The work of Hayes & Broadbent (1988) provided a method of inducing explicit or implicit processes in individuals who were learning a novel computer task.
Elucidating Need for Cognition Using a Procedural Learning Paradigm by Barry Radler
A dissertation submitted in partial fulfillment of requirements for the degree of
Doctor of Philosophy (Mass Communication)
at the University of Wisconsin - Madison 2000
i
Elucidating Need for Cognition Using a Procedural Learning Paradigm Barry Radler Under the supervision of Professor Robert P. Hawkins At the University of Wisconsin - Madison
Abstract The Need for Cognition (NFC) Scale (Cacioppo and Petty, 1982) is widely used to determine an individual’s disposition to engage in elaborative cognitive activity. While its validity and reliability are generally well established and it has been used to predict how people respond to persuasive messages (Cacioppo, Petty, Feinstein, and Jarvis, 1996; Petty and Cacioppo, 1986; Sadowski, and Gulgoz, 1992), there has been little examination of the processes at work which give rise to an individual being categorized as possessing high or low NFC. Low NFC in particular lacks a satisfying theoretical definition, being tautologically explained as the absence of high NFC. This paper attempted to examine the cognitive processes of high and low NFC individuals with an emphasis on elucidating low NFC processing. Current cognitive psychological theory gives much credence to the idea that information processing is an interaction between two typified cognitive subsystems, one explicit and one implicit (Sloman, 1996). This dichotic
ii conceptualization of thought seemed to provide a more complete and parsimonious conceptualization of both high and low NFC. The work of Hayes & Broadbent (1988) provided a method of inducing explicit or implicit processes in individuals who were learning a novel computer task. In an attempt to replicate their procedural learning experimental design, high and low NFC subjects interacted with a computer under either salient (explicit) or non-salient (implicit) conditions. In attempting to control a computer personality’s “mood,” high NFC individuals operating under salient conditions tended to perform optimally, but with noticeable deficits when required to process multiple variables. Low NFC individuals under non-salient conditions were most likely to improve their performance and reaction times as number and complexity of variables increased. Declarative and procedural verbal knowledge data indicated that the experimental condition of saliency was more responsible than NFC for an individual’s ability to make an accurate verbal explanation of his or her strategy. The results provide strong, albeit mixed, evidence that the absence of effortful cognitive activity is a less than adequate explanation of the processing of low NFC. Given the current experiment’s results, a better explanation of the phenomenon is that high NFC individuals can be described as tending to use a rule-based, systematic cognitive subsystem while low NFC individuals tend to use an associative, intuitive subsystem.
iii Table of Contents Page I.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
II.
Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Associative system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Rule-based system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Identifying systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
III. Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 IV. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Computer person task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 V.
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Performance data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Latency data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 Verbal knowledge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Additional analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
VI. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 VII. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 VIII. Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 IX.
Appendix B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
iv List of tables Table
Page
1.
Process Dichotomies (Abelson, 1994) . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.
Characteristics of Two Cognitive Subsystems (Sloman, 1996) . . . . . . . 13
3.
Experimental Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.
Average Number of Trials Correct . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.
Mixed Repeated-measures Summary for Performance . . . . . . . . . . . . . 44
6.
Performance T-tests for Each Experimental Group . . . . . . . . . . . . . . . 47
7.
Average Reaction Times in Seconds . . . . . . . . . . . . . . . . . . . . . . . . . . 49
8.
Mixed Repeated-measures Summary for Latency . . . . . . . . . . . . . . . . 51
9.
Declarative Knowledge Scores, First Session . . . . . . . . . . . . . . . . . . . . 53
10.
Declarative Knowledge Scores, Second Session . . . . . . . . . . . . . . . . . 53
11.
Crosstab of Procedural Knowledge, First Session . . . . . . . . . . . . . . . . 54
12.
Crosstab of Procedural Knowledge, Second Session . . . . . . . . . . . . . . 55
13.
NFC Crosstab of Procedural Knowledge . . . . . . . . . . . . . . . . . . . . . . . 56
14.
Saliency Crosstab of Procedural Knowledge . . . . . . . . . . . . . . . . . . . . 56
15.
Mixed Repeated-measures Summary for Performance Reversal . . . . . . 61
16.
Mixed Repeated-measures Summary for Latency Reversal . . . . . . . . . 63
v List of Figures Figure
Page
1.
Typical Relationship of High and Low NFC Subjects to Argument Quality . . . 5
2.
Performance Results from Hayes and Broadbents’ Second Experiment . . . . . . 25
3.
