOnnmglatiCJfl«
questions, that they strongly value military strength, an aggressive posture to- rk 1 planatron IS the contex ua 11
ward potential adversaries, and uncompromising opposition to communism. aware person~? T~~h~ ~i~s:xalthOUgh scarcely lacking in clarity,
"Doves" are persons who rejected these positions, preferring to emphasize ne- the two questlOns. .' d what they stand foro
ho the Contras are an
gotiations and accommodation with communism. Political awareness in the fig- pondents to k noW w ho are in general, poorly
nt will often go unmet a~ong personskw . 1985 indicated that only
ure is measured by simple tests of factual knowledge about politics. (Items used .. (C mercIaI surveys ta en m
in scale construction may be found in the Measures Appendix.) about politIcs.. om. hich side the United States was sup-
The left side of the figure shows that politicaIIy aware hawks and doves differ Mlf of th~ A~en~an ~ub}¡c knewT~e second question in Figure 2.~, by
greatly on the question of whether U.S. "aid to the Contras in Nicaragua" i~ in the ftghtmg m NIcaragua.) h t th alue implications of the Issue
. . makes clear w a e v
should be increased, decreased, or kept the same: Forty-two percent of the most ming commumsm, .' d h kish foreign policies to recog-
thercby enabling people mclmed towar aw
aware hawks, but only 3 percent of the most aware doves, favored increased
Contra aid. However, among persons in the middle third of the awareness scale, ancl support them. , I redispositions always depends on
hawks and doves differed only modestIy, and among persons at the bottom of the huso the impact of people s val~ec p tion needed to translate their val-
.. the contextua mlorma . f h
scale, there were no value-based differences at aH - a result that raises doubts j¡,llllethcr Cltlzens possess . . d'd t s and the possesslO n o suc
f (lar poJlcles or can I a e , .
whether the hawk-dove value dimension has any utiJity for understanding the \nlo support or par lCU. er be taken for granted. This contmgency
h'li',.,.. m",ricm can, as shown earlier, nev t e particular policies or candi-
views of poorIy informed persons. . . b values and suppor lor
However, the right-hand side of Figure 2.4 supports the conventionaI view of lile rclatronshlp etween sI have indicated, is to show
. . tudy whose purpose, a
the importance of vaJues. lt shows responses to a question about whether the tl.\lles underlies thls entlre s ,
 L.O
NATURE AND ORIGINS OF MASS OPINION lnformalion, predispositions, and opinion

how individuals use information from the political environment to translate th simple addition and multiplication, or a gre~t mathem~tici~n'l1
values and other predispositions into more specific opinion statements. 1 . enerate fluent written prose. A person who .IS ext~aordman y
A second shortcoming of the values literature arises from its failure, so far,
specify the nature of the theoretical relationship of different value continua
:nsion of intelJigence tends to be at Jeast falrly hlgh on othe~s;
. •. >a ears to be true for ideoJogy. It is unusuaJ to encountera ~~l'$:
one another and to political ideology. The problem arises from the fact tha
although there are numerous "value dimensions" between which there is'no o
~i~~aJ on one dimension of ideoJogy and extremeJy c.onservatIv~'i
here is a tendency, which is cJear but not overpower.mg , for pel!tP'éa
vious logical connection, many people nonetheless respond to different val Ut roughJy comparabJe positions on a series of seemmgJy unreC}Z:
dimensions as ifthey were organized by a common left-right dimensiono The
is, in other words, a tendency for people to be fairly consistently "left"
value dimensions. _ . .t . fre.
re two practica . l'Imp II'cations of thls vlew for the measuremen
. l 'p'.f' 2
o<P.<
"right" or "centrist" on such disparate value dimensions as economic ind .' . h' t d First one should whenever posslb e, use appra •.••..
s m t IS s U y . , , h l measute
vidualism, opinions toward communists, tolerance of nonconformists, raci ecific measures of political values, rather ~ an ~ .genera .t •.••......••.
as the operationaJ measure of citi~ens' predlspOSlh?n:h~~~~~:i();.
issues, sexual freedom, and religious authority. The correlations among thes
different value dimensions are never so strong as to suggest that there is on litical communications they recelve. The reaso.n IS . •. '.' .•.
and only one basic value dimension, but they are always at least moderatel J
genera measure of pe ople's Jeft-right tendencles, ISb t T . ~..
more bl<
strong, and among highly aware persons, the correlations are sometimes quit ions to a particuJar issue than is an indicator that has een al or
strong. 13 And, of course, there are also moderately strong correlations betwee
people's self-descriptions as liberal or conservative and their scores on the var-
ious values measures. 'ond implication of this analysis is that, si~ce va~ues a~e, to a
t organized by a person 's general ideologlcal onentatlOn, 0~:1s
What, then, is the nature of the relationship between "values," as examine ' I or omnibus ideology measures to capture peop '.••.. •.
in recent research, and "ideology," which an older generation of researcher ~~~;e~~; often it is necessary. Much of this book .f~cuses
on;
took so seriously? In view of the empirical covariation among measures of the n chan~e, but there are relatively few case.s of mass opml~ chan1~
two concepts, the question seems an obvious and important one. d b high-quaJity publicly aVallable surveys. ence
Let "values" be defined, as they normally are, as domain-specific organizing 11nuse of what y
capture JittJe good, data on opinion chan~~ are av ailable
' . [nsorner.
