Upload
kan
View
218
Download
0
Embed Size (px)
Citation preview
8/17/2019 FullText (23)
1/90
TitleMotor experience modulates perceptual representation ofobjects: the case of Chinese characterrecognition
Author(s) Tso, Van-yip, Ricky.;ù •im
.
Citation
Issued Date 2012
URL http://hdl.handle.net/10722/173880
RightsThe author retains all proprietary rights, (such as patent rights)and the right to use in future works.
8/17/2019 FullText (23)
2/90
Motor Experience Modulates Perceptual Representation of Objects: The Case of
Chinese Character Recognition
By
Ricky Van-yip Tso
BSocSc (HKU)
A thesis submitted in partial fulfillment of the requirements for
the degree of Master of Philosophy
at the University of Hong Kong
September 2012
8/17/2019 FullText (23)
3/90
8/17/2019 FullText (23)
4/90
c
Abstract of thesis entitled
“Motor Experience Modulates Perceptual Representation of Objects: The Case of
Chinese Character Recognition”
Submitted by
Ricky Van-yip Tso
For the degree of Master of Philosophy
at the University of Hong Kong
in September 2012
Holistic processing and left-side bias are both behavioral markers of expert face
recognition. In contrast, expertise in Chinese character recognition involves left-side bias but
reduced holistic processing (Hsiao & Cottrell, 2009). Here I hypothesized that this reduction
in holistic processing may be related to writing rather than reading experience. In
Experiment 1, I tested Chinese literates who could read and write Chinese characters
fluently (Writers), and Chinese literates who had limited writing practices and thus had
reading performance far exceeding their writing ability (Limited-writers). I found that
Writers perceived Chinese characters less holistically than Limited-writers. In contrast to
what previous research suggested, reduction in holistic processing in Chinese readers
8/17/2019 FullText (23)
5/90
d
depended on writing experience instead of reading performance. In addition, reading
performance was affected by font familiarity and context for Limited-writers but not Writers.
Writing experience seems to enhance analytic processing and awareness of orthographic
components of Chinese characters, which may in turn facilitate reading in unfamiliar fonts.
By contrast, both Writers and Limited-writers showed a similar level of left-side bias in
processing symmetric Chinese characters, suggesting that left-side bias is a consistent
expertise marker for orthographic processing uninfluenced by writing experience.
In Experiment 2, I investigate the developmental trend of holistic processing in
Chinese character recognition and its relationship with reading and writing abilities by
testing Chinese children who were learning Chinese at a public elementary school in Hong
Kong on these abilities. I found that the holistic processing effect of Chinese characters in
children was reduced as they reached higher grades; this reduction was driven by enhanced
Chinese literacy rather than age. In addition, I found that writing performance predicts
reading performance through reduced holistic processing as a mediator. Overall, the results
of this study suggest that writing hones analytic processing, which is essential for expert
Chinese character recognition, and in turn facilitates learning to read in Chinese. This study
is also the first to identify Limited-writers as a window onto basic processes of reading.
8/17/2019 FullText (23)
6/90
e
Motor Experience Modulates Perceptual Representation of Objects: The Case of
Chinese Character Recognition
By
Ricky Van-yip Tso
BSocSc (HKU)
A thesis submitted in partial fulfillment of the requirements for
the degree of Master of Philosophy
at the University of Hong Kong
September 2012
8/17/2019 FullText (23)
7/90
i
Declaration
I declare that this thesis represents my own work, except where due
acknowledgement is made, and that it has not been previously included in a thesis,
dissertation or report submitted to this University or to any other institution for a
degree, diploma or other qualifications.
Signed …………………………………………………
Ricky Van-yip Tso
8/17/2019 FullText (23)
8/90
ii
Acknowledgements
I must express my deepest gratitude to my advisor, Dr. Janet Hsiao. Janet has been guiding
me through my ups and downs with the greatest warmth, patience and encouragement one
could ever imagine. She has become more like a friend to me. Being her student is the most
fortunately thing that have happened to me throughout my postgraduate years. I must also
thank Prof. Terry Au for co-supervising my project, and for her continuous guidance and
supports. I am extremely grateful to have such wonderful supervisors and to have the
opportunity to learn from their expertise and wisdom.
I thank my fellow labmates, Kloser Cheung, Fanny Lam, Tianyin Liu and Yetta Wong.
Their unconditional supports through tough times have brightened every moment of my days.
I must also express deep gratitude to all the helpers and friends who gave me a helping hand
for my intensive school data collection. I must not forget to thank all the participants, parents
and schools that took part in this study. Without them this project would never have been
feasible.
I would like to express my greatest gratitude to my family who genuinely loved me and
cared for me. Last but not least, I would like to thank Amy Chen for her love and support.
She has always accepted me and supported me in every single moment. Her infinite
encouragement has given me the sunshine I need to get my work done.
8/17/2019 FullText (23)
9/90
iii
Author Note
Ricky Van-yip TSO, Department of Psychology, University of Hong Kong.
In this thesis, experiments reported in Chapter 2 were presented in the 33rd annual meeting of
the Cognitive Science Society, Boston, Massachusetts, USA, and was included in the
Proceedings of the 33rd Annual Conference of the Cognitive Science Society. Experiment 1
in Chapter 3 was presented in the 34th annual meeting of the Cognitive Science Society,
Sapporo, Japan, and was included in the Proceedings of the 34th Annual Conference of the
Cognitive Science Society
Correspondence concerning this thesis should be addressed to Ricky Van-yip Tso,
Department of Psychology, University of Hong Kong, Pokfulam, Hong Kong.
8/17/2019 FullText (23)
10/90
iv
Contents
Declaration .................................................................................................................................. i
Acknowledgements .................................................................................................................... ii
Author Note .............................................................................................................................. iii
Contents .................................................................................................................................... iv
List of abbreviations ................................................................................................................. vi
List of Figures .......................................................................................................................... vii
List of Tables .............................................................................................................................. x
Chapter 1 Introduction............................................................................................................... 1
1.1. Chinese Character Perception ......................................................................................... 3
1.2. Writing and reading in Chinese....................................................................................... 6
Chapter 2 Study 1: The Effect of Writing Experience on Chinese Character Perception ......... 9
2.1. Participants .................................................................................................................... 11
2.2. Materials and Procedures .............................................................................................. 13
2.2.1. Reading and writing performances ......................................................................... 13
2.2.2. Holistic processing ................................................................................................. 16
2.2.3. Left-side bias .......................................................................................................... 19
2.3. Results ........................................................................................................................... 21
2.3.1. Chinese reading and writing proficiency................................................................ 21
2.3.2. Holistic Processing ................................................................................................. 23
2.3.3. Copying Task .......................................................................................................... 25
8/17/2019 FullText (23)
11/90
v
2.3.4. Left-side bias .......................................................................................................... 26
2.3.5. Effects of word-embedment and font familiarity on character-naming performance
................................................................................................................................ 28
2.4. Discussion ..................................................................................................................... 29
Chapter 3 Study 2: The Effect of Learning to Read and Write in Chinese on Chinese
Character perception — A Developmental Investigation .......................................................... 32
3.1. Experiment 1: Test for holistic processing of Chinese characters in Chinese speaking
children ................................................................................................................................ 34
3.1.1. Participants ............................................................................................................. 34
3.1.2. Materials and Procedures ....................................................................................... 34
3.1.3. Results .................................................................................................................... 38
3.2. Experiment 2: Test for holistic processing of Chinese characters in non-Chinese
speaking children ................................................................................................................. 46
3.2.1. Participants ............................................................................................................. 47
3.2.2. Materials and Procedures ....................................................................................... 47
3.2.3. Results .................................................................................................................... 47
3.3. Discussion ..................................................................................................................... 49
Chapter 4 General Discussions and Conclusions .................................................................... 52
4.1. General Discussions ...................................................................................................... 52
4.1.1. Holistic Processing ................................................................................................. 52
4.1.2. Left-side bias .......................................................................................................... 57
4.2. Conclusions ................................................................................................................... 61
References ................................................................................................................................ 63
8/17/2019 FullText (23)
12/90
vi
List of abbreviations
ANOVA
fMRI
LH
RH
NCS
Analysis of Variance
Functional Magnetic Resonance Imaging
Left Hemisphere
Right Hemisphere
Non-Chinese Speaking
8/17/2019 FullText (23)
13/90
vii
List of Figures
Figure 1. Examples of chimeric face stimuli (adopted from Hsiao & Cottrell, 2009). Two left
halves of an original face (middle) were combined to form the left chimeric face
(left), and the two right halves formed the right chimeric face (right). ..................... 3
Figure 2. An example of a Ming font (a) and a Feng font (b) character. ................................. 14
Figure 3. Illustration of stimulus pairs in the complete composite paradigm and trial
sequences. In (a), it shows the four conditions used in the paradigm; the attended
components are shaded in grey. In (b), a 1,000 ms central fixation cross precedes
each trial, followed by a cue either below or above the cross to indicate which
halves (top or bottom) of the characters participants should attend to in the
following display. .................................................................................................... 17
Figure 4. Examples of the stimuli used and the test sequence in the left-side bias experiment.
