Content uploaded by Sofie Beier
Author content
All content in this area was uploaded by Sofie Beier on Aug 15, 2019
Content may be subject to copyright.
Contents lists available at ScienceDirect
Acta Psychologica
journal homepage: www.elsevier.com/locate/actpsy
Smaller visual angles show greater benefit of letter boldness than larger
visual angles
Sofie Beier
⁎
, Chiron A.T. Oderkerk
The Royal Danish Academy of Fine Arts, School of Design, Denmark
ARTICLE INFO
Keywords:
Stroke boldness
Visual angle
Font
Letter recognition
Legibility
ABSTRACT
Research has shown that fonts viewed at a smaller visual angle benefit from greater letter boldness. Since small
and large visual angles operate on different spatial frequencies, we examined whether the effect was dependent
on font size. By applying a paradigm of single-letter exposure across two experiments, we showed that fonts of
thinner letter strokes and of extreme boldness decreased recognition for all tested font sizes, and that there was a
positive effect of boldness at small visual angles which did not occur at large visual angles. The paper provides
evidence that bolder fonts are less effective at improving recognition at larger visual angles, and that over a scale
of font weights there is a drop-offat the lightest and the heaviest extremes at all tested font sizes.
1. Introduction
The act of reading words involves several levels of parallel proces-
sing, including the identification of letter components, whole letters,
and words. These processes are believed to involve a range of reciprocal
feedback loops (Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001;
McClelland & Rumelhart, 1981;Perry, Ziegler, & Zorzi, 2014). Letter
recognition is hence central to word recognition and most real life
reading situations.
To determine the most essential features for letter recognition, some
researchers have employed methodologies involving the degrading or
removing parts of the stimuli. Fiset et al. (2008) used the ‘bubbles
technique’to show that the letter stroke terminations of the Arial font
are the most important for letter recognition. In a lexical decision ex-
periment, Rosa, Perea, and Enneson (2016) found that removing mid-
segments of letter strokes was more detrimental to reading than re-
moving the junctions or the letter stroke terminations in the Minion
font; similar findings were shown for the Courier font in a single-letter
recognition task (Petit & Grainger, 2002). In contrast, Lanthier, Risko,
Stolz, and Besner (2009) demonstrated that removing the junctions of
the Arial font was more damaging than removing midsegments.
In spite of the diverse range of results in determining the most es-
sential letter features for identification, there appears to be a consensus
within cognitive psychology that feature detection is a vital aspect of
letter recognition (Finkbeiner & Coltheart, 2009;Pelli, Burns, Farell, &
Moore-Page, 2006;Sanocki & Dyson, 2012). The methodologies of the
above-mentioned experiments were employed to study components
within one font of regular weight. As extensive research has demon-
strated that font style has an effect on letter recognition (Beier & Larson,
2010;Beier, Starrfelt, & Sand, 2017;Pelli et al., 2006;Pušnik, Podlesek,
&Možina, 2016), it is likely that the specific choice of font may have
significantly influenced the results.
In this paper, we intend to show that letter boldness alone can in-
fluence letter recognition within the same font family. Furthermore,
due to the influence of spatial frequencies, we predicted that the effect
of letter boldness differs between different font sizes.
1.1. Spatial frequency
The effect that a test font has on a participant's reading greatly
depends on test methodology (Beier & Larson, 2010). One of many
reasons suggested for this is that fonts presented at larger or smaller
visual angles draw on different spatial frequency channels. In the per-
ceptual system, the spatial-frequency tuning of the visual neurons
varies in relation to both the size of the stimulus and the luminance
contrast (Alexander, Xie, & Derlacki, 1994;Chung, Legge, & Tjan,
2002); this mechanism is found to be largely similar in both foveal and
peripheral vision (Chung et al., 2002;Chung & Tjan, 2009). Majaj, Pelli,
Kurshan, and Palomares (2002) demonstrated that observers employed
only one channel at a time, the choice of which was dependent on the
https://doi.org/10.1016/j.actpsy.2019.102904
Received 14 December 2018; Received in revised form 24 July 2019; Accepted 4 August 2019
⁎
Corresponding author.
E-mail address: sbe@kadk.dk (S. Beier).
Acta Psychologica 199 (2019) 102904
0001-6918/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
T
stimulus and could not be selected by the observer. In other words,
small visual angles
1
require the use of lower spatial frequencies, which
results in letters being perceived as blurred images. The finer details
and edges that are perceived at higher spatial frequencies, are therefore
not available at small visual angles (Fig. 1). To facilitate greater letter
recognition at small visual angles, the focus should be on those visual
features that are visible when the observer makes use of the relevant
channels. One such feature is the distribution of letter boldness.
1.2. Letter boldness
The different shapes of the individual letter parts of the alphabet can
be viewed as different distributions of black and white surface areas —
this relationship between black and white changes with letter boldness
(Fig. 2). Extremely bold fonts cover a bigger black surface area, which
‘eats up’the surrounding white area, while the black surface area of
extremely light fonts is much smaller, which results in a bigger white
surface area inside and around the letter (Noordzij, 2005). This dif-
ference in the distribution of black and white surface area changes the
shapes of letter features between fonts of different boldness.
