Reading fiction gives at most a very small gain on theory-of-mind and social-cognition tasks, and the single-passage literary-fiction effect did not replicate¶

Claim · All claims
Evidence · 3 studies · 2 causal r?, 1 quant-synthesis r? · q3 · i0 negligible
Currency · newest study 2018 · 2 of 3 entries checked against their sources (automated) · not reviewed by a person

Kidd & Castano (2013) reported that reading a short passage of literary fiction — as distinct from popular fiction or non-fiction — temporarily improved performance on theory-of-mind (ToM) tasks, the ability to infer others' mental states. A three-lab replication with 792 adults (Panero et al. 2016) found no such advantage on the main outcome, and a meta-analysis of 14 experiments (Dodell-Feder & Tamir 2018) found fiction reading improved social-cognitive performance by g = .15–.16, below the conventional threshold for a small effect. What the evidence supports is a very small average effect of fiction reading, not a reliable effect of a single literary passage; no study recorded here measures empathy or lasting change.

Subclaims¶

q3 i0 Across 53 effect sizes from 14 experiments, reading fiction produced a small improvement in social-cognitive performance compared with reading nonfiction or not reading (g = .15–.16). The effect held up in sensitivity analyses and showed no sign of publication bias, but it falls below the d 0.2 threshold for a small effect. → Dodell-Feder & Tamir 2018

q3 i? A replication by three independent labs with 792 randomly assigned adults found no advantage on the Reading the Mind in the Eyes Test for reading a single literary passage over popular fiction, nonfiction or no reading. Lifetime exposure to fiction did predict scores in every condition. → Panero et al. 2016

q3 i? The original five experiments reported that a short literary-fiction passage temporarily improved affective and cognitive theory-of-mind scores compared with nonfiction, popular fiction or no reading. → Kidd & Castano 2013

Evidence¶

Dodell-Feder & Tamir 2018¶

Dodell-Feder, D., & Tamir, D. I. (2018). Fiction reading has a small positive impact on social cognition: A meta-analysis. Journal of Experimental Psychology: General, 147(11), 1713–1727. doi:10.1037/xge0000395

q3 · meta-analysis of experiments · i0 · negligible effect, g=.15–.16 · n=14 studies (53 effect sizes) · quant-synthesis · r?

A multilevel random-effects meta-analysis of published and unpublished experiments in which people were assigned to read fiction, to read nonfiction, or to read nothing, and then completed social-cognition tasks. Fiction reading produced a small but statistically significant improvement over both comparisons. The authors report that the effect survives sensitivity analyses and does not appear to come from publication bias. They call for stronger reading manipulations and for tests of whether the gain carries over into real-world social functioning, which no study had yet shown. (Read from the abstract only.)

Panero et al. 2016¶

Panero, M. E., Weisberg, D. S., Black, J., Goldstein, T. R., Barnes, J. L., Brownell, H., & Winner, E. (2016). Does reading a single passage of literary fiction really improve theory of mind? An attempt at replication. Journal of Personality and Social Psychology, 111(5), e46–e54. doi:10.1037/pspa0000064

q3 · multi-site randomised replication · i? · null result; no effect size for the condition contrast given in the abstract · n=792 participants · causal · r?

Three independent research groups randomly assigned 792 participants to one of four conditions: literary fiction, popular fiction, nonfiction, or no reading. Participants then took the Reading the Mind in the Eyes Test, the main outcome of Kidd and Castano (2013). Literary fiction gave no significant advantage over any other condition. Scores on the Author Recognition Test, which measures lifetime exposure to fiction, predicted scores on the eyes test in every condition. The authors conclude that the link between fiction and theory of mind probably reflects self-selection or gradual lifetime exposure rather than the effect of a single passage, and that verbal ability may play a part. (Read from the abstract only.)

Kidd & Castano 2013¶

Kidd, D. C., & Castano, E. (2013). Reading literary fiction improves theory of mind. Science, 342(6156), 377–380. doi:10.1126/science.1239918

q3 · five peer-reviewed experiments · i? · effect sizes not given in the abstract · n=5 experiments (participant n not given in the abstract) · causal · r?

In five experiments, adults read a short passage and then took tests of affective theory of mind (all five experiments) and cognitive theory of mind (experiments 4 and 5). Each passage was literary fiction, nonfiction, popular fiction, or nothing at all, depending on the experiment. Reading literary fiction led to better theory-of-mind performance than the comparison conditions, and the authors interpret the effect as temporary. This is the canonical study behind the claim. The replication by Panero et al. (2016) above did not reproduce its central eyes-test result. (Read from the abstract only.)

Discussion¶

Replication controversy. The claim originates from a widely cited set of five experiments (Kidd & Castano, 2013) reporting that brief assignments of literary fiction improved theory-of-mind performance relative to popular fiction, non-fiction, or nothing. The multi-site replication recorded here (Panero et al. 2016, 792 participants) found no advantage for literary fiction on the eyes test, and found that lifetime exposure to fiction predicted scores in every condition, suggesting self-selection or gradual exposure rather than a single-passage effect. The meta-analytic average across 14 experiments is very small (g = .15–.16). The single-passage effect should therefore be treated as unreplicated,

and any instructional design built on it should be piloted and measured locally.