principIes, such as economic individualism, where each value dimension lends . are exceJIent measures of po)¡t¡cal
e datasets, t here " h values
e Butthat in
structure to public opinions within a particular domain. "Ideology" may then
be defined as a more generaJ Jeft-right scheme capabJe of organizing a wide the value di:::s~~~; :hv~r;e::~:::1 ::a~~;:I~;v~I:;~r;entation,
range of fairly disparate concerns, where the concerns being organized inelude a survey may
I-conservative .
self-Identl 'f"IcatlOn, or measures
. of. value orientª~1
various value or issue dimensions or both.
:~ot particularly close to the opinion that IS unde~g~mg c;ange.
These definitions e10sely link the two concepts without, as far as I can see, $ kind I develop the best measure of generalle~t-ng t .ten e~cy t
violating the conventionaJ meaning of either termo There are, however, two sig- ' , . that one such measure may be, In practlce, amos
nificant novelties. First, the various value dimensions are no Jonger conceptually o ahead, hopmg loyed are always generalIy described in the
independent; rather, each is one among severaJ correlated dimensions of a mas- ther- (The measures emp 'b d' the Measures Appendlx. . ) Th e ;'lstitl.cal:10JLr~'r
l'

ter concept, ideology. Second, ideology is no longer the strictly unidimensional .ok and exactIy descn e m. , th notion that there is a general
concept that many discussions have considered it to be, but a constellation of ractice, beyo~d ~heer necesslt\;~u ~ rnany different value dimensions;
related vaJue dimensions. organizing pnnclple that runs t , g 'T'o the extent that general mea-
. , b 'ously a conservatlve one. l'
1is practlce IS o VI
The dimensionality of ideology may be analogous, in a certain respect, to the ' t ti n fail to capture a pred'IS position that is related to the
of value o~len a o
.
dimensionality of human intelligence. As a Jarge psychoJogical literature has . I tend to et weak or nonexistent relationshlps
shown and as common experience confirms, it is mistaken to say that there is a flf'lrllOn undergomg change, 1 wII l g h' s J have found appear weaker
A d . deed sorne of the re atlans Ip
singJe dimension of intelligence. Thus, we all know people who are better at values. n m " value measures were available,
Sorne kinds of tasks than others - mathematicaJ reasoning rather than verbaJ ex- I believe they would be If stronger . ts in this study J will describe
t .noJagy At sorne pOlO
pression, to take the most obvious case. Yet it is rare to find someone who is Finally, a note on errnl , ." J doing 1 will never mean to
. "'d uasa
ir!(IVI l s "liberal" or "conservatlve, n so , ..
very high on one dimension and very Jow on another _ a brilliant writer who
, , ho tend lo be conservative on economlc Issues
1'1 Tlle obvious exceptíons here are hbertanans, W ff"cienlly uncommon in lhe Unlted Statcs
13 For evidence of the breadth of altitudes apparently organized by the left-right dimension, see
Monroe, 1990; McClosky and Zaller. 1984: chapo 7.
' n I 'ssues But IIbertaflans are su I
:l1ld lIberal on I e-sty e l , , 1 d apparenlly safely ignore lhem.
Ihat studies of Ideology roulme y an
 Information. predispositions, and opinion
l~rU UKt AND ORIGINS OF MASS OPINION

imply that the people so designated are necessarily full-fledged, doctrinai Response stability in repeated interviews: Two examples
ideologues of the left or right. l will mean only that the people tend to b€ el . Oh
to the left or right pole of sorne particular value dimension, or eloser to one Opinion on level of govemroent servIces .
JanUilrJ 1980
the other poIe of the consteIlation of associated liberal-conservativevalu
Cut Middle Same DK
Thus, rather than say that a person is high on a measure of equalitariani
or high on a measure of hawkishness, 1 may say that the person is lib'eral june 1980
54% 38 19 34
conservative. But whichever term l use, the important point to remember is th 25 13 19 Cut
18 24 10 10
for purposes of this book, values and ideology have exactly the same theoretiC 24 17 16 Middle
11 25 59 15
41 60 18 Keepsame
status: They are indicators of predispositions to accept or reject particular p 17 14 13 41
11 10 47 Don't know
litical communications.
100% 100 100 100
100 100 100 138
362 122 208
153 238 74
WHA T IS AN OPINION?
