nf1263: okay right [0.3] so [0.6] moving on from the last few weeks where i've 
been [0.2] talking about other empirical work [0.3] that [0.3] relates to 
either navigation or things like that [0.6] so in previous weeks i've talked 
about either usability studies which sort of check out individual products [0.
7] or [0.5] er evaluations doing analyses that sort of thing [0.2] and research 
into different types of navigation [0.5] and last week presentation this one's 
going to focus very much on [0.2] reading from screen [0.7] okay [0.4] and the 
model for the [0.4] the second half of it that's the time [1.3] show you what i 
mean by the second half [0.4] the model for this part is definitely legibility 
studies so [0.3] if you can recall the sort of things that you might have read 
about in the first year [0.6] in terms of er legibility of print then this is 
the the equivalent in a way [0.3] although it takes a slightly different form 
[0.8] and i'm going to start off by [0.2] er sort of putting the bigger picture 
and also describing [0.5] what was [0.3] researched in the nineteen-eighties 
and that tended to be [0.4] making a comparison 
between reading from screen and reading from paper [0.8] er the second part is 
just reading from screen [0.6] and apologies because these case studies in 
reading from screen are my own work the reason being 'cause not very much has 
been done [0.3] er [0.9] it's a shame about that [1.2] okay [0.3] so [0.4] 
looking at reading from screen [0.4] compared with reading from paper [1.7] 
right [0.2] first it's useful to really remind ourselves about what the basic 
differences are [0.4] and that they are fairly obvious [0.3] er but they're 
listed here and i'll go through them and [1.0] they're here also because some 
research in the nineteen-eighties [0.2] was essentially trying to find out [0.
9] what might be [0.7] what might account for the differences that there are 
between reading from screen and reading from paper [0.6] so to pre-empt the the 
results essentially you read slower from screen [0.2] and so [0.4] a group of 
researchers based at I-B-M [0.5] wanted to find out [0.8] what could it be that 
is different about [0.3] a screen compared with [0.6] a piece of paper reading 
from print [0.4] that might be 
responsible for these differences so why is it harder to read from screen [1.1] 
so what we know are are different is the orientation because screens are most 
of the time upright in front of us whereas paper [0.6] can be held like that 
but a lot of the time it will be [0.4] down [0.8] on the table [0.2] so there's 
the orientation difference [1.4] er [0.2] the visual angle is different that's 
measuring from [0.3] the text or whatever on the screen [0.7] or on paper and 
how far we sit away from that and measuring that angle [1.2] we tend to sit 
further away [0.2] from screen [0.5] than we do [0.2] from [0.5] from paper [0.
6] er and therefore the visual angle's different so that might be a possibility 
[1.2] the aspect ratio is different in other words most screens tend to be [0.
2] it's true of this one [0.4] landscape rather than protrait [0.6] er [0.3] 
that's [0.4] that's generalizing and talking about the average sort of screen 
so you'll have a wide screen [0.3] but it won't be particularly deep so there's 
a difference [1.1] and we've got difference that re-, that relate to how the [0.
6] er how the 
text how the pixels are put onto the screen that's the dynamics so that the 
screen is refreshed rather than having something sort of solid in front of you 
[0.5] on paper [0.5] and you get flicker with screen [1.4] er the image 
polarity now that's [0.7] less the case now a difference it's less likely to be 
different now when we read from screen than print [0.5] but it used to be the 
case that there were all sorts of [0.6] colours i suppose used on screen [0.4] 
so now we're [0.2] m-, more likely to see black text on a white background on 
screen just like we have in print [0.5] but [0.3] people tried white text on a 
black background and green text on a [0.2] black background [0.7] er as i say 
now it's more likely that we're we're reading the same because they did find 
that black on white is better [1.6] and then we've got all sorts of typographic 
characteristics that that might [0.4] that once in the nineteen-eighties were 
different on screen compared with print [1.2] so actually what these 
researchers did was to [0.2] look at each particular [0.2] variable 
or each particular [0.2] difference and to try and isolate whether that one was 
the one [0.4] that was responsible [0.3] for it being more difficult to read 
from screen [0.7] and they found out that none of them seem to account for it 
[0.6] er but when you put them all together [0.2] then there is this [0.3] this 
problem [0.8] so they couldn't actually say [0.4] what it was in particular 
that accounted for that [0.5] and it's er [0.4] just like yeah namex you said 
that last week [0.2] didn't you when you put lots of things together [0.4] it 
improves the design but thing [0.6] individually those things might not make a 
difference it's sort of the same thing in reverse [0.7] you remember [0.6] you 
do [laughter] [0.7] right [0.4] along the bottom here [0.5] there is er [2.0] 
this this is going back to Jakob Nielsen who i've [0.2] mentioned quite a few 
times he has this sort of [0.2] column i suppose called Alert Box [0.6] it's [0.
