Professional Documents
Culture Documents
Useful Spanish Vocabulary
Useful Spanish Vocabulary
Thesis
Presented in Partial Fulfillment of the Requirements for the Degree Master of Arts
in the Graduate School of The Ohio State University
By
2009
2009
ABSTRACT
Foreign language learners (L2 learners) have often been observed to lack a natural
“rhythm” when reading aloud. This rhythm of speech, referred to as prosody, has been
oral reading may be due to a lack of comprehension of these syntactic structures. As little
research on prosody in L2 oral reading exists, the current study attempted to describe the
(JFL) learners by posing the research questions: 1) How do JFL learners apply prosody
when reading aloud? and 2) Does this prosodic phrasing match their interpretation of the
sentences?
experiment, JFL learners (experimental group) and native speakers (NS) of Japanese
(control group). Test sentences contained both unambiguous and ambiguous branching
modifiers, and the location of prosodic breaks in oral reading were the prosodic feature
under analysis.
Results indicated that prosodic preferences differed between the JFL learner and
modifiers contained in the test sentences. However, we found that the match between
answers to the comprehension questions were low. This indicated that although the JFL
learners were able to understand the target syntactic structures, their ability to produce the
correct prosody has not fully developed. Therefore, we concluded that production errors
were not necessarily due to the inability to comprehend the syntactic structures in the test
iii
Dedicated to my wife
iv
ACKNOWLEDGEMENTS
I would first like to express sincere thanks to my advisor Dr. Mineharu Nakayama
for his guidance and support throughout this project. His thorough comments and advice
personally feel that my knowledge of Japanese linguistics has been greatly enriched by
his guidance.
I would also like to thank my committee member Dr. Etsuyo Yuasa. Her
willingness to read my thesis drafts at short notice, and yet provide highly detailed
comments, surely improved the quality of this thesis. Her advice regarding syntax
State who participated in our experiment. Without their participation, this thesis would
not have been possible. I must also thank the students I have taught over the past two
academic experience, and has time and again reminded me that teaching is not limited to
a one-way exchange of knowledge. Thanks are also due to my fellow DEALL graduate
students for their friendship and the laughs we shared over the past two years.
v
I thank my family for their encouragement and support throughout the years. And
of course, I thank my wife Sayo for her love and support during these hectic years of
graduate school.
Finally, the experiment conducted in this thesis was made possible by funding
from the Department of East Asian Languages and Literatures and the Institute of
Japanese Studies at the Ohio State University. Their provision of the participant fees is
vi
VITA
December 2001….......B.A. East Asian Studies, University of North Carolina, Chapel Hill
FIELDS OF STUDY
vii
TABLE OF CONTENTS
Page
Abstract…………………………………………………………………………………... ii
Dedication……………………………………………………………………………….. iv
Acknowledgments…………………………………………………………………….… v
Vita……………………………………………………………………...……………… vii
List of Figures…………………………………………………………………………… xi
Chapters:
1. INTRODUCTION………………………………….……………………………. 1
1.1 Introduction……………………………………………………………...…… 1
1.4 Summary……………………………………………………………………... 6
viii
2.2 Prosodic Phrasing…………………………………...………………………. 10
2.5 Summary……………………………………………………………………. 24
3. THE EXPERIMENT…...............………………………………………………. 25
3.2 Participants………………………………………………………………….. 26
3.3 Materials……………………………………………………………………. 28
3.4 Procedure…………………………………………………………………… 33
3.5 Results………………………………………………………………………. 34
3.6 Summary…………………………………………………………………..... 45
ix
4.1.1 Prosody-syntax mapping in L2 Japanese…………………………. 47
4.3 Conclusion………………………………………………………………….. 55
References……………………………………………………………………….……… 77
x
LIST OF FIGURES
Figure Page
xi
LIST OF TABLES
Table Page
xii
SYMBOL AND ABBREVIATIONS
˄ : prosodic break
ADJ: adjective
LOC : location
TOP : topic
xiii
CHAPTER 1
INTRODUCTION
1.1 Introduction
who struggle to produce natural “rhythm” when reading aloud. Although their
stress is accurate enough, the rhythmic grouping of sentences above these individual
prosody, and can be observed in both spontaneous speech and oral reading (Beckman,
1996; Speer & Blodgett, 2006). Prosody has been shown to reflect surface syntax in
certain structures, such as relative clauses and branching modifiers, often manifesting as