In (a), two left halves of an original character (middle) were combined to form the
left chimeric character (left), and the two right halves formed the right chimeric
character (right; note that the chimeric characters are still legal Chinese characters).
In (b), participants were presented with an original character either on the left or
right of the screen in each trial (shown on the left of the screen here) and were
instructed to judge which of the right or left chimeric characters (above and below
8/17/2019 FullText (23)
14/90
viii
the arrow) looked more similar to the original image. ............................................ 20
Figure 5. Accuracy rate of Limited-writers and Writers for the dictation and word naming
task (***p < 0.001). ................................................................................................. 22
Figure 6. Response time (a) and A’(b) of Limited-writers and Writers in congruent and
incongruent trials of the holistic processing task (**p < 0.01). .............................. 25
Figure 7. Copying response time for real Chinese characters, real Korean characters, and
pseudo-Chinese characters for Limited-writers and Writers (** p < 0.01). ............ 26
Figure 8. Preference for left chimeric characters in Writers and Limited-writers in Ming and
Feng fonts. ............................................................................................................... 27
Figure 9. Response time for naming characters and words in Ming and Feng fonts for Writers
(a) and Limited-writers (b) (* p < 0.05; ** p < 0.01). ............................................. 29
Figure 10. Predicted mediation effect of reduced holistic processing between Chinese writing
and reading performance. ........................................................................................ 33
Figure 11. Examples of Chinese characters with left-right configuration (left) and top-bottom
configuration (right). ............................................................................................... 37
Figure 12. A' of congruent and incongruent trials for first, third and fifth graders in the
holistic processing task (* p < 0.05; ** p < 0.01; *** p < 0.001). ............................. 41
Figure 13. Partial mediation effect of reduced holistic processing on dictation and word
naming performances (* p
8/17/2019 FullText (23)
15/90
ix
Figure 14. A' of congruent and incongruent trials for non-Chinese speaking first, third and
fifth graders in the holistic processing task ............................................................. 47
8/17/2019 FullText (23)
16/90
x
List of Tables
Table 1. Means of reading, writing and copying performance in first, third and fifth
graders……………………………………………………………………..………39
Table 2. Correlations Among Age, Rapid-naming speed, Character naming accuracy,
Character naming response time, Word naming accuracy, Word naming response
time, Copying response time, Dictation accuracy, and Holistic A'………………...42
Table 3. Hierarchical regression analysis among holistic processing and reading, dictation and
copying performance…………………………………………………………….…43
8/17/2019 FullText (23)
17/90
1
Chapter 1
Introduction
Holistic processing is the tendency to process separate features of an object as a single
whole unit (Richler, Wong, & Gauthier, 2011), and it is shown to be a behavioral marker of
face recognition expertise. One common paradigm to test for holistic processing in face
recognition is the Complete Composite Paradigm (see Richler, Wong, & Gauthier, 2011).
In each trial, participants are instructed to judge whether two corresponding halves of two
stimuli are the same or different. The irrelevant halves will give the same responses in
congruent trials, while the giving different responses in incongruent trials. Holistic
processing is measured by the performance difference between congruent and incongruent
trials. Another paradigm to test for holistic processing is the part-whole paradigm (see
Joseph & Tanaka, 2002; Tanaka & Farah, 1993; Tanaka et al ., 1998). In this design,
participants were first presented with a face sample; immediately after this presentation,
participants were tested either in the whole-face condition or the isolated-part condition. In
the whole-face condition, participants were presented with the sample face and a novel
face that differ from the sample face by only one feature (e.g. nose), and were asked to
judge which face appeared in the previous trial. Alternatively in the isolated-part
condition, participants were presented with an isolated feature from the sample face (e.g.
mouth) and a novel feature and were asked to judge which feature appeared in the sample
8/17/2019 FullText (23)
18/90
2
face. Participants would perform better in the whole-face condition if they processed faces
holistically. Some have speculated that holistic processing applies to other types of expert-
level object recognition because it facilitates within-category object discrimination by
incorporating featural and configural information beyond individual parts (Bukach,
Gauthier, & Tarr, 2006; Gauthier & Bukach, 2007; Wong, Palmeri, & Gauthier, 2009; but
for a contrasting view, see McKone, Kanwisher, & Duchaine, 2007). For example, training
participants to recognize novel artificial symmetric objects (“Greebles”), Gauthier and
colleagues (1998) found a positive correlation between holistic processing and expertise in
within-category object recognition. Consistently, Wong, Palmeri and Gauthier (2009)
showed that participants had an increase in holistic processing when trained to
individualize an artificial object type (“Ziggerins”).
Left-side bias is another phenomenon consistently reported in face perception; it
refers to the effect that a chimeric face made from two left half-faces is usually judged
more similar to the original face compared with one made from two right half-faces from
the viewer’s perspective (Brady, Campbell, & Flaherty, 2005; Gilbert & Bakan, 1973;
Figure 1). Consistent with this perceptual bias effect, in eye movement studies, viewers
also have a tendency to look at the left side of the face more often than the right side when
processing faces (Leonards & Scott-Samuel, 2005; Mertens, Siegmund, & Crusser, 1993).
The left-side bias effect may well be due to the involvement of the right hemisphere (RH)
8/17/2019 FullText (23)
19/90
3
in face recognition (Hsiao, Shieh, & Cottrell, 2008; Burt & Perrett, 1997).
Figure 1. Examples of chimeric face stimuli (adopted from Hsiao & Cottrell,
2009). Two left halves of an original face (middle) were combined to form the
left chimeric face (left), and the two right halves formed the right chimeric face
(right).
1.1. Chinese Character Perception
Chinese characters, with their many shared visual properties with faces, were
hypothesized to induce a similar processing effect in expert readers (Hsiao & Cottrell,
2009; McCleery et al ., 2008). More specifically, the Chinese writing system is
logographic; while words in most alphabetic languages are linear in structure and consist
of letter series of varying lengths, Chinese characters have a more homogenous, square
configuration, and each character is a grapheme that maps onto a morpheme (Shu, 2003;
Wong & Gauthier, 2006). The basic units of a Chinese character are strokes, which
8/17/2019 FullText (23)
20/90
4
combine to form more than a thousand different stroke patterns in the Chinese writing
system (Hsiao & Shillock, 2006); these stroke patterns in turn form the characters. A
typical literate recognizes 3,000 to 4,000 characters. In addition, Chinese characters are
generally recognized regardless of variations in font and handwriting style, similar to face
recognition regardless of differences in facial expressions (Hsiao & Cottrell, 2009), and
experts recognize Chinese characters individually like faces (Wong & Gauthier, 2006).
Indeed, as in the case of expertise in face recognition, Hsiao and Cottrell (2009)
showed that expert Chinese readers demonstrated left-side bias when viewing mirror-
symmetric Chinese characters, whereas novices did not. They found that expert Chinese
readers significantly chose the left-chimeric image of a Chinese character to be more
similar to the original image, and had an overall tendency to look at the left side of
characters. Hsiao and Cottrell’s (2009) finding suggests that left-side bias is an expertise
marker for Chinese character recognition and was consistent with research suggesting a
RH involvement in Chinese orthographic processing (e.g., Tzeng, Hung, Cotton, & Wang,
1979; Yang & Cheng, 1999). This is also consistent with ERP studies that showed more
RH lateralized or bilateral processing in the occipitotemporal region (Liu & Perfetti, 2003;
Hsiao, Shillcock, & Lee, 2007), in contrast to the LH lateralization classically observed in
English and alphabetic word processing (see Dahaene & Cohen, 2011). Similar findings
were shown in fMRI studies (e.g. Tan et al ., 2000; Tan et al., 2001; see also Tan, Laird, Li,
8/17/2019 FullText (23)
21/90
5
& Fox, 2005).