3] it's quite an interesting thing to read so i would just recommend er casting 
your eye over that so you'd be able to get to that from the web 
pages it's just quite a good reference [0.5] it feels sort of a bit naive in 
relation to print the first part of it but as you [0.5] get lower down it's 
really pointing out some of the differences which is a [0.4] think there's some 
useful things just to to remind you so that might be worth [0.5] following up 
[2.5] okay so the conclusions of this were and and should repeat it was in the 
n-, late nineteen-eighties so we were talking about [0.4] different sorts of 
screens to the sorts of screens we have now [0.6] but you can't find a single 
factor responsible and if [0.4] if the quality is improved there there ought to 
be [0.3] smaller differences between reading from screen and from print [2.7] 
right so if we move on to some of the research that's been done that has [0.7] 
compared [0.9] how we read from screen [0.6] er or the outcomes when we read 
from screen compared with paper [3.0] people have divided up [0.2] the way in 
which you measure this into two categories [0.4] and these are outcome measures 
and process measures so i'll start with the outcome 
measures [0.7] i suspect this is all terribly familiar to you namex isn't it 
'cause you've done an essay on this yeah [1.2] there's a review article [0.3] 
by [0.2] Andrew Dillon [0.2] that i just noticed just now is not on your 
reading list but it's a good one that does actually summarize all of this [0.4] 
what i've put on your reading list is the book that [0.5] er is much more 
comprehensive [0.3] so [1.2] right so [0.8] this er these are sort of summaries 
of or [0.3] bringing together er the various studies that have been done [0.6] 
and it seemed to be the case that [0.4] all the research that was done [0.5] 
initially [0.4] er in relation to reading from screen [0.4] were making 
comparisons with print it seemed that that was the only interest you know why 
is it different let's always make this comparison [0.9] so most of the [0.3] 
studies that were done found that [0.2] people read more slowly from screen 
than they do from paper [0.3] or from print [0.5] and it's about twenty to 
thirty per cent slower so that's that's reasonably substantial [1.8] er [0.2] 
there were [0.6] or there was an odd study er as in the 
case of just one study it wasn't particularly odd [0.5] er that found that 
reading speed wasn't affected so there wasn't a difference between print and 
between [0.2] and screen [1.0] and it was [0.3] in in bringing them together 
and seeing that this particular study was different it's probably the case that 
[0.2] the texts were too short [0.5] because if you don't give [0.2] a long 
enough time period for people to read then you're not actually going to pull 
out these differences [0.4] so i would say it's a pretty reliable finding that 
we read slower [0.3] from screen [1.9] terms of accuracy one of the ways that 
you can see there the accuracy is measured is by giving a proofreading task [0.
4] which is a different type of accuracy to saying [0.2] well can you 
acculatery accurately remember what you've read [0.4] er it is obviously 
looking for errors on screen [0.9] and [0.8] those studies found that it you 
weren't people weren't finding it more difficult to proofread from screen [0.5] 
than from paper [0.7] and i don't know about you but that seems to me a rather 
surprising finding [0.6] because how many times do you [0.3] look at something 
on screen print it out and then find that there are typos [0.6] now [0.3] 
either we [0.4] don't put the same sort of [0.2] attention into [0.6] checking 
on screen [0.4] or [0.3] it is more difficult [0.5] now i i'm not saying that 
that study is er [0.9] that there's anything wrong with that study [0.5] er the 
result still [0.2] or the results still stand but i suspect that the particular 
task wasn't sensitive enough to pick out [0.3] the differences that i think [0.
5] are probably there [1.1] 
sm1264: [0.5] all of them the results were so dependent on the task [0.4] 
nf1263: yeah 
sm1264: some of them require them to read [0.4] the screen and then write an 
essay on it [0.5] 
nf1263: yeah 
sm1264: and the task and that will be dependent on the person's er ability to 
write an essay you see and have different results [0.6] just from that [0.3] 
nf1263: you mean they didn't [0.4] compare that with them reading from print 
and writing an essay they just [0.2] 
sm1264: yeah they did it both ways but often the tasks in the experiments were 
completely different from each other so it really depended on individual's 
ability [0.3] 
nf1263: yeah 
sm1264: ability of the task as well [0.3] 
nf1263: yes right so maybe the [0.2] the [0.4] the tasks are questionable then 
in themselves they're 
sm1264: that's the er thing [0.2] i think in that they found the tasks were 
more [0.3] 
nf1263: yeah 
sm1264: problems [1.8] 
nf1263: yeah can [0.4] 
sf1265: i think also you find that er [0.5] when you [0.2] if you're [0.6] on 
you know you're on screen [0.5] people rely on the [0.3] spellchecker [0.7] 
nf1263: that's true 
sf1265: to [0.2] for everything whereas they actually [0.8] and it's only when 
you've [0.2] got it you know in your hand reading it that you realize you 
haven't covered all the [0.7] you know grammar 'cause they don't [0.2] they 
can't pick it up [0.2] all the time with spellcheckers [0.4] 
nf1263: yeah so that suggests that there's probably [0.2] both things going on 
we're not [0.5] necessarily [0.2] trying so hard i mean maybe 
sf1265: lets you think the spellchecker will pick it up 
nf1263: because we think it pick it up [0.2] but maybe also i mean [0.6] the 
main thing is that the resolution is poorer [0.2] isn't it and so it's it's 
harder [0.8] i would say that has to make it slightly harder to read and 
because of that [0.7] er i suspect there's there's an influence [0.3] so if we 
get to screens of [0.5] six-hundred dots per inch [0.2] then maybe [0.4] we 
would expect not to have the problems [0.3] but we would still have them if [0.