. . . t ~ s to try very hard tog et
John MueIler begins his study of War, Presiden/s, and Public Opinion (197 was: "Some people feel. It IS .Importan or u hard to et along with
with a series of caveats that ought to appear on the opening pages of every boo ssia. Others feel it is ~ blg ~Ist~fv~: ~-point s;ale. Points 1,
em
on public opinion, but which rarely do. He writes: copie were as~~ to p aC~,.t 4 middle' 5, 6, and 7 as "tougher."
en counted as cooperate, as , h Id provide fewer
"S le thinle the govemment s ou d'
The interview situation is an odd social experience. The respondent, on his doorstep o n was: ome peop I h and education in order to reduce spen .mg.
in his living room, is barraged with a set of questions on a wide variety of subjects by ~n in areas such as hea tth ment to continue the services it now
stranger, usually a rather well-educated woman over 30, who carefully notes each r efeel it is important for. e ~ovem. "
en if it means no reductlon In spendmg ...
sponse on a sheet of papero Few people are accustomed to having their every utteranc
faithfully recorded and many find the experience f1attering. And, aware that their view 80 NES panel survey.
are being preserved for the ages, they do not wish to appear unprepared at that momento
Under these circumstances it is not surprising to find respondents pontificating in a seem c . •••.. A record instance of the effects of changes i~
ingly authoritative, if basically "truthful," manner on subjects about which they know 981: p. 277). . 011 in 1983 which found that pubhc SUIlDOrl
nothing or to which they have never given any thought whatsoever.... (p. 1) :my be a New York Times p roduction at that moment a topic of

The consequences of asking uninformed people to state opinions on topics to

e:~~~~o:e~~~~:a;;;:~~::t~nd 83 per~ent, depending on how the
. 15
which they have given little if any previous thought are quite predictable: Their
ed. h few analysts of public opinion fol1ow
sy lto.thin~~~~~~~i: ~a~lOgue of horrors in their o~ening ?ages.
opinion statements give every indication of being rough and superficial. The
opinion statements vacilIate randomly across repeated interviews of the same pJp e ~n ex . dismissed out of hand, losmg thelr au(iiellce
people (see Table 2. 1; also see Converse, 1964; Achen, 1975; Dean and Moran, Víous IS that they fear bemg 'f he too candidly reveal the dubious na c
1977; Erickson, 1979; Feldman, 1989; ZalIer, 1990); entirely trivial changes in ny argument has been made, 1 t Y
questionnaire construction, such as switching the order in which questions are . .' h' t d depends
be data on whlch t en s u y . . 'th t no one knows quite what to
. t ason l beheve IS a .
asked or response options are listed, can easily produce 5 to 10 percentage point more Importan re. ' s o i~ion. M~!.I:!!Y bell~:ve~hat
shifts in aggregate opinion, and occasionally double that (Schuman and Presser, the multlple vaganes of ~as .p.. h . ' dicated by the evidence Just
1981; Bishop, Oldendick, and Tuchfarber, 1984; Tourangeau et al., 1989); and .' bstantIal entIty.t .an.lS.m .. II
e opin~o~l~.~.~?~e~l!:.: . : . . are aH too real. Being unable to square.a
different ways of phrasing questions regularly have large effects on measured . ¡úldy~t tboe gl()~!!1!,.lndlcatl?ns sim ly puts aside the troublmg
levels of public support for an issue. For example, Rasinskí (1989) reports that, with what one believes IS true, one p lo ists to work out, and
. b' g leaving it to survey meth.o<:lüg...
over several surveys, 68 percent of Americans felt too little money was being for t~e time em: . inion one does understand.
spent on "haIting the rising crime rate," but that only 55 percent felt too little about those aspect~ of p~bhc op h 'that it conceals information from
was being spent on "Iaw enforcement." Similarly, 68 percent felt that too little An obvious problem wlth thls .approac IS rtunity of roakíng realistic
h er . that it relmqUlshes the oppo .
was being spent on "protecting social security," but only 53 percent felt this I'caC!cr. Anot h IS ass opinion, in all of its elusiveness, forros and changes.
way about spending on "social security." Or, in another type of case, 45 percent iíwtemcnts about ow ro
of Americans would "not allow" a communist to give a speech, whereas only
20 percent of Americans would "forbid" the same behavior (Schuman and ¡', Judit" Miller. 30 May 1982. p. Al.
30 NA TU RE AND ORIGINS OF MASS OPINION Information, predispositions, and opinion

In view of these considerations, the present study will make no effort to hid
or ~nderp1ay the types of prob1ems with opinion data that have just bd:n d
scr.lbed. Ind~ed, it will make a theory of why the problems exist an integral pa
of ltS analysls. The theory is more simplistic than I would like, but it will at leas
address the problems head on.
This approach is a gamble. Placing at the center of the book a theory 'of the
nature of public opinion - a subject that neither I nor anyone else fully under~i
sta~ds - ties its ~ntire argument to sorne weak reeds, giving critics an oppor
tumty to complam, correctly, that its foundations are uncertain. Whether th
returns on this risky strategy, the opportunity to sketch a unified and realisti
treatment of the dynamics of public opinion, have been worth their cost will b
up to the reader to decide.
Let me begin this part of my argument by recounting in more detail the va .
garies of mass political opinion to which I have alluded; the relatively narrow
gauge theories by which scholars have sought to explain sorne of these phenom •
~na; and the m~re general theory that I will use to explain these findings and t
mtegrate them lOto a model of the effect of elite discourse on mass opinions.
Problems with mass opinion reports: Over time instability
Table 2.1 gives two typical examples of response instability over time. The firs
que~ti?n, from a sample of respondents who were interviewed in January an
agam m lUDe of 1980, asks whether the United States should try harder to co
operate with the Soviet Union, our Cold War adversary, or whether we shoul
'sponse instability of the type shown in Table 2.1 prf~domilnarltIy
le sort of chance variation. But what sort of chance valrialtioln;a.l
eh of it?
