Our sources

We only publish numbers we can trace to a primary source: the original study, official statistics, or the organization's own reporting. Secondary retellings, influencer polls, and unattributed "studies show" claims are rejected at the fact-gate.

Primary sources currently cited

• Kajonius and Johnson, Europe's Journal of Psychology (2019)Peer-reviewed paper assessing the structure of the five-factor model with the public-domain IPIP-NEO-120. Quoted on the page: 'Personality is most frequently measured with the five factor model (FFM; McCrae, 2010). This represents regularities of thoughts, feelings, and behaviors in individuals expressed in five broad trait domains: N = Neuroticism, E = Extraversion, O = Openness, A = Agreeableness, C = Conscientiousness.' Use as the citation that introversion is a continuous trait dimension in mainstream personality research — which is the sentence that lets /introvert-vs-shy/ refuse MBTI typology without hand-waving. The only number here is 'five'; there is no percentage to quote.
• National Comorbidity Survey Replication (NCS-R), reported by Ruscio and colleagues in Psychological Medicine (2008)Same study and sample as ncsr-social-phobia-lifetime: WMH-CIDI diagnostic interviews, n=9,282, 2001–2003. Twelve-month prevalence.
• Pew Research Center (2023)Same survey: n=6,034 U.S. adults, fielded July 5–17, 2022. 'Ever used' figures. Subgroup n's are not published on this page; do not compute differences between groups.
• Pew Research Center (2023)Same report and survey: n=6,034 U.S. adults, July 5–17, 2022. Read verbatim on the report page.
• Pew Research Center (2023)Pew Research Center, 'The experiences of U.S. online daters', February 2, 2023; American Trends Panel Wave 111, 6,034 U.S. adults, July 5–17, 2022. Read verbatim on the chapter page.
• Pew Research Center (2023)Pew Research Center, 'The who, where and why of online dating', February 2, 2023; American Trends Panel Wave 111, 6,034 U.S. adults, July 5–17, 2022. The report states the 18–29 figure as 'one-in-five' — write it as 'about 20%' or 'one in five', never as a decimal.
• YouGov (2023)Same poll: online, March 14–16, 2023, n=1,000, weighted, MoE ~4%. This measures PERCEIVED advantage (public opinion), not outcomes. Do not phrase it as 'extroverts do better at parties'.

What we rejected, and why

This list matters as much as the one above. These are figures that circulate widely in writing about widowhood and that we could not trace to a primary source, so they do not appear anywhere on this site.

