Scott-Martin: "What do you believe are the specific elements and conditions of type design and typography that would become the subjects for scientific investigation? And, if you could make a list of them, how would you consider their interactions?"
Both Tinker and Luckiesh studied size of type, measure, leading, and typeface. Luckiesh also studies boldness, and other variables as well. You study these as anything else, by varying one and keeping the others fixed. When you have gone through all the permutations, you can get some meaningful results, if you are actually on to something. I think the blink rate tests can be done again much better today, because of much better equipment. The new equipment can measure also how long the eyes are closed, the amplitude of blinks, and do eye tracking on the text at the same time.
The ability to learn from experiment is hugely influenced by how good the theory you are testing. In effect, theories are searchlights, as Popper put it, and you experiments are only as good as the searchlight you are using. In the case of Semmelweiss, whose tragic story I mentioned above, he got it right: he rightly identified a key causal factor.
So the ability to get interesting results in readability research depends partly on how good the theory is going in—as this will strongly influence how experiments are set up.
As to a list, perhaps the best thing is for me to quote the bit of our article on W.A. Dwiggins's reaction after reading Luckiesh's report on his research: **** Dwiggins’s assessment of the Luckiesh and Moss work was both admiring and critical. He felt they had made important discoveries, but that their conclusions over-reached their test results. Dwiggins began his assessment by agreeing with Luckiesh that ‘in the case of any given size of letter, there is certainly an optimum weight for that letter and size, and it’s good to have a way for finding it.’ But Dwiggins also observed that there were optimums for other variables, too, and that the problem therefore was ‘how to hitch it [boldness] up with other “optimums” and make a team – that is what is needed to make the boldness findings valuable in the case of new designs.’ Dwiggins then listed the other four other variables, recommending that their optimums should also be determined: 1. the ratio of stem weight (breadth of the vertical stroke) to white space (i.e. the area of white space within a letter’s counter plus the area white space of its side bearings); 2. the triple ratio of areas: stem (boldness), to counter, to side bearings; 3. the side bearings; and 4. the thick-thin contrast, or ‘modeling’ (‘One would like to know: whether monotone and no modeling; or whether modeling, and if so how much or how little.’)
For Dwiggins, none of these variables, boldness included, operated independently; instead, ‘all these factors interplay; and the investigator has to keep all the balls in the air at once, as I see it, because each variable influences all the others. . . . Optimum weight alone is not enough to go ahead with’. He thought that the research should continue since the basis for an ideal text type was far from clear.
[Dwiggins:] I don’t want any of this to make it seem that I am blowing cold on the laboratory end of the game. I’m for getting all the facts via eye-blinks that a feller can get together. . . . It’s simply that I feel a little shaky about the L. & M. findings because I find the investigators so eager and willing to find a Super-Textype on ‘boldness’ alone: ‘Now we’ve got a sure basis to work on!’ They haven’t. They’ve built one corner of the foundation very nicely. ***
The variables that Dwiggins lists, and their interaction can be tested the same way this is always done: hold one fixed and vary the others, or hold two fixed and vary the others. It is work, but if you find something significant it would be very informative.
A lot more could be tested now. For example I think density of pixels per square inch on the screen, and its impact on readability would be very interesting to study with blink rate, as well as by checking for number of proof reading errors and speed of reading.
At at any rate, I am comfortable with being on Dwiggins's side—not Shinn's—here, being both critical and appreciative of past work, and in favor of further research work on readability, along the lines that Luckiesh started.
“Hold one variable fixed and vary the others.” That is what I did in the design of the Neology experiment. Action, Bill, along the lines spoken of by Dwiggins, not merely being in favor of it.
Unless scientists start doing this kind of thing in their reading research, I will continue to consider the worth of lab tests for type design to be little better than product testing.
That’s my challenge to the neuroculture crew.
The type design problem is to satisfy the requirements of a style that will function in an experiment, involving the end reader, and yet also work as an attractive design tool for typographers. Neology is, I suspect, so far too strange, too “wrong font”-y, too postmodern to do that. But at least it has alerted me to this issue, and got me thinking about developing this line of design.
So thanks for your discussion on this thread, everyone.
You said earlier that you hold that readability can not be measured, that that's "your philosophy," Now you're saying that you agree with Dwiggins, who wanted further testing along the lines of Luckiesh—who measured reader experience in more ways than anyone before or since. Ok then.
There’s a big difference between agreeing with someone that work needs to be done, and actually doing it.
I said that I had taken action along the lines suggested by Dwiggins—namely to hold one variable (in the case of Neology, text color) and vary others (geo vs. grot). The two readability outcomes were a straight comparison of texts set in geo vs. grot, and a wrong-font, pseudo-random mixture of them both.
I was attracted to pseudo-random because I’m interested in how this kind of a puzzle might be solved. The task of making suitable design specifications and adjustments to the two components is implicit, the means is drawing and designing to make everything fit nicely and smoothly, so that it’s easy to read and has some style.
