99 — Sources¶

Annotated, grouped by the design question they answer. Links were resolved in September 2026. Where a number is quoted in this knowledge base it comes from the source listed here; where a secondary source (review, handbook chapter, summary page) was used rather than the primary paper, that is stated, because several of these effect sizes are widely repeated and worth verifying against the primary before we lean hard on them.

Retrieval practice / testing effect¶

• Pan, S. C. & Rickard, T. C. — Transfer of test-enhanced learning: Meta-analytic review and synthesis. Reports transfer of testing benefits at about d = 0.40. https://www.researchgate.net/publication/324995852_Transfer_of_Test-Enhanced_Learning_Meta-Analytic_Review_and_Synthesis
• The Use of Retrieval Practice in the Health Professions: A State-of-the-Art Review (2025) — a recent applied synthesis; collects the g ~ 0.50 laboratory and classroom estimates. https://pmc.ncbi.nlm.nih.gov/articles/PMC12292765/
• Effects of retrieval practice on retention and application of complex educational concepts (Learning and Instruction, 2025) — retrieval practice with genuinely complex material. https://www.sciencedirect.com/science/article/pii/S0959475225001434
• Glaser, J. & Richter, T. (2025) — The testing effect in the lecture hall: does it transfer to content studied but not practiced? Relevant to how narrowly our retrieval items must target. https://journals.sagepub.com/doi/10.1177/00986283231218943

Spacing¶

• Latimier, A., Peyre, H. & Ramus, F. (2021) — A meta-analytic review of the benefit of spacing out retrieval practice episodes on retention, Educational Psychology Review. g = 0.74 for spaced vs. massed retrieval; expanding vs. uniform schedules g = 0.034. https://link.springer.com/article/10.1007/s10648-020-09572-8 (PDF: http://www.lscp.net/persons/ramus/docs/EPR20.pdf)
• Cepeda et al. — distributed practice meta-analysis (254 studies, >14 000 participants); the 10–20 % gap-to-retention-interval rule of thumb derives from this line of work. Summarized in https://files.eric.ed.gov/fulltext/ED536925.pdf
• The spacing effect stands up to big data (Behavior Research Methods, 2019) — large-scale confirmation outside the lab. https://link.springer.com/article/10.3758/s13428-018-1184-7

Study strategies, ranked¶

• Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J. & Willingham, D. T. (2013) — Improving students' learning with effective learning techniques, Psychological Science in the Public Interest. The reference ranking: practice testing and distributed practice = high utility; elaborative interrogation, self-explanation, interleaving = moderate; summarization, highlighting, rereading = low. https://pubmed.ncbi.nlm.nih.gov/26173288/ (PDF: https://www.whz.de/fileadmin/lehre/hochschuldidaktik/docs/dunloskiimprovingstudentlearning.pdf)
• Dunlosky, J. (2013) — Strengthening the student toolbox, American Educator. The readable version. https://www.aft.org/ae/fall2013/dunlosky

Generation, desirable difficulties, fluency¶

• Does text generation improve learning from expository text? A conceptual replication attempt (Cognitive Research: Principles and Implications, 2025) — a caution against assuming word-list generation effects carry over to expository text. https://pmc.ncbi.nlm.nih.gov/articles/PMC12185794/
• Undesirable difficulty effects in the learning of high-element-interactivity materials — the boundary condition that matters for dense derivations. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6099118/
• Avoiding illusions of learning: strategies for improving self-regulated learning (Psychonomic Society) — the fluency illusion and premature stopping. https://featuredcontent.psychonomic.org/avoiding-illusions-of-learning-strategies-for-improving-self-regulated-learning/

Interleaving (and why we mostly do not use it)¶

• Brunmair, M. & Richter, T. (2019) — Similarity matters: a meta-analysis of interleaved learning and its moderators, Psychological Bulletin. Overall g = 0.42; paintings g = 0.67; expository texts g = 0.21 (n.s.); words g = -0.39. https://www.psychologie.uni-wuerzburg.de/fileadmin/06020400/2019/Brunmair_Richter_in_press__2019_META-ANALYSIS_OF_INTERLEAVED_LEARNING.pdf
• Firth, J. et al. (2021) — A systematic review of interleaving as a concept learning strategy. https://bera-journals.onlinelibrary.wiley.com/doi/10.1002/rev3.3266

