50Dispositional Differences in Cognitive Motivation

John T. Cacioppo, Richard E. Petty, Jeffrey A. Feinstein and W. Blair G. Jarvis, "Dispositional Differences in Cognitive Motivation: The Life and Times of Individuals Varying in Need for Cognition", Psychological Bulletin 119(2), pp. 197–253, 1996

Linked to 14 chapters: supports 7 · contradicts 2 · mechanism 7 · same claim 2 · pairing 2

Psychology in the 1990s carried two accounts of the ordinary thinker side by side. One held that people search for information and solve problems effortfully as they steer through the world. The other held that most judgment runs off automatic processes and cursory attention to whatever is in front of them. Research had largely treated the switch between the two as a property of the situation: raise the personal stakes and people think, lower them and they coast. Cacioppo, Petty, Feinstein and Jarvis review more than a hundred empirical studies to argue that a second source of variance sits alongside the situation and is stable in the person. Some people act as cognitive misers in circumstances that make almost everyone else think hard; others think hard where almost everyone else is lulled into repose. The review traces the term from its origin in the 1950s to the redefinition Cacioppo and Petty gave it in 1982, sets out the scale and the evidence that it measures one thing, separates the disposition from intelligence, and assembles what high and low scorers do differently when they attend, remember, elaborate, are persuaded and resist persuasion. The picture that comes out is consistent, and in the size of its differences, modest.

The term and what it came to mean

The name is older than the construct. Cohen, Stotland and Wolfe introduced need for cognition in 1955 as "a need to understand and make reasonable the experiential world", and argued that a stronger need leads a person to see a situation as ambiguous even when it is relatively structured, because the standard for cognitive clarity is higher. That is an outcome: what the person wants is a world without loose ends, and the tension relieved is the tension of not knowing. The study that first supported it, in 1959, expected the high scorer to reach an ordered picture of the world by using heuristics and taking the advice of experts. Cohen's group never published their scale, and by the time Cacioppo and Petty came to the problem in 1982 no copies of it survived.

Cacioppo and Petty kept the term out of deference to that earlier work and changed what it referred to. Need for cognition in their hands is a stable tendency to engage in and enjoy effortful cognitive activity: a process rather than an outcome, what a person does with a problem rather than the state of the world they want at the end of it. The continuum is bipolar and single, low need for cognition being defined as the relative absence of the motivation that defines high need for cognition, as cold is the relative absence of heat. It is not a need in the drive-reduction sense at all; the authors call it a stable though not invariant intrinsic motivation, one that can be developed or lost. Taylor's cognitive miser gives them one end of the scale and the chronic cognizer the other. The image the paper uses is magnetic: information in daily life as fields of iron filings, people as magnets, and need for cognition as the strength of the field.

The redefinition reverses a sign. Where Cohen's high scorer wanted structure and would take an expert's word to get it, the review's high scorer is the one who scores lower on need for structure, need for closure, preference for order, preference for predictability, closed-mindedness and causal uncertainty. The two constructs share a name and little else, and the closest modern relatives of Cohen's are the tolerance-of-ambiguity and need-for-closure scales, which sit on the opposite side of the correlation table.

Scaling the difference

The scale was built from known groups. Cacioppo and Petty wrote 45 items about reactions to demands for effortful thinking across a range of settings, deliberately excluding reverie, daydreaming, religious experience and artistic rumination, which are cognitive without being effortful. They gave the pool to members of a university faculty as the high group and to assembly-line workers as the low group, on a nine-point scale, anonymously and embedded among other items, because face-valid items invite people to answer as they would like to be. Screening for ambiguity, irrelevance and internal consistency left 34 items, and a principal-components analysis with a scree test found one dominant factor, taking 30.1 per cent of the variance against 6.8 and 5.4 for the next two. A cross-validation on 419 undergraduates gave a narrower range of scores and a smaller first factor, 20.0 per cent, but the item loadings on that factor correlated .76 across the two studies. Neither study found a difference between men and women, and none of the many studies since has found one either.

Two years later Cacioppo, Petty and Kao found that Cronbach's alpha stopped rising once the 18 highest-loading items were in. The short form correlates .95 with the long one, has an alpha of .90, and yields a first factor accounting for 37 per cent of the variance, a cleaner single-factor solution than the scale it was cut from. Those 18 items are what most of the literature uses, half of them reverse-scored: one asks whether the respondent would prefer complex problems to simple ones, another whether thinking is their idea of fun.

Replications carry the structure outside the undergraduate pool. Perri and Wolfgang mailed 16 of the items to a random sample of patients of four physicians; the 94 who returned them produced one dominant factor at 38.8 per cent and an alpha of .88. Pieters, Verplanken and Modde translated the short form into Dutch and found a dominant first factor twice. Alphas at or above .85 are the norm. The one serious challenge has not held. Tanaka, Panter and Winborne proposed three subscales, cognitive persistence, confidence and complexity, from the original 45-item pool answered true or false; across their own two samples the loadings on those factors correlated −.01, .70 and .28, against .76 for the single factor in the original studies, and no subscale has been shown to predict anything the total score does not.