Our hypothesis is that the influence of boldness on letter identifi-
cation varies depending on the visual angle. In this study, we aim to
identify the ideal letter-stroke boldness at different visual angles.
Bold fonts play a central role in headlines and for text emphasis, as
the darker surface area makes them stand out from the page relative to
regular weights (Bateman, Gutwin, & Nacenta, 2008;Dyson & Beier,
2016). Furthermore, bolder fonts can also facilitate font legibility in
certain reading scenarios. Measuring visual acuity, Sheedy, Subbaram,
Zimmerman, and Hayes (2005) showed that Franklin Gothic Book was
less legible than Franklin Gothic Medium, Demi, and Heavy at small
font sizes. This indicates that small font sizes require a minimum stroke
weight to maintain their readability (Fig. 3). These findings matched
earlier findings by Kuntz and Sleight (1950). In a study involving
contrast threshold and visual acuity of numerals that had stroke width/
height (SW/H) ratios ranging from 1:4.2 to 1:16.0, Kuntz and Sleight
(1950) found that the font that delivered the best performance was the
one that had a SW/H ratio of 1:6.0 (this ratio equals Franklin Gothic
Medium in Fig. 3). In a follow-up experiment, the researchers tested a
smaller range of ratios and here demonstrated similar performances for
fonts in ratios of 1:4.0 to 1:6.0.
There are also indications that fonts of bold weights facilitate
reading in low-luminance conditions. Using a visual search task that
involved eye-tracking, Burmistrov, Zlokazova, Ishmuratova, and
Semenova (2016) found a general disadvantage of lighter-weight fonts,
showing longer search time, increased fixation duration, and lower
saccadic amplitude when testing different boldnesses of Helvetica Neue.
The study further demonstrated a positive effect of the bold font on
fixation duration, albeit only when there was a low-luminance contrast
between foreground and background. These results were in line with
earlier findings by Luckiesh and Moss (1940), who measured contrast
threshold and found that the light-weight version of the font family
Memphis was inferior to all the heavier weights tested.
In reading scenarios that involved regular text sizes and high-lu-
minance contrast, however, letter boldness has not been found to have
any real effect in either lexical decision tasks (Dobres, Reimer, &
Chahine, 2016;Dyson & Beier, 2016) or reading speed tasks (Bernard,
Kumar, Junge, & Chung, 2013;Tinker, 1964).
A look at the studies above suggests that the effect of letter boldness
on legibility is dependent on the reading situation and the level of the
spatial frequency. Bold fonts have been demonstrated to provide an
advantage at small visual angles and low-luminance contrasts, which
suggests that low-frequency channels benefit from bold fonts. Spatial
frequency tuning also results in the perceptual phenomenon that the
same font weight appears lighter in small sizes than in large sizes
(Fig. 4). As edges and details disappear in small sizes, the stroke be-
comes less defined and hence appears lighter. Since the early days of
printing, typographers have been aware of this effect and have used
optical scaling of their fonts to make letters of small sizes heavier,
wider, and with lower stroke contrast than in larger font sizes (Ahrens &
Mugikura, 2014).
However, the different processing of frequency channels might in-
dicate that those features that enhance letter recognition at small visual
angles could be diminished when replicated at large visual angles. In
the present study, we tested the hypothesis that letter recognition
benefits more from boldness at small visual angles than at large visual
angles. Thus, the goal of the experiments was to examine the varying
effect of letter-stroke at larger and smaller visual angles.
2. Experiment 1
2.1. Material and methods
2.1.1. Participants
The experiment was advertised through a participant recruitment
website (Forsoegsperson.dk). A total of 21 participants aged from 19 to
52 years (M
age
26.9 years, SD = 7.5 years, 15 women) took part.
Participants received a gift card of DKK 150 upon completion of the
experiment. All reported normal or corrected-to-normal vision. Written
informed consent was obtained from each participant after the experi-
ment was explained. The research followed the tenets of the Declaration
of Helsinki and The Danish Code of Conduct for Research Integrity.
2.1.2. Test material
Research has shown that reading rate is independent of whether the
test fonts are widely used or new to the participant, as long as the fonts
have common letter shapes (Beier & Larson, 2013). The five test fonts
originate in the font family Ovink and were designed for this experi-
ment. This made it possible to choose the weights that fit our
Fig. 1. Spatial frequency. The two images are identical. Left: at larger visual
angles the higher-frequency channels show details and edges, and thus, the
viewer sees the sharp ‘F’more than the blurred ‘D’. Right: at small visual angles,
the lower-frequency channels show letter weight and proportions, and thus, the
viewer sees the blurred ‘D’more clearly than the sharp ‘F’. One can further
experiment with viewing the large-size image from a larger reading distance.
This will make the letter ‘D’stand out instead of the ‘F’.
1
Following this, a sign showing 120-point type viewed at 4 m' distance would
have the same visual angle as a sign showing 12-point type viewed at a normal
reading distance of 40 cm.
