The most famous psychology study of the 1990s — Bargh's elderly walking study — failed to replicate with d = 0.002 across multiple labs. Kahneman himself admitted he was too credulous. But not everything collapsed.
The Collapse
The Many Labs projects tested priming effects across dozens of laboratories with thousands of participants. The results were devastating for behavioral priming:
Bargh's elderly walking study — People primed with old-age words supposedly walked slower. Multi-lab replication: d = 0.002. Effectively zero (Klein et al., 2018).
Professor priming — Thinking about professors supposedly made you smarter on trivia. 23 labs, 4,493 participants: d = 0.002. Zero again (O'Donnell et al., 2018).
Overall scorecard — Approximately 75% of multisite social/behavioral priming replications failed across Many Labs 1, 2, and 3.
The pattern is clear: when you remove experimenter expectation effects and publication bias, behavioral priming — the idea that subtle word cues change complex behavior — disappears.
What Actually Replicates
Not everything was fake. Four environmental effects survived the replication crisis with strong evidence:
1. Mere Exposure Effect — Repeated exposure increases liking. Bornstein's 1989 meta-analysis: r = .26 across 200 experiments. Updated by Montoya et al. 2017. Decades of evidence, multiple labs, consistent results.
2. Anchoring — First numbers influence subsequent judgments. Effect size: d = 0.82 — one of the largest in all of judgment and decision-making research. Successfully replicated across 36 labs in Many Labs 1.
3. Processing Fluency — Things that are easy to read feel more true, more beautiful, more trustworthy (Reber, Schwarz & Winkielman, 2004). Clean typography beats clever typography. Simple language beats complex language. Well-replicated across multiple paradigms.
4. First Impressions (50ms) — Lindgaard et al. (2006) found that 50-millisecond website judgments predicted longer deliberated judgments at r = 0.97. Your landing page is judged before visitors consciously process a single word.
What Separates Real from Fake
The pattern that separates replicated effects from failed ones:
• Real effects have direct cognitive mechanisms (perception → judgment). Fake effects claimed indirect paths (words → complex behavior).
• Real effects have moderate-to-large effect sizes (d > 0.4). Fake effects had small effects inflated by publication bias.
• Real effects show clear dose-response relationships. Fake effects showed no systematic pattern.
Why This Is Good News
The collapse of behavioral priming is actually good news for honest marketers and course creators.
The things that actually work — mere exposure, anchoring, fluency, first impressions — are transparent design choices, not hidden manipulation. You don't need subliminal tricks. You need:
• Consistent presence (mere exposure, r = .26)
• Honest price framing (anchoring, d = 0.82)
• Clean, readable design (processing fluency)
• A landing page that looks trustworthy in the first 50ms (r = 0.97)
The science of environmental influence is real. It just isn't what pop psychology promised. The mechanisms that survived the replication crisis are the ones that respect how brains actually work — through perception, familiarity, and cognitive ease, not through subliminal word games.
Key Statistics
• Behavioral priming (elderly walking): d = 0.002 — FAILED (Klein et al., 2018)
• Professor priming (intelligence): d = 0.002 — FAILED (O'Donnell et al., 2018)
• Mere exposure: r = .26 — REPLICATED (Bornstein, 1989; 200 experiments)
• Anchoring: d = 0.82 — REPLICATED (Many Labs 1, 36 labs)
• 50ms website judgments: r = 0.97 — REPLICATED (Lindgaard et al., 2006)
• Social priming multisite failure rate: ~75% (Many Labs 1-3 aggregate)