However, unlike face perception, the expertise marker for Chinese character
recognition turned out to be reduced holistic processing (Hsiao & Cottrell, 2009). Using
the Complete Composite Paradigm, Hsiao and Cottrell (2009) showed that experienced
Chinese readers engaged in less holistic processing than novices in perceiving Chinese
characters. Perhaps experienced Chinese readers are more sensitive to the constituent
components of Chinese characters and can more readily ignore some configural
information unimportant for character recognition, such as exact distances between
features (Ge, Wang, McGleery, & Lee, 2006). Such constituent components may not look
easily separable to novices, probably because novices are less able to distinguish
individual features and components in Chinese characters (Chen, Allport, & Marshall,
1996; Ho, Ng, & Ng, 2003; Hsiao & Cottrell, 2009). Hsiao and Cottrell (2009) have
therefore suggested that holistic processing is not a general expertise marker for object
processing; it depends on the features of the stimuli and the tasks typically performed on
the stimuli (see also Wong et al ., 2009).
Note however that the experience with learning to read Chinese characters is different
from typical face recognition in one apparent way — while a typical Chinese reader can
read and write characters proficiently, one is not expected to draw out all the familiar faces
seen every day. Thus, it is possible that the reduced holistic processing effect in expert
8/17/2019 FullText (23)
22/90
6
Chinese character processing, in contrast to expert face processing, is related to expert
readers’ writing rather than reading experience. Unlike writing alphabetic words, which
only requires recalling a few dozens of letters in an alphabet together with the specific
combinations corresponding to their sounds, writing Chinese characters requires retrieving
more than a thousand pieces of script information from long term memory. One may have
to attend analytically to detailed stroke patterns of individual Chinese characters in order
to memorize and write them. Perhaps expert Chinese readers in Hsiao and Cottrell’s
(2009) study had reduced holistic processing because they were concurrently experienced
character writers. Consistent with our speculation on the relationship between
writing/motor experience and holistic processing in expert Chinese character recognition,
Zhou, Cheng, Zhang, and Wong (2012) recently found that artists with face drawing
experiences had reduced holistic face processing compared with ordinary observers.
Never theless, Hsiao and Cottrell’s (2009) study offered a window on holistic processing
and left side bias in relation to expertise of complex object recognition by showing that
holistic processing may not be a general expertise marker and that it does not always
concur with RH lateralization.
1.2. Writing and reading in Chinese
In Hong Kong, although the internal structures of Chinese characters are not explicitly
emphasized in formal lessons, Chinese children acquire better orthographic awareness as
8/17/2019 FullText (23)
23/90
7
they progress to higher grades (Ho et al ., 2003). One explanation has to do with motor
programming through extensive copying and reading at school (Guan, Liu, Chan, Ye, &
Perfetti, 2011; Tan, Spinks, Eden, Perfetti, & Siok, 2005). Copying performance
significantly predicts reading ability (Chan, Ho, Tsang, Lee, & Chung, 2006; McBride-
Chang, Chung, & Tong, 2011; Tan, et al ., 2005), and dictation performance is correlated
with reading performance (McBride-Chang, Chung, et al ., 2011; Tse, Kwan, & Ho, 2010).
It has been proposed that writing performance can predict reading performance because
children may consolidate knowledge of orthographic structures of characters with
graphomotor memory of strokes as they copy the stroke sequences (Tan et al., 2005; Tse et
al., 2010). Consistent with this speculation, learning to write seems to strengthen Chinese
character recognition (Guan et al., 2011). Other research also suggested that writing
experience plays an important role in shaping the neural representation specialized for
reading (e.g. James & Atwood, 2009; Longcamp, Anton, Roth, & Velay, 2003). A neural
pathway linking the Broca’s area and the supplemental motor area was activated during
silent reading of Chinese pinyin (Romanized transcription of Mandarin pronunciation) in
an fMRI study (He et al., 2003). The left middle frontal gyrus, an area just anterior to the
premotor area, was activated in normal but not dyslexic Chinese readers when reading
(Siok, Perfetti, Jin, & Tan, 2004). These results consistently suggest a close relationship
between increasing sensory-motor integration through writing practice and the
8/17/2019 FullText (23)
24/90
8
development of reading skills.
This thesis aims at investigating the effect of enhanced Chinese literacy, particularly
writing experiences, on our perceptual representation of Chinese characters. In Chapter 2, I
will present a study that examines perceptual differences of Chinese characters between
Chinese readers who can read and write fluently and Chinese readers who can read
fluently but have limited writing performances (I will discuss this phenomenon in details
in the next chapter). I will then present my findings on the effect of enhanced Chinese
literacy on Chinese character perception in children in Chapter 3. Chapter 4 will discuss
the significance of the findings in this research and its educational implications.
8/17/2019 FullText (23)
25/90
9
Chapter 2
Study 1: The Effect of Writing Experience on Chinese Character Perception
Typically, fluent readers of a language can also write fluently; however, there exist some
Chinese readers who have high reading proficiency but far poorer writing ability – whom I
will call “Limited-writers”. They are usually students or graduates of international schools
who have learned to “write” in Chinese using computer software that converts input in a
phonic alphabet (e.g., the Pinyin system) into Chinese characters, expatriates living in
Chinese speaking countries, or overseas Chinese immigrants who learned to read in
Chinese from environmental prints including Chinese mass media. Writing Chinese
characters requires retrieving more than a thousand pieces of script information from long
term memory. Because writing in Chinese orthography is more complex and resource-
intensive than writing in an alphabetic language (Chan et al ., 2006; Chung & Ho, 2010;
Tse et al ., 2010), marked discrepancy between reading and writing performance in Chinese
is possible. With limited writing practice but plenty of reading experience, Limited-writers
may recognize the holistic structures of characters similarly to face recognition, with
limited analysis of the constituent structures. Thus, the cognitive processes involved in
Chinese reading for Limited-writers may be different from readers who have received
8/17/2019 FullText (23)
26/90
10
intensive character writing training (Writers). To date, virtually no research has recognized
such a Chinese community can be a window onto basic processes in reading.
In Study 1, I aim to examine whether these rather proficient Chinese readers who have
limited experience in writing Chinese characters (i.e., Limited-writers) process Chinese
characters differently from those who can both write by hand and read proficiently (i.e.,
Writers). I first investigate reading and writing performance differences between Writers
and Limited-writers. I then examine whether Writers perceive characters less holistically
than Limited-writers, and whether the reduced holistic processing effect is related to their
reading and writing performance. Since writing practice may enhance orthographic
awareness of characters and de-emphasize configural information in character recognition,
I predict that Writers will perceive characters less holistically than Limited-writers, and
this effect will be related to their difference in writing rather than reading performance –
contrary to what the research literature suggests. The ability to perceive characters
analytically (less holistically) may be the underlying mechanism for how writing
experience enhances Chinese character recognition.
We also examine whether Limited-writers as well as Writers have a similar left-side
bias effect in Chinese character perception. Brady et al . (2005) showed that the left-side
bias effect in face perception was stronger when viewing familiar faces compared with
unfamiliar faces; this phenomenon suggests that the left-side bias effect may be related to
8/17/2019 FullText (23)
27/90
11
familiarity with the stimuli. Since both Writers and Limited-writers are proficient readers
and thus are familiar with Chinese characters, I predict that Writers and Limited-writers
will have a similar degree of left-side bias in perceiving Chinese characters.
We also investigate the effect of word embedment of a character on its naming speed
and accuracy. Experienced Chinese readers can name characters embedded in a high
frequency word faster than when the characters were presented alone (e.g., Cheng, 1981,
Chu & Leung, 2005; Zhang & Peng, 1992). If Limited-writers indeed acquire Chinese
characters from the context where characters are seldom read alone, this word embedment
effect should be more prominent in Limited-writers than Writers who have extensively
practiced writing individual characters. In addition, I explore whether differences in
writing experience between these two groups would lead to performance differences in
reading different fonts. Limited-writers are usually exposed to printed rather than
handwritten Chinese characters and often find reading hand-written characters difficult.
Writers, by contrast, are more experienced in deciphering different handwritings and
perhaps as a result are less affected by font variations in reading printed Chinese words.
Thus, here I also examine the effect of font familiarity on character and word naming
speed.
