Companion to “Re(: )Measuring Political Sophistication”John BullockPolitical Science 353October 16, 2004Abstract: The paper I will present on Thursday describes different ways in which political knowledge is measured and shows that some are betterthan others. This companion offers context: it defines sophistication,knowledge, and the relationship between the two, all in more detail thanThursday’s paper provides. It also summarizes the findings in Thursday’spaper. It borrows heavily from Luskin (1987), to which the reader isreferred for more background information.As a variable in studies of mass behavior, political sophistication is widely used but notwell understood. The term refers, roughly, to the number and connectedness of one’s politicalthoughts. But it is now measured almost exclusively through knowledge-holding—that is,through counts of correct answers to factual questions. The current practice gives rise to a gooddeal of confusion, much of it epitomized by a common refrain: “What does it matter if a votercan’t name the Chief Justice of the Supreme Court?” This is not a bad question under any circumstances. But it may reflect misunderstandingof the purpose of knowledge measures, which is not—or should not be—simply to measureone’s stock of knowledge. Rather, knowledge-holding is used as a proxy for politicalsophistication (hereafter just “sophistication”), which is a more complex variable. The chiefpurposes of this document are to provide a brief introduction to sophistication and to explainwhy it might plausibly be proxied by knowledge-holding. A secondary purpose is to provide abrisk introduction to some of the ways in which Robert Luskin and I have studied sophistication.Sophistication lurked in the shadows of political science throughout the first half of thetwentieth century. We saw it fleetingly in Bryce (1910), Lippmann (1922), Berelson, Lazarsfeld,and McPhee (1954), and Downs (1957). But it did not receive full consideration until TheAmerican Voter (1960) and Converse (1964) thrust it center stage. It remains there todaybecause of the strong effects that have since been attributed to it. We now know the variable as amover in campaign contributions (Delli Carpini and Keeter 1996), direct participation incampaigns (Leighley 1991), and turnout (Neuman 1986). It furthers and is furthered by interestin politics (Luskin 1990). It makes people more likely to judge political messages by theircontent rather than solely by their sources (Iyengar and Valentino 2000). The sophisticated aremore likely to resist weak or specious counterarguments (Sniderman and Jackman 2004), moresusceptible to priming effects (Krosnick and Brannon 1993); more likely to vote along partylines (Zaller 1992), and more likely to decide on a candidate months in advance of an election(Delli Carpini and Keeter 1996). They also have different, generally more liberal policypreferences (Althaus 2003, Delli Carpini and Keeter 1996) and they may choose differentcandidates (Bartels 1996, Luskin and Globetti 2003)—but I will return to this latter possibility atthe end of this paper. To boot, many of these already-sizable effects have been underestimatedbecause of poor measurement at the high end of sophistication scales (Bullock 2004).1 The variable we use today is the one Campbell et al. and Converse discussed at length:the measurement has changed, but the concept remains the same. Now, as then, sophisticationrefers to the breadth, depth, and organization of political cognitions. The simplest cognitionsare bits of memory, things that can be recalled: the fundamental distinction between cognitionsand noncognitions is that only the former can be called into working memory. More complexcognitions link and contain simpler ones in much the way that sentences link and contain words.Beliefs—almost all of them—perform this linking function, as is clear when we consider theforms that beliefs typically take: x is a y; x causes y; if x, then y. In political science, it is typicalto speak of preferences and attitudes, which for our purposes are beliefs.22The terms used here have little currency in psychology. Our “sophistication” is often replaced byexpertise, with which it is synonymous (Anderson 2000, but see Krosnick 1990 for a more catholic1The citations here are to explicitly political articles. But in some cases, similar points were madeearlier in apolitical contexts: for example, compare Iyengar and Valentino (2000) to the review in McGuire(1985), or Sniderman and Jackman (2004) to Petty, Wells, and Brock (1976).The political dilettante who knows a little about many things is said to possess breadthbut not depth. The specialist who knows much about only one thing possesses depth but notbreadth. In practice, though, the distinction between breadth and depth has been abandoned: it isnow more common to speak of “differentiation” or “scope,” which encompass both qualities.The assumption is that breadth and depth are highly correlated in politics. It is hard, by theselights, to find someone who knows a fair amount about one thing and nothing about other things. Organization is the extent to which one’s cognitions are associated. Associationsbetween cognitions are the previously mentioned “complex cognitions” that link simpler ones;they are the “beams and girders of cognitive organization” (Luskin 1987). There is no absolutedistinction between cognitions and associations: as Scott (1963, 266) has it, “though a structureconstitutes a relation among elements, it may itself form an element in some more superordinatestructure.”In principle, cognitions can be organized—that is, associated—without the aid ofhigh-order abstractions like (but by no means limited to) “liberal” and “conservative.” Inpractice, abstraction and sophistication are inextricably intertwined. Campbell et al. suggest whythis taxonomic approach to association between cognitions should be so common:…any cognitive structure that subsumes content of wide scope and high diversity mustbe capped by concepts of a high order of abstraction.…Perceived events and states aregiven meaning because they may be coded into classes. The wider the range of objectsso classified, the more remote and general the concept that is necessary to capture theirsimilarity. (193)In practice, this is exactly how we often think and speak of politics. We recognize, for