6] say we we're relying on [0.2] er [0.4] on automatic checks that aren't [0.2] 
going to be infallible [0.2] you know [1.3] yeah [0.4] so i i [0.2] it's also 
'cause we mentioned fatigue as well y-, it's also the case i would [0.2] i mean 
i would feel that it's it is more tiring [0.6] reading for a long time on 
screen [0.4] but can you put [0.2] people in an experiment and try and simulate 
the sort of natural tasks i'm not sure that you can [0.9] so [0.7] er [0.2] so 
there's suggestions there [0.3] that if you [0.7] carry on reading for longer 
maybe there are [0.3] er [0.2] fatiguing effects [0.4] and if you have a 
demanding 
task then you're less accurate but er [1.7] who knows so those are are outcome 
measures because they're measuring sort of performance how quickly you do 
something or whether it's got [0.5] there are errors or things like that think 
i've got a few more [0.2] yeah continue with the outcome measures [0.4] another 
outcome measure is whether [0.7] you can recall [0.2] and understand what 
you've read [0.8] and again the studies have shown that there isn't a 
difference so we're equally as good at understanding what we've read on screen 
and on paper according to that [0.7] but there is always in this sort of work 
the possibility that [0.4] we trade-off how quickly we read with [0.3] how well 
we take it in this is the speed-accuracy trade-off [1.0] so that for example we 
might decide to slow down our reading from screen [0.5] to make sure that we [0.
3] take it all in because we're going to be asked questions afterwards [0.5] or 
if we speed up reading we might actually get worse in terms of comprehension [1.
5] so it's important in in sort of assessing whether 
something [0.7] is legible i suppose we're saying i-, is it [0.3] is the screen 
legible but you [0.2] you bring you take into account not only how quickly you 
read but whether people [0.6] also understand what they've read [0.7] and 
that's true of all legibility research [1.4] and then people were asked for 
preferences and this is where in a a very strange sort of experiment it seems 
to be the [0.7] the outcome is that if you use high quality you've got high 
quality screens high quality paper [0.2] then no preference [0.3] but people 
were preferring paper because it was a higher quality [0.5] er output [0.7] so 
[0.7] don't think that's surprising at all that says nothing about the 
differences between [0.4] screen and paper [0.4] just a preference for higher 
quality higher resolution [1.6] 
right so the other comparisons are the the process measures and we've got fewer 
of those [0.3] these are harder to measure but it's as it sounds you're [0.5] 
they're looking at how people are doing something rather than what the end 
result of that doing is [0.5] so as you probably recall eye movement studies 
have been done for reading print [1.1] looking at [0.3] er how big the the 
saccades are the [0.2] the way wh-, when we jump from one part of a word to 
another part [0.5] er how long we fixate for whether we need to go back if 
we've missed something [1.0] er so those sorts of studies have been done for 
reading from screen as well but not found a a big difference [0.5] er i suspect 
i wouldn't have predicted that because i s-, [0.2] i guess we're [0.3] so 
practised at reading print that this sort of automatic process will [0.6] er [0.
2] take effect from reading from screen as well but [0.6] that's my guess looks 
like there's no 
differences there [1.5] er and then some fairly obvious things the manipulation 
is less direct that's really just pointing out what we know anyway that when 
you've got paper [0.4] you turn over a page by [0.4] er sort of picking it up 
or whatever [0.3] whereas we've got the mouse as the intervening variable or 
the keyboard or whatever [0.2] that's in between [0.5] us and the action [0.3] 
taking effect [1.4] er [0.2] but basically if the appropriate tools are there 
then there's not a [0.2] particular problem [0.6] and i guess we're getting 
past the time when [0.6] er [1.0] you come across people [0.5] well when you 
come across a lot of people that aren't able to to use a mouse [0.6] i mean i i-
, [0.2] y-, don't have to go back too far in teaching in this department to 
have [0.4] people not able [0.7] initially to use a mouse but i don't think 
we're going to [0.2] get that sort of thing any more [0.9] er but actually in 
the experiments that i'm going to describe in just a moment we didn't use the 
mouse because we wanted to [0.4] have someone [0.3] come in to do our 
experiment and not have to 
say well you know when you move this up the cursor will go up sort of thing [0.
4] so we [0.7] only let people use the keyboard [0.6] 'cause there is that 
initial learning [0.5] do any of you remember when you first used a mouse [0.4] 
what it felt like [1.9] [laughter] [0.2] 
sf1266: [0.8] 
nf1263: did you have any trouble [0.5] did you [0.3] 'cause some people are er 
are sort of [0.6] like this [0.3] yeah [0.2] when they s-, first start because 
[0.8] because you don't realize you can [0.2] if you get to the edge of the 
table you can pick the mouse up and [0.2] put it back where it was and still go 
down again [0.5] you don't have to have the same amount of movement as as you 
have on the screen [0.6] none of you did that did you i don't know [0.5] i've 
seen quite a lot of people doing that [0.9] you're all very quiet as well [0.9] 
[laughter] [0.9] er [1.2] and then finally what i've talked about quite a lot 
is the is the navigation [0.2] that it's more complicated than [0.3] in [0.2] 
in books we've got i suppose more things that we can do moving forwards 
backwards previous [0.2] pages and things like that [0.4] 
and we might get lost [0.3] so those are just sort of summarizing some of the 
[0.3] the main differences [1.4] okay so those [0.2] that is what [0.4] people 
were interested in making the comparisons between [0.5] reading from screen and 
reading from print [0.3] this is moving into the second [0.7] section which i-, 
[0.4] are [0.3] it's basically two case studies of [1.1] reading from screen [1.
