Pith. sign in

REVIEW 3 major objections 5 minor 51 references

Thoth: Improved Rapid Serial Visual Presentation using Natural Language Processing

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Per-word timing based on word familiarity can make RSVP speed reading faster and more comprehensible.

desk verdict An honest tool paper whose central speed/comprehension claim is asserted, not demonstrated; the open-source implementation is real, but there is no user study to back any of the results. read the letter →

arxiv 1908.01699 v1 pith:SL7DDKLR submitted 2019-08-05 cs.CL cs.HC

David Awad This is my paper · ORCID
classification cs.CLcs.HC
keywords rapidserialvisualpresentationRSVPspeedreadingnaturallanguageprocessingreadabilityformulaswordfamiliaritycomprehensiontext
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that rapid serial visual presentation (RSVP) speed-reading tools are held back by uniform pacing: showing every word for the same fixed time ignores the fact that familiar and unfamiliar words take different amounts of processing. It presents Thoth, an open-source RSVP reader that scores each word with readability formulas and extends the display duration of words flagged as unfamiliar by a fixed multiplier. The central claim is that this word-by-word timing lets people read low- and medium-fidelity content faster while keeping more context and comprehension than conventional RSVP tools. Most of the supporting evidence is drawn from reading science; the paper's own user study is listed as future work.

What carries the argument

The mechanism is a readability-weighted timing rule: a readability formula's familiar-word list marks each token, and the RSVP engine multiplies the default display duration for any word not on the list by a factor of 1.5. The argument for why this should work rests on the visual word form dictionary, the brain's stored picture-like representations of known words, which make familiar words fast to recognize and unfamiliar words slow. What the rule does is convert a whole-text readability score into a per-word scheduling decision.

What would settle it

Run a controlled experiment where matched readers see the same passages under Thoth, a fixed-rate RSVP reader, and ordinary static text, then take comprehension tests at matched reading speeds; the central claim collapses if Thoth is neither faster at equal comprehension nor better at comprehension at equal speed. A cheaper proxy: eye-tracking would show whether unfamiliar-word labels actually predict longer fixations under RSVP.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that RSVP does not have to treat text as a flat sequence of equal units. Thoth combines several readability measures to estimate a text's required grade level, uses one of them to classify each word as familiar or unfamiliar, and assigns display times accordingly; unfamiliar words are shown roughly 1.5 times as long as familiar ones. Because familiar words can be recognized as whole images while unfamiliar words demand extra labor, the tool spends the limited resource of screen time where it is needed. The paper concludes that this approach yields faster average reading and better retention for the medium- and low-fidelity documents that people skim.

Load-bearing premise

The load-bearing premise is that a word marked as unfamiliar truly needs more display time, and that stretching it by 1.5 times is the correct amount; the paper adopts that factor as a fixed assumption, with no measurements behind it.

Editorial extensions

If this is right

  • If per-word timing works, RSVP tools can be tuned from text statistics alone, without eye tracking or user calibration.
  • Skimmers of long documents could keep comprehension close to normal while reading faster than current fixed-rate readers.
  • Readability formulas gain a new role: not just grading whole texts but scheduling individual words.
  • An open-source implementation means the timing rule can be tested, improved, and extended by other developers.
  • The same timing logic could be reversed into a 'speed writing' mode that substitutes unfamiliar words with familiar synonyms before display.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The strongest test of the paper's logic is a direct A/B comparison of Thoth, a uniform RSVP reader, and static text on the same passages with comprehension checks; the paper does not report such a study, so the central claim is an engineering prediction rather than a measured result.
  • A graded difficulty signal such as word frequency or surprisal would likely outperform the binary familiar/unfamiliar split, and the 1.5x multiplier could be tuned per user or per text.
  • The same scheduling principle—give more time to predicted-hard items—generalizes to flashcard decks, subtitles, and captioning, where pacing is currently uniform.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper describes Thoth, an open-source rapid serial visual presentation (RSVP) speed-reading tool that uses natural language processing and multiple readability formulas (Dale-Chall, Flesch-Kincaid, SMOG, Spache, Coleman-Liau, etc.) to estimate word-level familiarity and assign per-word display times. The central claim, stated in the abstract and in Section 7, is that readability-based per-word timing improves reading speed and comprehension relative to conventional RSVP tools. The manuscript describes the tool's architecture, lists related work, and identifies future directions, including tunable parameters and user research. However, Section 5, titled 'Results,' contains no experimental measurements, and Section 8.3 explicitly defers user studies to future work; the effectiveness claim is therefore asserted rather than demonstrated.