changes in pitch or pauses in Japanese (e.g. Kubozono, 1993; Venditti, 1994). Prosodic
errors noted in the oral reading of language learners may be due to the lack of
prosody and syntax in oral reading, we might gain insight into the source of these
1
rhythmic errors, which would be a great asset to researchers and teachers of Japanese. In
order to do so, concrete evidence of learners’ use of prosody when reading aloud is
(henceforth L2) learners. In the current case study, I will examine how Japanese as a
foreign language (JFL) learners apply prosody during oral reading and whether this
Features such as length, rhythm, accent and stress that give spoken language its
natural rhythm are termed prosodic features, also known as suprasegmentals due to their
presence above individual segments (Fox, 2000). Prosody also interacts with
paralinguistic and pragmatic features of speech, for example, when a speaker places
emphasis on a certain word or phrase, or contrasts between two statements. The effective
a crucial ability for L2 learners as well. These prosodic features, once considered a
neglected area of study both in L1 and L2 language instruction, are now widely
recognized as a vital part of language proficiency, both in spoken language and oral
reading (e.g. Dowhower, 1991; Eda, 2004; Lems, 2006; Schwanenflugel, 2004).
considered a key component in oral reading, albeit one whose role is difficult to
2
characterize (Mathson et al., 2006). A description of how L2 learners apply prosody in
oral reading would help us to better understand and identify the source of prosodic errors.
Researchers in phonetics and phonology have long studied the relationship between
mapping of sound and syntax (e.g. Kubozono, 1993; Pierrehumbert & Beckman, 1988;
Venditti, 1996).
comprehension, or does the application of prosody to a given text precede, and therefore
2006). Some research supports the claim that prosody applied while reading aloud is the
outcome of reading after comprehension has occurred, then it would seem at least
tentatively that without comprehension, correct prosody would be difficult for the reader
to produce. However, evidence supporting this claim is strikingly scarce and inconclusive,
with the majority of related studies focusing on reading in L1 by children. Shibata &
Japanese, but their task focused exclusively on production ability. Dowhower (1991),
summing up several older studies, identified features which hinder prosodic reading and
are characteristic of dys-fluent reading, including pausal intrusions within syntactic units,
Eagan (1975), young readers who scored lower on silent reading comprehension tests
displayed more of these prosodic errors during oral reading activities. Similar studies
3
found that prosodic reading ability increases as children progress in grade levels,
appropriate prosody (Kowal et al., 1975). Although these studies examine prosody and
reader’s ability or inability to generate prosody in oral reading. The majority of these
early studies were concerned with prosody as a component of oral reading fluency, that is,
the ability to read aloud with appropriate expression and intonation. However, phonetic
generally not included in the data. In the current study, I will attempt to identify specific
structures in Japanese where the interpretation of its constituents might directly affect the
way in which prosody is applied. This will provide us with insight as to how sentence
syntactic structure, although the mapping is not universally one-to-one (e.g. Kubozono,
1993; Selkirk & Tateishi, 1991; Venditti, 1994; Venditti et al., 1996). Structures
phrases (NPs) and relative clauses. Branching NPs are structures with configurations such
4
as [X[YZ]] and [[XY]Z], in which an adjective or noun modifies a head noun. Below,
English examples are from Venditti (1994) and Japanese from Selkirk & Tateishi (1991):
The brackets show us that in (1) both the men and women are described as being
old, whereas in (2) only the men are described as old. In the Japanese sentences (3) and
(4), although again structurally ambiguous, the brackets indicate that Aoyama modifies
Tateishi found that speakers produce different prosodic phrasing depending on the
location of syntactic boundaries. For example, in (4) the double brackets before
this prosodic phrasing may be phonetically realized as either a pause or a rise in F0,
proposed varying accounts of the prosodic realization in similar structures, but the
consensus seems to be that prosody varies depending on the modification relationship, i.e.