A motivation, not an ability

The claim that most needed defending is that this is a difference in wanting, not in being able. Intelligence and need for cognition should covary, since people who are good at thinking collect more competence feedback for doing it, and they do: the review's pooled correlations are .24 with verbal intelligence, .26 with college entrance scores, .26 with high-school grades and .17 with college grades. None of these is large, and with abstract reasoning ability the correlation is −.03. The functional test matters more than the statistical one. When students read a persuasive message, need for cognition predicted how many of its arguments they could recall even after the variance explained by measured intelligence had been removed. The analogy the authors offer is the motivation to take physical exercise, which is related to physical ability and is not the same thing.

Where the disposition sits among other traits is mapped by correlations with something close to a hundred scales. It goes with attributional complexity (.36 to .51), with basing beliefs on evidence (.47), with curiosity (about .57 across scales), and most strongly with openness to ideas (.78), less with openness to actions, feelings and values, and not at all with openness to fantasy. It runs against dogmatism (−.23 to −.24), attention to social comparison cues (−.41), and the whole family of structure-seeking measures. It is unrelated to sociability, extraversion, shyness, loneliness and trait anger, so the high scorer is not the cold one; high scorers in fact score lower on alexithymia, reporting less difficulty in identifying and communicating what they feel. What the disposition does go with, besides thinking, is agency: private rather than public self-consciousness, self-esteem (.38), desire for control (.48), and lower neuroticism and social anxiety. Education correlates .25 and age about −.17, and whether education raises the disposition or the disposition selects people into education cannot be told apart from the data available.

Engagement in effortful cognitive endeavors

Everything after this rests on whether an eighteen-question form predicts behaviour, and the first test was recall. Cacioppo, Petty and Morris gave students an editorial containing six strong or six weak arguments for requiring seniors to pass a comprehensive examination, then five minutes to list the arguments. High scorers produced about two thirds of them and low scorers about half, whatever the quality of the arguments. Twenty-two further tests followed, on trial evidence, faces, brand names, events in a video, traits about the self and behaviours ascribed to strangers, and the pooled difference is d = .355. It disappears where the material is irrelevant to the task at hand, and grows where the material is task-relevant and inconsistent with what the person expected, which is the case that takes work to resolve.

The second test was sensitivity to argument quality. Students high and low in need for cognition read four strong or four weak arguments for raising their own tuition; the high scorers' attitudes moved with the quality of the arguments and the low scorers' moved less. Eleven tests of that interaction pool to d = .312, and across the five studies that asked participants to rate the message directly, the ratings of high scorers tracked the manipulated quality far better (d = .539). Low scorers are not incapable of scrutiny; they need a reason to bother. They examined the arguments when the source was untrustworthy and not when he was trustworthy, and examined them when the arguments arrived framed in a way they had not expected and not otherwise. The attitudes of high scorers tracked argument quality in every condition.

What moves the low scorer instead is the cue. Participants heard a speaker announce that he would give either ten reasons or two, and then give six either way; only the low scorers agreed more with the speaker who had claimed ten. Only low scorers were swayed by how warmly the audience was said to have applauded, and then only when the topic had little personal relevance; when it mattered to them, the cue moved nobody. This is the ground the elaboration likelihood model occupies, and the model also explains where the interaction fails to appear. A variable such as the source's credibility, or the listener's mood, can work as a simple cue when nobody is thinking and as a bias on thinking when everybody is; it then shifts both groups' attitudes for different reasons, and no interaction with need for cognition shows up at all.

Thought-listing gives the same answer in a subtler form. High scorers list somewhat more task-relevant thoughts (d = .205), but the difference the authors find more telling is in the profile rather than the count: the balance of favourable and unfavourable thoughts of a high scorer follows the quality of what has just been read, and that of a low scorer does not. The judgments then follow the thoughts. In one advertising study, high scorers' attitudes correlated .39 with the favourable thoughts they listed and −.58 with the unfavourable ones, while low scorers' attitudes correlated with neither; for a low scorer the better predictor is the sheer number of arguments recalled. Downstream, high scorers list more about presidential candidates, do better on a test of trivia (d = .734), and are aware that they know more; they perform better at mental arithmetic, anagrams and coursework, with the gap widest where mastering the course requires thought. Their attitudes persist longer, resist counterattitudinal attack better, and predict behaviour better, which is what the elaboration likelihood model says an attitude built out of thinking should do.