. ous paper on "The nature of belief systems in mass publics," Co
argued that opinion instability is due mainly to individuals who'lal?
gs on the given issue but nevertheless indulge interviewers byp
ing as best they can between the response options put in fro~t
often choosing in an essentially random fashion. "[L]arge portI()
rate," he suggested, "simply do not have meaningful beliefs, ev
at have formed the basis for intense political controversy a
pstantial periods of time" (1964: p. 245).
usion has been strongly challenged by scholars who contend t
le's "survey responses" f1uctuate greatly, citizens have und~
itudes" that are overwhelmingly stable (Achen, 1975, 1983; ID
,!977; Erikson, 1979; Judd and Milburn, 1980; Judd, Milburn,
1; Feldman, 1989; Zaller, 1990; an exception is Krosnick, 19
views see Kinder and Sears, 1985 and Smith, 1984). The flu
pear i~ people's overt opinion statements are. attribute~ t~ "
r " where such error is said to stem from the mherent dlffIcult
s'preexisting opinions onto the unavoidably vague language of s
of measurement error has an especially attractive implication.
that attitudinal variables have been measured with large arnlounts fj)t
r it follows that their correlations with other variables will be
I~ted. And if this is so, it is legitimate to reinflate these delJre:sse cl
i

get tougher. As can be seen, 60 percent of those who favored a tougher stand i
by means of standard psychometric techniques,. ,,:hi~h r~searc~~l"s
lanuary still took this position in lUDe; the rest were scattered across the other
. Thus, attitudinal variables which, in fact, exhlblt hlgh mstablhty
~hr~~ options (greater cooperation with Russia, a middle position, or "no opin- ud low correlations with other variables are made, by means of cor-
lO.n ): Of t?ose who took a neutral middle position in January, only 24 percent
cedures, to appear almost perfectly stable and highly correlated with
stdl dld so m lune, with most of the rest now favoring either more cooperation
bIes. In this way, the problem of response instability is rendered not
or less. Altogether, only 50 percent of the respondents took the same position in
lUDe that they had taken in lanuary. (If everyone were simply guessing each time ous but invisible.
-Converse and measurement error approaches to response instability
he were asked the question, roughly 32 percent would be expeeted by chance .
have deficiencies. Converse's thesis, which takes any instabili~y as ~v­
alo~e to state the same opinion on successive interviews, given my recoding of
"nonattitude," was an extreme claim intended to charactenze opm-
the Item.) The same tendencies are apparent in the seeond question, whieh con-
on eertain highly abstraet issues. On more typieal issues, as Co~verse
cerns the proper level of government services. Here, sorne 55 percent of the sur-
tkllS (1979) argue, people's opinions may be more or les~ "erystalhze~"
vey respondents managed to state the same opinion on successive interviews.
, os a result of this, more or less stable. But this only raIses the que~t~on
O~e obv~o~s interpretation of these f1ip-flops is that many people undergo
l cxactly crystallization consists of. Since no one has ever sald: ?pmlon
genume opmlOn change between interviews. The evidence, however, fails to
Ii:lation remains more a metaphor than a testable theory of opmlOn sta-
s~pport this int~rpretation. When the same respondents are asked the same ques-
tlon on three dlfferent occasions, one can typically predict their opinion on the lZrosnick and Schuman, 1988).
ncwcr "measurement error" theory of response instability seems equally
third interview as weIl from the first interview as the second. If changes be-
pccificd at its theoretical coreo When, as all estimates agree, mea~ure­
t~een the first and second interviews represented systematic opinion change,
'\:1'1'01'" normally eonstitutes one-half to three-quarters 01 the vananee
thls would not be possible. The generally accepted conclusion, therefore, has
 32
NATURE AND ORIGINS OF MASS OPINION
010pinion items, one naturally wonders what this chance "error" consists
o.f and how it ?as been generated. Yet researchers have been remarkably Jncu-
nous about thls problem. In a large majority of cases in which it is acknowl-
edged, anal~sts ma~e a statistical correction for it and move quickly on to
whatever thelr study IS mainly about. As a result, "measurement error" is closer
to being a euphemism for "unexplained variance" than it is to being a ~ell­
understood phenomenon (see, however, Schuman and Presser, 1981; Krosnick
and Berent, 1992).
Problems with mass opinion reports: "Response effects"
A second embarrassment to the conventional view of opinions has been the dis-
co~e:y of substantial amounts of nonrandom or "systematic error" in people's
opmlOn reports. Many respondents react to the context in which a question is
asked, to th~ order in .w?ich alternative responses are presented, and to wholly
nonsubstantIve and tnvlal alterations in questions. The systematic effects of
such seemingly irrelevant features of the interview process are known as "re-
sponse effects."
