Dispositional differences in cognitive motivation: the life and times of individuals varying in need for cognition

Psychological Bulletin 119(2): 197–253, doi 10.1037/0033-2909.119.2.197. Read by researcher R5b (change with reasons), 5 October 2026, from a scanned publisher PDF with an OCR text layer, hosted on the second author's website. Provenance: papers/change/cacioppo1996_need_for_cognition.provenance.json.

What was read

• Read in full: all 57 pages: text, Tables 1–2, the Appendix (the 18-item scale), references.
• The rotated Tables 3–5 (printed pages 218–228, 234–235 and 240–242) have an unusable OCR layer. They were rendered at 150 dpi, rotated and read as images; the Table 3 notes were also checked at 300 dpi.
• Conversion note: the read file is pdftotext in reading order; a few paragraph blocks fall out of column order, and OCR errors (e.g. "Caciolapo") were read through. Figure 1 (a schematic) was read from its caption.
• Errors in the article noted:
  • Table 3's notes print negative averages for belief listing (−.534), knowledge (−.734) and NC × AQ × source (−.435), although every contributing entry is positive and the text gives belief listing +.534 and knowledge +.734;
  • the text gives annoyance d = −.447, Table 5's note −.477.

What it is

• A narrative review with meta-analytic averages of more than 100 studies (1982–1995) of need for cognition (NC): "an individual's tendency to engage in and enjoy effortful cognitive endeavors", a continuum from "cognitive misers" to "cognizers".
• The construct: a motivation, not an ability. It is proposed to come from past experience of competence and satisfaction in thinking, sustained by accessible memories and behavioural histories.
• The scale: the 34-item NCS, and the 18-item short form (r = .95 with the long form; α = .90; one factor with 37% of variance).
• Effect sizes are expressed as r or d. Studies' non-significant results without statistics were entered as d from p = .5. Averages are weighted d with one-tailed p.

Main results (verified)

Measurement and stability

• One dominant factor in nearly every analysis; proposed three-subscale structures did not replicate (cross-sample loading correlations −.01, .70, .28). α typically > .85.
• Test–retest: r = .88 over 7 weeks (18-item NCS, 71 undergraduates; Sadowski and Gulgoz 1992b) and .66 over about 8 months (6 Dutch items, 289 residents; Verplanken 1991).
• No gender difference (e.g. 708 women, 510 men; Sadowski 1993).
• Education: average r = .25 (.13 to .46, larger with wider ranges). Age: average r = −.17.

Where NC sits among other traits (Table 2)

• With abilities: verbal intelligence average r = .24; ACT .26; high-school GPA .26; college GPA .17; abstract reasoning −.03. Related to ability, but not the same thing; NC predicted argument recall beyond verbal intelligence.
• Open-mindedness and closure:
  • openness to ideas r = .78 (also actions .39, feelings .31, values .29, fantasy .07);
  • closed-mindedness −.34; need for closure −.26; dogmatism −.23 to −.24;
  • objectivism .47; attributional complexity .36–.51; curiosity about .57.
• Ambivalence −.41 and −.33; self-esteem .38; social anxiety −.35; neuroticism −.24; external locus of control −.32.
• Unrelated to sociability, extraversion and affect intensity; less alexithymia (−.42 to −.52).

Persuasion: arguments against cues (Tables 3–4)