Cognitive load, transiency, modality¶

• Leahy, W. & Sweller, J. (2011) — Cognitive load theory, modality of presentation and the transient information effect, Applied Cognitive Psychology. https://onlinelibrary.wiley.com/doi/abs/10.1002/acp.1787
• Singh, A.-M., Marcus, N. & Ayres, P. (2012) — The transient information effect: investigating the impact of segmentation on spoken and written text, Applied Cognitive Psychology. Segmentation removes the disadvantage of spoken text. https://onlinelibrary.wiley.com/doi/10.1002/acp.2885
• Wong, A. et al. (2012) — Cognitive load theory, the transient information effect and e-learning, Learning and Instruction. https://www.sciencedirect.com/science/article/abs/pii/S0959475212000369
• Cognitive load theory and the effects of transient information on the modality effect (Instructional Science, 2016). https://link.springer.com/article/10.1007/s11251-015-9362-9

Multimedia design principles¶

• Mayer, R. E. — Multimedia Learning (3^rd ed.) and The Cambridge Handbook of Multimedia Learning. Chapter on segmenting / pre-training / modality: https://www.cambridge.org/core/books/abs/cambridge-handbook-of-multimedia-learning/principles-for-managing-essential-processing-in-multimedia-learning-segmenting-pretraining-and-modality-principles/DD24C2F48B9B1277CE59F78276110258 Chapter on personalization / voice / image / embodiment: https://www.cambridge.org/core/books/abs/cambridge-handbook-of-multimedia-learning/principles-based-on-social-cues-in-multimedia-learning-personalization-voice-image-and-embodiment-principles/3841340D8AD820C26DBCD39AE664BCEC
• Mayer, R. E. & Moreno, R. — Applying the science of learning: evidence-based principles for the design of multimedia instruction. Source of the commonly quoted effect sizes: coherence 0.97, signaling 0.52, segmenting ~0.98, personalization median d = 1.11, voice median d = 0.74. https://www.researchgate.net/publication/23478495_Applying_the_Science_of_Learning_Evidence-Based_Principles_for_the_Design_of_Multimedia_Instruction
• Mayer, R. E. & Fiorella, L. — Evidence-based principles for how to design effective instructional videos (2021). https://www.sciencedirect.com/science/article/abs/pii/S2211368121000231

Prequestions / pretesting¶

• Pan, S. C. & Carpenter, S. K. (2023) — Prequestioning and pretesting effects: a review of empirical research, theoretical perspectives, and implications for educational practice, Educational Psychology Review. https://link.springer.com/article/10.1007/s10648-023-09814-5
• Guessing as a learning intervention: a meta-analytic review of the prequestion effect (Psychonomic Bulletin & Review, 2023). Specific effect g = .66; general effect absent. https://link.springer.com/article/10.3758/s13423-023-02353-8
• The effect of prequestions on learning: a multilevel meta-analysis (Educational Psychology Review, 2025). Specific g = 0.54; general g = 0.04. https://link.springer.com/article/10.1007/s10648-025-10075-7

Narrative, seductive details¶

• Memory and comprehension of narrative versus expository texts: a meta-analysis (Psychonomic Bulletin & Review, 2020). 150 effect sizes, >33 000 participants. https://link.springer.com/article/10.3758/s13423-020-01853-1
• Lehman, S., Schraw, G. et al. — Processing and recall of seductive details in scientific text, Contemporary Educational Psychology. https://www.sciencedirect.com/science/article/abs/pii/S0361476X06000300
• Informative narratives increase students' situational interest in science topics (Learning and Instruction, 2024) — interest up, comprehension not. https://www.sciencedirect.com/science/article/pii/S0959475224001002

Dual coding, concreteness, and a usable lexical resource¶

• Paivio, A. — dual coding theory; overview: https://www.sciencedirect.com/topics/neuroscience/dual-coding-theory
• Brysbaert, M., Warriner, A. B. & Kuperman, V. (2014) — Concreteness ratings for 40 thousand generally known English word lemmas, Behavior Research Methods. Freely downloadable norms; this is the dataset behind our abstractness scoring. Also relevant: Kuperman et al. age-of-acquisition norms, and the Brysbaert word-prevalence norms.
• Effects of concreteness and semantic relatedness on composite imagery ratings and cued recall (Memory & Cognition). https://link.springer.com/article/10.3758/BF03214222

