The Influence of Typography on Algorithms that Predict the Speed and Comfort of Reading
Round 1
Reviewer 1 Report
This manuscript presents new, original and sound data on how autocorrelation algorithms can predict the speed and comfort of text reading. And the ms is well written : I have therefore a limited number of comments.
However my main concern is about presentation of the study: there are a few weak points about the “shortness” of the text in some parts, the lack of clarity of some figures, and the lack of many references to highly relevant parts of the literature that we might call the “psychophysics of reading” (following the huge influence in this respect of Gordon Legge’s work). The point is that the present work gives the impression that this clear psychophysical framework is not acknowledged.
Moreover, it is a pity that the authors do not link their interesting work to its potential role in Low Vision reading. This could perhaps be the unexpected influence of the ms as this work is very relevant to many questions bearing on the psychophysics of low vision. The main difficulty of low vision persons is reading and many psychophysical studies have investigated how low level visual aspects of text can affect reading performance. And autocorrelation might be an important factor to investigate in low vision reading.
The points that will improve the ms are presented below:
a/ In the Introduction : p. 1: l. 28: about the role of “x-height” and typography, it would be interesting to cite this very influential paper: (Legge, G.E., Bigelow, C.A., 2011. Does print size matter for reading? A review of findings from vision science and typography. J Vis 11, 1–22. https://doi.org/10.1167/11.5.8). Similarly, the book of Legge (both on normal and low vision) might be a useful starting reference : Legge, G.B., 2007. Psychophysics of Reading in Normal and Low Vision. Lawrence Erlbaum Associates, Mahwah , New Jersey, London.
More generally, to introduce the low vision framework, the Introduction should mention that these typographic, and therefore low level visual, aspects are key factors when trying to understand low vision reading.
Notably, the effect of line spacing (mentioned by the authors in line 28) on reading speed has been carefully investigated in a large group of AMD patients and non intuitive results emerged ( Calabrèse, et al., 2010. Small effect of interline spacing on maximal reading speed in low-vision patients with central field loss irrespective of scotoma size. Investigative ophthalmology & visual science 51, 1247–54. https://doi.org/10.1167/iovs.09-3682.). In sum, the psychophysics of low vision reading has often produced more investigations of low level visual factors than the psychophysics of normal reading. Actually, many studies of low vision reading have produced data both for normal and low vision reading.
b/ p. 4: Figure 2
This figure really needs to be improved: the text for the x and y axes is wrong, the data points are AC peak values !
A legend with the different font names (or at least a few of them) would allow a much more comfortable and efficient understanding of the figure . The reader is not supposed to read back and forth between table 1 and figure 2 to look for the “best” font ! not in our times when free softwares (e.g. R) allow us to create very clear and elegant plots.
The legend of the figure is also unclear and too short.
c/ p. 9 l. 282: it would be instructive to have a figure showing a typical fit of the 1/f “cone” function. This is an important part of the paper and this would give a “feeling” of the data leading to correlation values.
Line 284: is it “luminance” contrast? If yes, please make it explicit.
Line 285: is “weight” a standard name in this context? If yes, a reference should be given.
d/ p. 12. The General Discussion is a bit short and frustrating given the potential of this study, especially for low vision patients. I notably predict that the autocorrelation measure used on word might be fruitfully injected in the few models of word reading that take into account both crowding and spatial uncertainty of letter position (cf. confusion matrices). This is clearly shown in a recent model of word reading : Bernard, J.-B., Castet, E., 2019. The optimal use of non-optimal letter information in foveal and parafoveal word recognition. Vision Research 155, 44–61. https://doi.org/10.1016/j.visres.2018.12.006. (as you probably know crowding increases with eccentricity which strengthens its role in low vision persons who have to read with their peripheral retina). There is simply a change of scale from text to letters, through words, when applying autocorrelation. I would like to have the authors’ opinion on these theoretical links with autocorrelation measures and see this point discussed in the ms.
e/ line 354 – 360: What the authors describe here, i.e. changing for instance the ascenders of some visually similar letters, is exactly what was systematically done in a study that created a new font; Eido, to improve low vision reading (Bernard, et al. 2016. A New Font, Specifically Designed for Peripheral Vision, Improves Peripheral Letter and Word Recognition, but Not Eye-Mediated Reading Performance. PLOS ONE 11, e0152506. https://doi.org/10.1371/journal.pone.0152506). This reference is clearly relevant here, as well as in the Conclusion: “… font design has the potential to affect reading fluency and comfort appreciably”. This sentence is especially true for low vision reading in my opinion. The reference to the Eido font paper will also help readers understand the text of the authors (lines 354-360) and thus the link between crowding and autocorrelation.
f/ The experiment performed with iBooks should also be emphasized in the context of low vision. This is because low vision patients read more and more with electronic devices such as tablets, smartphones and ebooks (Crossland, M.D., S. Silva, R., Macedo, A.F., 2014. Smartphone, tablet computer and e-reader use by people with vision impairment. Ophthalmic Physiol Opt 34, 552–557. https://doi.org/10.1111/opo.12136).