• 50.7% of the U.S. population are introverts (49.3% extraverts) — the Myers-Briggs 'estimated frequency' figure — Could not be read on any primary page: the Myers & Briggs Foundation 'How Frequent Is My Type' URL returned 404 on every attempt, and every source repeating the number is a blog or an MBTI-industry aggregator, not a reputable secondary reporting a published study. The methodology also fails on its own terms: the figure is described as compiled from MBTI results collected 1972–2002 in the CAPT / CPP / SRI databanks — i.e. self-selected people who took (usually paid for) a personality instrument, over three decades, with no probability sampling of the U.S. population and no published margin of error. It additionally forces a continuous trait into a two-box split, which the five-factor literature does not support. Not usable at any confidence. Where the site needs 'how many introverts', use yougov-2023-introvert-selfid (39% SAY they are introverts) and say plainly that it is self-description.
• Any figure for the true share of the population that 'is' introverted (as a measured trait rather than self-report) — No such figure exists to quote, because extraversion is measured as a continuous dimension with no agreed cut-point — there is no published threshold above which a person 'is an introvert'. Every population percentage in circulation is either self-identification (YouGov) or an MBTI dichotomisation (rejected above). /statistics/ must state this limitation rather than pick a number.
• 'Social battery' has a measured capacity, average size, or documented drain rate — Searched PubMed and the peer-reviewed literature: 'social battery' is not a scientific construct — there is no validated scale, no published norm, no measured unit. The only sources offering figures are consumer-health and niche blogs. This absence is itself the honest fact /social-battery/ and /social-battery/recharge/ must state ('a metaphor, not a measured quantity'), exactly as the content-spec already frames it. Related constructs (extraversion in the FFM; Eysenck's arousal theory) exist, but they do not license a number.
• It takes N hours (or 'a day', or 'X minutes of solitude') to recharge after social contact — No measurement found in any peer-reviewed or authoritative source. Every recovery-time figure surfaced came from introvert-lifestyle blogs with no study behind them. /social-battery/recharge/ must answer 'how long does it take' by saying no measurable norm has been established — which the spec already requires — and must not print hours.
• About 78–79% of dating app users report dating-app burnout (Forbes Health survey) — Commercial survey run for a publisher's affiliate-adjacent vertical; no probability sample, no published weighting or margin of error, panel and question wording not disclosed at the standard required here. Rejected as an engagement-driven survey rather than a primary research instrument. The nearest defensible figures are Pew's, and they measure something narrower: 36% at least sometimes overwhelmed by message volume (pew-2023-overwhelmed-by-messages) and 46% describing their overall experience as negative (pew-2023-positive-vs-negative-experience). Do not label either of those 'burnout'.
• Most people now prefer texting to phone calls — The only measurement from an authoritative source is Pew Research Center's 2011 survey (n=2,277 adults, April 26 – May 22, 2011), and it points the other way: 31% preferred texts while 53% preferred a voice call. Fifteen years stale and contrary to the framing the pages wanted. Pages that lean on 'introverts prefer text-first formats' must argue it from the format's demands, without a statistic.
• Introverts are more likely to partner with other introverts (YouGov 2021 introvert/extrovert dating poll) — The YouGov 2021 article was not retrievable — every candidate URL 404'd or redirected to a dead path — so no figure was read on a primary page. Untraceable at fetch time, therefore out. /dating/two-introverts/ and /dating/introvert-relationships/ must run on described mechanics, not on a pairing percentage.
• Introverts make up X% of dating-app users, or introverts have worse/better dating outcomes — No source found that crosses personality measurement with dating-app usage in a representative U.S. sample. Pew does not ask about introversion; YouGov's introvert polls do not ask about app usage. Any such number would be an invented cross-tab. The site may not state, imply, or estimate one.
• "Most people sit somewhere in the middle" / "most people are ambiverts" — any majority claim about the middle of the extraversion dimension — Added 2026-08-12 during fact-verification of /dating/introvert-vs-shy/, where the draft asserted this twice and attributed it to the YouGov split. It is not in any approved figure and cannot be read on a primary page. The only measured split is YouGov 2023 SELF-IDENTIFICATION: 39% introvert, 31% equal mix, 22% extrovert — the middle category is 31%, a minority, so 'most' is not what the poll shows. Reaching a majority requires summing the 'more introverted than extroverted' (29%) and 'more extroverted than introverted' (15%) cells into a 'middle' bucket YouGov never defined: an invented cross-tab. Separately, no trait-measurement source licenses a claim about where 'most people' fall, because there is no published cut-point on the dimension. Safe rephrase, no number: 'extraversion is measured as a continuous dimension, so being between the poles is an ordinary place to be, not a special type.'
• "Most people asking this question find three of those four have moved at once" — any counted share of people whose social load rose on N of the four audit dimensions — Added 2026-08-12 during fact-verification of /social-battery/why-so-low/. This is a majority claim ('most people') carrying a count ('three of those four') about a four-item audit list the site invented on the page. Nothing measures it: there is no instrument covering 'unscheduled contact', 'contact that continues after you leave', 'contact that requires performance' and 'lost recovery' as dimensions, and no population estimate of how many of them move together. It reads as a survey finding and is an editorial guess, which is precisely the substitution this gate exists to block. Safe rephrase, no number: 'It is rarely one of those four on its own — they tend to move together, which is why a week feels heavier than the sum of its arrangements.'

The rules

1. Named source in the same sentence as the number, with the year.
2. Primary source only — we link the origin, not an article about it.
3. Dated verification: every statistic carries its verification date on the statistics page.
4. Unverifiable → unpublished. No exceptions.