Social influence on risk perception during adolescence

Psychological Science 26(5): 583–592, doi 10.1177/0956797615569578; open access (CC BY 3.0), PMC4426139. Read by researcher R5a, 5 October 2026.

Read in place of Steinberg and Monahan (2007), the brief's candidate for age and resistance to peer influence. Its PMC author manuscript sits behind a proof-of-work check; Europe PMC withholds its XML at the publisher's request; Europe PMC's PDF route returned a Cloudflare challenge; no author copy was found. A copy on a third-party advocacy site was not used. Provenance: papers/change/knoll2015_social_influence_risk.provenance.json.

What was read

• Read in full: the Europe PMC XML, converted with tools/pmc2txt.py (self-check passed): abstract, introduction, method, results, Tables 1–2, discussion, conclusion, acknowledgements, footnotes, references.
• Not read: Figures 1–2 as images (values in the text); the online Supplemental Material (the scenario list and model details), which is not in the XML.

What they did

• Participants. 563 visitors to the Science Museum, London (April 2013), aged 8–59 (mean 23.4); 63 excluded. Five age groups:

  Group	Ages	n
  Children	8–11	106
  Young adolescents	12–14	68
  Mid-adolescents	15–18	52
  Young adults	19–25	136
  Adults	26–59	201

• Task. 18 everyday risky scenarios, drawn from 79 (for example crossing on a red light), each rated for riskiness on a 0–10 slider.

  • Then, for 2 seconds, a rating "from a group of teenagers" or "from a group of adults" was shown. It was actually random. In control trials the participant's own rating was shown.
  • The participant then rated again, immediately.
  • Participants were told there were no right answers. There were no social consequences: the group was never to be met.

• Analysis. A linear mixed model of the second rating on the first rating and Δrating (provided minus first), with age group and source interactions and per-person random slopes.

Main results (verified)

• Share of the gap moved toward the provided rating (all groups significant):

  Group	Share moved
  Children	33%
  Young adolescents	27%
  Mid-adolescents	18%
  Young adults	12%
  Adults	7%

• People differ. The random-effect SD of the Δrating slope is 0.14, as large as or larger than the adult means. The residual SD is 1.02.

• Source:

  • children, young adults and adults moved more toward "adults" (the adults' own test at p = .051);
  • mid-adolescents did not differ;
  • young adolescents moved more toward "teenagers" (t = −2.33, p < .05).

• Baseline risk: children rated situations riskier (6.20) than every other group (5.12–5.57).

• The authors' reading: the influence is probably informational, since there were no social consequences. The decline with age may reflect confidence from experience. The young adolescents' pull toward peers may add a normative part, which they mark as speculative.

Limits

• Format. Immediate re-rating after one exposure, from an anonymous fictitious group. There is no delayed retest. It measures risk perception, not risk-taking.
• Design. Cross-sectional; museum visitors.
• Inconsistencies found:
  • The adults' source effect is called significant at p = .051.
  • "p < .002" in the text against .002 in Table 1.
  • The same df (9605.25) is printed for two different follow-up tests.

What it means for Kurisutina (inference)

• A human reference for weighting other people's bare judgements (no reasons, no social cost):
  • adults move about 7–12% of the gap;
  • young people 18–33%;
  • with large spread between people.
• For B7, susceptibility depends on:
  • age: it declines with age;
  • the source's perceived competence: adults move people more;
  • in-group status in early adolescence: young adolescents follow teenagers.
• Not the same quantity as M1's s. s is a probability of yielding in a dialogue; this is a fractional weight on an estimate. Do not calibrate s from it directly.