Significance. If the central claim were supported, the idea of adapting RSVP presentation rates to lexical familiarity would be a useful contribution to reading technology and human-computer interaction. The paper has some concrete strengths: the tool is open source and publicly hosted, the design integrates several established readability measures, and the mechanism is described clearly enough to be implemented and tested by others. However, the paper's main claim is an empirical causal claim about human reading speed and comprehension, and no user data, baseline comparison, or measurement of any kind is provided. The plausibility of using word familiarity to modulate display time does not substitute for evidence, and as it stands the paper is a system description without validation of its central assertion.

major comments (3)
  1. [§5, §7, §8.3] The central claim that Thoth 'has enabled users to read through medium and low fidelity content faster on average while retaining more context and comprehension' is not supported by any experimental evidence. Section 5, titled 'Results,' contains no measurements, participants, or comparisons; it only reports tool availability and uniqueness. Section 8.3 explicitly lists user research as future work, confirming that the required evaluation has not been performed. To support the conclusion, the paper would need a controlled study measuring reading speed and comprehension for Thoth against at least one baseline condition (e.g., conventional RSVP or normal reading); no such study is present.
  2. [§8.1] The fixed 1.5x display-time multiplier for unfamiliar words is a load-bearing parameter of the proposed mechanism, but the paper provides no empirical justification, user study, or cited prior result for this specific value. If this mapping is incorrect, the claimed comprehension benefit does not follow. The paper itself acknowledges that 'it is possible we are losing time by simply scaling the display time of each unfamiliar word by 1.5,' which underscores that this parameter remains unvalidated.
  3. [§6] The statement that 'it doesn't seem to make a significant different which dictionary is used' is presented as a finding, but no analysis or data supporting it is given. The sentence also conflates 'significant' as a statistical term with 'significant' as a substantive judgment, and the claim should either be removed or supported with a formal comparison of the dictionaries under consideration.
minor comments (5)
  1. [Abstract, §3, §4] The possessive 'its' is repeatedly written as 'it's' (e.g., Abstract 'It's largest insight,' §3 'it's ease of use,' §4 'it's presentation'); these should be corrected.
  2. [§6] The phrase 'significant different' should be 'significant difference.'
  3. [References] The reference list contains irrelevant or unexplained entries (e.g., #13 'What is the amplitude of a wave?' and #22 'Effects of the Seasons and of Bright Light ...') and duplicates (#10 and #27 are the same Dehaene et al. citation; #11 and #35 are the same Deheane book). Several in-text citations do not match the reference list format (e.g., 'Gelzer et. al, 2015' appears as 'Glezer, L., et al.' in the list).
  4. [Figure 1] The caption says 'Source: Rayner, K. sagepub.com' but no complete citation for this figure is provided in the reference list.
  5. [§5] The opening sentence 'The results have been clear' is misleading because no results are presented; consider retitling the section to 'System Availability' or similar.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's benefit claim is unsupported assertion, not a derivation that reduces to its inputs.