Hypothesized Results with NFC as a Single Process . . . . . . . . . . . . . . . . . . . . 29
4.
Hypothesized Results with NFC as a Dual Process . . . . . . . . . . . . . . . . . . . . . 33
5.
Factor Scree Plot for 18-item NFC Scale, First Six Factors . . . . . . . . . . . . . . 39
6.
Mean Number of Correct Trials During Last 10 Trials of Each Phase for Salient/non-salient Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
7.
Mean Number of Correct Trials During Last 10 Trials of Each Phase for NFC Condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
8.
Mean Number of Correct Trials During Last 10 Trials of Each Phase for Experimental Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
9.
Mean Reaction Times During Last 10 Trials of Each Phase for Experimental Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
10. Mean Number of Correct Trials in 10-trial Increments for 2nd and 3rd Phases . . 60 11. Mean Reaction Times in 10-trial Increments for 2nd and 3rd Phases . . . . . . . . . 62 12.
Mean Number of Correct Trials for First Session in 10-trial increments . . . . . . 64
13.
Mean Reaction Times for First Session in 10-trial increments . . . . . . . . . . . . . 65
1
Introduction Since Cacioppo and Pettys’ (1982) seminal paper on the Need for Cognition (NFC), social science research has widely employed the NFC scale in an effort to measure people’s propensity for effortful thought. Perhaps because of Cacioppo and Pettys’ work on the Elaboration Likelihood Model (ELM), most work using NFC has been concentrated in persuasion. In this field, NFC has proved useful as a predictor of the amount of thought people tend to expend when thinking about persuasive messages. Yet while it has become a pragmatic tool for prediction, its use has become almost cavalier and its contribution to the development of human information processing theory has languished. One reason for this is that many studies explain NFC through correlation and not causation, through product and not process. In order to further inform human information processing theory, NFC must find a place within it. Doing so will not only contribute to the knowledge-base of cognitive and social psychology, but allow NFC to be more usefully and appropriately applied in the field.
2
Literature review One of the most popular theoretical rubrics in cognitive psychology for the past 20 years, and one which the ELM is based on, is the idea that information processing consists of the workings of two distinct but interacting subsystems (Abelson, 1994; Sloman, 1996;). The ELM makes reference to these two subsystems in its language of central and peripheral processing, but similar lines of research have variously described processes typified by the thoughtful reasoning of a “high” end and the less discriminating “low” end. The “high” end subsystem has been referred to as elaborative, deliberative, rule-based and systematic. In the ELM, this subsystem is that purported to be engaged when a person processes information centrally. In contrast, the “low” end subsystem has been referred to as associative, unselective, unintentional and heuristic-based. This subsystem is that purported to be engaged when a person processes information peripherally. In commenting on the plethora of dual-process theories developed in the past 20 years, Abelson (1994) stated that the social cognition field has realized a higher level of realistic complexity to explain the enormous variability with which people process information, be it person perception, message processing, decision-making, etc.. He refers to the hypothesized existence of coacting central subsystems, two qualitatively distinct cognitive (or motivational)
3 activities that are carried out in parallel1. Table 1 shows the spate of studies that Abelson refers to as proposing such a dichotomy.
Table 1: Process Dichotomies (Abelson, 1994). Source
System A
System B
Zajonc (1980)
Judgmental
Affective
Langer (1989)
Mindful
Mindless
Devine (1989)
Self-conscious
Habitual
Dovido and Fazio (1992)
Deliberate
Spontaneous
Schneider and Shiffrin (1977)
Controlled
Automatic
Uleman and Bargh (1989)
Intended
Unintended
Banaji and Greenwald (1993)
Explicit
Implicit
Self-interested
Symbolic
Instrumental
Expressive
Bruner (1986)
Logical
Narrative
Nemeroff and Rozin (1990)
Rational
Metaphorical
Berry and Broadbent (1988)
Selective
Unselective
Kinder and Sears (1985) Katz (1960)
While it is generally accepted that the information processing high NFC individuals exhibit corresponds to that typified by the studies under System A in Abelson’s table, the opposite is not necessarily true. A problem exists in placing NFC into such a dichotic, dual-process theory of cognition. First, NFC is purported to be an intrinsic motivation to engage in effortful cognition,
1
The parallel-processing controversy is not taken up here. The purpose of this paper is to use NFC to explicate the typical processes of these two subsystems.
4 and research has convincingly demonstrated this tendency (Thompson, Chaiken, and Hazlewood, 1993). Sadowski and Gulgoz (1992) found substantial test-retest reliability over a seven-week time span (r69 = .88, p