2.1. Participants
40 Cantonese native-speaking Chinese readers (17 males and 23 females) in Hong Kong
8/17/2019 FullText (23)
28/90
12
participated in our study. They had similar college-level education background. Half of
them had always attended conventional local schools and reported to have fluent reading
and writing proficiency (i.e., Writers), whereas the other half had either studied overseas or
at international schools and had not received formal Chinese lessons that prepared students
for the local public Chinese examinations (i.e., Limited-writers). All Limited-writers
reported having studied either at international schools or overseas and had little Chinese
writing experiences. They all reported having fluent reading ability. Writers’ and Limited-
writers’ reading and writing abilities were further tested by a word -naming task and a
dictation task respectively (which will be described in 2.2.1.). We differentiated our
Limited-writers from Writers by their performances in the word-naming and dictation task,
i.e., Limited-writers were expected to have the same performance level in the word-
naming task as Writers, but have poorer performance in the dictation task (see 2.3.1.). The
average age of Limited-writers was 21.45 (S.E. = .57) and Writers was 21.47 (SE = .50).
The two groups did not differ in age, F(1, 38) = .001, n.s. They all had normal or
corrected-to-normal vision.
8/17/2019 FullText (23)
29/90
13
2.2. Materials and Procedures
2.2.1. Reading and writing performances
Four tests were administered to test for reading and writing performances: 1. Character
naming task, 2. Word naming task, 3. Character copying task, and 4. Word dictation task.
Tasks 1 and 2 assessed participant’s reading ability, while Tasks 3 and 4 assessed their
copying and word recalling/writing ability respectively. Tasks 2 and 4 were compared to
examine the discrepancy between word naming and word recalling/writing performances.
1. Character naming task:
Participants were presented with 84 medium to high frequency Chinese characters of
similar visual complexity one at a time and asked to read aloud as quickly and as
accurately as possible. The characters had an average frequency of 443.3 per 663,424
characters (SE = 45.4) and stroke number of 10.9 (SE = .16; Ho & Kwan, 2001). Half of
the stimuli were presented in Ming font (a common font in print) and the other half were
presented in Feng font (an unfamiliar font that simulates handwriting; Figure 2). Each
character was approximately 1.5 x 1.5 cm2 in size and each participant had a viewing
distance of 55 cm. Each character spanned about 1.6 degree of visual angle. Each trial
started with a central fixation cross for 500 ms, followed by the character presentation.
After a participant had responded, the screen would turn blank and the experimenter would
press a button to record the accuracy and to start the next trial. The response time was
8/17/2019 FullText (23)
30/90
14
measured as the time difference between the stimulus onset and the onset of the
pronunciation, detected by a microphone.
Figure 2. An example of a Ming font (a) and a Feng font (b) character.
2. Word naming task:
Participants read aloud 40 medium to high frequency two-character words (average
frequency = 194.7 per 530,452 words, SE = 19.6; Taiwan Ministry of Education, 1997) as
quickly and accurately as possible. Half of the stimuli were presented in Ming font and the
other half in Feng font. Each word was approximately 1.5 x 3.0 cm2 in size and each
participant had a viewing distance of 55 cm. Each word spanned about 3.1 degree of visual
angle. The procedure was the same as that of Task 1. The response time was measured as
the time difference between the stimulus onset and the onset of the pronunciation of the
first character.
3. Character copying task:
Participants copied 60 characters (20 real characters, 20 pseudo-characters, and 20 Korean
characters) as quickly and as accurately as possible. The Chinese characters were high
frequency characters that were randomly selected from the characters used in Task 1 with
an average stroke number of 10.7 (SE = .48). The pseudo-characters were orthographically
8/17/2019 FullText (23)
31/90
15
legal but non-sense characters. Each trial started with a central fixation cross for 500 ms,
followed by the character presentation. Each character was approximately 1.5 x 1.5 cm 2 in
size and each participant had a viewing distance of 55 cm. Each character spanned about
1.6 degree of visual angle. I used Korean and pseudo-characters to see whether
participants’ performance in copying Chinese characters can be generalized to copying
novel but configurally similar characters (e.g., top-bottom, left-right, top heavy, bottom
heavy). After copying a character, a participant would press a button to make the screen
turn blank and start the next trial. The response time was recorded.
4. Word dictation task:
Participants wrote down 40 two-character words as quickly and as accurately as possible
when they heard from a computer each word read aloud by a female native-speaker of
Cantonese (Words instead of single characters were used here to reduce ambiguity due to
the many homophonic characters in the Chinese lexicon). The words were the same as
those used in Task 2. Each trial started with the words “Get ready” on the screen for
500ms. After hearing the word, participants pressed corresponding buttons to indicate
whether they could recall the word or not, before they started writing. After they finished
writing, the experimenter would press a button to record accuracy and to reveal the next
word.
8/17/2019 FullText (23)
32/90
16
2.2.2. Holistic processing
To examine holistic processing effects, procedures were adopted from Hsiao and Cottrell
(2009). 80 pairs of medium to high frequency Chinese characters in Ming font with a top-
bottom configuration as adopted by Hsiao and Cottrell (2009) were chosen. Participants
were asked to attend to only half (either top or bottom) of each character in a pair on each
trial and make a same-different judgment. Twenty pairs were presented in each of the four
conditions (Figure 3a): same in congruent trials, different in congruent trials, same in
incongruent trials, and different in congruent trials. A complete composite paradigm
(Gauthier & Bukach, 2007) was adopted; in congruent trials, the attended and irrelevant
halves of the characters led to the same response (i.e., both were the same or different),
while in incongruent trials, the attended and irrelevant halves led to different responses 1. If
a stimulus was processed holistically, there should be interference from the irrelevant
halves in matching the attended halves in incongruent trials; holistic processing thus could
be assessed as the performance difference between incongruent and congruent trials. I
adopted this design to avoid influence from response biases that may occur in the partial
composite design, in which the irrelevant halves are always different (see Gauthier &
Bukach, 2007; Robbins & McKone, 2007). In addition, holistic face processing measured
1 Unlike the partial composite paradigm, in which a misaligned condition must be administered (i.e. the top
halves and bottom halves of the stimuli are misaligned from each other; e.g., Hole, 1994; Robbins &
McKone, 2007), holistic processing can be indicated by the performance difference between the congruent
and incongruent trials using the complete composite design without a misaligned condition (Gauthier &
Bukach, 2007; Hsiao & Cottrell, 2009).
8/17/2019 FullText (23)
33/90
17
with the complete composite paradigm was positively correlated with face recognition
performance; in contrast, this relationship was obscured when using the partial composite
design (Richler, Wong, & Gauthier, 2011; see also Konar, Bennett, & Sekuler, 2010).
Figure 3. Illustration of stimulus pairs in the complete composite paradigm and
trial sequences. In (a), it shows the four conditions used in the paradigm; the
attended components are shaded in grey. In (b), a 1,000 ms central fixation
cross precedes each trial, followed by a cue either below or above the cross to
indicate which halves (top or bottom) of the characters participants should
attend to in the following display.
8/17/2019 FullText (23)
34/90
18
Each character was approximately 1.5 cm x 1.5 cm in size and each participant had a
viewing distance of 55 cm. Each character spanned about 1.6 degree of visual angle.
During the experiment, the stimuli presented on the screen were of relatively low contrasts
to avoid ceiling effects. In each trial, after 1,000 ms of central fixation, participants were
cued with a symbol that indicated which half (top or bottom) of each character they should
attend to. The pair of characters was then presented, with one above and one below the
initial fixation, 5.7 degree of visual angle away from each other. During the 500 ms
presentation time, participants looked at each character once and responded as quickly and
accurately as possible, pressing corresponding buttons to judge if the character parts were
the same or different (Figure 3b). Accuracy and reaction time were collected. I measured
participants’ discrimination sensitivity A' as:
Where H and F are the hit and false alarm rate, respectively. A' is a bias-free
nonparametric measure of sensitivity; d' was not used here because response biases may
affect its measurement when assumptions of normality and homogeneity of variance are
not met (Stanislaw & Todorov, 1999). The A’ or response time difference between
incongruent and congruent trials measures holistic processing — a larger difference
indicates stronger holistic processing.
8/17/2019 FullText (23)
35/90
19
2.2.3. Left-side bias
To compare the left-side bias effect between Writers and Limited-writers, I adopted the
procedure from Hsiao and Cottrell (2009). 80 Chinese mirror-symmetric characters of high
frequency were selected (average frequency = 385.0 per 530,452 words, SE = 67.8; Ho &
Kwan, 2001). There were a total of 160 trials with each character presented twice in Ming
font and Feng font respectively. For characters presented in each font, mirror images were
used in half of the trials. If a character was presented in Ming font, then the mirror image
of the original character was presented in Feng font, and vice versa; this was to
counterbalance any differences between the two sides of each character. For each character
image, one chimeric character was created from two left halves (left chimeric character)
and another one from two right halves of the character (right chimeric character; Figure
4a), similar to chimeric faces.