full rationale

The paper contains no derivation chain in which a purported prediction is equivalent to its inputs by construction. The central design rule — using Dale-Chall word familiarity to scale display time by a fixed 1.5 multiplier for unfamiliar words — is defined independently of the claimed outcome of faster reading with retained comprehension. That outcome is asserted in the conclusion ('Thoth has enabled users to read through medium and low fidelity content faster on average while retaining more context and comprehension') rather than derived from the display-time rule, so there is no fitted parameter renamed as a prediction and no self-definitional loop. The paper's own Section 8.3 lists user research as future work, confirming that the central effectiveness claim has not been tested, but absence of evidence is a correctness risk, not circularity. There are no load-bearing self-citations and no imported uniqueness theorems. Accordingly, the appropriate circularity score is 0.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central claim rests on an unvalidated mapping between readability-sourced word familiarity and optimal display duration, plus an arbitrary 1.5 multiplier. No new entities are introduced.

free parameters (1)
  • Unfamiliar word display time scaling factor = 1.5
    Section 8.1 states 'simply scaling the display time of each unfamiliar word by 1.5'. This is a fixed, author-chosen multiplier with no user data or prior result supporting it.
assumptions (3)
  • domain assumption Dale-Chall word familiarity is a valid predictor of the processing time a reader needs for a word.
    Section 4 says Thoth uses Dale-Chall to determine how familiar a word is, and Section 4 states unfamiliar words should be displayed longer. This mapping is assumed without empirical validation.
  • ad hoc to paper Showing an unfamiliar word for 1.5x longer improves comprehension.
    Section 8.1 admits 'fixed assumptions about how much longer to display a word that is unfamiliar' and mentions 1.5 as the scaling factor. The benefit is asserted, not measured.
  • domain assumption The top 1000 words comprise 80% of English language text.
    Section 4 states this without a citation; it motivates word familiarity timing but is not sourced in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Thoth: Improved Rapid Serial Visual Presentation using Natural Language Processing." pith.science (2026). https://pith.science/paper/SL7DDKLR

@misc{pith2026190801699,
  author       = {Pith},
  title        = {Pith review of: Thoth: Improved Rapid Serial Visual Presentation using Natural Language Processing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SL7DDKLR}},
  note         = {Machine review of arXiv:1908.01699}
}
read the original abstract

Thoth is a tool designed to combine many different types of speed reading technology. The largest insight is using natural language parsing for more optimal rapid serial visual presentation and more effective reading information.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages

  1. [2]

    R., Masson, M

    Rayner, K., Schotter, E. R., Masson, M. E. J., Potter, M. C., & Treiman, R. (2016). So Much to Read, So Little Time: How Do We Read, and Can Speed Reading Help? Psychological Science in the Public Interest, 17(1), 4–34. ​ https://doi.org/10.1177/1529100615623267

    work page doi:10.1177/1529100615623267 2016

  2. [3]

    Fletcher (2006) Measuring Reading Comprehension, Scientific Studies of Reading, 10:3, 323-330, DOI: 10.1207/s1532799xssr1003_7

    Jack M. Fletcher (2006) Measuring Reading Comprehension, Scientific Studies of Reading, 10:3, 323-330, DOI: 10.1207/s1532799xssr1003_7

    work page doi:10.1207/s1532799xssr1003_7 2006

  3. [4]

    A., & Jetton, T

    Alexander, P. A., & Jetton, T. L. (2000). Learning from text: A multidimensional and developmental perspective. In M. L. Kamil, P. B. Mosenthal, P. D. Pearson, & R. Barr (Eds.), Handbook of reading research (Vol., 3, pp. 285–310). Mahwah, NJ: Erlbaum

    work page 2000

  4. [5]

    I., Fischo, V

    Brown, J. I., Fischo, V. V., & Hanna, G. S. (1993). Nelson-Denny Reading Test. Chicago: Riverside. Cook, T. D., & Campbell, D. A. (1979). Quasi-experimental design: Design and analysis issues. Chicago: Rand McNally

    work page 1993

  5. [6]

    R., Francis, D

    Foorman, B. R., Francis, D. J., Davidson, K. C., Harm, M. W., & Griffin, J. (2004). Variability in text features in six grade 1 basal reading programs. Scientific Studies of Reading, 8, 167–197

    work page 2004

  6. [7]