S. Beier and C.A.T. Oderkerk Acta Psychologica 199 (2019) 102904
2
experimental aims and avoid dependency on font weights designed by
others. To generate the interpolation of the intermediate fonts in the
Glyphs software, the two extreme fonts were assigned weight values of
1 (Light) and 5 (Bold), with the three intermediate fonts assigned
weight values of 2 (Regular), 3 (Medium), and 4 (Semi Bold). Thus, the
boldness increased linearly across the five test fonts.
As is customary in professional font design, the heavy weights are
perceptually adjusted in the junctions, so that when a round shape
meets a stem, the letter stroke thins, and for letters like ‘a’and ‘e’, which
have many details in small spaces, the middle part is thinner than the
rest of the letter. This results in a different kind of letter contrast in the
bolder fonts compared with the lighter fonts; however, it also ensures
that the findings of the experiment can be directly transferred into real-
life usage (Fig. 5). This way of adding a perceptually comparable
amount of weight to the stroke follows the tradition of sans serifs fonts.
The tradition of serif fonts, which adds weight by increasing the stroke
contrast in such a way that the main bulk of the weight is placed at
vertical strokes (Noordzij, 2005), is left for future investigations.
We tested the following stroke width/height ratios: Light (1:20.0),
Regular (1:10.0), Medium (1:6.4), Semi Bold (1:4.7), Bold (1:3.8).
Stroke width was measured from the horizontal width of the lowercase
stem, and height was measured from the baseline to the top of the as-
cending lowercase letters. Thus, the letter ‘h’with a stroke width of
20 units and a height of 200 units has a ratio of 1:10.
2.1.3. Apparatus
The stimuli were displayed on a 12.3-inch LCD monitor in a dimly
lit room (refresh rate = 60hz, resolution = 3000 × 2000). Experiments
were created using the software OpenSesame 3.2. (Mathôt, Schreij, &
Theeuwes, 2012). Stimuli were presented as black text (#000000) on a
light grey background (#DADADA). However, stimuli were presented
without anti-aliasing.
2
Therefore, in order to ensure that the resolution
of the stimuli was comparable for all three sizes participants were se-
ated far away from the monitor. The distance between the participant
and the monitor was determined at the beginning of the experiment,
such that the distance between the participant's eyes and the monitor
was 200 cm when the participant was seated in their preferred position
with their back to the chair.
2.1.4. Procedure
Fonts of different boldness vary in the amount of inter-letter spacing
so that bolder fonts have smaller inter-letter spacing compared to
lighter fonts. To eliminate an unwanted variable of inter-letter spacing,
which would have shown in letter-string presentations, we presented
the stimuli as single lowercase letters.
We applied a method of short exposure at 3.15° left or right side of
the fixation circle. We tested three size conditions (measured from the
top to the bottom of the descender): Small 0.08° (equals x-height
3
of
0.06 cm at a reading distance of 40 cm = 3.5 points), Medium 0.14°
(equals x-height of 0.01 cm at a reading distance of 40 cm = 6 points),
and Large 0.20°(equals x-height of 0.14 cm at a reading distance of
40 cm = 9 points), and five weight conditions: Light (1), Regular (2),
Medium (3), Semi Bold (4), and Bold (5).
In each trial, the stimulus consisted of one of sixteen lowercase
letters (a, d, e, f, g, h, k, m, n, o, p, r, s, t, u, y), presented individually
for a short exposure time. Stimuli were presented on the left or right
side of the fixation circle, 496 ms after the initiation of a trial.
Fig. 2. The bolder the letter, the smaller the letter counter which is the area not covered by the letter stroke (marked in red). Demonstrated in the font family Avenir
Next. (For interpretation of the references to colour in this figure legend, the reader is referred to the digital version of this article.)
Fig. 3. Sheedy et al. (2005) tested four weights of the font family Franklin Gothic and found Franklin Gothic Book (far left) to be inferior to all the other weights in a
visual acuity experiment. The stroke width/height ratio (SW/H) from left: Franklin Gothic Book (1:9.0), Gothic Medium (1:6.0), Demi (1:4.7) and Heavy (1:3.5).
Fig. 4. Top row: all letters are set in the font weight Ovink Regular. Bottom
row: The smallest letter is set in Ovink Bold, the following is set in Ovink
Medium, while the three larger ones are set in Ovink Regular. In the top row,
the weight appears increasingly lighter in smaller sizes, while the weight in the
bottom row appears more even across the different sizes.
2
Anti-aliasing is a technique that smooths the edges of strokes so they do not
appear jagged.
3
The x-height is the height of the lowercase ‘x’.
S. Beier and C.A.T. Oderkerk Acta Psychologica 199 (2019) 102904
3
Participants were instructed to maintain fixation on the fixation circle
while the stimulus was being presented. In order to remove any possible
after-image effects, the stimuli were followed by a mask for 496 ms in
the form of a Gaussian noise patch of variable size. Following a stimulus
presentation, participants were asked to name the stimulus letter if they
were able, after which it was recorded by the experimenter.
In each block, each of the five fonts was shown for each of the three
sizes on both the left and the right side of the fixation circle for a total of
60 trials per block. Over the course of an experimental session, each
participant engaged in one practice block and eight test blocks.
