File size: 10,414 Bytes
b4ff819 914fe82 b4ff819 914fe82 b4ff819 ec6a360 b4ff819 914fe82 b4ff819 914fe82 b4ff819 914fe82 b4ff819 c4b9198 b4ff819 c4b9198 b4ff819 c4b9198 58b4fa7 b4ff819 58b4fa7 b4ff819 58b4fa7 b4ff819 ec6a360 58b4fa7 ec6a360 58b4fa7 ec6a360 58b4fa7 ec6a360 58b4fa7 b4ff819 914fe82 58b4fa7 b4ff819 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 | # Brain–language-model alignment: ds006239 (whole-brain)
Wang et al. 2025 — word-level phonological and semantic reading tasks in children and adolescents aged 10–17.
- Paper: https://www.sciencedirect.com/science/article/pii/S2352340925009692
- Data: https://openneuro.org/datasets/ds006239/versions/1.0.5
- Generated: 2026-09-07
- Pipeline: https://github.com/suchirsalhan/cdl-representations-brains-babylms
## Read this first: does the measurement work?
Every alignment number in this dataset is only as meaningful as the brain
RDMs it was computed against. So before any model result, the same
pipeline is asked whether *anything* stimulus-driven correlates with those
RDMs — stimulus duration, intensity, word length, frequency, phoneme and
syllable counts, an acoustic model of the audio where the stimuli are
audio, and the study's own condition contrast — each tested by a
permutation test that shuffles stimulus identity.
**GATE: FAILED. 0/38 stimulus tests are significant** after Holm
correction — not the acoustic model of the audio the children actually
heard, not the study's own experimental contrast.
**The alignment numbers below are therefore uninterpretable as
evidence about language models.** They measure a representational
geometry that does not demonstrably encode the stimuli. They are
published for completeness and for whoever fixes the estimator, not as
a result. Do not cite them as evidence that models fail to align with
the developing brain.
Measured cause, from `control/`:
- RDM effective rank: **57** of 84 stimuli
Note that this is NOT ds003604's failure mode. There, the RDM
effective rank was ~3 of 40-48 stimuli -- near-degenerate betas
that could not express stimulus-level structure at all. The rank
recorded above is a large fraction of the stimulus count, so these
RDMs do carry stimulus structure and the control failing here means
the specific controls tested did not reach significance, not that
the measurement is uninterpretable. Check `control/` for which
controls ran: an acoustic or visual control needs the dataset's
stimulus files present, and reports zero features if they are not.
## What was built
40 task × session cells, each an RDM over the stimuli
shared by that cell's subjects, with voxel patterns z-scored **within
run** before aggregation (without that, the RDM measures scanner drift
rather than language) and an inter-subject noise ceiling.
| task | session | n_stim | ceiling_lower | ceiling_upper | ceiling_n |
|:---------|:----------|---------:|----------------:|----------------:|------------:|
| Orth | ses-11+ | 96 | 0.552402 | 0.707302 | 5 |
| Orth | ses-11 | 96 | 0.508439 | 0.645119 | 7 |
| Phon | ses-11+ | 96 | 0.552402 | 0.707302 | 5 |
| Phon | ses-11 | 96 | 0.508439 | 0.645119 | 7 |
| Sem | ses-11+ | 72 | 0.370744 | 0.591736 | 5 |
| Sem | ses-11 | 72 | 0.2928 | 0.489077 | 7 |
| SemLocal | ses-11+ | 48 | 0.265494 | 0.526554 | 5 |
| SemLocal | ses-11 | 48 | 0.235427 | 0.477001 | 6 |
| Orth | ses-11+ | 96 | 0.527509 | 0.760826 | 3 |
| Orth | ses-11 | 96 | 0.395701 | 0.694777 | 3 |
| Phon | ses-11+ | 96 | 0.527509 | 0.760826 | 3 |
| Phon | ses-11 | 96 | 0.395701 | 0.694777 | 3 |
| Sem | ses-11+ | 72 | 0.313565 | 0.658019 | 3 |
| Sem | ses-11 | 72 | 0.185478 | 0.586945 | 3 |
| SemLocal | ses-11+ | 48 | 0.208521 | 0.601744 | 3 |
| SemLocal | ses-11 | 48 | 0.216461 | 0.607091 | 3 |
| Orth | ses-11+ | 96 | 0.527509 | 0.760826 | 3 |
| Orth | ses-11 | 96 | 0.395701 | 0.694777 | 3 |
| Phon | ses-11+ | 96 | 0.527509 | 0.760826 | 3 |
| Phon | ses-11 | 96 | 0.395701 | 0.694777 | 3 |
| Sem | ses-11+ | 72 | 0.313565 | 0.658019 | 3 |
| Sem | ses-11 | 72 | 0.185478 | 0.586945 | 3 |
| SemLocal | ses-11+ | 48 | 0.208521 | 0.601744 | 3 |
| SemLocal | ses-11 | 48 | 0.216461 | 0.607091 | 3 |
| Orth | ses-11+ | 96 | 0.527509 | 0.760826 | 3 |
| Orth | ses-11 | 96 | 0.395701 | 0.694777 | 3 |
| Phon | ses-11+ | 96 | 0.527509 | 0.760826 | 3 |
| Phon | ses-11 | 96 | 0.395701 | 0.694777 | 3 |
| Sem | ses-11+ | 72 | 0.313565 | 0.658019 | 3 |
| Sem | ses-11 | 72 | 0.185478 | 0.586945 | 3 |
| SemLocal | ses-11+ | 48 | 0.208521 | 0.601744 | 3 |
| SemLocal | ses-11 | 48 | 0.216461 | 0.607091 | 3 |
| Orth | ses-11+ | 96 | 0.527509 | 0.760826 | 3 |
| Orth | ses-11 | 96 | 0.395701 | 0.694777 | 3 |
| Phon | ses-11+ | 96 | 0.527509 | 0.760826 | 3 |
| Phon | ses-11 | 96 | 0.395701 | 0.694777 | 3 |
| Sem | ses-11+ | 72 | 0.313565 | 0.658019 | 3 |
| Sem | ses-11 | 72 | 0.185478 | 0.586945 | 3 |
| SemLocal | ses-11+ | 48 | 0.208521 | 0.601744 | 3 |
| SemLocal | ses-11 | 48 | 0.216461 | 0.607091 | 3 |
Model grid: **15 families**, 10480 alignment rows across 8 cells.
