Self-reference and cognitive effort: source memory for affectively neutral information is impaired following negative compared to positive self-referential processing

ABSTRACT

Previous research suggests a close relationship between self-reference and emotional valence. The present study investigated potential differences in cognitive resources required for positive vs. negative self-referential processing by examining how self/other-referential processing of positive/negative information affects memory for subsequently presented items. On each encoding trial, participants first judged whether a positive or negative trait adjective described themselves or another person. Then, they were shown a neutral noun and indicated its screen location. Subsequent memory tests showed better memory for self-referenced than other-referenced trait adjectives, and the size of this self-reference effect was not modulated by emotional valence. Although memory for nouns was not affected by preceding positive/negative self/other-referential processing, memory for their associated contextual features was significantly impaired following negative vs. positive self-referential processing. Our findings suggest that negative self-referential processing requires more cognitive resources than positive self-referential processing, thereby leaving relatively less cognitive resources to encode subsequently presented information.

KEYWORDS:

• Self-reference
• emotional valence
• cognitive effort
• memory
• self-positivity

Acknowledgments

We thank research assistants in the Memory, Cognition, and Self (MCS) Lab at Wesleyan University for their help with data collection and helpful discussions.

Data availability statement

The data that support the findings of the present study have been made publicly available via the Open Science Framework and can be accessed at https://osf.io/zvebs.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Tom Hanks was chosen as the “other” referent in consultation with undergraduate research assistants. Tom Hanks was deemed to be a celebrity that most study participants would be familiar with, given his appearance in a number of movies within a few years preceding the data collection for the present study.

2 To take into account participants’ responses to trait adjectives (i.e. positive or negative endorsement), mean RTs were also analyzed using a 2 (Referent: self, other) × 2 (Valence: positive, negative) × 2 (Response: yes, no) repeated-measures ANOVA. It should be noted that because this additional 2 × 2 × 2 repeated-measures ANOVA and its follow-up simple effects analyses reported in this footnote were conditionalized by particular types of responses with some participants’ having no response of a given type (e.g. no positive endorsement of negative adjectives in the self-referent condition), degrees of freedom between analyses sometimes differed. The 2 × 2 × 2 repeated-measures ANOVA yielded a number of significant effects (Referent [F(1, 77) = 9.13, p = .003, η_p² = .11, BF_Inclusion = 22.81], Response [F(1, 77) = 11.88, p = .001, η_p² = .13, BF_Inclusion = 4.78], Valence × Response [F(1, 77) = 208.26, p < .001, η_p² = .73, BF_Inclusion = 6.92 × 10³⁷]) that were qualified by a significant Referent × Valence × Response interaction, F(1, 77) = 4.70, p = .033, η_p² = .06, BF_Inclusion = 2.60. Follow-up analyses revealed that while participants were faster to positively endorse positive adjectives and slower to positively endorse negative adjectives in both the self-referent (Positive adjectives: Yes [M = 1157.43, SD = 217.66] vs. No [M = 1514.44, SD = 348.29]; Negative adjectives: Yes [M = 1474.01, SD = 311.43] vs. No [M = 1248.23, SD = 213.41]) and other-referent conditions (Positive adjectives: Yes [M = 1274.05, SD = 254.73] vs. No [M = 1521.51, SD = 359.02]; Negative adjectives: Yes [M = 1489.60, SD = 307.67] vs. No [M = 1321.75, SD = 234.70]), all ts ≥ 5.49, all ps < .001, all ds ≥ 0.57, all BF₁₀s ≥ 3.63 × 10⁴, the size of the RT difference between positive vs. negative endorsement was significantly larger in the self-referent condition than in the other-referent condition for positive adjectives, t(89) = −2.67, p = .009, d = 0.28, BF₁₀ = 3.29, but not for negative adjectives, t(88) = 1.26, p = .21, BF₁₀ = 0.25.

3 Source memory accuracies from all 2 (Referent) × 2 (Valence) combinations of the conditions were significantly above chance level (.50), all ts ≥ 3.00, all ps ≤ .003, all ds ≥ 0.29, all BF₁₀s ≥ 7.34.