1] er [0.5] so i decided that i didn't want to compare [0.5] reading print and 
reading from stree-, [0.2] screen because i didn't actually see the point of 
that [0.4] because yes you find it's different [0.2] but really it seems to me 
[0.3] what we want to do [0.3] is to find out how we can [0.3] make it better 
or easier for us to read from screen [0.6] rather than worry about how it's 
different from print [0.8] and as i said at the beginning there wasn't [0.8] or 
there still isn't very much done on [2.1] typographic [0.2] variables or things 
that i would call typographic in relation to to reading from screen [0.5] so 
we've got all this stuff on the legibility of print [0.2] admittedly a lot of 
it done by Tinker [0.2] and coworkers [0.8] 
but there's not the volume that's of research for for reading from screen and 
so you have to think about why and i suppose one obvious reason is that [0.7] 
it's more recent [0.6] and so maybe there hasn't been time for people to do the 
research [1.2] or [0.7] maybe [0.2] people think well we know it all because 
we've got the the legibility of print research so [0.3] we'll just use that [1.
0] or [0.6] maybe [0.2] people get [0.5] er frustrated by the fact that screens 
are changing and so they think well what's the point of doing this research 
because in five years time [0.4] it'll be obsolete because there'll be a 
different screen [0.6] and certainly [0.2] the the s-, [0.2] some of the stuff 
that has been done [0.4] again from the nineteen-eighties early nineties [0.6] 
doesn't use [0.2] d-, [0.2] the same sort of screens as we have now so you do 
think [0.3] well is it [0.4] is that useless because [0.2] it's a different 
sort of screen [1.8] or [0.4] i've got one more [0.2] up my sleeve as to why 
people don't do this sort of research [0.5] people don't think [0.3] typography 
is important coming back to our theme of last 
week but is it this is a variation on it [0.7] because [1.1] people that work 
on [0.5] computer stuff or human-computer [0.2] interaction stuff [0.4] tend 
not to be typographers they're not okay [0.3] so they might be computer 
scientists or they might be psychologists or whoever [0.9] and i do have some 
evidence that they they don't [0.2] really think that the sort of factors that 
we're interested in i hope like let's say line length 'cause that's what we're 
going to be talking about [0.6] they sort of think well yes so what [0.6] er so 
this is sort of the same theme as last week [0.6] so you're not going to find 
people that are going to put together it seems to me the typography with [0.2] 
the [0.3] the screen stuff [0.2] so [0.3] i should say well great that's lucky 
because i'm doing that and [0.4] there's not many [0.2] many more people doing 
it [0.5] but when people say the sort of so what it's a bit disheartening [0.9] 
well that's enough of the preamble though [0.5] so [0.2] i wanted to find out 
really [0.6] er answers to questions about the legibility of text on screen [0.
4] and i 
phrased the first question in a fairly general way by saying what amount of 
text [0.5] really is best for maximum legibility [0.3] and that's 'cause i'm 
talking about layout but i'm talking about things like the line length or the 
number of columns [0.5] so there is a different volume of text on screen [0.2] 
if you've got a long line length [0.3] as compared to a short one [0.7] now i 
know you know what line lengths look like but i will be showing you the 
material in a bit [0.4] so that was the first question [0.3] and that was 
combined with [0.2] er what i've called method of movement or the person i 
worked with called [0.4] and all of that is either scrollling [0.4] going down 
line by line through the text or paging [0.4] which is [0.3] going from one 
screenful of text pressing a button and getting a second screenful [0.3] so 
it's equivalent to turning over the page yeah so you don't gradually scroll 
through [0.7] so those two things er were varied and i looked at [0.2] in the 
first [0.7] experiments that i did [3.2] right so [0.6] what was i measuring 
and basically i 
was trying to measure lots of things because [0.3] as i sort of mentioned last 
week it can be very difficult to find [0.6] any differences [0.3] so if you [0.
2] have quite a broad spectrum [0.3] of er measures then you hope that 
somewhere something will come out of it [0.7] and i wouldn't dream of comparing 
typefaces 'cause i don't think anything would come out of it [0.8] so [0.3] got 
people to read from screen [0.2] and timed how long it took them to silently 
read from screen [0.4] and the way that we did that was basically the the 
computer started the clock at the beginning [0.3] and when they got to the end 
it stopped the clock [0.5] so [0.3] i didn't even have to sort of sit there and 
watch them [1.6] i [1.3] checked their comprehension and i wouldn't say i 
tested their comprehension because [0.5] it was [0.4] not a particularly 
sensitive way to measure comprehension [0.6] what i did was er borrowing from 
some other people [0.8] i asked [0.3] or i put a set of questions [0.3] and 
said [0.2] could you answer these questions from what you've read [0.6] so they 
didn't have to answer them [0.5] but they could 
they just had to say yes or no [0.6] so there would be a question like you know 
did [0.5] Fred fall down the hill [0.4] now if that had been included in what 
they had read then they could say yes i could answer that [0.3] if it hadn't 
been they would say no [0.5] now it might sound an extremely obscure way of of 
checking comprehension [1.0] but the reason to do it this way is that is does 
check whether someone's read because if they get fifty per cent right [0.2] 
then [0.2] they're guessing [0.2] so they haven't [0.5] or they haven't taken 
it in properly [0.8] and the other thing is that it's very easy to mark [0.2] 
because if you [0.5] ask the question er [0.4] what did Fred do [0.9] then you 
have to decide whether what someone answers is correct or not because they're 
not going to use exactly the right wording [0.4] so it's essentially just 
easier to do this sort of thing [0.6] 
sf1267: do they know they're being timed [0.3] [0.2] 
nf1263: yeah they did yeah [0.2] yeah that's a good point to make though [0.4] 
i think we sort of said they also knew they had to answer questions so that th-,
you got to [0.5] think are they going to 
speed through or [0.3] and think that's more important or [0.3] or sort of [1.
