Novel and operationalizable judgment construct
The paper derives the retrospective gambler’s fallacy from the law of small numbers and translates it into open-ended estimates of unobserved prior trials.
↳ Introduction §1; Study 1 §2.1
Abstract
Abstract The gambler’s fallacy (Tune, 1964) refers to the belief that a streak is more likely to end than chance would dictate. In three studies, participants exhibited a retrospective gambler’s fallacy (RGF) in which an event that seems rare appears to come from a longer sequence than an event that seems more common. Study 1 demonstrates this bias for streaks, while Study 2 does so with single rare events and shows that the appearance of rarity is more important than actual rarity. Study 3 extends these findings from abstract gambling domains into real world domains to demonstrate the generalizability of the effects. The RGF follows from the law of small numbers (Tversky & Kahneman, 1971) and has many applications, from perceptions of the social world to philosophical debates about the existence of multiple universes.
Key Claims
Primary findingSupported
In three studies, participants exhibited a retrospective gambler’s fallacy (RGF) in which an event that seems rare appears to come from a longer sequence than an event that seems more common.
convergent experimental contrasts support the direction, though small samples and the heterogeneous vignette study limit certainty
Abstract; General discussion, p. 331
Replication outlook: fragile
SecondarySupported
Study 2 does so with single rare events and shows that the appearance of rarity is more important than actual rarity.
the three-condition dice comparison separates perceived representativeness from objective probability, but alternative scenario features remain possible
Abstract; Study 2a, §3.2 Results
Replication outlook: fragile
SecondaryTentative
Study 3 extends these findings from abstract gambling domains into real world domains to demonstrate the generalizability of the effects.
selected vignettes include contrary-direction items and outcomes where prior behavior may be diagnostically informative
Abstract; Study 3, §5.2 Results and discussion
Replication outlook: fragile
SecondarySupported
the difference in estimated number of trials observed between versions is mediated by the difference in average estimated likelihood
mixed-model coefficient changes support mediation, though the causal-steps analysis does not directly estimate the indirect effect
Study 3, §5.2 Results and discussion; Table 2, Model 3
Derived from the full evaluation — not a separate score.
Strengths · Limitations
Strengths
Convergent multi-design experimental evidence
The predicted direction appears in a between-subject coin task, dice tasks, and a counterbalanced within-subject replication, reducing dependence on a single paradigm.
↳ Studies 1–2b, §§2–4
Appropriate crossed-effects analysis
Study 3 models random intercepts for participants and stories, log-transforms skewed estimates, and examines perceived likelihood as a mediator.
↳ Study 3 §5.2; Table 2
Limitations
Real-world generalization exceeds item pattern
Several Study 3 stories have means opposite to the predicted effect, including the motivating pregnancy scenario, yet the discussion describes the reasoning as broadly prevalent.
↳ Study 3 Table 1; §5.2; General discussion §6
Everyday scenarios permit diagnostic inference
For several outcomes, prior exposure, skill, opportunity, or stable risk may reasonably inform estimates of prior trials, weakening their interpretation as stochastic fallacies.
↳ Study 3 Method, Table 1, and §5.2
Reported counts contain minor inconsistencies
Study 3's stated design and exclusions imply 433 retained observations rather than 435, while Study 1's outlier removal is difficult to reconcile with t(106).
↳ Study 1 §§2.1–2.2; Study 3 §5.2
Nabu Commentary
The coin-flip and dice studies provide coherent evidence that apparently rare outcomes increase estimates of unobserved prior trials, including a counterbalanced within-subject replication. Study 3 adds a suitable crossed mixed-effects analysis and a perceived-likelihood mediation sequence, but Table 1 shows meaningful item heterogeneity. Several everyday outcomes may also carry legitimate information about stable risk, skill, opportunity, or prior exposure, so they do not cleanly reproduce the independence structure of the gambling tasks. The resulting contribution is novel and credible at its core, while its claims of broad real-world prevalence and application require substantial narrowing.
Author Note / Field Feedback
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Dimension breakdown
Study Quality
Sound3.1
Contribution
3.2
The paper defines a novel retrospective judgment effect and supports its core direction across coin-flip, dice, and everyday-story tasks. The broader claim of real-world generalizability exceeds the mixed item-level pattern in Study 3.
Abstract; Study 1 §2.2; Study 2a §3.2; Study 3 Table 1 and §5.2
“participants exhibited a retrospective gambler’s fallacy”
Methodological Rigour
3.1
Randomized between-subject comparisons, a counterbalanced within-subject replication, log transformation, and crossed random-intercept modeling fit the experimental structure. Small convenience samples and unexplored exclusions reduce precision and external validity.
Studies 1–3 Methods and Results; Table 2
“we modelled random intercepts for the 15 remaining stories and the 30 remaining participants”
Reporting
3.4
The argument proceeds coherently from theory through progressively broader studies, and the statistical distinction between sampled trials and occurrence somewhere in a sequence is clearly explained. Abstract and discussion language presents speculative applications and vignette generalization more strongly than the evidence permits.
Abstract; Study 2b §4.3; Study 3 §5.2; General discussion §6
“to demonstrate the generalizability of the effects”
Positioning
2.8
The paper engages the law of small numbers, representativeness, gambler’s-fallacy, and inverse-gambler-fallacy literatures. It does not adequately consider that prior behavior, skill, exposure, or stable risk may be diagnostically related to several Study 3 outcomes.
Introduction §1; Study 3 Table 1 and §5.2; General discussion §6
“this reasoning generalizes and is prevalent in reasoning about rare and common events more broadly”
Reliability
Caveats4 of 4 checks
Internal CoherenceNoted
Several numerical and interpretive inconsistencies warrant caution but do not overturn the convergent direction of the core experimental findings. The clearest issue is the mismatch between Study 3's stated counts and its reported 435 observations.
Research ConductClean
Funding is disclosed, and the absence of an ethics statement, preregistration, and data or code availability statements is not independently concerning under the reporting norms applicable to these 2009 minimal-risk vignette studies.
Flags: 1 declared / 5 total
Reference IntegrityNoted
21 of 25 checkable references verified
29 references in manuscript 4 are books, websites or datasets — counted, but not index-checkable
Post-Publication RecordClean
No retraction notice found in Retraction Watch.
Last checked: 24 Aug 2026
Sources: Retraction Watch ✓
Impact Potential
Low2.1
Relevance
1.8
The paper primarily addresses an academic question and names only speculative connections to victim blaming, planning, memory, and cosmology. It identifies no practitioner, organization, or implementation process.
General discussion §6, pp. 332–333
“there are also potential real world applications of the phenomenon”
Readiness
1.8
The evidence remains at the controlled behavioral-observation stage. No intervention, assessment instrument, operational validation, or demonstrated decision outcome is presented.
Studies 1–3; General discussion §6
“The RGF offers a new tool in the repertoire for studying such topics.”
Generalizability
2.4
The effect is tested across streaks, single dice outcomes, within-subject judgments, and 15 everyday stories. Transfer is constrained by Stanford and Princeton convenience samples and by contrary-direction item means in Table 1.
Studies 1–3 Methods; Study 3 Table 1
“Study 3 replicated the effects of Studies 1 and 2 in a variety of real-world domains.”
Trajectory
2.7
The RGF paradigm connects directly to established research on representativeness, sample-size reasoning, and reconstructive memory. Its prospective development is primarily academic because the proposed applied links are not operationalized.
Introduction §1; General discussion §6, pp. 332–333
“Investigations into the RGF could lend new insight to this problem.”
AI-generated, human-governed. Something look off? Contact us to request a review.