In order to ensure performance comparability between participants,
the stimulus exposure duration was determined for each participant
separately using a staircase procedure adapted from the accelerated
stochastic approximation by Kesten (1958) (Treutwein, 1995). The trial
outline for this staircase procedure was similar to the later testing
session, although participants were only presented with Medium-sized
stimulus letters of Medium (3) boldness over a series of trials. After an
initial exposure duration of 160 ms on the first trial, the exposure
duration during the following trials would increase or decrease by 64
ms, depending on the accuracy of the participant's report in the pre-
ceding trial. This meant that the exposure duration would continue to
increase by steps of 64 ms if the participant continually failed to report
the correct letter. When the exposure duration was long enough for the
participant to make their first correct reports, the step size by which the
exposure duration changed between trials would decrease after a pre-
determined number of reversals of accuracy (i.e., a correct report fol-
lowed by an incorrect report, or vice versa). Step sizes decreased to
48 ms after seven reversals, to 32 ms after 13 reversals, and to 16 ms
after 18 reversals. The staircase procedure was terminated after 25
reversals, at which time the final exposure duration to be used during
that participant's following test session was the average of the exposure
durations of the final six reversals.
2.2. Results
Using a 3 (size condition: Small, Medium, and Large) × 5 (weight
condition: Light, Regular, Medium, Semi Bold, and Bold) repeated-
measures ANOVA on mean accuracy, we found large main effects of size
condition, F(2, 40) = 410.31, p< .001, ω
2
= 0.81, and of weight
condition, F(4, 80) = 49.06, p< .001, ω
2
= 0.21. However, the size of
the stimuli influenced the accuracy of the responses differently across
the five weight conditions, resulting in a small significant interaction
effect, F(8, 160) = 2.55, p= .012, ω
2
= 0.02.
Using 30 planned comparison, corrected for multiple comparisons
using the Bonferroni method, showed that the interaction resulted from
a significant difference between the mean accuracy in the Regular (2)
and Semi Bold (4) weight, t(20) = 5.48, p= .001, dz = 1.20, which
only occurred in Medium-sized stimuli. Conversely, there was no effect
between Regular (2) and Semi Bold (4) weight conditions at the Small, t
(20) = 2.49, p= .648, dz = 0.54, or Large sizes, t(20) = 1.80,
p= .999, dz = 0.39.
Accuracy in the Light (1) weight condition was impaired in all sizes.
Specifically, in the Small size mean accuracy in the Light (1) weight was
significantly lower than Medium (3), t(20) = 4.98, p= .002, dz = 1.09,
Semi Bold (4), t(20) = 5.12, p= .002, dz = 1.19, and Bold (5), t
(20) = 5.40, p= .001, dz = 1.18, though the difference between mean
accuracy of Light (1) and Regular (2) did not reach significance, t
(20) = 3.12, p= .164, dz = 0.68; in the Medium size mean accuracy
for Light (1) was significantly lower than Regular (2), t(20) = 4.93,
p= .002, dz = 1.08, Medium (3), t(20) = 6.83, p< .001, dz = 1.49,
Semi Bold (4), t(20) = 8.59, p< .001, dz = 1.86, and Bold (5), t
(20) = 6.52, p< .001, dz = 1.42; and in the Large size mean accuracy
for Light (1) was near-significantly lower than Regular (2), t
(20) = 3.49, p=.070, dz = 0.76, and significantly lower than Medium
(3), t(20) = 4.61, p= .005, dz = 1.01, Semi Bold (4), t(20) = 6.13,
p< .001, dz = 1.34, and Bold (5), t(20) = 3.81, p= .033, dz = 0.83.
No other comparisons reached significance (all p's > 0.117) (Fig. 6).
2.3. Discussion experiment 1
Supporting our hypothesis that the effect of boldness was dependent
on size, Experiment 1 showed a small significant interaction effect be-
tween size and weight, which indicated that boldness had a stronger
effect on the Medium font sizes, which did not occur in the Small or the
Large font size. Furthermore, we found that recognition improved in
nearly all weight conditions, relative to the lightest weight. Measuring
reading speed with rapid serial visual presentation of words, Bernard
et al. (2013) found that performance dropped when reading the boldest
font of their experiment. We, however, did not replicate this in Ex-
periment 1, as our boldest font (Ovink Bold (5)) resulted in generally
good performances. The boldest font of Bernard et al. (2013) was much
bolder than any of our fonts. By adding an extreme bold font to our
Experiment 2, we were interested in seeing if the experimental para-
digm of single letter recognition, could similarly result in a performance
drop with extreme letter boldness.
Fig. 5. The sixteen letters tested in experiment 1, set in the five test fonts. From the top: Ovink Light (1), Ovink Regular (2), Ovink Medium (3), Ovink Semi Bold (4),
and Ovink Bold (5). The numbers to the right are the stroke width/height ratio.
S. Beier and C.A.T. Oderkerk Acta Psychologica 199 (2019) 102904
4
3. Experiment 2
We were interested in replicating the interaction between Regular
(2) and Semi Bold (4) at Medium and Large sizes, as well as seeing if,
like the lightest condition, there was a drop in performance with very
heavy weight.