    W., McGrew, K

    Woodcock, R. W., McGrew, K. S., & Mather, N. (2001). Woodcock–Johnson III Tests of Achievement. Itasca, IL: Riverside

    work page 2001

  7. [9]

    C., & Ortony, A

    Anderson, R. C., & Ortony, A. (1975). On putting apples into bottles – a problem of polysemy. Cognitive Psychology, 7, 167–180

    work page 1975

  8. [10]

    W.; Ventura, P., Nunes F

    Dehaene, S., Pegado, F., Braga, L. W.; Ventura, P., Nunes F. G., Jobert, A., Dehaene-Lambertz, G., Kolinsky, R., Morais, J ; Cohen, L., Dehaene, S, How Learning to Read Changes the Cortical Networks for Vision and Language. Science (New York, N.Y.), December 3, 2010, Vol.330(6009), pp.1359-1364. 10

    work page 2010

Show all 51 references
  1. [12]

    ​ Saccade size in reading depends upon character spaces and not visual angle​

    Morrison & Rayner, (1981). ​ Saccade size in reading depends upon character spaces and not visual angle​ . Retrieved May 23, 2019, from https://link-springer-com.prx.library.gatech.edu/content/pdf/10.3758%2F BF03206156.pdf

    1981

  2. [13]

    ​ What is the amplitude of a wave? Retrieved May 23, 2019, from ​ http://cse.ssl.berkeley.edu/light/measure_amp.html

    Unknown, (2019). ​ What is the amplitude of a wave? Retrieved May 23, 2019, from ​ http://cse.ssl.berkeley.edu/light/measure_amp.html

    2019

  3. [14]

    W., Morrison, R

    Rayner, K., Inhoff, A. W., Morrison, R. E., Slowiaczek, M. L., & Bertera, J. H. (1981). Masking of foveal and parafoveal vision during eye fixations in reading. Journal of Experimental Psychology: Human Perception and Performance, 7(1), 167–179. https://doi-org.prx.library.gat...

    1981 doi

  4. [15]

    ​ Who is the Fastest Reader in the World? Retrieved May 23, 2019, from https://www.toptenreviews.com/software/articles/who-is-the-fastest-read er-in-the-world/

    Unknown, (2019). ​ Who is the Fastest Reader in the World? Retrieved May 23, 2019, from https://www.toptenreviews.com/software/articles/who-is-the-fastest-read er-in-the-world/

    2019

  5. [16]

    D., McClelland, J

    Jackson, M. D., McClelland, J. L. (30 July 1975). ​ Sensory and Cognitive Determinants of Reading Speed​ . Retrieved May 23,

    1975

  6. [17]

    The Influence of Reading Speed and Line Length on the Effectiveness of Reading from Screen

    Dyson, Mary C, and Mark Haselgrove. “The Influence of Reading Speed and Line Length on the Effectiveness of Reading from Screen.” International Journal of Human-Computer Studies​ , vol. 54, no. 4, 2001, pp. 585–612., doi:10.1006/ijhc.2001.0458

    2001

  7. [18]

    American Optometric Association, Retrieved May 23, 2019, from https://www.aoa.org/patients-and-public/resources-for-teachers/how-you r-eyes-work

    2019

  8. [19]

    Shiel Jr., (December 12 2018)

    William C. Shiel Jr., (December 12 2018). ​ Definition of Optic Nerve​ . https://www.medicinenet.com/script/main/art.asp?articlekey=4653

    2018

  9. [20]

    ​ Protecting Sight​

    Unknown, (2019). ​ Protecting Sight​ . Sightsavers. Retrieved May 23, 2019, from ​ https://www.sightsavers.org/protecting-sight/the-eyes/

    2019

  10. [21]

    O'Brien, B., Mansfield, J., & Legge, G. (2005). The effect of print size on reading speed in dyslexia. Journal of Research in Reading, 28(3), 332-349

    2005

  11. [22]

    V E Palchikov, D Y Zolotarev, K V Danilenko & A A Putilov (1997) Effects of the Seasons and of Bright Light Administered at Different Times of Day on Sleep EEG and Mood in Patients with Seasonal Affective Disorder, Biological Rhythm Research, 28:2, 166-184, DOI: 10.1076/brhm.2...

