# Audit: The correlation between human brain volume and intelligence is r = 0.4

- Generated: 2026-03-27
- Reader summary: [proof.md](proof.md)
- Proof script: [proof.py](proof.py)

---

## Claim Specification

| Field | Value |
|-------|-------|
| Subject | Pearson r between total in-vivo brain volume (MRI) and intelligence (IQ/g-factor) |
| Property | Meta-analytic correlation coefficient |
| Operator | within |
| Threshold | 0.40 |
| Tolerance | 0.05 |
| Operator note | r = 0.4 is interpreted as r within ±0.05 of 0.40 (0.35 ≤ r ≤ 0.45). Two sub-claims evaluated: SC1 (unconditional overall estimate) and SC2 (conditional: healthy adults, high-quality tests). 'Brain volume' = total in vivo MRI. 'Intelligence' = psychometric IQ or g-factor. |

---

## Fact Registry

| ID | Key / Description |
|----|-------------------|
| B1 | `pietschnig_2015` — Pietschnig et al. (2015): 88 studies, 8000+ subjects; weighted r = .24 |
| B2 | `pmc_2022` — Nave et al. (2022): largest meta-analysis (N=26k+); r = 0.24, range 0.10–0.37 |
| B3 | `wiki_conditional` — Wikipedia Neuroscience & Intelligence: r ≈ 0.4 for healthy adults, high-quality tests |
| A1 | SC1-A: \|r_Pietschnig − 0.40\| |
| A2 | SC1-B: \|r_PMC2022 − 0.40\| |
| A3 | SC2: \|r_conditional − 0.40\| |
| A4 | Cross-check: Pietschnig 2015 vs PMC 2022 overall r agreement |

---

## Full Evidence Table

### Type A (Computed) Facts

| ID | Fact | Method | Result |
|----|------|--------|--------|
| A1 | SC1-A: \|r_Pietschnig − 0.40\| | abs(0.24 − 0.40) | 0.1600 |
| A2 | SC1-B: \|r_PMC2022 − 0.40\| | abs(0.24 − 0.40) | 0.1600 |
| A3 | SC2: \|r_conditional − 0.40\| | abs(0.40 − 0.40) | 0.0000 |
| A4 | Cross-check: Pietschnig 2015 vs PMC 2022 overall r | cross_check(0.24, 0.24, tol=0.01, mode='absolute') | Agreement |

### Type B (Empirical) Facts

| ID | Fact | Source | URL | Quote | Status | Method | Credibility |
|----|------|--------|-----|-------|--------|--------|-------------|
| B1 | Overall r = .24 | Pietschnig et al. (2015), Neuroscience & Biobehavioral Reviews — PubMed | https://pubmed.ncbi.nlm.nih.gov/26449760/ | "Our results showed significant positive associations of brain volume and IQ (r=.24, R(2)=.06) that generalize over age (children vs. adults), IQ domain (full-scale, performance, and verbal IQ), and sex." | Verified | full_quote | Tier 5 (government) |
| B2 | Overall r = 0.24 | Nave et al. (2022), Royal Society Open Science — PMC | https://pmc.ncbi.nlm.nih.gov/articles/PMC9096623/ | "Brain size and IQ associations yielded r = 0.24, with the strongest effects observed for more g-loaded tests and in healthy samples that generalize across participant sex and age bands." | Partial (50% fragment) | fragment | Tier 5 (government) |
| B3 | Conditional r ≈ 0.4 | Wikipedia — Neuroscience and intelligence | https://en.wikipedia.org/wiki/Neuroscience_and_intelligence | "In healthy adults, the correlation of total brain volume and IQ is approximately 0.4 when high-quality tests are used." | Verified | full_quote | Tier 3 (reference) |

---

## Citation Verification Details

**B1 — Pietschnig et al. (2015), PubMed**
- Status: verified
- Method: full_quote
- Fetch mode: live
- Impact: Primary evidence for SC1. Full quote verified.

**B2 — Nave et al. (2022), PMC**
- Status: partial (fragment match, 50% word coverage)
- Method: fragment
- Fetch mode: live
- Impact: Corroborating evidence for SC1. Partial quote verification, but the key data value (r = 0.24) was independently confirmed live via `verify_data_values` (found: true). The partial match likely reflects minor HTML/whitespace formatting differences in the full-text PMC article vs. the quote extracted from the abstract. The numerical conclusion is independently supported by B1 (full verification, same r = 0.24).

**B3 — Wikipedia — Neuroscience and intelligence**
- Status: verified
- Method: full_quote
- Fetch mode: live
- Impact: Primary evidence for SC2. Full quote verified. Wikipedia cites Gignac & Bates (2017, *Intelligence*) for this figure.

