This is a large-scale (N=356) multi-laboratory replication of DeLong, Urbach & Kutas (2005), testing whether readers pre-activate the phonological form of upcoming nouns during sentence comprehension. Participants read sentences word-by-word (RSVP, 2 words per second) that contained indefinite articles (a/an) preceding either highly expected or unexpected nouns (based on cloze probability), while EEG was recorded.
Nine laboratories in the UK collected data following a pre-registered replication protocol (https://osf.io/eyzaq). The original study by DeLong et al. reported N400-like effects on the indefinite articles (larger negativity for unexpected articles). Nieuwland et al. found reliable N400 effects on the target nouns but no statistically significant effect on the preceding articles, challenging strong prediction accounts.
- 356 total participants (222 women / 134 men)
- All right-handed, native English speakers
- Age 18–35 years (mean 19.8)
- Normal or corrected-to-normal vision
- Free from known language or learning disorders
- 89 reported a left-handed parent or sibling
After applying the paper's quality threshold (<60/80 article or noun trials),
334 subjects were retained in the statistical analyses. In this BIDS release
we include ALL subjects for which raw data is available, with an
included_in_paper flag in participants.tsv so users can filter themselves.
| Lab (paper #) | Institution | Format | Sfreq | Channels |
|---|---|---|---|---|
| BIRM (1) | University of Birmingham | BrainVision | 500 Hz | 64 EEG |
| BRIS (2) | University of Bristol | BrainVision | 1000 Hz | 32 EEG |
| EDIN (3) | University of Edinburgh | BioSemi BDF | 512 Hz | 64 EEG + 8 EXG |
| GLAS (4) | University of Glasgow | BioSemi BDF | 512 Hz | 128 EEG + 8 EXG |
| KENT (5) | University of Kent | BrainVision | 500 Hz | 64 EEG + HEOG/VEOG |
| LOND (6) | University College London | BioSemi BDF | 512 Hz | 32 EEG + 8 EXG |
| OXFO (7) | University of Oxford | BioSemi BDF | 2048 Hz | 64 EEG + 8 EXG |
| STIR (8) | University of Stirling | Neuroscan CNT | 250 Hz | 64 EEG + EOG |
| YORK (9) | University of York | BrainVision | 500 Hz | 64 EEG + HEOG/VEOG |
- Word-by-word RSVP: 200 ms word duration + 300 ms blank (2 words/sec)
- 80 Delong replication sentences + 80 control sentences
- Comprehension questions on a subset of trials (yes/no button response)
- Two counter-balanced stimulus lists (list 1 / list 2)
task-delong: Main experiment (all subjects, all labs)task-control: Control grammaticality experiment (BRIS subjects, LOND 1-2)
Delong experiment: a_expected — article "a", expected (high cloze) context an_expected — article "an", expected (high cloze) context a_unexpected — article "a", unexpected (low cloze) context an_unexpected — article "an", unexpected (low cloze) context noun_expected — target noun, expected condition noun_unexpected — target noun, unexpected condition final_expected — sentence-final word, expected condition final_unexpected — sentence-final word, unexpected condition
Control experiment: control_correct — grammatically correct article control_incorrect — grammatically incorrect article
General: cloze_marker — cloze probability marker (trigger 1-100 or 200) item_marker — stimulus item marker (trigger 101-180) question — comprehension question onset filler_word — any other (non-critical) word in sentence unknown_trigger — trigger code not matched to any known category
Each event in events.tsv is enriched (when applicable) with:
- sequence_id, item_number, list, task_type, condition
- expected_article / unexpected_article (a or an)
- expected_noun / unexpected_noun (strings)
- expected_cloze / unexpected_cloze (0-100)
- plausibility_expected / plausibility_unexpected (1-7 Likert)
- sentence_context / sentence_ending (strings)
- has_question, question_text, question_answer
These come from the authors' REPLICATION_ITEMS.xlsx file on OSF.
participant_id — sub- lab — birm/bris/edin/glas/kent/lond/oxfo/stir/york lab_number — 1-9 (paper's numbering) institution — full institution name list — stimulus list (1 or 2) accuracy — % correct on comprehension questions (from OSF) n_article_trials — article trials kept (out of 80) n_noun_trials — noun trials kept (out of 80) included_in_paper — True if >=60/80 trials (paper's threshold) exclusion_note — e.g. "random_answers", "non_native", "low_trials" hand — R (all right-handed) age_range — 18-35 (all participants) native_language — English (all participants) recording_system — manufacturer + model
- Original raw data is kept — no filtering, no resampling, no artifact rejection
- Channel types: EEG, EOG, and misc (peripheral) channels are labeled
- For BDF labs, channels EXG1-8, GSR1/2, Erg1/2, Resp, Plet, Temp are marked misc
- GLAS has a 128-channel BioSemi montage (biosemi128)
- STIR data is read with a custom Neuroscan CNT parser (MNE's built-in reader has a bug with the corrupted total_samples header field)
- OXFO has 3 subjects recorded with BrainVision instead of BDF
Nieuwland, M.S., Politzer-Ahles, S., Heyselaar, E., Segaert, K., Darley, E., Kazanina, N., ..., Huettig, F. (2018). Large-scale replication study reveals a limit on probabilistic prediction in language comprehension. eLife, 7, e33468. https://doi.org/10.7554/eLife.33468
Appelhoff, S., Sanderson, M., Brooks, T., Vliet, M., Quentin, R., Holdgraf, C., Chaumon, M., Mikulan, E., Tavabi, K., Höchenberger, R., Welke, D., Brunner, C., Rockhill, A., Larson, E., Gramfort, A. and Jas, M. (2019). MNE-BIDS: Organizing electrophysiological data into the BIDS format and facilitating their analysis. Journal of Open Source Software 4: (1896).https://doi.org/10.21105/joss.01896
Pernet, C. R., Appelhoff, S., Gorgolewski, K. J., Flandin, G., Phillips, C., Delorme, A., Oostenveld, R. (2019). EEG-BIDS, an extension to the brain imaging data structure for electroencephalography. Scientific Data, 6, 103.https://doi.org/10.1038/s41597-019-0104-8