Considero a ~ell-kn?wn experiment during the 1970s on Americans' opinions
to:v~rd SovIet ]ournahsts. In a split-half sample, 37 percent of respondents were
wllhng to allow Communist reporters in the United States. Yet when, in the
other half-sample, respondents were first asked whether U.S. reporters should
be al.lowed in Russia (which most favored), the percentage agreeing to allow
Russlan reporters into the United States nearIy doubled to 73 percent. The ex-
planation for this huge difference, as Schuman and Presser (1981) suggest, is
that when .people ~re asked the Communist reporters item alone, they respond
on the basls of antI-Communist opinions. When, however, the question is pre-
ceded by one about American reporters working in Russia,
a norrn of reciprocity i~ irnrnediately rna~e salient and a substantial nurnber of respon-
dents feel bound to provlde an answer that IS consistent with their previous response....
!he c~ux of the. rnatter seerns to b~ that the reporter questions have two rneanings, one
mvolvmg an attltude toward an obJect and another involving an attitude toward a norrn.
(p. 28)
Note that this explanation implicitly abandons the notion that individuals
possess .a ~i~gle, fixed opinion toward the rights of Communist reporters.
Rather, mdlvlduals are assumed to have at least two considerations one involv-
ing Communists and the other involving the norm of fair play, a~d to answer
the question according to whichever consideration has been made salient by
the questionnaire.
~here are numerous other findings of this type: People are less likely to de-
scnbe themselves as interested in politics just after they have been asked about
obscure issues (Bishop et al., 1984); people's opinions toward abortion are af-
fected by the kinds of items (concerning, for example, religion or women's
rights) that precede it (Tourangeau and Rasinski, 1988; Tourangeau et al. 1989);
lnformation, predispositions, and opinion
'> !!.ive quite different answers to open-ended questions than to q~:~tl~~~[~i~~~¡r'it!
.~ them to choose among a series of prespecified options (S(:hulllma,ií
1987).
Question-wording effects
. 'common for a change in question order to shift public opinion by
or 15 percentage points, and many shift opinion by smaller amounts
11. Changes in the substantive wording of questions can, on the ot
roduce much larger effects on political opinions and can do so mu
liably. Yet these changes are not normally considered ~ither worriso
especially interesting. The feeling seems to be that .dlfferentiy wo.r
ns should get different answers, since they change elther the emO!l
of the issue 01', in sorne cases, what the respondent is being asked¿tb
ot clear, however, that this complaisance is warranted. It is, for exalli'
tablished that adding the endorsement of a prominent politician to a ..
stion - as in, "Do you favor 01' oppose President X's policy of ... ?,l
to change the public's response to that issue, depending on the p.0~
. President X. But if, as conventional opinion models assume, cltl
reexisting "tme attitudes" that they merely reveal to the inqu'
such "endorsement effects" should not occur. The fact tha! t
Uf, and quite reliably, indicates that many respondents are making
pinions - 01' at least editing and modifying them - as they go thro··
estionnaire.
sider another type of question wording effect. In his study of
orean War, Mueller (1973) found that people were more likely to eXIJr~~s .•.•
rt for the war if the antiwar option required them to confess that
try had made a mistake by entering the war. SimilarIy, support for the
consistently 15 to 20 percentage points greater if the war was .
ssary to stop communism. When both factors were at w.ork, thelr ¡VIILll.II}''''·'
on opinion was considerable. Thus, in one poli taken m the fall of
about 38 percent of the public said that "the Korean War has been.
ing"; but in another poli taken at about the same time, 64 percent sald ..
United States "did the right thing in sending troops to stop the Commul1lst
vas ion " (Mueller: table 3.1). As Mueller remarked,
data suggest sornewhat conflicting observations. On .the one ha~d, support for the
was c1early tied to the anti-Cornrnunist spirit in Arnenca at the time. To ?,e~eratc a
01' war fcver, one rnerely had to toss the words, "CornrnuOlst I~vaslon: ~nto the
(1rNcussion. On the othcr hand, the Cornrnunist clcrnent was not entlrely bUllt I~to th.e
íI;r;ponse to the war because Arnericans had tob~ rerninded of it before thclr antl-
C'ornmunism was fully activated .... (Mucller 1973, pp. 46-8)
So, we again find out that a sizable fraction of survey respondents ap~ear to
jorm their opinions during the interview on the ~asis of t~e ~dea,s, made s~hentt.(~
drem by the question, rather than simply reveahng preexIstmg true altItudes.
 Information, predispositions, and opinion
34 NATURE AND ORIGINS OF MASS OPINION

·.f· d' ore than methodological curiosities. They

The counterargument to this conclusion - that different questions were i "¡'In mgs are m . s of "Vvh,lttlil
volved and should therefore be expected to produce different answers ev~n - e conventional view that surveys are passlve measure fr2lcti~ñ:i;':i;r;r;;s~:fii
By believes." More ominously, they raise the fear that a l~rg"en"¡"f'_"
people did have preexisting opinions - does not seem to me credible. The iss
that people were addressing - the appropriateness ofthe U.S. response to an i pinion researchers measure is either .random OT :~tesm:~~IOgists ha
vasion of South Korea by the Communist government of North Korea - was th ast ~ifteen years, survey meth~do~O;~~~ i~~~~: ha~/tended to aba~
same whether or not the survey question used the critical phrase, "Comm~ni d thls problem ~nd sought to ea fixed o inions that they simpl:
invasion." Thus, anyone who had a fixed opinion on the war should have be onventional notton that people possess . P d the "questiÓ,0
d have begun to concentrate mstea on . .' .