Each character was approximately 6 x 6 cm2 in size and each participant had a
viewing distance of 55 cm, under this viewing distance, each character spanned about 6.7
degree of visual angle. In each trial, after 1,000 ms of a central fixation, the original
character was presented randomly either on the left or the right side of the computer
screen, at about 7.2 degree of visual angle away from the center. The left and right
chimeric characters were presented along with the original image, with one above and one
below an arrow at the center; the arrow directed the location of the original character at
8/17/2019 FullText (23)
36/90
20
which participants were told to look first. Each character was about 3 degree of visual
angle away from the center. The stimuli stayed on the screen until participants made a
response. Participants judged which of the two chimeric characters looked more similar to
the original one by pressing the corresponding buttons (Figure 4b). I measured the left-side
bias effect as the percentage of trials in which the left chimeric character was selected.
These experiments were all conducted using E-prime v2.0 (Psychology Software
Tools, Pittsburgh, PA).
Figure 4. Examples of the stimuli used and the test sequence in the left-side bias
experiment. In (a), two left halves of an original character (middle) were
combined to form the left chimeric character (left), and the two right halves
formed the right chimeric character (right; note that the chimeric characters are
8/17/2019 FullText (23)
37/90
21
still legal Chinese characters). In (b), participants were presented with an
original character either on the left or right of the screen in each trial (shown on
the left of the screen here) and were instructed to judge which of the right or left
chimeric characters (above and below the arrow) looked more similar to the
original image.
2.3. Results
2.3.1. Chinese reading and writing proficiency
ANOVA was used for the analyses. For the reading performance, the variables were font
(Ming vs. Feng fonts) and group (Writers vs. Limited-writers). The results showed that
Writers and Limited-writers did not differ in word naming accuracy, F (1, 38) = .471, n.s.,
η p 2 = .02 ( M = .98, SE = .006, and M = .98, SE = .007 respectively), suggesting that both
groups had high reading proficiency for high frequency words. Nevertheless, Writers had
significantly shorter response times in word naming ( M = 450 ms, SE = 13) than Limited-
writers ( M = 593 ms, SE = 36), F (1, 38) = 12.365, p < .01, η p 2 = .26. In character naming,
Writers outperformed Limited-writers in both accuracy, F (1, 38) = 4.806, p = .05, η p 2 =
.097 ( M = .98, SE = .004, and M = .97, SE = .007 respectively), and response time, F (1,
38) = 14.45, p < 0.01, η p 2 = .315 ( M = 478 ms, SE = 16, and M = 682 ms, SE = 32
respectively. I will report the font effects in reading performance in a later section). As for
8/17/2019 FullText (23)
38/90
22
writing performance, Writers had shorter response times overall in character copying ( M =
3881 ms, SE = 173) than Limited-writers ( M = 5601 ms, SE = 350), F (1, 38) = 15.39, p <
.01, η p 2 = .347. In the dictation task, Writers were significantly more accurate ( M = .86,
SE = .01) than Limited-writers ( M = .35, SE = .04), F (1, 38) = 140.53, p < .001, η p 2 = 604.
Figure 5 contrasts the discrepancy between dictation (word writing) and word naming
accuracy in Writers and Limited-writers (i.e., they had similar word reading accuracy but
differed in dictation/writing accuracy)2.
Figure 5. Accuracy rate of Limited-writers and Writers for the dictation and
word naming task (***p < 0.001).
2 We did not report accuracy for the copying task as it required additional inter-rating efforts to judge
objectively the accuracy of the stimuli copied by the adults. We also did not report the response time for
the dictation task as it is difficult to consider both the recalling time and writing time of the words.
8/17/2019 FullText (23)
39/90
23
2.3.2. Holistic Processing
Repeated-measures ANOVA was used to investigate holistic processing effects
(condition: congruent vs. incongruent trials x group: Writers vs. Limited-writers). For
response time, I found a main effect of condition, F (1, 38) = 26.14, p < .001, η p 2 = .408,
and an interaction between condition and group, F (1, 38) = 15.6, p < .001, η p 2 = = .291,
but no main effect of group, F (1, 38) = 1.20, n.s., η p 2 = .031. Limited-writers responded
significantly more slowly in incongruent trials than in congruent trials, t (19) = 5.489, p <
.001, d = 1.30, while Writers recorded similar response times in congruent and incongruent
trials, t (19) = 0.875, n.s. , d = .21. Similarly, on A’, I found a main effect of condition, F (1,
38) = 10.66, p < .01, η p 2 = .219, and an interaction between condition and group, F (1, 38)
= 8.02, p < .01, η p 2 = .174, but no main effect of group, F (39) = .434, n.s., η p
2 = .011.
Limited-writers had a significantly smaller A’ in incongruent trials than in congruent trials
t (19) = 3.592, p < .01,, d = .84, while this difference was not significant for Writers, t (19)
= 0.390, n.s., d = .08. These results suggest that Writers perceived Chinese characters less
holistically than Limited-writers3 (Figure 6).
Since Writers and Limited Writers differed in some reading performance measures in
3 Since our two participant groups were all typically developed adults in similar ages, their performance
difference in the incongruent trials was unlikely to be due to difference in inhibition control. In addition,
Hsiao and Cottrell (2009) showed that the holistic processing effect observed in Chinese character
processing disappeared when character halves were misaligned, suggesting that the effect was due to
inability to selectively attend to character halves when they were aligned.
8/17/2019 FullText (23)
40/90
24
addition to their difference in writing experience, to examine whether their difference in
holistic processing was due to their difference in writing or reading abilities, I analyzed the
holistic processing effect in response time with their reading and writing performance
measures put as covariates (ANCOVA). The difference in holistic processing between
Writers and Limited Writers was still significant even when character naming response
time, F (1, 38) = 6.978, p
8/17/2019 FullText (23)
41/90
25
Figure 6. Response time (a) and A’(b) of Limited-writers and Writers in
congruent and incongruent trials of the holistic processing task (**p < 0.01).
2.3.3. Copying Task
Repeated-measures ANOVA was used for the analysis on copying RT (group x
character type: real Chinese vs. real Korean vs. pseudo-Chinese). I found a main effect of
character type, F (2, 37) = 79.68, p < .001, η p 2 = .172, and a main effect of group, F (1, 38)
= 19.73, p < .001, η p 2 = .342, but no interaction between group and character type, F (2,
37) = 1.94, n.s. Real characters were copied faster than Korean characters, t (39) = 9.736, p
< .001, d = 1.53, and pseudo-characters, t (39) = 11.985, p < .001, d = .45, while pseudo-
characters were copied faster than Korean characters, t (39) = 2.827, p < .01, d = 1.89.
When I investigated copying performance for different character types separately, Writers
copied significantly faster than Limited-writers for Chinese characters, F (1, 38) = 15.39, p
= .003, η p 2 = .359, Korean characters, F (1, 38) = 10.257, p = .003, η p
2 = .289, and
pseudo-characters, F (1, 38) = 14.706, p = .001, η p 2 = .332 (Figure 7). These results
8/17/2019 FullText (23)
42/90
26
suggest that writers are able to generalize their Chinese character copying ability to
copying novel characters that have similar structures.
Figure 7. Copying response time for real Chinese characters, real Korean
characters, and pseudo-Chinese characters for Limited-writers and Writers (** p
< 0.01).
2.3.4. Left-side bias
Repeated-measures ANOVA (font x group) was used for the analysis on the left-side bias
effect. I found a significant main effect of font, F (1, 38) = 11.558, p < .01, η p 2 = .233:
participants had a stronger preference for left chimeric characters when the characters
were presented in Ming font than when they were presented in Feng font. There was no
main effect of group, F (1, 38) = .470, n.s., or an interaction between group and font, F (1,
8/17/2019 FullText (23)
43/90
27
38) = .024, n.s.,. This showed that both Writers and Limited-writers had a similar degree
of left-side bias in perceiving Chinese characters in either font.
