The 1998 Physiology & Behavior study tested Kemper's hypothesis that fans experience vicarious hormonal dominance responses. In Study 1 (8 male Georgia fans at the neutral-site GeorgiaβGeorgia Tech game), winners' fans rose and losers' fans fell: team Γ time interaction F(1,6)=9.94, p<0.05, partial Ξ·Β²=0.62, means 6.3β6.5 vs lower postgame in losers. In Study 2 (12 Brazilian and 9 Italian fans watching the 1994 World Cup final on TV), interaction F(1,19)=13.01, p<0.01, partial Ξ·Β²=0.42, overall means 7.7β7.3 ng/dL . The soccer effect was notable because the win reversed the normal 30β40% diurnal evening decline .
Later work replicated the pattern in a different domain: male voters supporting the 2008 election winner maintained testosterone (83.4β76.5 pg/mL across the night) while losers declined, with no changes in women .
Strengths: naturalistic field design, within-subject pre/post sampling, single-batch duplicate radioimmunoassay, and the two studies converging despite different venues and sports. The authors honestly note the effect likely occurred in the final seconds of each game.
Limitations: samples are tiny (n=8 and 21 analyzed males), all-male, self-selected bar/arena attendees, with $10 payment and no randomization. The soccer study lacked blinding and control for alcohol, food, nicotine, and group atmosphere at taverns; diurnal timing differences between pre and post samples were not matched. Italian fans dropped out disproportionately before the next-morning sample (ΟΒ²(1, n=37)=3.85), a selective-attraction risk. Because the observed changes (Β±0.2β0.4 ng/dL group means) are small relative to within-group SDs (2.7β3.5), individual-level inference is weak, and no raw by-person data are published. The authors correctly frame findings as consistent with the biosocial status model but admit the data say nothing about downstream behavioral effects. Hypothesizing-after (Kemper's prediction) is acknowledged, reducing HARKing risk but limiting theoretical precision.
What would change the conclusion: null team Γ time interactions in a larger, pre-registered, sex-inclusive design with matched sampling times, controlled substances, and published individual-level data. Confidence in the qualitative direction is moderate; confidence in magnitude and generalizability is low.
Deeper questions remain open, e.g., whether female fans show equivalent responses despite different testosterone production pathways β the authors explicitly flag this as untested .
Know what changed, what holds up, and what remains uncertain. Every Friday. No ads.