able to express it whether communism was mentioned or not. surveys'"abn which individuals conslruct opinion reports m respon
A c1ear demonstration that changes in question wording can change people' g process Y
responses even when the underlying issue remains exactly the same may b rticular stimulus tha~ confronts th~7t~on and Hodges (1991) deserf
found in Tversky and Kahneman's (1982) case of the rare Asian disease. Thes cent example of thls resear~h, t" II a "file drawer" mod¡tl
ional view of attitudes as bemg essen la y
two psychologists put the following questions to a sample of college students:
Imagine the U.S. is preparing for the outbreak of an unusual Asian disease, which is ex : mething such as legalized abortion,t
pected to kili 600 people. Two alternative programs have been proposed. Assume that th pIe are asked how they ~eel about so bl the~ consult a mental file contain
exact scientific estimates of the consequences of the programs are as follows: rry, or anchovies on a plZZ~, pres~~a" Ybortion " or "Uncle Harry," or
Vation. They look for the f¡le mar.e a ,
If program A is adopted, 200 people will be saved. and report the evaluation it contams. . . .
If program B is adopted, there is a one-third probability that 600 people will be saved,
and a two-thirds probability that no people will besaved.
ilson andHod~es (.1991)reje~t tht~:::~:~I;a~~;~t~~~ht~e~~~~~~ts
eonstruct thelr attltudes, rat erb d the ideas in a large but in
Which of these twoprograms do you favor? ·le." Th7~r attitude,~eports a~e ~~:tti~:de" that is reported ~t a gi
This problem requires respondents to choose between a certain fixed loss and I1flicted database, s~ th:,t t h'ch depends on peculiarities of the pr6'...
a gamble having an identical expected loss. In this case, 72 percent chose the a "temporary constructlon w l.
certain loss, that is, saving 200 persons via program A. Yet when a comparable which a person has constructed 11. . .h
sample was asked in a differently worded question to respond to precisely the d Hodges are not the only researchers to e~ter~am sUacballd()lle~!1
on an l' geau and Rasmskl also
S, In a prominent 1988 paper, ouran
same dilemma, the results were radicalIy different. In the second sample, the rep1 it with a
alternatives were described as follows: d.'itionaI view of political attitudes and proposedht? h ace -OlnstruC:t'~:t-'
. " ocess by w IC people l;'
If program e is adopted, 400 people will die. ode! of the '(~qU~~lt:;;t~~~~:~n:it:~heir modellater on.) Similar
If program D is adopted, there is a one-third probability that nobody will die, and a statements. WI b Schuman and Presser (1981),
two-thirds probability that 600 people will die. modest steps have dbeKe~ dtake(~9i7) BarteIs (1988), Kinder and ~a[Il!c:;l-1i
(1984), Iyengar an m er ,
In this situation, only 22 percent chose the certain loss of 400 lives - a re-
), and popkin (1991). the new " uestion-answering
fortunately, most research on d has ha~ nothing to say about .,.;tlhPr-iiíl,n. i
duction of 50 percentage points. The conclusion to be drawn from this example,
as from earlier ones, is that differences in the wording of questions can deter- ,d to focus on response effects, an b which people form
mine how people think about and hence respond to issues even when, as here, . bTt the larger process y
response msta. II Yor . I d from the political environment.
the denotative meanings of the competing wordings are exactly the same. in response to mformatlon g eane . d I
' . b d r question-answenng mo e .
lcrcfore needed IS a roa e . h" b k L"ke a fair number of survey
The need for a model of the survey response t ovide m t IS 00" I .
This is what I attempt o pro don the conventional but implauslblc
There are, to reiterate, two types of evidence that weigh against the conventional ifi1(~t.ho,do)ogísts and psycholog , I aban . d "on every issue about which
lStS
. . 11 ossess "true attltu es .. 1
notion that individuals have preexisting attitudes that they simply reveal in re- that citlzens typlca y p 'd"'. a mode! of how individua s
. . e and mstea propose
sponse to survey questions. The first is the widely replicated finding that 50 to pollster may happen to m~U1r , th particular stimuli that confronl
. . orts ID response to e . . f' .
75 percent of the variance in typical opinion items is random "error" - an c()I1Slruct oplmon rep t of how people acqUlre III 01-
. d l 't . one theory an accoun
(hcm. This mo e um es m
amount that is too large to be comfortably ascribed to the effects of vaguely , r . s sketched .m" t he f'flTSt part
mallon about po ItICS, a .
of this chapter, with an ac-
. ",
late responses to tYPlcal SU! vcy
worded questions. The second is the evidence of large amounts of syslematic
counl of how they use that mformatlOn to OTmu
"error" arising from the effects of question order and question wording.
 lnformation, predispositions, and opinion
NATURE AND ORIGINS OF MASS OPINION

'lOt feature of the model is that people who a.re mor~ ,~~~i~;~;ldk;~"i~:;~t~';¡r;j
items. The model is consistent with - indeed, it seeks to provide explanati .' 1 . bout the "information" that they mternahze -
for - the vagaries of the mass survey response, as outlined in these pag~s. re se ecuve a . . -tent
'U be more likely to reject ideas that are mc~nsls .