To further understand this phenomenon, I conducted post-hoc t-tests for Writers and
Limited-writers separately. I found a significant preference for left chimeric characters in
Ming font in both Writers (.54), t (19) = 2.378, p < .05, d = .53, and Limited-writers (.56),
t (19) = 2.271, p < .05, d = .51, whereas the preference for left chimeric characters in Feng
font was not significant in either Writers (.48, n.s.,) or Limited-writers (.49, n.s.) This
result suggested that participants exhibited left-side bias for Chinese characters only in a
familiar font (Ming) but not in an unfamiliar font (Feng; Figure 8).
Figure 8. Preference for left chimeric characters in Writers and Limited-writers
in Ming and Feng fonts.
8/17/2019 FullText (23)
44/90
28
2.3.5. Effects of word-embedment and font familiarity on character-naming performance
Repeated-measures ANOVA (font x word-embedment: word vs. character x group)
was used for the analysis on naming response time. I found a main effect of font, F (1, 18)
= 37, p < .001, η p 2 = .242, a main effect of group, F (1, 38) = 18.79, p < .001, η p
2 = .323,
and a main effect of word-embedment, F (1, 38) = 4.486, p < .05, η p 2 = .106. I also found
an interaction between font and group, F (1, 38) = 6.072, p < .05, η p 2 = .139, between font
and word-embedment, F (1, 38) = 5.933, p < .05, η p 2 = .135, and among font, word-
embedment, and group, F (1, 38) = 4.293, p = .05, η p 2 = .093. These effects were not
observed in the accuracy data. To further understand these interactions, I analyzed data for
Writers and Limited-writers separately.
For Limited-writers, the results revealed a main effect of font, F (1, 19) = 8.686, p <
.01, η p 2 = .341. Although no main effect was found for word-embedment (word vs.
character), F (1, 19) = 1.484, n.s., an interaction between font and word-embedment was
found, F (1, 19) = 4.962, p < .05, η p 2 = .211. Post-doc t-tests found that Limited-writers
named words faster than characters in Ming font, t (19) = 3.802, p < .05, d = .87, but this
difference was not found in Feng font, t (19) = .664, n.s. Limited-writers named Ming font
words faster than Feng font words, t (19) = 2.684, p < .05, d = .60, but their naming speed
between Ming font and Feng font characters did not differ, t (19) = .437, n.s.. For Writers,
the results revealed neither an interaction between word-embedment and font, F (1, 19) =
8/17/2019 FullText (23)
45/90
29
1.174, n.s., nor a main effect of font, F (1, 19) = 2.574, n.s., although a main effect of
word-embedment was found, F (1, 19) = 12.846, p < .01, η p 2 = .212. Post-doc t-tests found
that Writers named words faster than characters regardless of whether words were in Ming
font, t (19) = 3.427, p < .01, d = .53, or Feng font, t (19) = 2.460, p < .05, d = .45. As shown
in Figure 9, these data suggest that Limited Writers’ character recognition depends on
familiarity with both font and context (word-embedment), whereas Writers can generalize
their recognition ability to unfamiliar fonts.
Figure 9. Response time for naming characters and words in Ming and Feng
fonts for Writers (a) and Limited-writers (b) (* p < 0.05; ** p < 0.01).
2.4. Discussion
The results from Study 1 suggest that Limited-writers processed Chinese characters more
holistically than Writers. ANCOVA showed that this effect could be strongly accounted for
by their performance difference in the dictation (word writing) task, but importantly could
8/17/2019 FullText (23)
46/90
30
not be as strongly accounted for by their differences in any of the reading performance
measures, including character and word naming accuracy and response times. These
results suggest that the holistic processing effect in Chinese character recognition depends
mainly on writing experience, or more specifically, the ability to recall and write down
Chinese characters rather than reading performance. Perhaps writing experience enhances
the ability to analyze the orthographic structures and components of Chinese characters,
which leads to reduced holistic processing. Complementing Hsiao and Cottrell’s (2009)
study, here I showed that the reduced holistic processing effect observed in expert Chinese
readers uniquely depended on writing experiences, or more specifically, the ability to
recall and write Chinese characters.
Both Writers and Limited writers had similar accuracy in the word naming task.
Though the dictation task and the word-naming task used the same words, Writers
outperformed Limited in the dictation task. In addition, Writers also outperformed
Limited-writers in the copying task that involved Korean characters (Hangul), each of
which typically consists of several phonetic symbols arranged in configurations
commonly found in Chinese characters. The Writers' ability to analyze Chinese characters
may have generalized to Korean characters, thereby facilitating Korean character copying.
Study 1 also showed that Limited-writers recognized a character embedded in a word
of Ming font faster than when it is alone or of Feng font, while Writers named two-
8/17/2019 FullText (23)
47/90
31
character words faster than single characters regardless of font type (Figure 7). It suggests
both context (word-embedment) and font familiarity helps Limited-writers recognize
Chinese characters.
Consistent with Hsiao and Cottrell’s (2009) findings, our experiment on left-side
bias in Chinese characters showed that both Writers and Limited-writers — both are expert
Chinese readers — have a similar preference for left chimeric characters. This left-side bias
was only found for characters in Ming font but not Feng font. It seems that left-side bias is
a consistent expertise marker for Chinese character recognition and is uninfluenced by
writing experience.
8/17/2019 FullText (23)
48/90
32
Chapter 3
Study 2: The Effect of Learning to Read and Write in Chinese on Chinese
Character perception — A Developmental Investigation
Although previous studies have suggested a close relationship between Chinese writing
and reading performance in children, the underlying mechanism remains unclear (e.g.
Guan et al ., 2011; McBride-Chang et al ., 2011)5. Study 1 suggests that writing experience
reduces HP of Chinese characters, which marks expert-level character recognition; here I
hypothesize that writing experience enhances character recognition performance by
modulating the perceptual system, allowing readers to identify Chinese characters more
analytically. The modulating effect of writing experience on our perception of Chinese
characters has never been studied before in Children who are learning to read and write
Chinese characters. Thus, I investigate whether children in upper grades perceived
characters less holistically than children in lower grades in an elementary school where the
Chinese language is taught. I also examined their Chinese reading and writing
performance to see what can predict Children’s reduced HP of Chinese characters. I
predict that upper-grade children (who should have better Chinese literacy than lower-
grade children) will process Chinese characters more analytically (i.e., less holistically)
5 Although some studies have proposed that writing enhances graphomotor memory of Chinese characters,
which in turn facilitates reading (see e.g. Tan et al ., 2005; Tse et al ., 2010). Yet, this hypothesis has not been
statistically tested.
8/17/2019 FullText (23)
49/90
33
than children in lower grades. I hypothesize that children’s reduction in HP can be
predicted by writing performances across grades. Since writing performance also strongly
correlates with reading abilities in Chinese (Guan et al ., 2011; McBride-Chang et al .,
2011; Tan et al ., 2005), I hypothesize that HP mediates between Chinese reading and
writing abilities in children. More specifically, I predict that writing experience leads to
reduced holistic processing in Chinese characters, which in turn enhances reading abilities
in Chinese (Figure 10).
Figure 10. Predicted mediation effect of reduced holistic processing between
Chinese writing and reading performance.
The results from Study 1 also suggest that writing experience may facilitate reading
Chinese characters in an unfamiliar font, as Limited-writers had difficulty reading words
in the Feng font (a font that mimics handwriting and was unfamiliar to the participants)
whereas this effect in Writers was minimal. Hence here I also examine the possible effect
of enhanced Chinese writing proficiency on naming characters and words in familiar and
unfamiliar fonts.
8/17/2019 FullText (23)
50/90
34
3.1. Experiment 1: Test for holistic processing of Chinese characters in
Chinese speaking children
Experiment 1 tested for HP and left-side bias of Chinese characters in children from first,
third, and fifth grades. They were all learning to read and write in Chinese at school and
were all tested for reading and writing performances in Chinese. This experiment aims at
testing the effect of enhanced Chinese literacy on HP and left-side bias of Chinese
characters as children progress to higher grades.
3.1.1. Participants
56 first grade (mean age = 5.88 years, SE = .051), 73 third grade (mean age = 7.90, SE =
.056), and 88 fifth grade (mean age = 9.89, SE = .047) Chinese children from an
elementary in Hong Kong participated in our study. They were all Cantonese native-
speaking and were all receiving regular Chinese language curriculum at school. All of
them had normal or corrected-to-normal vision.
3.1.2. Materials and Procedures
3.1.2.1. Reading and writing performances
Four tests were administered: 1. Character naming task, 2. Word naming task, 3. Character
copying task, and 4. Word dictation task. Tasks 1 and 2 assessed participant’s reading
ability, while Tasks 3 and 4 assessed their copying and word recalling/writing ability
8/17/2019 FullText (23)
51/90
35
respectively6.