Minor point:
- 7: line 209 ; It IS (missing) therefore …
Author Response
Thank you for your very helpful comments.
However my main concern is about presentation of the study: there are a few weak points about the “shortness” of the text in some parts, the lack of clarity of some figures, and the lack of many references to highly relevant parts of the literature that we might call the “psychophysics of reading” (following the huge influence in this respect of Gordon Legge’s work). The point is that the present work gives the impression that this clear psychophysical framework is not acknowledged.
Gordon Legge's work, and that of Denis Pelli is now acknowledged
Moreover, it is a pity that the authors do not link their interesting work to its potential role in Low Vision reading. This could perhaps be the unexpected influence of the ms as this work is very relevant to many questions bearing on the psychophysics of low vision. The main difficulty of low vision persons is reading and many psychophysical studies have investigated how low level visual aspects of text can affect reading performance. And autocorrelation might be an important factor to investigate in low vision reading.
I suspect that some filtering might be valuable before autocorrelation in low vision reading, and this is now mentioned.
The points that will improve the ms are presented below:
a/ In the Introduction : p. 1: l. 28: about the role of “x-height” and typography, it would be interesting to cite this very influential paper: (Legge, G.E., Bigelow, C.A., 2011. Does print size matter for reading? A review of findings from vision science and typography. J Vis 11, 1–22. https://doi.org/10.1167/11.5.8). Similarly, the book of Legge (both on normal and low vision) might be a useful starting reference : Legge, G.B., 2007. Psychophysics of Reading in Normal and Low Vision. Lawrence Erlbaum Associates, Mahwah , New Jersey, London.
Both these are now referenced.
More generally, to introduce the low vision framework, the Introduction should mention that these typographic, and therefore low level visual, aspects are key factors when trying to understand low vision reading.
This point is made, but in the Discussion.
Notably, the effect of line spacing (mentioned by the authors in line 28) on reading speed has been carefully investigated in a large group of AMD patients and non intuitive results emerged ( Calabrèse, et al., 2010. Small effect of interline spacing on maximal reading speed in low-vision patients with central field loss irrespective of scotoma size. Investigative ophthalmology & visual science 51, 1247–54. https://doi.org/10.1167/iovs.09-3682.). In sum, the psychophysics of low vision reading has often produced more investigations of low level visual factors than the psychophysics of normal reading. Actually, many studies of low vision reading have produced data both for normal and low vision reading.
We accept this point, but the methods we describe are probably not directly applicable in low vision reading without alteration.
b/ p. 4: Figure 2
This figure really needs to be improved: the text for the x and y axes is wrong, the data points are AC peak values !
The figure has been improved.
A legend with the different font names (or at least a few of them) would allow a much more comfortable and efficient understanding of the figure . The reader is not supposed to read back and forth between table 1 and figure 2 to look for the “best” font ! not in our times when free softwares (e.g. R) allow us to create very clear and elegant plots.
The legend of the figure is also unclear and too short.
These shortcomings have been rectified.
c/ p. 9 l. 282: it would be instructive to have a figure showing a typical fit of the 1/f “cone” function. This is an important part of the paper and this would give a “feeling” of the data leading to correlation values.
Such a figure is now included.
Line 284: is it “luminance” contrast? If yes, please make it explicit.
This has been done
Line 285: is “weight” a standard name in this context? If yes, a reference should be given.
The meaning in typography and in the current context have been described in a phrase.
d/ p. 12. The General Discussion is a bit short and frustrating given the potential of this study, especially for low vision patients. I notably predict that the autocorrelation measure used on word might be fruitfully injected in the few models of word reading that take into account both crowding and spatial uncertainty of letter position (cf. confusion matrices). This is clearly shown in a recent model of word reading : Bernard, J.-B., Castet, E., 2019. The optimal use of non-optimal letter information in foveal and parafoveal word recognition. Vision Research 155, 44–61. https://doi.org/10.1016/j.visres.2018.12.006. (as you probably know crowding increases with eccentricity which strengthens its role in low vision persons who have to read with their peripheral retina). There is simply a change of scale from text to letters, through words, when applying autocorrelation. I would like to have the authors’ opinion on these theoretical links with autocorrelation measures and see this point discussed in the ms.