The argument, first made in a 1984 convention papel' (Zaller, 1984b), e:~lt of this selectivity, the ideas they internahze are ~ore mter"
roughly as follows: People are continuously exposed to a stream of politi , more consistent with their values. Responses o..
news and information, much of it valenced so as to push public opinion in . ~~:s~10ns reflect thi: by exhibiting greater over time stablhty
direction or the other. But, owing to the generaIly low Jevels of attenÜon
ogical consistency wlth one another.
politics in this country, most people on most issues are relatively uncriti
about the ideas they internalize. In consequence, they filJ up their minds w
large stores of only partially consistent ideas, arguments, and consideratio Background of the question-answering model ,
When asked a survey question, they call to mind as many of these ideas as . uestion-answering approach to underst~ndl
immediately accessible in memory and use them to make choices among the ~Ug~~~~~~e~~~h~ the public opinion field, it is ap?roachmg ,
tions offered to them. But they make these choices in great haste - typically ~~:Xl in sorne psychologic~l circles. lt is instruclive to re
the basis of the one 01' perhaps two considerations that happen to be at the "t y , h h' ergmg orthodoxy rests,
of the head" 16 at the moment of response. esearch on whlc t IS em mathematical psychologists ano
The basic claim of the model, thus, is that survey responses are a function of this research, conducted ?y t f bin filled with múlti
1' rds the mmd as a sor o
immediately accessible "considerations," where the f10w of information in eli l'oblem so vmg, rega ,h t f "stimulus elements" (raw s
.•. t cts Confronted Wlt a se o ,
discourse determines which considerations are salient. The reason for respon ~ons r~. , dividuals stochastically search thelr .
instabiJity, on this view, is that different considerations happen to be saJient slons) m a probJem, m f the stimuli (Atkinson, Bower,
l them to make sense o "
different times, which causes people's survey responses to differ ayer repeate hat enab e , chosen, determmes one s
1965) An interpretlve construct, once , ;'lclguler¡l§it
interviews. Ch~mges in question order or question wording can bring about sy ' , s onse to it A perso n s l'
tematic changes in the considerations immediately salient to people, and henc ~~fq~i~es~~:~~l~n:nh~~~~~en:sSt~:tPhappe~ to' come to the top
systematic changes in their survey responses. , 'h blem-solving exerCIse.
By way of illustrating the operation of the model, consider how typicaJ citi~ ritical pomts I~ ~ e pro n amon social psychologists and
zens might have responded to a question about the proper leveJ of U.S. defense . l' research .tradltlo~, .commo ongthe organization of ideas in
spending during the cold war. Most would have heard a fair amount about the ¡icd with social cogmtl~n, focusehs. " hema" a term that has
t' thls researc IS sc ,
issue without ever having had the occasion to answer a survey question about it. A central concep m h ' a cognitive structure that
,, chology A sc ema IS
They might have been upset about reports of Pentagon waste and mismanage- from cogmuve psy . , d a central value 01' idea, "'"1 ." . . ;0.

, prior information and expenence, aroun ' nce

ment, but they might aJso have worried about the Soviet threat and America's . ' f ew informatlOn an d expene .
capacity to contain it - all without thinking about 01' even recognizing the trade- the interpretatlOn o n ' h t people typically have several
offs between these competing concerns. When unexpectedly asked on a survey ritical point about, schemat~::nt ~enomenon. For example, an .
for their opinion on defense spending, they must have, in just a second 01' two, )Ie for understandmg any g P ~ " would react quite .' Ir
introduced to a "forty-y~ar-old pro es~?r e ty-year-old mother of three,"
somehow puIled these and other thoughts together into a "survey response" on , t ad mtroduced as a lor .. h
defense spending. In doing so, they did not fuIly canvass their minds for aIl áme person were ms e . d different qualitles of t e per-
't' s would come to mm , ,
relevant thoughts. Rather, if they had happened the night before to see a news is different assocta Ion , Id be drawn from the person s
, . d d'ff ent concluSlO ns wou
program on a major defense procurement scandal, they might answer the ques- )would be notlce, 1 er , , "attitude" toward the perso n
. d f th In short the percelver s
tion on the basis of that consideration. But if, on the other hand, they had t'lnensms, an so or ' , 78) in statements that nicely capture a cen-
ulc\ be different. Thus Tesser (19, '.
been reminded by an earlier question about the threat of Soviet aggression they . g suggested wntes.
might instead answer that defense spending should remain high. And if, in a fcature of the mo de I bem ' , And
. . . . the result of a constructlve process. : .' .'
follow-up survey sometime Jater, the survey questions were asked in a different '1llitude at a particular pomt m time IS b' b t' 'r"ather any number of attltudes de-
, . d t ard an o 'leet u" .
order, 01' if they had seen a different TV program the night before, they might is not a single attltu e ow '1 ble for thinking about the ob]ects.