Firstly, we predicted that the new font, Ovink Ultra Black (6), would
impede performance in regardless of stimulus size. Secondly, we hoped
to replicate the size and weight interaction from Experiment 1. Namely,
we predicted that in the Medium sized letters the increased boldness of
Ovink Semi Bold (4) would facilitate greater letter recognition relative
to Ovink Regular (2), only in the Medium sized letters, while there
would be no such significant difference between the same two weights
for the Large letters.
At the time of testing, presenting stimulus letters with anti-aliasing
in OpenSesame 3.2 was dependent on the x-height of letters on the
monitor being no smaller than 0.73 cm; this was not the case for any of
the sizes in Experiment 1. In order to ensure anti-aliasing, participants
in Experiment 2 were seated yet further from the monitor, such that the
smallest size included in Experiment 2, the Medium font size, had the
same visual angle as in Experiment 1 of 0.14° and an x-height of
0.73 cm.
3.1. Material and methods
3.1.1. Participants
Participant recruitment was the same as in Experiment 1. 15 par-
ticipants (M
age
27.33 years, SD = 5.09 years, 11 women) took part.
Each received a DKK 150 gift card in remuneration upon completion of
the experiment. All reported normal or corrected-to-normal vision.
3.1.2. Test material
The fonts Ovink Regular (2) and Ovink Semi Bold (4) are identical
to the same fonts in Experiment 1 (Fig. 7). The weight of the new font
Ovink Ultra Black (6) was chosen based on the premises of including a
font that is as heavy as possible without the letter counters closing up.
3.1.3. Apparatus
Stimuli in Experiment 2 were displayed on a backlit 17-inch IBM/
Sony CRT monitor (refresh rate=85 hz, resolution = 1024 × 768) in a
darkened room. Distance between the participant and the monitor was
maintained through the use of a chin rest.
3.1.4. Procedure
With a few exceptions, Experiment 2 was identical to Experiment 1.
Though the distance between the participant and the monitor was in-
creased to 300 cm, the sizes of the Medium and Large stimuli were kept
at the same visual angles as in Experiment 1, at 0.14° and 0.20°, re-
spectively. Given the increased distance and the limited width of the
monitor, stimuli in Experiment 2 were presented 2.80° left or right of
the fixation circle. A new weight condition - Ultra-Black (6) - was
added, while Light (1), Medium (3) and Bold (5) weights were ex-
cluded. Stimulus letters were, therefore, presented in either Regular (2),
Semi Bold (4), or Ultra Black (6), at Medium or Large sizes. Contrary to
Experiment 1, participants were first given a chance to learn the task in
the practice block, before the stimulus exposure duration was calibrated
to their performance in the staircase block. The stimulus exposure
duration during the practice block and the first trial of the staircase
block was set to 180 ms. The step sizes with which the exposure
duration increased or decreased during the staircase block was changed
to match the refresh rate of the monitor. The initial step size was set to
48 ms, after which it decreased to 36 ms after seven reversals, to 24 ms
after 13 reversals, and to 12 ms after 18 reversals. As in Experiment 1,
all stimuli in the staircase block were presented at Medium visual angle
and at Medium (3) boldness. Lastly, participants recorded their own
unspeeded responses on a keyboard.
3.2. Result
Using a 2 (size condition: Medium, and Large) × 3 (weight condi-
tion: Regular (2), Semi Bold (4), and Ultra Black (6)) repeated-measures
ANOVA on mean accuracy, we found large main effects of size, F(1,
14) = 157.78, p< .001, ω
2
= 0.45, and of weight, F(2, 28) = 136.54,
p< .001, ω
2
= 0.46, as well as a small but significant interaction ef-
fect of size and weight, F(2, 28) = 5.38, p= .011, ω
2
= 0.02.
This interaction appears to result from the effect of weight, which
only facilitated recognition of the Medium sized stimuli. Specifically,
planned comparisons, corrected for multiple comparisons using the
Bonferroni method, showed a significantly lower mean accuracy in the
Regular (2) than in the Semi Bold (4) weight condition of the Medium-
sized stimuli, t(14) = 4.80, p= .002, dz = 1.24, while this same com-
parison did not reach significance for the Large stimuli, t(14) = 1.24,
p> .999, dz = 0.32.
Conversely, excessive weight appeared detrimental to performance
regardless of size. Mean accuracy of the Ultra Black (6) font was
Fig. 6. Mean accuracy of the responses
across size and weight conditions. The blue
bars represent Small size conditions, the
green bars represent Medium size condi-
tions, and the red bars represent Large size
conditions. Numbers on the x-axis represent
the weight conditions Light (1), Regular (2),
Medium (3), Semi Bold (4), and Bold (5).
Comparisons marked with * were sig-
nificantly different. (For interpretation of
the references to colour in this figure le-
gend, the reader is referred to the digital
version of this article.)