5
the syntactic structure of the sentence (Kubozono, 1993). Listeners were also able to
phrasing in controlled experiments (e.g. Blodgett, 2004; Snedeker & Trueswell, 2003).
Studies examining the mapping of prosody and syntax in this type of structure will be
Relative clauses have also been tapped by researchers due to the ambiguities that
arise in the attachment of NPs (“high attachment” vs. “low attachment”) (Jun, 2003).
language, have been shown to be crucially utilized by both the speaker and listener in
expressing and interpreting meaning (Kjelgaard & Speer, 1999; Kang & Speer, 2003).
Kang and Speer (2003) provided evidence that prosodic boundary information can
provide reliable clues to the structure of spoken sentences in Korean. In a later study,
Kang, Speer & Nakayama (2006) demonstrated that this relationship between syntax and
prosody also maintains in Japanese, and they concluded that prosodic phrasing which
1.4 Summary
The wealth of recent research on prosody and its relationship with syntactic
structure in sentence processing in L1 will provide the background for the current case
study. Research describing L2 learners’ use of prosodic phrasing is limited, and empirical
data would provide a springboard for further research. Specifically, the current study will
6
examine prosody in oral reading and its relationship to sentence comprehension.
Assuming that a correspondence between prosodic phrasing and syntax exists in certain
sentence structures, such as branching modifiers, I propose to test similar structures with
JFL learners. Evidence suggesting that prosody can be used as a reliable indicator of
sentence structure. Very little research has specifically examined prosodic phrasing in L2
above the word level, and even more scant is research focusing on JFL learners. Shibata
& Hurtig (2008) conducted research on prosody acquisition in L2 Japanese, which tested
study is their examination of prosodic production in oral reading of left- and right-
similar theoretical basis as in the current study with regards to the prosodic structure of
Japanese. However, their test sentences in the oral reading activity were presented along
with sentence meanings, removing the task of sentence comprehension by the learner and
activity, I propose to expand on their study by allowing learners to read test stimuli first
silently to measure comprehension, then aloud. By doing so, it will enable us to examine
the relationship between syntactic parsing during silent reading comprehension and
7
In this thesis, I propose the following two research questions relating to prosody
in L2 Japanese:
modifiers?
which examine the mapping of prosody to syntax. Next I will present a case study which
learners. Chapter 3 will detail the design and procedure of the experiment, as well as
present the relevant data findings. In Chapter 4, I will provide the discussion based on the
results of the experiment and draw conclusions from this data in an attempt to provide a
8
CHAPTER 2
elements at one level grouped together into larger elements at the next higher level”
(Venditti, 1996: 288). Pierrehumbert & Beckman (1988: 21) proposed a hierarchically
grouped prosodic tree, which contains the utterance as its highest unit, followed by
Prosody is reflected in speech not only as features such as length, accent, and
stress, which concern individual words, but also at the phrasal and sentence levels. The
current study examines prosodic structure at the phrasal level in Japanese, but a
description of the prosodic features from the lexical level upwards is necessary to
understand the constituents that form prosodic phrases. As I will examine how L2
Japanese, both at the lexical and phrasal levels is necessary. Below is an overview of the
9
2.2 Prosodic Phrasing
In order to examine prosody and its relation to other language structures, a system
Blodgett (2006) describe two primary methods in which prosody is “visually” analyzed,
those being phonetic analyses of duration and fundamental frequency (F0), and
intonational labeling systems, chiefly the ToBI (Tones and Break Indicies) annotation
examining the waveform of an utterance with computer software. Amplitude and duration
of pauses are the key features which show accent, pitch and sentential breaks of the
spoken signal. As we will see later, both pitch height and break duration are critical
timed language such as English (Kubozono, 1988). According to Venditti (2006: 209),
1
Developed for English by Silverman et al. (1992), this system has also been applied to various other
languages, including Japanese, where it is called J_ToBI, and has helped scholars of language to label
features of spoken language such as rhythm, pitch, stress and accent (Venditti, 1995). The ToBI system
represents the sound string of a sentence as a series of tonal events, and these are labeled with a notational
system. The three tonal events described by this notation include an initial boundary tone (L%), a high tone
(H-), and another (L%) phrase-final boundary tone. The phrase intonation would be the contour between
these three points (Venditti, 1995, 2006).
