BGPT: Paper Review: Second-to-fourth digit length ratio as a measure of harm avoidance

Explore by Goal

Quick Explanation Copied

Core finding (with strong caveats)

In a sample of 728 adults, the right-hand 2D:4D ratio showed a sex difference (women higher than men) and, for women only, higher 2D:4D was associated with greater harm avoidance—specifically fear of uncertainty and shyness with strangers after partialing other harm-avoidance subscales ().

Major critique: the reported effects are small-to-modest correlations in a single-country community sample, with multiple-testing risk, measurement choices (right hand only; direct caliper), and correlational design limiting causal inference.

Long Explanation

Paper: Second-to-fourth digit length ratio as a measure of harm avoidance

DOI: 10.1016/j.paid.2016.03.017 • Journal: Personality and Individual Differences • Publication date metadata: 2016

Data leveraged here are taken directly from the provided full-text excerpt (tables/abstract/results/methods).

Figure 1 — Sample structure & outcomes targeted

Jeon et al. test correlations between right-hand 2D:4D and TCI temperament/character, with the key significant harm-avoidance subscales reported for women.

Figure 2 — Right-hand 2D:4D sex difference

Reported means (right hand): men 0.9518 ± 0.0308, women 0.9763 ± 0.0283.

Figure 3 — Effect sizes and significance: harm-avoidance subscales in women

Jeon et al. report women-only partial regression associations for two harm-avoidance subscales: fear of uncertainty (r_partial=0.206, p=0.028) and shyness with strangers (r_partial=0.252, p=0.024).

Figure 4 — Raw correlation signals mentioned (women-only p-values)

The abstract also reports positive correlations for women in harm-avoidance subscales (fear of uncertainty, shyness with strangers).

Methods audit (what was measured, and where bias can enter)

Population & inclusion/exclusion: community health centers; adults aged 20–45; exclusions include self-reported chronic physical/mental diseases and digit injuries to the 2nd/4th digits; written consent and IRB approval are described.
Digit ratio measurement: direct measurement of right-hand 2D and 4D lengths using a digital Vernier caliper, measured in 0.01 mm increments between the ventral proximal crease and fingertip; two independent researchers; ICC=0.87 reported; retest in 39 subjects (1–2 week interval) showing no significant difference.
Instrument for temperament/character: Temperament and Character Inventory (TCI) with temperament scales including novelty seeking (NS), harm avoidance (HA), reward dependence (RD), persistence (P), and character scales (self-directedness, cooperativeness, self-transcendence).

Skeptical interpretation: even with good measurement reliability, digit ratio is treated as a stable “biological marker,” while temperament is measured by self-report. That combination can still produce spurious correlations via unmeasured confounding (e.g., demographic, experiential, or measurement artifacts) because the design is cross-sectional and correlational.

Results map: where the paper’s strongest claims come from

The paper’s strongest statistical statements are:

Sexual dimorphism in right-hand 2D:4D (women higher than men) reported for total and age subgroups.
Women-only associations with harm avoidance temperament subscales: fear of uncertainty and shyness with strangers showing significant partial regression coefficients.
Temperament vs character emphasis: authors argue 2D:4D correlates more with temperament than character, consistent with the hypothesis that 2D:4D is fixed early.

Visualization warning: the summary bar uses mixed statistics (t-statistic for sex difference and r_partial for correlations). This is only to show where the paper reports statistical signals; it is not a common effect-size metric.

Critical appraisal (biases, limitations, and “what could change my mind”)

1) Multiple-testing and statistical fragility

The authors themselves note multiple testing across many TCI scales/subscales and discuss partial regression to address multicollinearity.

Skeptical read: partial regression reduces shared variance among harm-avoidance subscales, but it doesn’t eliminate false positives from the broader model-space explored across all TCI dimensions unless an a priori analysis plan and robust correction strategy are fully implemented (not fully verifiable from the excerpt).

2) Limited screening and residual confounding

The authors state that subjects were not screened based on a standardized instrument (not just unmeasured confounding—specifically standardized screening).

Skeptical read: even “healthy” participants can vary in subclinical anxiety or traits that influence both self-reported harm avoidance and potentially digit measurements (via measurement conditions, health-related growth variance, etc.). The paper is correlational, so this matters.

3) Measurement scope choices

The paper measures only the right hand. It also excludes people with injuries to 2nd/4th digits and uses direct caliper measurement.

Skeptical read: restricting to right hand can miss left-hand effects or change the magnitude/direction of correlations due to digit-ratio distribution differences. This doesn’t invalidate the result, but it limits generalizability.

4) Directionality & causality

The study interprets 2D:4D as reflecting prenatal testosterone exposure and proposes a mechanistic link to temperament and harm avoidance, but the design is observational/correlational.

What would disprove/seriously change the conclusion? A robust replication that shows the same harm-avoidance subscales associate with 2D:4D (preferably with pre-registered primary outcomes, correction strategy, and balanced sex analyses) would increase confidence; finding no association or inconsistent sex-specific effects would weaken the claim.

Reproducibility checklist (based on what the excerpt provides)

Sample size, sex counts, age range, and enrollment window are stated.
Digit measurement procedure is described, including caliper increment resolution and reliability (ICC).
TCI version and Korean validation reliability statistics are mentioned (Cronbach alpha ranges; test-retest correlations from cited source in excerpt).
But: explicit raw data availability is not shown in the excerpt; “data availability statements” are empty in the supplied metadata.

Author reviews (click for deeper, author-specific critique)

Feedback:

Updated: March 31, 2026