0] or [0.2] try and make sure they understood the questions [0.2] no could 
answer the questions [1.2] er the other thing we did i've called it record of 
keystrokes was again the computer [0.4] sort of recorded how they moved through 
the document [0.5] so [0.4] we took the mouse away and they had to use the 
cursor keys so you go down with the pointer [0.4] and and back up with the 
pointer [0.4] and every time they pressed a key [0.2] there was a record of 
when they pressed it and the fact that they had pressed it [0.5] so what you 
can map out is whether someone is [1.2] scrolling through fairly slowly so 
perhaps moving a little bit [0.2] reading a little bit moving a little bit [0.
4] or perhaps whether they're reading a whole screenful [0.5] and then 
scrolling [0.2] through a whole screenful [0.6] for the page [0.6] pages that's 
not particularly interesting but it is for the scrolling [0.6] you can't tell 
when someone is reading but you can tell when they're not scrolling and you 
might guess that they're probably 
reading if if they're not scrolling [0.4] 'cause otherwise they'd just be 
looking out the window [0.5] which is a possibility of course we can't help 
that [1.5] and then finally we we got the s-, the s-, [0.2] in a sense the 
preference although that wasn't quite the question but [0.3] after they'd done 
the reading [0.5] er then [0.6] they were shown [0.2] pairs of things they'd 
read and said which do you think is easiest to read [0.3] and they get all of 
the combinations of pairs [0.4] so you might think that they would base that on 
[0.5] their [0.4] feelings about having read it or maybe they'd just base it on 
what they [0.4] s-, th-, see in front of them [0.2] but certainly they had 
experienced the reading at that point [1.1] so these these are sort of methods 
for for this lot of work [0.5] so [0.9] first one looking at line length [0.4] 
and [0.5] trying to show you what it looks like [2.7] this is the shortest line 
length you'll be pleased to hear [0.5] and i can't simulate it exactly 'cause i 
don't know how to get rid of the two er menus at the top [0.3] but essentially 
there were no menus at all [0.4] 
so you've got er a screen that only has the text on it [0.6] and it was in this 
typeface and the paragraphs were like this so this is approximately twenty-five 
characters per line the narrowest [0.8] and [0.3] i can't also ca-, show you 
the scro-, the er the paging but if you were scrolling [0.9] mm then you would 
go down [0.3] line by line and then when you got to the [0.2] bottom [0.6] it 
would start to to move up [0.3] but in fact in the paged version it would jump 
[0.2] from the top to the bottom [0.6] 
so this one's [1.0] the narrow one [0.2] there were six line lengths i'm so i'm 
showing you the narrowest [0.2] one vaguely in the middle [0.5] oh that's if i 
can find it [0.4] er [3.3] right that's twenty-five the one in the middle and 
then i'll show you the the the extreme one the longest line length [0.3] so 
that's fifty-five characters per line so you're probably feeling a bit more 
comfortable with this one [0.8] er [1.1] and then [0.7] the the widest one 
which essentially was done to fill the whole screen [0.7] okay [1.7] so [3.0] 
if we get back [10.4] right so there were six line lengths and they 
could either be in a scrolling version or in er a paged version [0.5] and did 
you expect this that the one that was read the fastest was that one that filled 
the screen [0.5] one-hundred characters per line [1.9] i didn't [1.0] er [0.4] 
to be fair er [0.8] although it was read [0.4] it was the fastest i think if 
you [0.2] 'cause you know a little bit about statistics it was only [0.8] 
significantly different from the very narrowest one so er [0.3] but it was 
still if you looked at at [0.2] looked right across the six of them it was the 
[0.3] the fastest one [0.4] so i've thought ah well maybe i've got something 
here [0.8] er [0.3] but when people [0.4] making their judgements about what's 
easiest to read they did not consider that the easiest to read they thought it 
was [0.6] the least [0.2] easiest to read so they probably had the sort of 
reaction that you would have i think that er [0.3] that really looks quite 
overwhelming [1.6] er the check on comprehension ju-, didn't show any 
differences because of course there is the possibility of this speed-accuracy 
trade-off [0.5] you see [0.6] a screen full 
of text and you think ergh i'm just going to try and get through that ge-, [0.
2] er and then you you don't actually take it in [0.5] but it appeared not to 
be the case [1.3] er [0.4] now needless to say you know sort of [0.3] 
colleagues didn't [0.6] think this was [0.2] the result that i should find or 
it's the sort of thing that you think know can you really go around 
recommending these long line lengths on screen [0.4] certainly before i did 
this research [0.2] i was giving tutorials saying [0.3] use quite short line 
lengths on screen probably shorter than in print because [0.2] it's more 
difficult to read from the screen [1.0] er [0.3] so you look for reasons why 
you might get something like this was there something about the experiment [0.
4] and if you can i won't go back to it but if you can think back to that 
narrow line length there was a screen [0.6] the screen was all white the 
background was white and certainly some of the people found it quite glaring [0.