| | |
|---|---|
| mean noise ceiling | 0.359 |
| best alignment anywhere | 0.0913 |
| as a fraction of ceiling | 42.2% |
| families equivalent to zero (TOST ±0.05) | 15/15 |
| Pythia scale trend | ρ = -0.190, p = 0.24 |
### Per family
| family | n_checkpoints | rsa_mean | rsa_sd | rsa_abs_max | frac_of_ceiling_abs_max | p_equivalence_tost |
|:----------------------|----------------:|-----------:|---------:|--------------:|--------------------------:|---------------------:|
| pico-decoder-tiny | 21 | 0.0093 | 0.0104 | 0.0698 | 0.3227 | 0 |
| pico-decoder-small | 21 | 0.0047 | 0.0219 | 0.0913 | 0.4218 | 0.0003 |
| pico-decoder-large | 21 | 0.0011 | 0.0134 | 0.0677 | 0.3127 | 0 |
| pico-decoder-medium | 21 | 0.0005 | 0.0116 | 0.0776 | 0.3585 | 0 |
| pythia-70m-full | 21 | 0.0004 | 0.0113 | 0.0596 | 0.2755 | 0 |
| babylm-gpt2 | 9 | 0 | 0.0155 | 0.0494 | 0.2281 | 0 |
| pythia-410m-full | 21 | -0.0011 | 0.0122 | 0.0603 | 0.2787 | 0 |
| beetle-humanscale-eng | 18 | -0.0034 | 0.0083 | 0.0401 | 0.2164 | 0 |
| pythia-1b-full | 21 | -0.0048 | 0.0067 | 0.0498 | 0.2303 | 0 |
| beetle-fineweb3-eng | 19 | -0.007 | 0.0071 | 0.0576 | 0.2662 | 0 |
| pythia-160m-full | 21 | -0.0096 | 0.0104 | 0.0573 | 0.2649 | 0 |
| pythia-1.4b-full | 21 | -0.0097 | 0.0083 | 0.048 | 0.2519 | 0 |
| babylm-gpt2-7 | 9 | -0.0257 | 0.0293 | 0.062 | 0.1462 | 0.0256 |
| babylm-gpt2-5 | 9 | -0.0262 | 0.0291 | 0.0636 | 0.1478 | 0.0268 |
| babylm-gpt2-3 | 9 | -0.0263 | 0.0296 | 0.0649 | 0.1513 | 0.029 |
## Dataset-specific notes
Contains **LocalSem**, the only genuinely run/stimulus-CROSSED language cell across all four datasets in this project: its stimuli recur across runs, so run identity and stimulus identity are separable and the scanner-run confound that invalidated the first ds003604 analysis cannot arise. Per-subject age is NOT recoverable from the release — participants.tsv has birthdate but no scan date and there are no *_scans.tsv files — so this dataset is cohort-level only and cannot carry the developmental axis as published.
## Files
| path | what | present here |
|---|---|---|
| `alignment_by_checkpoint.csv` | every model × checkpoint × cell, with ceiling | ✓ |
| `alignment_by_family.csv` | per family, with equivalence tests | ✓ |
| `alignment_by_cell.csv` | per task × session | ✓ |
| `ceilings_ds006239.csv` | noise ceiling per cell | ✓ |
| `control/` | the positive control and RDM dimensionality — the gate | ✓ |
| `scale_ladder.csv` | the Pythia 70M→1.4B scale test | ✓ |
| `fig_*.pdf, fig_*.png` | figures | ✓ |
## Method
Representational similarity analysis. For each cell, a brain RDM over
stimuli (correlation distance between per-stimulus GLM beta patterns,
within-run z-scored, aggregated across subjects) is compared by Spearman
correlation with a model RDM over the same stimuli, taken from each
checkpoint's hidden states. Alignment is reported raw and as a fraction of
the inter-subject noise ceiling, and judged against a null built from the
PARC suite — 18 models differing only by random seed, which is what 'no
effect' looks like on this measurement.
Null and fixation trials are excluded from the stimulus set. For paired
designs the stimulus identity is the pair, not either word alone. |