S. Beier and C.A.T. Oderkerk Acta Psychologica 199 (2019) 102904
5
significantly lower than the Regular (2) font in both the Medium size, t
(14) = 6.02, p< .001, dz = 1.56, and the Large size, t(14) = 10.87,
p< .001, dz = 2.81. Similarly, mean accuracy of the Ultra Black (6)
font was significantly lower than the Semi Bold (4) font in the Medium
size, t(14) = 11.26, p< .001, dz = 2.91, as well as the Large size, t
(14) = 14.91, p< .001, dz = 3.85 (Fig. 8).
3.3. Discussions experiment 2
As in Experiment 1, Experiment 2 showed that letter recognition
was significantly lower at the Regular (2) weight compared to the Semi
Bold (4) weight, though only in the Medium size. We further found a
decline in recognition performance in the Ultra Black (6) weight com-
pared to all other fonts at both font sizes. The implications of this will
be discussed in the following.
4. General discussion
Our goal was to study the influence of font size on the effect of
boldness on letter recognition. The data showed that boldness influ-
ences letter recognition in different ways for small and large sizes, and
that extreme weights caused lower letter recognition.
4.1. Results differ between font sizes
We found that participants were significantly better at recognising
bolder fonts –such as Ovink Regular (2) compared to Ovink Semi Bold
(4) –although this was only true for the Medium-sized letters in
Experiment 1. We then replicated this finding in Experiment 2 when
comparing Ovink Regular (2) and Ovink Semi Bold (4) at Medium and
Large sizes. Based on the theory of spatial frequency processing, which
holds that small visual angles are sensitive to boldness and proportions
with letters appearing blurred, while large visual angles are sensitive to
details and edges, we hypothesised that small visual angles would show
a greater benefit of letter boldness than large visual angles. In line with
our predictions, our results showed that letter recognition of fonts
viewed at a small visual angle (Medium size) benefitted more from
boldness than fonts viewed at a large visual angle (Large size).
Our findings are in line with previous studies on visual acuity and
luminance contrast (Burmistrov et al., 2016;Kuntz & Sleight, 1950;
Luckiesh & Moss, 1940;Sheedy et al., 2005), demonstrating that
boldness can enhance legibility. Specifically, our data supports the
Fig. 7. The sixteen letters tested in Experiment 2, set in the three test fonts. From the top: Ovink Regular (2), Ovink Semi Bold (4), and Ovink Ultra BLack (6). The
numbers to the right are the stroke width/height ratio.
Fig. 8. Mean accuracy of the responses across size and weight conditions. The green bars represent Medium size conditions, and the red bars represent Large size
conditions. Numbers on the x-axis represent the weight conditions Regular (2), Semi Bold (4), and Ultra Black (6). (For interpretation of the references to colour in
this figure legend, the reader is referred to the digital version of this article.)
S. Beier and C.A.T. Oderkerk Acta Psychologica 199 (2019) 102904
6
work of Sheedy et al. (2005), who found that Franklin Gothic Book
(which has a similar stroke-width ratio to Ovink Regular (2)) had to be
read at a larger visual angle than Franklin Gothic Heavy (which has a
similar stroke-width ratio to Ovink Bold (5)). Sheedy et al. (2005)
further found Franklin Gothic Book to be inferior to all heavier weights
tested.
4.2. Extreme weights impair recognition
At all tested visual angles in Experiment 1, Ovink Light (1) resulted
in a lower recognition rate. This finding that very thin letter strokes
impeded performance in all sizes suggests that thin lines not only blur
out in small point sizes but also cause edges and details to become in-
sufficiently visible in large sizes. In extension of this, the heaviest
weight tested, Ovink Ultra Black (6), was inferior to all other test fonts
at both visual angles in Experiment 2. This follows earlier findings by
Bernard et al. (2013), who demonstrated that extremely heavy font
weights had a negative effect on reading speed, as there is a limit to
how much weight a stroke can carry before the inside of the letter
counter fills out completely. Our collective data suggests that for all
font sizes, there is an optimal level of boldness with drop-offs at both
extreme ends, as for both small and large sizes there is evidence for a
significant drop in performances at the two extremes of Ovink Light (1)
and Ovink Ultra Black (6).
4.3. Visual cues and letter recognition
Prior research aimed at identifying the most important letter com-
ponents for recognition only tested one font of regular weight. Some
identified the midsegment of the letter stroke to be the most important
feature (Petit & Grainger, 2002;Rosa et al., 2016), while others iden-
tified the junctions (Lanthier et al., 2009) or the stroke terminations
(Fiset et al., 2008) to be most important. However, visual representa-
tion of letters are not generic. A given letter will always be visualized in
a specific font style and boldness. Findings from one font cannot ne-
cessarily translate into the reading of other fonts. By testing different
letter weights within one font family, we demonstrated both that letter
recognition is enhanced by letter boldness, and that this effect is de-
pendent on size. It could be that our finding that letter boldness en-
hanced recognition on small visual angles resulted from boldness en-
hancing the visibility of all the letter components and thus enhancing
important visual cues needed for letter recognition.
The experiment adds to the existing body of knowledge by de-
monstrating that the positive effect of letter boldness on recognition can
be found in bold weights if the size of the letters is small.