Boundary conditions. Where effects have been found, they are typically short-lived (measured immediately after reading), small in magnitude, and dependent on the text genuinely requiring inference about characters' mental states — literary fiction's defining feature in this literature is its demand that readers construct characters' minds from sparse cues [~W]. Texts that do not impose this inferential demand (formulaic popular fiction, expository non-fiction) are not expected to produce the effect [-W]. This makes the selection of text the critical design variable: a "fiction reading" intervention using plot-driven genre fiction is unlikely to engage the mechanism at all.

Mechanism. The proposed mechanism is that literary fiction functions as a low-stakes simulation of social cognition: readers practice inferring motives, emotions, and beliefs of characters, which primes the same processes measured by ToM tasks [~W]. This connects to broader claims about narrative and empathy — see Building empathy improves intergroup attitudes. — though empathy interventions and ToM training are distinct constructs and should not be conflated. The inferential work readers do — filling gaps about character intention from sparse cues — also parallels the constructive inference described in Analogical reasoning improves transfer..

Design implications. If used instructionally, the effect is best treated as a short-term activation or priming phenomenon rather than durable training: pair reading with immediate social-inference discussion or perspective-taking tasks rather than expecting lasting ToM gains [~W]. Discussion formats that require learners to articulate and defend interpretations of characters' mental states align with the mechanism better than silent reading alone — see Discussion-based learning and Act it out. Role-play and drama-based formats (Acting-role-play) plausibly engage the same social-inference processes more directly than reading, though head-to-head comparisons are lacking.

Open questions. Whether repeated exposure produces durable gains (rather than momentary priming), whether effects transfer beyond laboratory ToM tests to real social behavior, and whether effects differ by reader age or reading skill all remain unresolved.

Merged from “Reading fiction produces a very small improvement in social-cognitive (theory-of-mind) task performance, which has not been shown to be a gain in empathy” (fiction-reading-improves-empathy): Mechanism. The most commonly proposed mechanism is that literary fiction demands theory-of-mind inference: because characters are not fully described, readers must simulate their intentions, beliefs, and emotions. This is consistent with the broader claim that building empathy improves intergroup attitudes and that analogical reasoning improves transfer — narrative situations may serve as analogs readers map onto social situations.

Boundary conditions and open questions. Effects reported in this literature tend to be small and short-lived [~W], raising the question of whether a single reading session changes dispositional empathy or only momentary performance on state-based measures. Genre matters in most accounts: literary fiction with complex characters is hypothesized to drive the effect, whereas plot-driven genre fiction may not [~W]. Whether repeated reading over time produces durable dispositional change, and whether effects transfer to real-world prosocial behavior rather than laboratory tasks, remain open. Practitioners should treat fiction reading as one component of empathy-building alongside interactional approaches such as role-play and acting it out, perspective-taking activities, and structured discussion-based learning.

Measurement concerns. Much of the evidence rests on brief state-based instruments administered immediately after reading, which are sensitive to priming and demand characteristics. Correlational studies showing that lifelong fiction readers score higher on empathy cannot establish direction — more empathic people may simply choose to read more fiction [~W]. Replications with pre-registered designs and behavioral outcomes are needed before strong instructional claims are warranted.

Implications for learning design. Where fiction is used in curricula, pair reading with activities that make mental-state inference explicit — discussion-based learning, acting it out, or case-based learning — rather than assuming reading alone produces durable empathy gains [~W]. Select texts with psychologically complex characters over purely plot-driven works when the goal is perspective-taking [~W].

Related Claims¶

• Building empathy improves intergroup attitudes. — related affective outcome of narrative engagement, but a distinct construct from ToM
• Analogical reasoning improves transfer. — characterizing the inferential processes readers may exercise during literary reading
• Cognitive flexibility theory: multiple cases. — multiple ill-structured cases as a route to flexible knowledge, parallel to fiction's varied social scenarios
• Discussion-based learning. — a natural instructional pairing for converting momentary priming into articulated social inference
• Act it out. — embodied perspective-taking that may exercise the same mental-state inference processes as literary reading
• False-belief training improves preschoolers' theory of mind scores but the review reports it has no effect on language skills — related
• Brief intervention improves empathic discipline and cuts suspensions. — empathy can be shifted by brief interventions in educational settings
• Case-based learning improves exam performance. — narrative cases as a vehicle for perspective-taking and situated understanding
• Perspective taking increases the creativity of dyadic interactions — related
• High-quality pretend play is an important facilitator of perspective taking and later abstract thought — related
• Expressively oriented literary programs improved ToM, emotion understanding, and self-reported insight in children and adolescents — related
• Lifetime exposure to narrative fiction correlates with ToM and empathy, but correlational designs preclude causal conclusions — related