Listening vs. reading, speech rate¶

• Rogowsky, B. A., Calhoun, B. M. & Tallal, P. (2016) — Does modality matter? The effects of reading, listening, and dual modality on comprehension. https://journals.sagepub.com/doi/10.1177/2158244016669550
• Rubin, D. et al. (2021) — A lingering question addressed: reading rate and most efficient listening rate are highly similar. https://pubmed.ncbi.nlm.nih.gov/34516216/
• A speech-rate intelligibility threshold for speeded and time-compressed connected speech. https://link.springer.com/content/pdf/10.3758/BF03199702.pdf
• The effects of speech rate on comprehension. https://digitalcommons.lindenwood.edu/cgi/viewcontent.cgi?article=1215&context=psych_journals

Verbalizing non-prose content (accessibility + TTS engineering)¶

• DIAGRAM Center — Image Description Guidelines, including the specific guidance for graphs and charts. http://diagramcenter.org/table-of-contents-2.html, http://diagramcenter.org/specific-guidelines-e.html (note: the site currently serves a mismatched TLS certificate; fetch over plain HTTP or use the Internet Archive copy)
• SIGACCESS — Describing Figures: the concrete rule set we implement (title first, general to specific, do not repeat the text, axes and units, colour only when meaningful, expand abbreviations, tabulate dense data). https://www.sigaccess.org/welcome-to-sigaccess/resources/describing-figures/
• W3C WAI — Complex Images tutorial https://www.w3.org/WAI/tutorials/images/complex/ and Tables tutorial https://www.w3.org/WAI/tutorials/tables/
• WebAIM — Creating accessible tables https://webaim.org/techniques/tables/
• MIT Visualization Group — Rich screen reader experiences for accessible data visualization. Useful model for hierarchical (overview then drill-down) narration of charts. https://vis.csail.mit.edu/pubs/rich-screen-reader-vis-experiences/
• Frankel, L., Brownstein, B. & Soiffer, N. (2016) — Development and initial evaluation of the ClearSpeak style for automated speaking of algebra, ETS Research Report. https://files.eric.ed.gov/fulltext/ED570857.pdf
• Soiffer, N. — A comparison of different styles of speech for mathematics. https://scholarworks.calstate.edu/downloads/5t34sv64c
• Isaacson, M. — MathSpeak; overview and integration: https://jsesd.github.io/web-articles/JSESD-Volume-14/1015-Isaacson/
• Zhang, H., Sproat, R. et al. (2019) — Neural models of text normalization for speech applications, Computational Linguistics 45(2). Why hand-written grammars still guard the numbers. https://aclanthology.org/J19-2004/

Software we may reuse rather than reimplement¶

• Speech Rule Engine (Volker Sorge) — the open-source implementation of MathSpeak/ClearSpeak used by MathJax; Node-based, callable from Python as a subprocess.
• Brysbaert concreteness / AoA / prevalence norms — plain CSV, no licence obstacle for research use; check the terms before redistribution.
• GROBID — machine-learning extraction of structured TEI from scholarly PDFs (sections, references, figures); the strongest option for papers specifically.
• PyMuPDF, docling, marker — general PDF-to-structure; ebooklib for EPUB.
• Awesome-Chart-Understanding — a maintained bibliography of chart QA, chart captioning and chart-to-table work, including the benchmarks that measure how often a vision model gets a chart wrong (CHOCOLATE, CharXiv) and the caption corpora that could score ours (SciCap, Chart-to-Text). https://github.com/khuangaf/Awesome-Chart-Understanding
• VoiceStudio — a fully local desktop application wrapping sixteen TTS engines, which exposes a local REST API with OpenAI-compatible audio endpoints. Interesting for M7 not as a dependency but as a shape: an adapter written against the OpenAI speech endpoint reaches VoiceStudio and several other local servers at once, which is a better target than a Piper-specific adapter. Licensing needs care and mimem must not bundle it — the application is AGPL-3.0 and its default voice weights are CC-BY-NC, so it is a thing a user runs, not a thing mimem ships. https://github.com/debpalash/VoiceStudio