    1997 doi

  12. [23]

    Kong, Seo, & Zhai. (2018). Comparison of reading performance on screen and on paper: A meta-analysis. Computers & Education, 123, 138-149

    2018

  13. [25]

    Tarasov, Sergeev, & Filimonov. (2015). Legibility of Textbooks: A Literature Review. Procedia - Social and Behavioral Sciences, 174, 1300-1308

    2015

  14. [26]

    Mangen, Anne, Walgermo, Bente R., & Bronnick, Kolbjorn. (2013). Reading Linear Texts on Paper versus Computer Screen: Effects on Reading Comprehension. International Journal of Educational Research, 58(C), 61-68

    2013

  15. [27]

    W.; Ventura, P., Nunes F

    Dehaene, S., Pegado, F., Braga, L. W.; Ventura, P., Nunes F. G., Jobert, A., Dehaene-Lambertz, G., Kolinsky, R., Morais, J ; Cohen, L., Dehaene, S, How Learning to Read Changes the Cortical Networks for Vision and Language. Science (New York, N.Y.), December 3, 2010, Vol.330(6...

    2010

  16. [28]

    Tulving, E. (2002). Episodic Memory: From Mind to Brain. Annual Review of Psychology., 53(1), 1-25

    2002

  17. [30]

    ​ Memory Loss and Gains of Older Adults​

    Unknown, (2017). ​ Memory Loss and Gains of Older Adults​ . Sentab. Retrieved May 29, 2019, from https://s3-eu-west-1.amazonaws.com/www-documents/Memory+in+older +age_v2.pdf

    2017

  18. [31]

    Nai Ding, Lucia Melloni, Hang Zhang, Xing Tian, & David Poeppel. (2015). Cortical tracking of hierarchical linguistic structures in connected speech. Nature Neuroscience, 19(1), 158-15864

    2015

  19. [32]

    Dehaene, S

    Nelson, M., El Karoui, I., Giber, K., Yang, X., Cohen, L., Koopman, H., . . . Dehaene, S. (2017). Neurophysiological dynamics of phrase-structure building during sentence processing. Proceedings of the National Academy of Sciences of the United States of America, 114(18), E3669-E3678

    2017

  20. [33]

    and Loper, E., (2019) ​ Natural Language Processing with Python​ , Retrieved May 30, 2019, from https://www.nltk.org/book/ch08.html 12

    Bird, S., Klein, E. and Loper, E., (2019) ​ Natural Language Processing with Python​ , Retrieved May 30, 2019, from https://www.nltk.org/book/ch08.html 12

    2019

  21. [34]

    Humphrey, N., & Weiskrantz, L. (1969). Size Constancy in Monkeys with Inferotemporal Lesions. Quarterly Journal of Experimental Psychology, 21(3), 225-238

    1969

  22. [35]

    Deheane, S. (2009). ​ Reading in the Brain: The Science and Evolution of a Human Invention​ . New York: Viking Adult, Penguin Publishing

    2009

  23. [36]

    Cognitive Psychology: Mind and Brain

    Smith; Kosslyn (2007). Cognitive Psychology: Mind and Brain. New Jersey: Prentice Hall. pp. 21, 194–199,

    2007

  24. [37]

    M., Greene, J

    Frankland, S. M., Greene, J. D., (2015), ​ Structured meaning in superior temporal cortex, ​ Proceedings of the National Academy of Sciences Sep 2015, 112 (37) 11732-11737; DOI: 10.1073/pnas.1421236112

    2015 doi

  25. [38]

    Foster, J. (2002). THE OXFORD HANDBOOK OF MEMORY. Brain, 125(2), 439-441

    2002

  26. [40]