The review is careful not to let this become a story about accuracy. Thinking more is not thinking impartially. A positive mood made high scorers' thoughts about a foster-care proposal more favourable and their attitudes followed their thoughts, while the same mood shifted low scorers' attitudes directly, without passing through anything they thought. When the message was recast as the costs of not complying rather than the benefits of complying, mood reversed its effect on high scorers and left low scorers untouched. Priming the concept of winning or losing changed how much high scorers bet on a roulette wheel and left low scorers' bets alone. Where the high scorers differ is at the next step: once a biasing factor is made obvious they are the ones who spend the effort to correct for it, discounting a dispositional attribution in the light of the situation. A correction is not the same as an accurate judgment, the authors add, since nobody knows the true size of the influence being adjusted for.

Enjoyment, and where the difference disappears

Half the definition is enjoyment, and the cleanest demonstration of it is the dullest task in the review. Cacioppo and Petty gave participants tables of 3,500 random digits and ten minutes, and told half of them to circle every 1, 5 and 7, and the other half to circle every 3, any 6 that came before a 7, and every other 4. High scorers reacted more positively to the complex version and low scorers to the simple one. Neither group enjoys tasks more in general; they enjoy different tasks. Pooled across studies, high scorers report less annoyance, frustration and tension at demanding tasks (d = −.447) and more enjoyment of them (d = .306), and the studies that break the pattern are the ones whose tasks were too easy to test anything.

Two experiments treat the disposition as an intrinsic motivation and test it as one. Participants brainstormed with or without an extrinsic reward and were then left alone with extra brainstorming puzzles. The high scorers played with the puzzles more when they had not been paid and the low scorers more when they had, so that the undermining effect of reward appeared only in the group for whom the activity was intrinsically rewarding to begin with. In the second, people brainstormed either alone or as part of a group sharing responsibility for the result. Low scorers produced fewer ideas in the group, the familiar social loafing; high scorers produced as many either way. When the task was screwing and unscrewing nuts and bolts, both groups loafed. High scorers also seek out more information (d = .457) and, in a sample of 233 residents of one Florida city, read newspapers and magazines for news and watched less television.

The review's summary figure is an admission of limits. The difference between high and low scorers is largest when the situation's demand for thought is neither very low nor very high, because at the bottom nobody thinks and at the top everybody does. Eight weeks before a presidential election, high scorers reported thinking about it more than low scorers; by election day the two groups reported the same. Where a task was made personally relevant, low scorers reported slightly more effort than high scorers, which the authors read as unfamiliarity rather than diligence: effort feels heavier to someone who does not often exert it. There are also conditions in which a high scorer should think less rather than more, because the thinking is already done — they answered questions about a message's unstated implications faster than low scorers, and judged whether a trait described them faster while judging whether the same word was a noun no faster at all.

Two questions are left open at the end. The first is where the disposition belongs in the taxonomy of personality. The authors place it under the fifth factor, openness or intellect, note that its agentic side points at conscientiousness as well, decline to choose, and observe that the five-factor account describes how trait words covary rather than what causes anything. The second is where the disposition comes from. The theory says it should grow out of competence feedback, so that children who find they can settle their problems by reasoning rather than by force come to enjoy reasoning, and that it can be worn away by surveillance, time pressure and payment that turns the thinker into a pawn. Neither claim has been tested. The review names what nobody has looked at — the metacognition of high and low scorers, the development of the disposition in childhood, any genetic contribution beyond the one running through intelligence, samples in which age and education are not confounded — and closes on the line that gives the construct its character: high scorers "view cognitive effort not as toil but as a way of life".

Since 1996

The eighteen-item scale outlived nearly everything else in the review. It has been translated many times over, cut down to six items for survey work, and used across education, health communication, consumer research, political science and law, while the review itself became one of the most heavily cited papers its journal has published. The single-factor structure has held. So has the separation from intelligence, confirmed since with better ability measures than the college entrance scores the review had to lean on. The question the authors left open about the five-factor model was answered roughly as they guessed it would be: openness, or intellect, is the largest correlate, with a smaller but real relation to conscientiousness.

Two parts have travelled less well. The first is the evidence on bias. Several of the demonstrations that high scorers are the more susceptible to primes belong to the social-priming literature of the 1980s and 1990s, much of which failed to replicate when it was retested after 2011, and the conclusion that thinking harder carries a subtle influence further into judgment rests on studies that would now have to be done again. The second is the architecture behind the persuasion results. The elaboration likelihood model's separation of a central from a peripheral route was challenged within three years by single-process accounts arguing that cues and arguments are both evidence and differ mainly in length and complexity, and the dual-process assumption the review treats as settled is now one position among several.

What has lasted is smaller and sturdier than either. The disposition is measurable, stable, largely independent of ability, and predicts, in differences of about a fifth to a third of a standard deviation, who will weigh the argument and who will read the room. The question the authors raised in their last pages, of how a person comes to enjoy thinking and what makes them stop, is still the least answered thing about it thirty years on.