• NC × argument quality on attitudes: d = .312 (11 tests, p < .001; homogeneity χ²(10) = 16.80, p = .07): high-NC people's attitudes track argument strength more. Individual tests 0 to .73, three null.
• NC × argument quality on ratings of message quality: d = .539 (5 tests).
• NC × peripheral (source) cues: d = .106, n.s. (4 tests). The authors: cue variables like credibility or mood can act as arguments or bias processing at higher elaboration, so no consistent interaction is expected.
• Low-NC people are moved by cues when not otherwise motivated:
  • a claimed number of arguments ("10 reasons" against "2", the same six arguments) moved low-NC but not high-NC people (Chaiken et al. 1985, as cited);
  • a consensus cue (audience applause) moved low-NC people only when personal relevance was low (Axsom et al. 1987).
• Low-NC people scrutinize arguments only when there is a special reason: when the source might be dishonest (Priester and Petty 1995, d .40 and .50) or the message frame was unexpected (Smith and Petty). High-NC people track argument quality regardless.
• Thinking and recall:
  • high NC recall more of what they are exposed to (d = .355, 19 tests; message arguments d = .312, 8 tests);
  • they generate more issue-relevant thoughts (d = .205, 10 tests), whose valence tracks argument quality more (negative thoughts d .38, positive .23);
  • overall they have more negative thoughts to the mostly counter-attitudinal messages (d .23).
• Thinking predicts attitudes and attitudes predict action more for high NC (Table 4):
  • thoughts predicting attitudes: average d 1.13 against 0.66;
  • message ratings predicting attitudes: 1.29 against 0.43;
  • attitudes predicting intentions: 1.54 against 0.90;
  • attitude predicting the vote: r .87 against .46 (Cacioppo et al. 1986).

Persistence and resistance (Table 3)

• Persuasion persisted for high NC but not low NC:
  • Haugtvedt and Petty 1992, Study 1: d .76;
  • Verplanken 1991, over 14 months and three waves: high NC d .48, low NC d .04.
• Resistance: after equal initial persuasion, a later challenge moved low-NC but not high-NC people (Haugtvedt and Petty 1992, Study 2, d 1.26).
• Thought itself polarizes or moderates:
  • with spontaneous thought (a belief survey), high-NC attitudes became more extreme (Smith, Haugtvedt and Petty 1994);
  • when explicitly told to reflect for 180 s, high NC moderated and low NC polarized (Leone and Ensley 1986);
  • Lassiter et al. (1994) reproduced both patterns by instruction.
• Order effects are mixed: primacy for high NC with two clear opposing sides (trial evidence; Kassin et al. 1990), but primacy for low NC with 16 sequential trait cues (Ahlering and Parker 1989).

Thinking is not the same as objectivity

• High-NC processing can be biased by mood (via thoughts and likelihood judgments) and by subtle primes (bets after "winning" primes).
• High-NC people correct for biases they notice (less correspondence bias, d −.64; correction of blatant primes). Corrections follow naive theories and need not be accurate.

Situation × person

• NC differences are largest when situational pressure to think is moderate. Under high personal relevance both groups think; under none, neither does (Figure 1).
• High NC enjoy effortful tasks more (negative affect d about −.45; enjoyment .31), seek more information (.46), and do not loaf on cognitive tasks.
• Antecedents are argued (intrinsic motivation, competence feedback) but untested; no genetic data.

Limits

• Pre-replication-era evidence: small student samples, many unpublished or in-press studies, one-tailed meta-analytic tests, coarse effect-size imputation for non-significant results. Many tests come from the review's own authors' laboratories.
• NC is a self-report; test–retest rests on two studies.
• The argument-quality manipulation uses pretested strong and weak arguments; "quality" is defined by the thoughts it elicits, not by logical validity.

What it means for Kurisutina (inference)

• NC is a stable person setting for updating on reasons: a test–retest correlation of .88 over weeks and .66 over months. It scales how much argument quality matters relative to cues (an interaction of d about .3), how long a change lasts, how well it resists later attack, and how closely it predicts behaviour.
• It is not a bias-free dial. High-NC people elaborate whatever is salient, including mood and primes, and correct only for biases they notice. A slot setting for "elaboration" should scale both sensitivity to argument quality and the reach of biased thought.
• The situation matters as much as the person. Personal relevance, surprise and suspicion of the source switch low-NC people into scrutiny. The rule of change needs an elaboration state (setting × situation), not a fixed trait weight.
• For M1: good and bad argument families can carry an exact ground truth for who discriminates: high-elaboration persons follow validity; low-elaboration persons follow cues (tie trust, number of arguments, consensus) unless an episode is surprising or the source is suspect.