4.4. Letter boldness and older age
The literature indicates that while sensitivity to low spatial fre-
quencies remains relatively constant throughout adulthood, healthy
ageing adults may suffer a loss of sensitivity in the higher and middle
spatial frequency regions (Derefeldt, Lennerstrand, & Lundh, 1979;
Owsley, Sekuler, & Siemsen, 1983;Wright & Drasdo, 1985). Older
readers may, therefore, struggle to identify the finer details of letters,
while the recognition of the overall letter proportions remains intact.
This suggests that findings concerning letter boldness are especially
relevant for this age group. As we did not include older participants in
the present investigation, it is likely that a replication of the experi-
ments with an ageing pool of participants would yield an even greater
effect of weight in the Small and Medium sizes.
4.5. Reading situations
Our findings relate to single lowercase letter recognition. The way
our results translate into the reading of letter strings and words depends
on the inter-letter spacing. As fonts read at small visual angles are
affected by a phenomenon known as crowding, where neighbouring
letters appear to merge (Hess, Dakin, & Kapoor, 2000), it is possible that
the tradition of adding a small amount of inter-letter spacing in bold
fonts will be counter-productive as narrow letter spacing is known to
induce letter crowding (Bouma, 1970).
The results from our experiments could suggest that letter re-
cognition of small font sizes will benefit from having text set in bold
fonts, while for letter recognition of larger font sizes the text can be set
in both regular and bold weights. To put this finding into a real-life
context, the Medium font size of 0.14° typically equals that used for
setting text for footnotes (6 point at a reading distance of 40 cm), while
the Large font size of 0.20°, lies within contemporary newspaper and
book font sizes (9 point at a reading distance of 40 cm, Legge &
Bigelow, 2011). As the critical print size for normal vision readers is
0.20°, which is the smallest font size before reading speed will rapidly
decline (Legge, 2006), and as studies of font boldness at large font sizes
found no effect of bold fonts similar in weight to our Ovink Semi Bold
(4) when compared to regular weights (Dobres et al., 2016;Dyson &
Beier, 2016), we would expect that in any font bigger than 9 point text
sizes read at normal reading distances will similarly fail to enhance
reading performance through increased boldness alone.
The letters presented in the small size were so small that it would
require participants with a visual acuity of under 0.0 logMAR (3.5 point
at a reading distance of 40 cm), which does not represent any real-life
reading situation. We did, however, include this font size in Experiment
1 as we expected the advantages of boldness to be the strongest here,
although this did not turn out to be the case.
Considering the many reading situations involving small visual
angles, our present findings provide evidence that under such reading
conditions, bolder weights facilitate letter recognition, and that both
light and ultra-black font weights should be avoided in any case where
letter recognition is a priority.
5. Conclusion
In all font sizes, the light and the ultra-black fonts were inferior to
all the fonts in the middle of the scale. The bolder weights in the middle
of the scale enhanced recognition in the Medium font size, while failing
to do so in the Large font size. We therefore suggest that, while boldness
enhances letter recognition at small visual angles for the tested sans
serif font family, it fails to do so at large visual angles, as these are
perceived via higher-frequency channels and consequently are more
affected by letter details and edges than by letter stroke weight and
proportions.
Acknowledgement
This work was supported by the Danish Council for Independent
Research [grant number DFF –7013-00039].
References
Ahrens, T., & Mugikura, S. (2014). Size-specific adjustments to type designs: An investigation
of the principles guiding the design of optical sizes. Just Another Foundry.
Alexander, K. R., Xie, W., & Derlacki, D. J. (1994). Spatial-frequency characteristics of
letter identification. JOSA A, 11(9), 2375–2382.
Bateman, S., Gutwin, C., & Nacenta, M. (2008). Seeing things in the clouds: The effect of
visual features on tag cloud selections. Paper presented at the proceedings of the nineteenth
ACM conference on hypertext and hypermedia.
Beier, S., & Larson, K. (2010). Design improvements for frequently misrecognized letters.
Information Design Journal, 18(2), 118–137.
Beier, S., & Larson, K. (2013). How does typeface familiarity affect reading performance
and reader preference? Information Design Journal, 20(1), 16–31.
Beier, S., Starrfelt, R., & Sand, K. (2017). Legibility implications of expressive display
typefaces. Visible Language (pp. 112–133). .
Bernard, J.-B., Kumar, G., Junge, J., & Chung, S. T. (2013). The effect of letter-stroke
boldness on reading speed in central and peripheral vision. Vision Research, 84,
33–42.
Bouma, H. (1970). Interaction effects in parafoveal letter recognition. Nature, 226,
177–178.
S. Beier and C.A.T. Oderkerk Acta Psychologica 199 (2019) 102904
7
Burmistrov, I., Zlokazova, T., Ishmuratova, I., & Semenova, M. (2016). Legibility of light
and ultra-light fonts: Eyetracking study. Paper presented at the proceedings of the 9th
Nordic conference on human-computer interaction.
Chung, S. T. L., Legge, G. E., & Tjan, B. S. (2002). Spatial-frequency characteristics of
letter identification in central and peripheral vision. Vision Research, 42(18),
2137–2152.