3] so there might have been a problem [0.6] with line lengths that don't fill 
the screen because 
it's bright and it's glaring [0.4] so we changed to grey [0.2] and just to show 
you as an image of what that looked like rather than the actual screen [1.5] so 
now we've just got the background as grey [0.4] and didn't go through all six 
line lengths just went through the twenty-five the s-, the shortest the middle 
and the longest [0.6] and er well it got a bit more blurry [0.3] er now er not 
the screen the results got er [0.3] got blurred [0.5] er and it wasn't so 
obvious that this long line length was speeding people up [0.7] what was 
certainly [0.4] obvious and it again is logically the case anyway [0.4] is that 
if you've got a very narrow line length you've got to scroll down [0.4] sort of 
screens and screens [0.2] for the same length as document if [0.4] if you've 
got a wide line length you've only got a few screenfuls [0.3] so it's actually 
quicker mechnically [0.5] to get through [0.5] er [0.4] a short fat [0.2] a 
sort of a wide [0.4] er not particularly tall document [0.2] so maybe that's 
all it it was i didn't think it was but that certainly was contributing [0.3] 
having to scroll 
through a long document [3.0] right it sort of [0.2] that never became totally 
clear but i moved on to something related although not exactly the same [0.4] 
which was to compare [0.9] columns of text [1.3] because and i forgot to 
mention this bit earlier [0.5] this work was funded by Microsoft Corporation 
because they really [0.4] wanted to find out about [0.2] how people read from 
screen or rather what's er useful layouts and so on and so forth [0.8] and i 
know last week we talked about sort of [0.4] who does experiments influencing 
er the results but fortunately they [0.4] there was [0.2] no way in which they 
they influenced these 'cause they couldn't really [0.3] except to say the sort 
of things they were interested in looking at [0.7] and about this time which 
was [0.7] probably about five years ago [0.5] er they were starting an online 
magazine web magazine called Slate and they were interested in whether to use 
single columns or multiple columns for that [1.3] now [0.9] my my view is [0.2] 
that you can't [0.7] put text on screen in three columns and have it scrolling 
[1.0] er [0.2] now i 
know thi-, this [0.5] is true on the web [0.2] but if you think about what 
you're having to do then you're having to read down [0.5] scroll down get to 
the bottom of a column [0.2] scroll back up [0.4] come down again scroll back 
up [0.5] so my reckoning having [0.2] three columns it's got to be in a paged 
format you've got to be able to read what's on screen [0.3] and then get the 
next [0.7] page as such the next screenful [0.4] 
sm1268: can you not have three columns and just wait for them [0.2] [0.6] [1.2] 
nf1263: you wa-, right yeah [0.4] [laughter] the trouble is i'm ass-, i'm 
assuming that everyone has the same size screen of course because if you've got 
a tiny one you're going to have to scroll down anyway [0.6] but [0.3] that was 
where i started from that three columns has got to be paged [0.7] since i was 
comparing it with a single column i then compared it with that column [0.3] 
being scrolled or being paged [0.2] so that finds out whether it's scrolling 
that's causing something or whether it's the columns [1.1] er [0.8] right [0.3] 
show you the sort of size that i was working with so there [0.3] gone 
back to the [0.3] the grey surround er [0.3] and [0.2] the three columns yeah 
[0.3] so for those that you probably did all get it by then i i think you can't 
sort of carry on [0.4] having to scroll down and come back up here [0.3] so you 
do all three [0.4] and then you press somethng and you get another screenful [1.
2] and the single column was about eighty characters per line so we're probably 
talking [0.4] yeah it covered the same sort of area [0.4] it's a pretty [0.4] 
pretty long line again [3.3] right [0.2] so what we found there was that the 
the [0.2] it was the single [0.5] column the fairly long line [0.3] length that 
was read fastest faster than three columns [0.4] er and it was faster in paged 
because i think you basically it's faster because you haven't got to scroll so 
you just save time on [0.3] having to do that mechanical thing [1.4] but we 
happened to be recording how old p-, [0.2] our [0.3] participants were [0.5] 
and most of them were within this age range we didn't take exact ages within 
the age range eighteen to twenty-four because we recruited from within the 
university [0.6] but we 
did have some that were [0.3] twenty-five [0.2] plus [exclamation] [0.5] and 
they showed a different pattern or rather the [0.4] the younger ones the 
eighteen to twenty-four year olds [0.5] were [0.8] showing this [0.4] this 
pattern and not everyone else [1.1] so why is it that a particular age group 
should read a single column faster [0.3] than [0.5] than everyone else [0.2] 
basically [0.4] so everyone else didn't show a difference between [0.3] 
everyone else being the older people [0.8] didn't show a difference between 
three columns and a single column [0.8] but eighteen to two-, twenty-four year 
olds did [1.5] i've only had [0.3] i've i've made one guess at this but i i've 
[1.2] you're eighteen to twenty-four year olds aren't you [0.7] mostly [0.9] 
does anyone squeeze out the top [2.5] 
sm1269: so what's your guess [0.4] 
nf1263: my guess is that they were more used to reading [0.5] web pages [0.2] 
and that [0.2] the [0.6] the text could have looked pretty similar to that [0.