What is starting to intrigue me more and more about what you did Nick, is that the traditional method in the the study of typographical factors is to try and control the cognitive load and then vary the typographic factors one expects might affect a measure (blink rate or reading speed). Cognitive load is an external factor which might introduce a confound.
Your method was to control what you think might be a decisive typographic factor (even colour) and then vary other typographical factors one expects might affect the measure. Even colour is an internal factor.
This might be a better way to evaluate how things “hitch up.”
Comments
Both Tinker and Luckiesh studied size of type, measure, leading, and typeface. Luckiesh also studies boldness, and other variables as well. You study these as anything else, by varying one and keeping the others fixed. When you have gone through all the permutations, you can get some meaningful results, if you are actually on to something. I think the blink rate tests can be done again much better today, because of much better equipment. The new equipment can measure also how long the eyes are closed, the amplitude of blinks, and do eye tracking on the text at the same time.
The ability to learn from experiment is hugely influenced by how good the theory you are testing. In effect, theories are searchlights, as Popper put it, and you experiments are only as good as the searchlight you are using. In the case of Semmelweiss, whose tragic story I mentioned above, he got it right: he rightly identified a key causal factor.
So the ability to get interesting results in readability research depends partly on how good the theory is going in—as this will strongly influence how experiments are set up.
As to a list, perhaps the best thing is for me to quote the bit of our article on W.A. Dwiggins's reaction after reading Luckiesh's report on his research:
****
Dwiggins’s assessment of the Luckiesh and Moss work was both admiring and critical. He felt they had made important discoveries, but that their conclusions over-reached their test results. Dwiggins began his assessment by agreeing with Luckiesh that ‘in the case of any given size of letter, there is certainly an optimum weight for that letter and size, and it’s good to have a way for finding it.’ But Dwiggins also observed that there were optimums for other variables, too, and that the problem therefore was ‘how to hitch it [boldness] up with other “optimums” and make a team – that is what is needed to make the boldness findings valuable in the case of new designs.’ Dwiggins then listed the other four other variables, recommending that their optimums should also be determined:
1. the ratio of stem weight (breadth of the vertical stroke) to white space (i.e. the area of white space within a letter’s counter plus the area white space of its side bearings);
2. the triple ratio of areas: stem (boldness), to counter, to side bearings;
3. the side bearings; and
4. the thick-thin contrast, or ‘modeling’ (‘One would like to know: whether monotone and no modeling; or whether modeling, and if so how much or how little.’)
For Dwiggins, none of these variables, boldness included, operated independently; instead, ‘all these factors interplay; and the investigator has to keep all the balls in the air at once, as I see it, because each variable influences all the others. . . . Optimum weight alone is not enough to go ahead with’. He thought that the research should continue since the basis for an ideal text type was far from clear.
[Dwiggins:]
I don’t want any of this to make it seem that I am blowing cold on the laboratory end of the game. I’m for getting all the facts via eye-blinks that a feller can get together. . . .
It’s simply that I feel a little shaky about the L. & M. findings because I find the investigators so eager and willing to find a Super-Textype on ‘boldness’ alone: ‘Now we’ve got a sure basis to work on!’ They haven’t. They’ve built one corner of the foundation very nicely.
***
The variables that Dwiggins lists, and their interaction can be tested the same way this is always done: hold one fixed and vary the others, or hold two fixed and vary the others. It is work, but if you find something significant it would be very informative.
A lot more could be tested now. For example I think density of pixels per square inch on the screen, and its impact on readability would be very interesting to study with blink rate, as well as by checking for number of proof reading errors and speed of reading.
At at any rate, I am comfortable with being on Dwiggins's side—not Shinn's—here, being both critical and appreciative of past work, and in favor of further research work on readability, along the lines that Luckiesh started.
Unless scientists start doing this kind of thing in their reading research, I will continue to consider the worth of lab tests for type design to be little better than product testing.
That’s my challenge to the neuroculture crew.
The type design problem is to satisfy the requirements of a style that will function in an experiment, involving the end reader, and yet also work as an attractive design tool for typographers. Neology is, I suspect, so far too strange, too “wrong font”-y, too postmodern to do that. But at least it has alerted me to this issue, and got me thinking about developing this line of design.
So thanks for your discussion on this thread, everyone.
I said that I had taken action along the lines suggested by Dwiggins—namely to hold one variable (in the case of Neology, text color) and vary others (geo vs. grot). The two readability outcomes were a straight comparison of texts set in geo vs. grot, and a wrong-font, pseudo-random mixture of them both.
I was attracted to pseudo-random because I’m interested in how this kind of a puzzle might be solved. The task of making suitable design specifications and adjustments to the two components is implicit, the means is drawing and designing to make everything fit nicely and smoothly, so that it’s easy to read and has some style.
Your method was to control what you think might be a decisive typographic factor (even colour) and then vary other typographical factors one expects might affect the measure. Even colour is an internal factor.
This might be a better way to evaluate how things “hitch up.”