    Oakley, B. (2015). Learning How We Learn: In an unlikely confluence, bioengineers are finding fertile ground in research and teaching about learning. IEEE Pulse, 6(2), 37-41

    2015

  27. [41]

    Miller, G., & Kintsch, Walter. (1994). The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information. Psychological Review, 101(2), 343-352

    1994

  28. [42]

    Schmandt-Besserat, D. (1986). The Origins of Writing: An Archaeologist's Perspective. Written Communication, 3(1), 31-45

    1986

  29. [43]

    The brain-targeted teaching model for 21st-century schools. (2012). Reference and Research Book News, 27(2), Reference and Research Book News, Vol.27(2)

    2012

  30. [44]

    Dehaene, & Cohen. (2011). The unique role of the visual word form area in reading. Trends in Cognitive Sciences, 15(6), 254-262

    2011

  31. [45]

    Glezer, L., Kim, J., Rule, J., Jiang, X., & Riesenhuber, M. (2015). Adding words to the brain's visual dictionary: Novel word learning selectively sharpens orthographic representations in the VWFA. The Journal of Neuroscience : The Official Journal of the Society for Neuroscie...

    2015

  32. [46]

    Johnson, R., & Eisler, M. (2012). The importance of the first and last letter in words during sentence reading. Acta Psychologica., 141(3), 336-351

    2012

  33. [47]

    McElree, B. (2000). Sentence comprehension is mediated by content-addressable memory structures. Journal of Psycholinguistic Research., 29(2), 111-123

    2000

  34. [48]

    Kincaid, J.P., Fishburne, R.P., Rogers, R.L., & Chissom, B.S. (1975). Derivation of new readability formulas (automated readability index, fog count, and flesch reading ease formula) for Navy enlisted personnel. Research Branch Report 8–75. Chief of Naval Technical Training: N...

    1975

  35. [49]

    Kincaid, J., & Delionbach, L. (1973). Validation of the Automated Readability Index: A Follow-Up. Human Factors: The Journal of Human Factors and Ergonomics Society, 15(1), 17-20

    1973

  36. [50]

    Dale, E., & Chall, J. (1948). A Formula for Predicting Readability. Educational Research Bulletin, 27(1), 11-28

    1948

  37. [51]

    Flesch, R. (1948). A new readability yardstick. Journal of Applied Psychology, 32(3), 221-233

    1948

  38. [52]

    Hedman, A. (2008). Using the SMOG Formula to Revise a Health-Related Document. American Journal of Health Education, 39(1), 61-64

    2008

  39. [53]

    Spache, G. (1953). A New Readability Formula for Primary-Grade Reading Materials. The Elementary School Journal, 53(7), 410-413

    1953

  40. [54]

    Coleman, M., & Liau, T. (1975). A computer readability formula designed for machine scoring. Journal of Applied Psychology, 60(2), 283-284. 10 APPENDIX Latest Version : The latest version of the source code as well as each of the branches for the milestones are freely availabl...

    1975

  41. [1975]

    a definition of effective reading therefore needs to differentiate between faster reading and more accurate recall of what is read

    Line Lengths and Screens impacts on reading—​ Lines ​ with a length of 55 cpl (characters per line) were found to support effective speed and comprehension. (Dyson, et. al, 2001). They also found that fast readers spend less time between moving from one line to the next. They ...

    2001

  42. [2011]

    processing time

    The eyes are pretty good at the process of reading, but what they’re doing isn’t exactly what we think of when we normally discuss reading. Our brains actually sees (and chunks) words we already know as a single picture instead of reading at the granularity of ​ individual ​ l...

    2015

  43. [2018]

    (Mangen, 2013)

    Another study on Norwegian school districts found that we should expect a significant impact on reading performance. (Mangen, 2013). Scrolling specifically seems to really impede reading performance. 3 Paper and Digital offer similar speeds.—​ Reading on paper is not significa...

    2013

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.