1. Character naming task:
Children were presented with 60 Chinese characters one at a time. Half of the stimuli were
presented in Ming font and the other half were presented in Feng font. The fonts used for
each character were counterbalanced across participants. They were instructed to read
aloud the characters as quickly and as accurately as possible. The characters had an
average stroke number of 9.08 (SE = .10) and were arranged from high to low frequency
(frequency information for primary students was obtained from Leung & Lee, 2001). The
trials stopped after 5 consecutive errors made7. The procedures and paradigm were
identical to the character naming task in Study 1.
2. Word naming task:
Children read aloud 30 two-character words arranged from high to low frequency
(frequency information for primary students was obtained from Leung & Lee, 2001) as
quickly and accurately as possible. Half of the stimuli were presented in Ming font and the
other half were presented in Feng font. The fonts used for each word were
counterbalanced across participants. The procedures and paradigm were identical to the
word naming task in Study 1.
3. Character copying task:
6 Because of the literacy level difference between adults and children, stimuli used for testing Chinese
proficiency in Study 1 and Study 2 were obtained from different corpuses.7 Similar experimental procedures have been used in Yeung and colleagues, 2011
8/17/2019 FullText (23)
52/90
36
Children copied 30 characters (10 real characters, 10 pseudo-characters, and 10 Korean
characters) as quickly and as accurately as possible. The Chinese characters were
randomly selected from the characters used in task 1. The procedures and paradigm were
identical to the character copying task in Study 1
4. Word dictation task:
Children wrote down 30 two-character words (the same words used in task 2) as quickly
and as accurately as possible when they heard each word said in a female voice presented
by a computer. The procedures and paradigm were identical to the word dictation task in
Study 1.
3.1.2.2. Test for rapid automatized naming
Because general symbol processing speed is related to reading ability in children (Ho &
Lai, 1999; Wolf & Bowers, 1999), we adopted a computerized rapid automatized naming
(RAN) task to measure this component in children (see Denckla & Rudel, 1976). The
digits 1, 2, 3, 4, 5, 6, 7, 8, and 9 were presented one at a time for 4 times on the computer
screen. Each trial started with a central fixation cross for 500ms, followed by the digit
presentation. The screen turned blank after a child had responded and the experimenter
pressed a button to record the accuracy and to start the next trial. Their response time was
measured as the time difference between the stimulus onset and the onset of the
pronunciation, detected by a microphone.
8/17/2019 FullText (23)
53/90
37
3.1.2.3. Test for holistic processing
To test for HP effects, I adopted the same procedures from Study 1. Here 160 pairs of
medium to high frequency Chinese characters in Ming font were chosen; 80 pairs had a
top-bottom configuration and 80 pairs had a left-right configuration (Figure 11). The
frequency information of the Chinese characters was obtained from Ho and Kwan (2001).
The top-down and left-right characters were matched in stroke number and frequency. In
each trial, children were presented with two characters and instructed to attend to only half
(either top or bottom for top-bottom characters, or left or right for left-right characters) of
each character and judge whether they were the same or different. Forty pairs were
presented in each of the four conditions (Figure 3a): same in congruent trials, different in
congruent trials, same in incongruent trials, and different in incongruent trials.
Figure 11. Examples of Chinese characters with left-right configuration (left) and
top-bottom configuration (right).
3.1.2.4. Left-side bias
We adopted the same procedures from Study 1 to compare the left-side bias effect of
Chinese characters between children from first, third and fifth grades. Here 40 Chinese
8/17/2019 FullText (23)
54/90
38
mirror-symmetric characters of high frequency were selected (average frequency = 385.0
per 530,452 words, SE = 67.8; Ho & Kwan, 2001). There were a total of 80 trials with
each character presented twice in Ming font and Feng font respectively.
These experiments were all conducted using E-prime v2.0 (Psychology Software
Tools, Pittsburgh, PA).
3.1.3. Results
3.1.3.1. Chinese reading and writing proficiency
Pearson’s correlation regression analysis showed a positive correlation between grade and
character-naming accuracy (r2 = .602, p < .001), word-naming accuracy (r2 = .512, p <
.001) and dictation accuracy (r2 = .707, p < .001); and a negative correlation between
grade and character-naming response time (r2 = .269, p < .001), word-naming response
time (r2 = .347, p < .001) and character-copying response time (r2 = .620, p < .001). These
results suggest that children had better Chinese reading and writing proficiency as they
reached higher grades. Table 1 summarizes the means of different Chinese proficiency
measures in each grade.
8/17/2019 FullText (23)
55/90
39
Table 1. Means of reading, writing and copying performance in first, third and fifth
graders
Grade 1 Grade 3 Grade 5Mean (SD)
Character-naming
accuracy
.38 (.23) .80 (.18) .93 (.06)
Character-naming
Response Time (ms)
1237 (506) 850 (271) 714 (217)
Word-naming
accuracy
.43 (.29) .87 (.16) .96 (.06
Word-naming
Response Time (ms)
1101 (429) 720 (162) 603 (129)
Copying RT (ms) 12162 (2289) 6618 (1481) 5727 (1189)
Dictation Accuracy .11 (.07) .42 (.20) .72 (.17)
3.1.3.2 Holistic Processing
Repeated-measures ANOVA was used to investigate HP effects (congruency: congruent
vs. incongruent trials x grade: Grade 1 vs. Grade 3 vs. Grade 5). I found a significant
effect of grade, F(2, 213) = 25.090, p < 0.001, η p 2 = .2, a significant effect of congruency,
F(1, 213) = 268.319, p < 0.001, η p 2 = .56, and an interaction between congruency and
grade F(1, 213) = 11.376, p < 0.001, η p 2 = .10. The main effect of congruency suggests
that across grades, children process Chinese characters holistically. While the main effect
of grade showed that the performance level increased with grade, the interaction between
grade and congruency suggests that children processed Chinese characters with varying
levels of congruency effect across grades (Figure 12).
Pairwise post-hoc t-tests showed that A' in congruent trials was larger than in
8/17/2019 FullText (23)
56/90
40
incongruent trials in first, t(55) = 12.01, p < .001, d = 1.60, third, t(72) = 7.838, p < .001,
d = .92, and fifth graders t(86) = 8.439, p < .001, d = .91. In congruent trials, first graders
had a smaller A' than third, t(127) = 3.226, p < .01, d = .57, and fifth graders, t(141) =
4.576, p < .001, d = .77, while third and fifth graders did not differ statistically in A',
t(158) = .994, n.s. In incongruent trials, first graders had a smaller A' than third, t(127) =
3.991, p < .001, d = .71, and fifth graders, t(141) = 7.878, p < .001, d = 1.33, and third
graders had a smaller A' than fifth graders, t(158) = 2.450, p < .05, d = .40. I also
conducted pairwise post-hoc t-tests on the A' difference between incongruent and
congruent trials (i.e., Holistic A') between children in the 3 grades. I found that first
graders had a larger Holistic A' than third graders, t(127) = 2.334, p
8/17/2019 FullText (23)
57/90
41
Figure 12. A' of congruent and incongruent trials for first, third and fifth graders in
the holistic processing task (* p < 0.05; ** p < 0.01; *** p < 0.001).
3.1.3.3. Pearson’s Correlation Analysis
Table 2 presents the correlations among age, rapid-naming speed, literacy measure, and
Holistic A'. Most of the correlations among the variables were statistically significant,
except for that between rapid-naming speed and Holistic A'. This shows that the
developmental change in HP of Chinese characters is least likely related to a general
improvement in orthography recognition.
8/17/2019 FullText (23)
58/90
42
Table 2. Correlations Among Age, Rapid-naming speed, Character naming
accuracy, Character naming response time, Word naming accuracy, Word naming
response time, Copying response time, Dictation accuracy, and Holistic A'
3.1.3.3. Hierarchical Regression Analysis
We investigated how HP could be uniquely predicted by literacy level by partialing out the
variance due to age. As summarized in Table 3, HP was predicted uniquely by reading and
writing performance and vice versa. The variance of HP can be significantly explained by
reading and dictation performances, but not copying performance, when partialing out the
variance due to age.