---

## Computation Traces

```
[✓] pietschnig_2015: Full quote verified for pietschnig_2015 (source: tier 5/government)
[~] pmc_2022: Only 15/30 quote words matched for pmc_2022 — partial verification only (source: tier 5/government)
[✓] wiki_conditional: Full quote verified for wiki_conditional (source: tier 3/reference)
[✓] B1.r_overall: '.24' found on page [live]
[✓] B2.r_overall: '0.24' found on page [live]
[✓] B3.r_conditional: '0.4' found on page [live]
  B1_r_overall: Parsed '.24' -> 0.24 (source text: '.24')
  B2_r_overall: Parsed '0.24' -> 0.24
  B3_r_conditional: Parsed '0.4' -> 0.4
  [✓] B1: extracted .24 from quote
  [✓] B2: extracted 0.24 from quote
  [✓] B3: extracted 0.4 from quote
  SC1 cross-check: Pietschnig 2015 vs PMC 2022 overall r: 0.24 vs 0.24, diff=0.0, tolerance=0.01 -> AGREE

  SC1-A: |r_Pietschnig(0.24) - threshold(0.40)| = 0.1600
  SC1-A: Pietschnig r within ±0.05 of 0.40: 0.16000000000000003 <= 0.05 = False

  SC1-B: |r_PMC2022(0.24) - threshold(0.40)| = 0.1600
  SC1-B: PMC 2022 r within ±0.05 of 0.40: 0.16000000000000003 <= 0.05 = False
  SC1: max unconditional r deviation within ±0.05 of 0.40: 0.16000000000000003 <= 0.05 = False

  SC2:   |r_conditional(0.40) - threshold(0.40)| = 0.0000
  SC2: conditional r within ±0.05 of 0.40: 0.0 <= 0.05 = True
```

---

## Independent Source Agreement (Rule 6)

| Description | Source A | Source B | Agreement | Tolerance |
|-------------|----------|----------|-----------|-----------|
| SC1: unconditional overall r | Pietschnig 2015: r = 0.24 | PMC 2022: r = 0.24 | Yes | 0.01 absolute |

Both meta-analyses used different study samples, different time windows, and different methodological approaches (Pietschnig: weighted by inverse standard error, 88 studies; Nave: combinatorial + specification curve, 86 studies). They independently converge on r = 0.24. These are independently published sources from different research groups with different base corpora.

---

## Adversarial Checks (Rule 5)

**Check 1: Does any major unconditional meta-analysis report r = 0.40?**
- Question: Does any major unconditional meta-analysis report r = 0.40 for brain volume vs. IQ?
- Searched: 'brain volume IQ meta-analysis r = 0.4 overall'; reviewed McDaniel (2005), Pietschnig et al. (2015), Nave et al. (2022).
- Finding: No. McDaniel (2005) r = 0.33; Pietschnig (2015) r = .24; Nave (2022) r = 0.24. Gignac & Bates (2017) report r ≈ 0.40 only conditionally (excellent-quality tests).
- Breaks proof: No

**Check 2: Could publication bias be deflating estimates below 0.40?**
- Question: Could publication bias be deflating the estimates below 0.40?
- Searched: Publication bias analyses in Pietschnig et al. (2015) and Nave et al. (2022).
- Finding: Both papers found publication bias *inflates* (not deflates) reported correlations. Corrected estimates remain ~0.24. This is the opposite direction needed to rescue the r = 0.40 claim.
- Breaks proof: No

**Check 3: Is the Wikipedia source for SC2 credible?**
- Question: Is the Wikipedia source for SC2 citing a credible peer-reviewed finding?
- Searched: Gignac & Bates (2017), *Intelligence* (Elsevier). Paper found corrected r of .23 (fair), .32 (good), .39 (excellent quality), concluding the association "is arguably best characterised as r ≈ .40."
- Finding: SC2 is supported by peer-reviewed research. Conditional r ≈ 0.40 is credible for healthy adults with excellent-quality tests.
- Breaks proof: No

---

## Source Credibility Assessment

| Fact ID | Domain | Type | Tier | Note |
|---------|--------|------|------|------|
| B1 | nih.gov | government | 5 | PubMed — U.S. National Institutes of Health |
| B2 | nih.gov | government | 5 | PMC — U.S. National Institutes of Health |
| B3 | wikipedia.org | reference | 3 | Established reference source; SC2 conclusion backed by Gignac & Bates (2017, peer-reviewed) |

No sources have tier ≤ 2.

---

## Extraction Records

| ID | Value | Found in Quote | Quote Snippet | Extraction Method |
|----|-------|----------------|---------------|-------------------|
| B1 | 0.24 | Yes | "…brain volume and IQ (r=.24, R(2)=.06)…" | `parse_number_from_quote(".24", r"([.\d]+)", "B1_r_overall")` → float(".24") = 0.24 |
| B2 | 0.24 | Yes | "Brain size and IQ associations yielded r = 0.24…" | `parse_number_from_quote("0.24", r"([.\d]+)", "B2_r_overall")` → float("0.24") = 0.24 |
| B3 | 0.4 | Yes | "…approximately 0.4 when high-quality tests are used." | `parse_number_from_quote("0.4", r"([.\d]+)", "B3_r_conditional")` → float("0.4") = 0.4 |

All values parsed programmatically from data_values strings derived from page content; none hand-typed.

---

## Hardening Checklist

| Rule | Status | Notes |
|------|--------|-------|
| Rule 1: Every empirical value parsed from quote text | ✓ | All r values parsed via `parse_number_from_quote` from `data_values` strings |
| Rule 2: Every citation URL fetched and quote checked | ✓ | B1 full, B2 partial (50%; data value independently confirmed), B3 full |
| Rule 3: System time for date-dependent logic | N/A | No date-dependent computations |
| Rule 4: Claim interpretation explicit with operator rationale | ✓ | `CLAIM_FORMAL` with `operator_note`, tolerance documented |
| Rule 5: Adversarial checks searched for counter-evidence | ✓ | 3 checks: unconditional r = 0.40 search, publication bias direction, SC2 credibility |
| Rule 6: Cross-checks from independent sources | ✓ | Pietschnig 2015 and PMC 2022 independently report r = 0.24; agreement confirmed |
| Rule 7: No hard-coded constants or unsafe formulas | ✓ | All comparisons use `compare()`; `cross_check()` for source agreement |