11 the number of schemas aval a
have had different ideas at the tops of their heads and hence made different sur- o . '011 of an object. Fee1ings vary ~epcnd-
vey responses. 1P!crsons do not have a sin!?,l: feelmg or ~:a.I~~~e in," (pp. 297-8, 307, emphasl s 111 lhe
\ng upon the particular cognttl ve schema
16 This useful descriplion, which will recur many limes in lhis book, is owed lo Taylor and
Fiske (1978). original)
 Information, predispositions, and opinion
.. "" UK~ AND ORIGINS OF MASS OPINION
. . u ested by KeIley and by Anderso~.
The key idea in these studies is that individuals do not typically possess "ju <.lisparate mformatlO n , as s gg d ., by averaging overan.·.·
. . d' .d Is make eCISlOns
one opinion" toward issues, but multiple potential opinions. The logital ne tissummg that m lVI ua s' derations, where the size of,.
· ample of relevant con I . ...
question is then how, in the face of this, people manage to come to decisio stoc h as t IC s I d large. The size andCOll}
at al!. I.ls.ideratio may vary b~tweenf antextual and individual moti
ns
:'. I d d on a vanety o con .:.
This turns out to be an immensely complicated issue, involving the encodi .• s. samp e epen . b de salient by the questIon na.l.
.. h what ldeas have een ma .
of incoming information, perception or interpretations of informatioh, ef rs, suc as Ily pays to the subject at hand. ..'
uch attention a person genera
ciency of memory search, and degree of motivation, among other things. Thu
psychologists attempting to come to grips with these various issues have pro
posed models that are fearfully complicated (e.g., Wyer and Srull, 1989). SUMMARY

.,._.. .tt:r:~ leo~~~~ .. ~I~~~~~.~f",7"~

One point that does, however, appear reasonably clear is that, in the course o on interactions
focuses .__.._ -..... ._.. _ ... among m eltte dlscourse, me
making decisions, including those involving political matters, individuals rarel .\ "ariatiQn .inJhe.. tnfQ(!!!tl!lQ.,! ~(l~r: . '. ·· ..d·.. ·•···..a -···=a·.. :~. •.:
~.......,. . '"'' ~_ . , . . h Id be evaluate , m· IVluUi:I
take the time to canvass their minds for all relevant thoughts. Life is too sho smL"';"'"'I''' d··..;..f.·ti·--·-:"""~:"'.·
t how.. . new lOfoXUlat1Q.Jl
ou...... ~ . rse and indivIdual-leve 1 erenc
and the human mind too fallible. Rather, they appear to make decisions "o
the top of the head" on the basis of whatever ideas are immediately acces
in . -._ -_ -:-~Q..JP1~ on
attentton. 9.1~.<;g!!
these ,vanables. · ' ·.· ·e.··.·fne·miX·()f"·
..' ··.·..·.·d·.·.··.eterTIlm... ..
ible in memory. Thus, as Taylor and Fiske (1978) note, numerous studies hay
lues. lnter~l::!!2!!§
-_.._..-..... . ... '!m_.. .
g. . .
l ' heads" . ,""" . . ·· ····W" lC'h
h " .. of .these consldera"
-....-
" that gets.lOto".pe.o..!l .es...· ~-"""'. etU
shown that the introduction or emphasis of a single piece of information - suc .' .- h . omentoLCQnfrontlOg~UrY.e.~ ..' 1"-.
~the. t0p·0f.the."heal!·"at-.~ e..tn d 1 based on these ideas lS
as the fact that a particular person is a woman or a lawyer - can greatly affec responses t~ the__.._.__
questtons. Am()
.,..... . e _ . ".....""~ .
subsequent expressions of opinion. Reviewing a variety of such evidence, Taylo
and Fiske maintain, in an argument quite novel for its time, that many people
he nexfSÍ1aptyr.
make social judgments by seizing on
a single, sufficient and salient explanation ... often the first satisfactory one that comes
along.... [I]nstead of employing base rate or consensus information logical!y, people
are more often influenced by a single, colorful piece of case history evidence.... In-
stead of reviewing al! the evidence that bears upon a particular problem, people fre-
quently use the information which is most salient or available to them, that is, that which
is most easily brought to mind. (p. 251)
On the basis of a much larger volume of evidence, Wyer and Srull (1989)
maintain that people are
unlikely to conduct an exhaustive search of memory for al! of the knowledge they have
accumulated that is relevant to a particular decision. Rather they retrieve and use only a
smal! subset of this knowledge, apparently assuming that its implications are represen-
tative of al! the knowledge they have acquired. (p. 81)
Yet, at the same time, much data in political science (Kelley, 1983), political
psychology (Lodge, McGraw, and Stroh, 1989), and cognitive psychology
(Anderson, 1974) make it cIear that individuals may often utilize many diverse
pieces of information in their decision making. For example, KeHey (1983)
shows that voters decide between presidential candidates as if they were sum-
ming up numerous "likes" and "dislikes" about each party and candidate and
choosing the one with the highest net total. Anderson's information-averaging
models, which have achieved wide recognition in psychology, likewise show
that individuals make use of a wide set of relevant cognitions in formulating
opinion statements.
The model 1 propose tries to accommodate both top-of-the-head decision
making, as suggested by Taylor and Fiske, and averaging across potentially large