Table 3. Hierarchical regression analysis among holistic processing and reading,
8/17/2019 FullText (23)
59/90
43
dictation and copying performance
Predicted Variable: Holistic A'
Age vs. Character Naming (Response Time & Accuracy)Steps Variables Δr
2
1 Age .110***
2 Character Naming .037*
1 Character Naming .141***
2 Age .005
Age vs. Word Naming (Response Time & Accuracy)
1 Age .099***
2 Word Naming .056**
1 Word Naming .147***
2 Age .008Age vs. Copying (Response Time)
1 Age .104***
2 Copying .001
1 Copying .067***
2 Age .038**
Age vs. Dictation (Accuracy)
1 Age .155***
2 Dictation .014
1 Dictation .118***
2 Age .0111 p < 0.1 * p
8/17/2019 FullText (23)
60/90
44
between dictation accuracy and word naming accuracy (z = 2.403, p < .05). The mediator
effect of Holistic A' was partial as the direct effect of dictation accuracy and word naming
accuracy remained statistically significant, β = .635, p < .001 (Figure 13)8.
Sobel’s β = .635***
Figure 13. Partial mediation effect of reduced holistic processing on dictation and
word naming performances (* p
8/17/2019 FullText (23)
61/90
45
This showed that first, third and fifth graders had a similar degree of left-side bias in
perceiving Chinese characters in either font.
To examine whether the children had a left-side bias effect in the perception of
characters, I conducted one-sample t-tests again the chance level (i.e., 50% left chimeric
character preference) for Ming and Feng Chinese characters. I did not find a preference for
left chimeric characters in Ming font (.51), t (214) = .957, n.s., nor Feng font in children
(.50), t (214) = .554, n.s. This result suggested that children had not developed left-side
bias for Chinese characters in either a familiar font (Ming) or an unfamiliar font.
3.1.3.6. Effects of font familiarity on character- and word-naming performances
Repeated-measures ANOVA (font: Ming vs. Feng x grade) was used for the analysis on
character-naming response time. I found a main effect of font, F(1, 181) = 16.9, p < .001,
η p 2 = .085, and a main effect of grade, F(2, 181) = 37.01, p
8/17/2019 FullText (23)
62/90
46
d = .37, and fifth graders, t(117) = 2.192, p < .05, d = .41, while third and fifth graders did
not differ statistically, t(128) = .135, n.s., d = .02. These results suggest that word-naming
performances depended greatly on font familiarity in lower grades than in upper grades.
This font-grade interaction was not found in character-naming.
3.2. Experiment 2: Test for holistic processing of Chinese characters in
non-Chinese speaking children
Since I could not totally rule out the effect of inhibitory control differences due to age on
HP in Experiment 1, Experiment 2 aimed to test HP of Chinese characters in non-Chinese
speaking (NCS) children that do not receive the local Chinese curriculum as a control
group. In Hong Kong, there are designated local schools that admit NCS students. They
learn under the same curriculum as typical local Chinese children except for reading and
writing Chinese, and their language of instruction is English because their native language
is not Chinese and their Chinese reading and writing proficiency is far below the levels of
children from mainstream schools that were teaching Chinese as a native language. Their
school adopted a school-based curriculum in which NCS children may only be taught
simple Chinese speaking skills and simple Chinese vocabulary. If reduction in HP is
mainly contributed by enhanced Chinese literacy, I predict that the HP of Chinese
characters of NCS children will remain the same as they progress to higher grades, since
their Chinese literacy level should be homogenous across grades.
8/17/2019 FullText (23)
63/90
47
3.2.1. Participants
30 first grade (mean age = 7.00 years, SE = .079), 30 third grade (mean age = 9.33, SE =
.151), and 22 fifth grade (mean age = 11.47, SE = .198) NCS children from an elementary
in Hong Kong participated in our study. They were all receiving the same curriculum as
typical Chinese children but were not receiving regular Chinese language curriculum at
school, and the language of instruction at school was English. All of them had normal or
corrected-to-normal vision.
3.2.2. Materials and Procedures
To test for HP effects, I adopted identical procedures and materials from Experiment 1.
3.2.3. Results
Figure 14. A' of congruent and incongruent trials for non-Chinese speaking first,
8/17/2019 FullText (23)
64/90
48
third and fifth graders in the holistic processing task
Repeated-measures ANOVA was used to investigate HP effects (congruency: congruent
vs. incongruent trials x grade: Grade 1 vs. Grade 3 vs. Grade 5). I found a significant
effect of grade, F(2, 79) = 15.496, p < 0.001, η p 2 = . 282, a significant effect of
congruency, F(1, 79) = 66.244, p < 0.001, η p 2 = . 456. However, I did not find an
interaction between congruency and grade, F(2, 79) = 0.780, p = .462,. The main effect of
congruency suggests that across grades, NCS children also process Chinese characters
holistically. The main effect of grade showed that the performance level increased with
grade. The absence of interaction between grade and congruency suggests that children
processed Chinese characters with the same level of congruency effect across grades
(Figure 14).
3.2.3.1. Comparing with Experiment 1
To further understand the HP difference between Chinese and NSC children in each grade,
I defined holistic A’ as the A’ difference between congruent and incongruent trials and ran
a two-way ANOVA with group (Chinese vs. NCS children) and grade (grade 1 vs. grade 3
vs. grade 5) as the between-subject variables to test for effects in holistic A'. I found a
main effect of grade, F(2, 292) = 4.790, p < .01, η p 2 = .032, but no main effect of group,
F(1,292) = 1.112, n.s. I also found a marginal interaction between group and grade,
8/17/2019 FullText (23)
65/90
49
F(2,292) = 2.333, p = .099, η p 2 = .0169.
To further understand this marginal interaction effect, I ran post-hoc t-tests to
compare the HP effect (i.e., holistic A’) between Chinese and NCS children in each grade
(Levene’s test showed homogeneity of variance across groups). I found that Chinese
children were more holistic than NCS children in Grade 1, t(52) = 2.261, p < .05, d = 0.62.
However, both Chinese children and NCS children had similar HP in Chinese characters in
Grade 3, t(65) = .277, n.s, and Grade 5, t(28) = .240, n.s. These effects suggested that the
reduction in holistic processing observed in Chinese speaking children as they progressed
to higher grades was mainly due to a stronger holistic processing effect in grade 1
compared with NCS children10.
3.3. Discussion
Experiment 1 showed that as children who learned the Chinese language in a mainstream
school progress to higher grades, they processed Chinese characters less holistically.
Hierarchical regression analysis suggests that this reduction in HP could not be solely be
9 Note that in this two-way ANOVA, Levene’s test showed homogeneity of variance across groups, although
the analysis involved unequal sample sizes. In a separate analysis, I conducted a mixed ANOVA with group
and grade as the between-subject variable and congruency (congruent vs. incongruent) as the within-
subject variable, and Levene’s test failed to support the homogeneity of variance assumption; thus this
mixed ANOVA analysis was not reliable and not reported here.10
In grade 1, Chinese speaking and NCS children differed significantly in performance in congruent trials (t-
test with Welch’s correction for unequal variances t(42) = 3.223, p < .01) but did not differ in incongruent
trials (t(55) = 1.232, n.s.). Their performance in incongruent trials was also significantly above the chance
level (A’ = 0.5; t(85) = 4.334, p < .001;). Thus the weaker HP effect in NCS children was unlikely to be due to
a floor effect in the incongruent trials.
8/17/2019 FullText (23)
66/90
50
accounted for by age, but by an overall improvement in Chinese reading and writing
performance.
The mediation analysis showed that HP partially mediates the effect of dictation
accuracy and word naming accuracy (Figure 13). This is consistent with our hypothesis
that writing performance (i.e., the ability to recall and write down Chinese characters)
predicts the HP effect in Chinese character processing, which in turn predicts word
reading performance (Figure 10). Writing experience can perhaps enhance the ability to
analyze character components (i.e., reduced HP), which facilitates character recognition.
My control study in Experiment 2 suggests that without learning to read and writing
Chinese characters extensively, children do not have a reduction in HP as they reached
higher grades. Unlike in Experiment 1 where children learning the Chinese language at
school showed a reduction in HP as they progressed to higher grades, the NCS children in
Experiment 2 were shown to process Chinese characters with the same level of HP across
all 3 grades. Since the NCS children were not learning the typical Chinese curriculum,
reduction in HP of Chinese characters in higher grades in Chinese-speaking children was
likely a result of enhanced Chinese literacy but not age.
Post-hoc t-tests showed that first graders who are learning the Chinese language in a
mainstream school processed Chinese characters more holistically than NCS first graders.
The Chinese proficiency of the Chinese first graders in the current study was much higher