3] and not in [0.2] three columns 
sm1270: were they eighteen to twenty-four year olds 
nf1263: eigh-, twenty-four four year olds were more used to than the 
older people [0.3] yeah [0.3] and so it was kind of a 
sm1270: well 
nf1263: familar-, familiarity [0.2] thing [1.1] 
sm1270: and they read less newspapers [0.9] 
nf1263: yeah d-, if it if it [0.3] does generalize from print yeah you would 
think that the column thing would come out yeah [1.4] it [0.2] it [0.3] sligh-, 
[0.2] gets slightly more [1.0] complicated er [0.2] no i haven't put it on an 
overhead [0.5] but if you [2.1] if you also what i did was to divide the group 
into [0.2] faster and slower readers basically just because you've got the 
times and looked at whether there were differences [0.6] er there [0.3] make 
sure i get this right 'cause it's complicated [0.8] er [1.4] faster readers 
were showing although generally there were no differences in comprehension 
faster readers were [1.1] showing better comprehension when they were reading 
the three columns [2.3] so the-, there's two funny things happening here 
there's an age differential [0.5] and faster readers seem to be [0.6] showing 
something a bit different to slower readers as well [0.8] er now those were 
sort of indications and i i just put them in now because 
they did er feed in to what i went on to do after that [0.4] but just to 
conclude this part before moving on to the second case study [1.1] i would say 
that these together have suggested that [0.4] longer line lengths may be okay 
on screen there is this problem that people are definitely saying [0.5] well i 
don't think it's easier to read [0.4] but they are reading them fairly fast and 
they don't apparently ap-, [0.3] be [0.5] reading them without taking in what 
they read because they appear to be able to [0.4] to answer questions [0.7] 
however my second point there may be different levels of comprehension [0.3] i 
don't think what i wa-, [0.4] the way i was measuring comprehension was 
particularly sensitive [0.5] so my feeling is that [0.6] maybe [0.6] er with a 
long line length there's a possibility that [0.3] they're they're skimming it 
because it's rather uncomfortable but i can't pick that up because of the way 
i'm testing it [1.3] and i'm suggesting maybe different reading patterns 
according to age or according to how fast you read sort of from the last 
results there [0.7] but what seems to be absolutely certain is that we can't 
say we've done all this research in print let's use those results [0.3] we 
don't know what the optimum line length is for reading from screen [0.4] my 
research hasn't shown it but my research has questioned [0.4] that [0.3] fifty-
five characters or seventy characters the sort of thing that that comes out of 
print is the right line length for screen [0.6] so we shouldn't rely on that [1.
1] 
er but the reason to go on with the next s-, sort of series of studies was [0.
6] c-, i suppose my uncomfortableness with this one-hundred characters per line 
and and reco-, and recommending that [0.3] and also [0.2] other people being 
uncomfortable [0.8] other colleagues [0.4] so my thoughts were [0.4] that [4.0] 
long line lengths [0.6] can be read as far [0.4] can be read as fast but 
they're perhaps that we definitely not perceived as easy to read so perhaps 
there is a problem with comprehension that's really just summarizing already [0.
5] and perhaps there's something happening [0.5] with [0.3] faster reading [0.
2] 
because we're finding some sort of differences there [1.4] so [1.9] what i went 
on to do was to [0.7] essentially look at different reading rates so [0.2] to 
get people to first of all read at their normal speed [0.5] and then to [0.2] 
try and speed them up [0.3] er this is the fast reading rate [0.7] and the way 
that we did that was essentially to train people to read faster than they do 
normally [0.4] so we're not trying to get everyone to a certain reading speed 
but we tried to get people [0.3] to a speed that was twice as fast as they 
normally read [1.2] so [0.2] get them to read something and then you say okay 
try reading the next document faster than that and you see how fast they read 
it [0.3] and then say okay well try reading a bit faster than that if they 
haven't got up to [0.5] twice the speed [0.6] and what we found [0.3] perhaps 
not too surprising it is very difficult to read at twice the speed [0.4] that 
you normally read at [0.3] if you're still trying to take in [0.2] what you're 
reading [0.5] so we actually stopped at [0.2] allowing people to read [0.8] at 
seventy seventy per cent faster not twice as fast [0.3] sounds like it's more 
[0.2] but they wouldn't have to speed up quite as much as that [1.5] er still 
interested in line lengths 'cause i had some sort of idea that different line 
lengths might be [0.5] more useful at different speeds and maybe it's handy to 
have a long line length when you read fast but it's not when you're reading at 
a normal speed something like that [1.3] but because i thought well this 
comprehension thing isn't really working [0.2] that i set out to measure 
comprehension in a different way [0.6] and [0.2] to try and to tease out 
perhaps different things that might be [0.4] lost [0.2] when you read fast [0.
5] so there were various different types of questions [0.9] one referring to 
the title one to a main idea [0.2] and i'll show you what those are it's 
probably [0.4] easier than er [1.0] course you don't know what the text is all 
about so this is [0.2] slightly abstract here [0.7] er multiple choice 
questions which [0.2] believe me are incredibly hard to set because fair enough 
you can 
do the right answer [0.2] but [0.3] there's a real problem in getting [0.6] two 
alternatives that are [0.6] vaguely plausible [0.2] so essentially these [0.2] 
these two alternatives that are wrong have come from somewhere in the document 
[0.4] so that people might be fooled by them [0.8] so this was a document all 
about Monaco and [0.2] casinos and stuff like that [0.3] so you there's a qu-, 
one question which says which is the best title which i-, which is best fits 
the text [0.3] so you've got to have got a sort of overview of what you've read 
to be able to answer that 'cause it doesn't exist within the document [1.3] 
then there's a question that's a main idea so what the sort of gist is of the 
document [0.6] er [0.3] and [0.2] Monaco is either dedicated to the principle 
of fabulous excess motor sport or strength in the continuity of the monarchy [0.