Thank you. We agree that the Discussion was deficient. It has been improved and extended.
e/ line 354 – 360: What the authors describe here, i.e. changing for instance the ascenders of some visually similar letters, is exactly what was systematically done in a study that created a new font; Eido, to improve low vision reading (Bernard, et al. 2016. A New Font, Specifically Designed for Peripheral Vision, Improves Peripheral Letter and Word Recognition, but Not Eye-Mediated Reading Performance. PLOS ONE 11, e0152506. https://doi.org/10.1371/journal.pone.0152506). This reference is clearly relevant here, as well as in the Conclusion: “… font design has the potential to affect reading fluency and comfort appreciably”. This sentence is especially true for low vision reading in my opinion. The reference to the Eido font paper will also help readers understand the text of the authors (lines 354-360) and thus the link between crowding and autocorrelation.
The Eldo font paper is now referenced.
f/ The experiment performed with iBooks should also be emphasized in the context of low vision. This is because low vision patients read more and more with electronic devices such as tablets, smartphones and ebooks (Crossland, M.D., S. Silva, R., Macedo, A.F., 2014. Smartphone, tablet computer and e-reader use by people with vision impairment. Ophthalmic Physiol Opt 34, 552–557. https://doi.org/10.1111/opo.12136).
This has been referenced.
Minor point:
- 7: line 209 ; It IS (missing) therefore …corrected. Thank you!
Reviewer 2 Report
Review of ms “The influence of typography on algorithms that predict the speed and comfort of reading”
This manuscript examined how the characteristics of fonts may affect reading. The manuscript offers some good ideas and I found it interesting. At the same time, I don’t think the manuscript reached its potential. The literature background on previous studies on typographical effects in reading was incomplete (I’ve indicated a few references below, but there are many more), and for an experimental psychologist, the approach followed here was a bit mixing apples and oranges. For instance, if one wants to compare whether serifs influence reading, the best option is to have two fonts that are exactly the same, except for the presence/absence of serifs (e.g., as done in doi:10.7334/psicothema2012.141) rather than comparing many fonts (with/without serifs) varying in all sort of features. Incidentally, there is no effect of serifs during sentence reading (as shown in the above-cited paper). Also, how informative are the effects offered here in a normal reading scenario?
In sum, I think the ms has potential, and I’d happy to review a revised version, but the manuscript needs some work—and also it needs to indicate its limitations (in case no further studies are included).
McGowan, V. A., White, S. J., & Paterson, K. B. (2015). The effects of interword spacing on the eye movements of young and older readers. Journal of Cognitive Psychology, 27(5), 609-621.
Perea, M. (2013). Why does the APA recommend the use of serif fonts? Psicothema, 25, 13-17. doi:10.7334/psicothema2012.141
Perea, M., Giner, L., Marcet, A., & Gomez, P. (2016). Does extra interletter spacing help text reading in skilled adult readers? Spanish Journal of Psychology, 19, e26, 1–7. DOI: 10.1017/sjp.2016.28
Schneps M. H., Thomson J. M., Sonnert G., Pomplun M., Chen C., & Heffner-Wong A. (2013). Shorter lines facilitate reading in those who struggle. PLoS ONE, 5, e71161 http://dx.doi.org/10.1371/journal.pone.0071161
Slattery, T. J., Yates, M., & Angele, B. (2016). Interword and interletter spacing effects during reading revisited: Interactions with word and font characteristics. Journal of Experimental Psychology: Applied, 22, 406–422. doi:10.1037/xap0000104
Author Response
This manuscript examined how the characteristics of fonts may affect reading. The manuscript offers some good ideas and I found it interesting. At the same time, I don’t think the manuscript reached its potential. The literature background on previous studies on typographical effects in reading was incomplete (I’ve indicated a few references below, but there are many more), and for an experimental psychologist, the approach followed here was a bit mixing apples and oranges.
Thank you. We agree. The deficiencies in the background have been rectified.
For instance, if one wants to compare whether serifs influence reading, the best option is to have two fonts that are exactly the same, except for the presence/absence of serifs (e.g., as done in doi:10.7334/psicothema2012.141) rather than comparing many fonts (with/without serifs) varying in all sort of features. Incidentally, there is no effect of serifs during sentence reading (as shown in the above-cited paper). Also, how informative are the effects offered here in a normal reading scenario?
We have now followed the referenced article and compared Lucida bright and Lucida sans and shown that the presence of serifs has little effect on the autocorrelation. We have referenced the paper.
In sum, I think the ms has potential, and I’d happy to review a revised version, but the manuscript needs some work—and also it needs to indicate its limitations (in case no further studies are included).
Further studies have been included, and better reference made to the literature. We hope the manuscript is now acceptable.
Round 2
Reviewer 2 Report
The authors did a fine job. I only have a very minor concern.
On lines 415-16
And there are no algorithms available with electronic
books to guide users’ choice of font and spacing.
i would say “Currently, there are no....”
Author Response
Thank you for this suggestion. I will change the ms.
