How Much Mightier Is the Pen than the Keyboard for Note-Taking? A Replication and Extension of Mueller and Oppenheimer (2014)

Experiment 1 — 193 recruited, 187 analysed (62 longhand, 64 laptop, 61 eWriter). Experiment 2 — 222 undergraduates across four groups (longhand, laptop, eWriter, and a NO-NOTES control of 32). Roughly three times the per-cell sample of the original.

Undergraduates at a large Midwestern university (Kent State), watching TED talks in individual cubicles. Laboratory, not a real course — the same population and setting as the original, which is what makes it a direct rather than a conceptual replication.

A direct replication using Mueller & Oppenheimer's own videos, instructions, question types, distractor interval and scoring, extended in three ways that turn out to matter more than the replication itself. (1) An eWriter arm (write with a stylus on an electronic tablet) to separate the physical act of handwriting from the paper. (2) A NO-NOTES control in Experiment 2 — the condition the original never ran. (3) A delayed test in Experiment 2 preceded by seven minutes of studying one's own notes, which tests the STORAGE function of note-taking rather than only the encoding function. Random assignment to method, video and test group throughout. Notes were transcribed and scored for word count, verbatim overlap, idea units and test relevance. WHY C AND NOT B: the samples are adequate (n = 187 and 222, about three times the original's per cell) and this is a genuine direct replication, but every outcome is still a researcher-designed quiz on TED talks in a lab cubicle, with no standardized measure, no course grade and no stake for the participant — the measure-type discount METHODOLOGY.md applies keeps it off B. Overall performance in Experiment 1 was near floor, which the authors acknowledge and address in Experiment 2 by using only the two easiest videos. Independent academic work; no funder with a stake in either medium. SCOPE NOTE: this pair (this file and the original) is filed under ed-tech rather than with the archive's handwriting sources on purpose. The question here is not whether handwriting builds letter knowledge or transcription fluency — it is whether a DEVICE in a lecture costs learning, which is the argument schools actually invoke when writing a laptop policy.

The famous conceptual-question advantage of longhand did not replicate. In Experiment 1 the immediate factual difference favoured longhand and reached the authors' one-tailed threshold (t(61) = 1.79, p = .04, d = 0.45) but the CONCEPTUAL difference — the original's headline — was d = 0.13, p = .31. In Experiment 2 the trend on conceptual questions actually REVERSED toward laptops (d = -0.35, p = .09), and no delayed-test difference between methods survived after students studied their notes. The most damaging extension is the one the original never ran: the NO-NOTES group performed as well as, and sometimes numerically better than, every note-taking group — so in this paradigm there is no detectable encoding benefit of taking notes at all, which makes a medium contrast between two ways of taking them hard to interpret. eWriter notes behaved like longhand and did not differ from either other group on performance. What did replicate cleanly is the MECHANISM without its consequence: laptop notes were longer (d = 0.44 and 0.58) and, in Experiment 1, more verbatim (d = 0.43; in Experiment 2 the overlap difference was not significant, d = 0.20). Pooling all four direct replications gives a small non-significant longhand advantage: conceptual d = 0.17, 95% CI [-0.07, 0.40]; factual d = 0.22, 95% CI [-0.01, 0.46].

Minimal (randomized allocation to note-taking method in both experiments).

Direction, stated so the flag is not misread: this paper IS the replication — of Mueller & Oppenheimer (2014; `edt-mueller-oppenheimer-2014-longhand-vs-laptop`) — and it is the reason that source is flagged `failed`. Whether this paper's own null has itself been independently reproduced we have not checked, so `unclear` rather than `unreplicated`, which would assert a fact about the literature we cannot source. What it does carry internally is a continuously cumulating meta-analysis pooling all four direct-replication experiments (Mueller & Oppenheimer's Studies 1 and 2 plus these two), which is the best single estimate available: conceptual d = 0.17, 95% CI [-0.07, 0.40], p = .16; factual d = 0.22, 95% CI [-0.01, 0.46], p = .06. The authors' own summary is that "concluding which method is superior for improving the functions of note-taking seems premature" — not that longhand is worse, and not that laptops are fine. The archive should carry that agnosticism rather than flipping the original's conclusion.