5] course you might be able to answer these without reading [0.6] there's 
always a danger without reading the document [1.2] structure question [0.2] 
that's that's an important one because it came out very very badly [0.4] and 
this is actually just asking the question as to [0.3] what happened when or 
what happened before something else [0.4] so what you've got to remember is the 
sequence of events in something that you've read [0.5] so who does the autor 
author interview just before blah blah blah [0.3] so these three people would 
have come up somewhere in that document [1.2] then you've got incidental sort 
of detailed questions h-, [0.2] er with some a number in that case [0.6] er [0.
2] and there was also one that didn't refer to a number [0.3] and then a main 
factual question which wasn't the main idea but was still something fairly 
major [0.6] and then finally [0.4] this is remembering precise wording so you 
have to say whether this precise wording was in something that you've read [1.
1] and the ones that weren't in there were taken from another part of the 
document so you've got the same writing style [0.8] so what i hoped to find was 
well i thought this must be really sensitive so maybe people'll lose some of 
these things they wi-, won't be able 
to answer some of them but they will be able to answer others [3.3] right [0.3] 
so [0.4] still found that short line lengths are read slower [0.2] this had [0.
3] twenty-five fifty-five and one-hundred again so your very short [0.3] your 
medium line length and your long line length [0.7] but the long ones [0.3] the 
long line length wasn't coming out much much better it was the same as the 
medium line length so i'm losing [0.2] my [0.7] my good effect of a long line 
length it just doesn't seem to be coming out in the same way as it did in the 
first experiment [1.7] but there were differences in comprehension [0.2] which 
was good because that was something that i hadn't found before so this was more 
sensitive [3.0] and [0.2] it's the middling line length that essentially is 
coming out [0.2] better [0.3] so comprehension's better [0.4] and those line 
lengths are read reasonably fast [2.8] er [1.5] when people read fast as you 
might expect their comprehension is poorer [0.3] so that that [0.5] you want to 
find that because it makes sure that what you're you actually got a sensitive 
measuring instrument that was found [0.6] and people recall less detail rather 
than the main facts so what goes is the detail [0.6] but the one that i sort of 
[0.2] mentioned just now [0.4] for some reason [0.2] people had [0.2] much more 
difficulty with remembering where something came within the text the structure 
question [0.6] than they did any of the other questions and that's regardless 
of how fast or or slow they were reading [1.3] and [0.5] again giving a stab at 
why this might happen [0.5] all of this was scrolling text and i think i've 
referred to this before what you lose in scrolling text that you have in books 
[0.5] is any [0.7] cue to where something sits on a page so whereas in a book 
you might think oh yes that idea came on [0.5] er a left page near the top [0.
4] you haven't got that [0.2] sort of indication when you've got [0.3] text 
that you just scroll through [0.4] and maybe that somehow influences you 
recalling [0.3] where certain things happened within the text [0.4] that's one 
of those guesses that i haven't [0.2] got any f-, more evidence for [1.4] and 
the final 
thing that i was trying to do by looking at [0.2] how people scrolled was to 
say [0.4] well is there [0.7] are there good ways of scrolling through text 
that either speed us up [0.2] so we can read faster if we want to do that 
'cause maybe we want to just skim read [0.3] or are there [1.1] other [0.3] 
maybe the same methods or different methods that enable us to understand better 
[0.7] so i was looking for [0.8] the characteristics of effective readers an 
effective reader might be a fast reader [0.3] or an effective reader might be 
someone that's [0.2] good at understanding what they've read [0.3] and maybe 
depending on the task you have [0.5] different different demands maybe you want 
to read quickly or maybe you want to [0.4] to [0.2] study it more and get more 
out of it [1.0] and there are different scrolling patterns that do match up but 
they are different 
depending on whether you want to read fast or understand [0.7] i've just got 
the fast one here [0.6] so [0.2] those people that are reading fast were [0.3] 
spending less time between scrolling [0.2] so they were also making only a few 
scrolling movements [0.2] so essentially they weren't [0.2] pausing for very 
long to read [0.3] they were keeping the text [0.5] moving [0.5] and that 
speeds up reading [0.6] whereas people that showed better comprehension were [0.
4] sort of doing the opposite they were pausing for longer [0.4] in between 
scrolling [1.1] so [0.2] what that might mean if i carry on and do some 
research is that [0.2] you can manipulate how people scroll through text and 
see whether it does improve [0.6] either their speed of reading [0.5] or [0.9] 
their [0.4] comprehension [0.3] because this doesn't do that that this just 
said [0.3] says those readers that did that [0.2] came out well [0.2] it 
doesn't 
say that if [0.2] if i ask you to do that [0.2] then [0.2] then [0.2] you'll [0.
2] you'll come out well from that [1.3] so in conclusion my always my 
conclusions are there's room for lots lots more research and [0.5] what i can 
say definitely from this and it's probably the only thing [0.4] is that i you 
cannot predict from [0.6] print legibility research [0.2] how people are going 
to read from screen [0.8] and [0.6] that [0.3] ought to be obvious because 
we've got a more interactive way of reading we can manipulate the text in 
different ways [0.5] but there seems to be something a bit more [0.9] perhaps 
basic in relation to the lengths of lines that [0.3] that affects how we read 
from screen [1.8] okay