Chung, S. T. L., & Tjan, B. S. (2009). Spatial-frequency and contrast properties of reading
in central and peripheral vision. Journal of Vision, 9(9), 1–19.
Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). DRC: A dual route
cascaded model of visual word recognition and reading aloud. Psychological Review,
108(1), 204–256.
Derefeldt, G., Lennerstrand, G., & Lundh, B. (1979). Age variations in normal human
contrast sensitivity. Acta Ophthalmologica, 57(4), 679–690.
Dobres, J., Reimer, B., & Chahine, N. (2016). The effect of font weight and rendering
system on glance-based text legibility. Paper presented at the proceedings of the 8th
international conference on automotive user interfaces and interactive vehicular applica-
tions.
Dyson, M. C., & Beier, S. (2016). Investigating typographic differentiation: Italics are
more subtle than bold for emphasis. Information Design Journal, 22(1), 3–18.
Finkbeiner, M., & Coltheart, M. (2009). Letter recognition: From perception to re-
presentation. Cognitive Neuropsychology, 26(1), 1–6.
Fiset, D., Blais, C., Ethier-Majcher, C., Arguin, M., Bub, D., & Gosselin, F. (2008). Features
for identification of uppercase and lowercase letters. Psychological Science, 19(11),
1161–1168.
Hess, R. F., Dakin, S. C., & Kapoor, N. (2000). The foveal “crowding”effect: Physics or
physiology? Vision Research, 40(4), 365–370.
Kesten, H. (1958). Accelerated stochastic approximation. The Annals of Mathematical
Statistics, 29(1), 41–59.
Kuntz, J. E., & Sleight, R. B. (1950). Legibility of numerals: The optimal ratio of height to
width of stroke. The American Journal of Psychology, 63(4), 567–575.
Lanthier, S. N., Risko, E. F., Stolz, J. A., & Besner, D. (2009). Not all visual features are
created equal: Early processing n letter and word recognition. Psychonomic Bulletin &
Review, 16(1), 67–73.
Legge, G. E. (2006). Psychophysics of reading in normal and low vision. CRC Press.
Legge, G. E., & Bigelow, C. A. (2011). Does print size matter for reading? A review of
findings from vision science and typography. Journal of Vision, 11(5), 1–22.
Luckiesh, M., & Moss, F. K. (1940). Boldness as a factor in type-design and typography.
Journal of Applied Psychology, 24(2), 170–183.
Majaj, N. J., Pelli, D. G., Kurshan, P., & Palomares, M. (2002). The role of spatial fre-
quency channels in letter identification. Vision Research, 42(9), 1165–1184.
Mathôt, S., Schreij, D., & Theeuwes, J. (2012). OpenSesame: An open-source, graphical
experiment builder for the social sciences. Behavior Research Methods, 44(2),
314–324.
McClelland, J. L., & Rumelhart, D. E. (1981). An interactive activation model of context
effects in letter perception: I. an account of basic findings. Psychological Review, 88(5),
375–407.
Noordzij, G. (2005). The stroke. London: Hyphen Press.
Owsley, C., Sekuler, R., & Siemsen, D. (1983). Contrast sensitivity throughout adulthood.
Vision Research, 23(7), 689–699.
Pelli, D. G., Burns, C. W., Farell, B., & Moore-Page, D. C. (2006). Feature detection and
letter identification. Vision Research, 46(28), 4646–4674.
Perry, C., Ziegler, J. C., & Zorzi, M. (2014). When silent letters say more than a thousand
words: An implementation and evaluation of CDP++ in French. Journal of Memory
and Language, 72,98–115.
Petit, J.-P., & Grainger, J. (2002). Masked partial priming of letter perception. Visual
Cognition, 9(3), 337–353.
Pušnik, N., Podlesek, A., & Možina, K. (2016). Typeface comparison−does the x-height
of lower-case letters increased to the size of upper-case letters speed up recognition?
International Journal of Industrial Ergonomics, 54, 164–169.
Rosa, E., Perea, M., & Enneson, P. (2016). The role of letter features in visual-word re-
cognition: Evidence from a delayed segment technique. Acta Psychologica, 169,
133–142.
Sanocki, T., & Dyson, M. C. (2012). Letter processing and font information during
reading: Beyond distinctiveness, where vision meets design. Attention, Perception, &
Psychophysics, 74(1), 132–145.
Sheedy, J. E., Subbaram, M. V., Zimmerman, A. B., & Hayes, J. R. (2005). Text legibility
and the letter superiority effect. Human Factors: The Journal of the Human Factors and
Ergonomics Society, 47(4), 797–815.
Tinker, M. A. (1964). Legibility of print. Iowa State University Press.
Treutwein, B. (1995). Adaptive psychophysical procedures. Vision Research, 35(17),
2503–2522.
Wright, C. E., & Drasdo, N. (1985). The influence of age on the spatial and temporal
contrast sensitivity function. Documenta Ophthalmologica, 59(4), 385–395.
S. Beier and C.A.T. Oderkerk Acta Psychologica 199 (2019) 102904
8