10
strong syllables or words at the sentence level.” Rather than accent being determined by
word structure alone, English speakers may choose to place stress on a given word for
property of individual lexical items themselves (Kubozono, 1988). That is, a word is
Accented words display a rise in pitch on the accented mora, followed by a sharp
decrease (HL: high to low) in pitch at the end of the accented mora. Words that are
unaccented exhibit a similar initial rise in pitch (LH), but the sharp drop present in
accented words is not found in unaccented words. Both accented and unaccented words
undergo a rise in pitch on their second mora, which is termed a “phrase accent”
Next, we will examine prosodic structure above the lexical level and introduce
2
These words can be distinguished in written form by Chinese characters: ‘owner’ and
‘buyer.’
11
2.2.2 Accentual Phrase
Prosody above the word level is typically grouped into accentual phrases (AP)
and intonation phrases (IP). These units of phrasing refer to the way that speakers
rhythmically group the words in an utterance (Blodgett, 2004). Venditti (2006: 211)
refers to an AP as “[a] grouping of words delimited by three tones: a low boundary tone
at the start of the phrase, a high phrase tone near the second mora and another boundary
tone at the end of the phrase.” In terms of word constituents, APs tend to be composed of
‘Mr. Tanaka-NOM’. Venditti (1994) also provides examples in which unaccented words
phrase together with adjacent unaccented and accented words to form longer APs.3
Accentual phrases then group into a higher level of phrase, the intonation phrase, which
is the highest level of prosodic grouping in Japanese (Speer, 2004; Venditti, 1994).
Intonation phrases (IPs) are also defined by tonal characteristics similar to those
found in the AP. In the hierarchy of prosodic structure, IPs form the highest level and are
composed of groupings of APs (Venditti, 2006). Since IPs often comprise entire
sentences, an optional pitch movement indicating a question, for example, at the right
3
Venditti (2006) notes that in continuous speech, individual unaccented words may be de-phrased into a
sentence-length APs.
12
found at the IP level is the phonological process of downstep. Downstep occurs when
(Kubozono, 1993; Selkirk & Tateishi, 1991; Venditti, 1994, 2006). As long as the
accented APs are adjacent within the same IP, downstep has been shown to occur
(Kubozono, 1993; Pierrehumbert & Beckman, 1988; Selkirk & Tateishi, 1991). A break
in the IP would cause pitch to “reset” to an initial value, effectively stopping the
downward trend of pitch movement. Boundaries between IPs have also been shown to
appear as prosodic breaks or pauses in speech. Lovric et al. (2001) found that in relative
reflected syntactic attachment. In the current study, prosodic breaks, which are assumed
indicators as to how subjects parse sentences and apply prosody in an oral reading task.
Japanese. The following section will consider studies which examine the mapping
psycholinguistics has focused on the mapping between prosodic structure and underlying
syntax (for Japanese: Selkirk & Tateishi, 1991; Venditti, 1994; for Korean: Schafer &
Jun, 2002; for English: Blodgett, 2004; Kjelgaard & Speer, 1999). Determining the nature
of this mapping relationship would give researchers a clearer picture as to how speakers
13
use prosody to convey meaning, as well as how prosody can influence listener judgment
(Fox, 2000). Even in L2 speech, as will be seen in the current study on L2 oral reading,
establishing a mapping relationship between prosody and syntax would provide insight
into how learners are processing certain structures, such as branching pre-nominal
modifiers.
Studies specifically investigating this topic in Japanese L1 are thorough and have
shown salient evidence linking prosody and syntactic structure. Pierrehumbert &
Beckman’s (1986, 1988) studies on prosodic phenomena such as downstep pioneered the
investigation of the effect of syntactic structure on prosodic phrasing. Below I will briefly
mapping relation.
makes a strong case for the phonetic realization of syntax in certain structures. The
structures in mention are left-branching (LB) and right-branching (RB) sentences, shown
4
The symbol ˄ indicates a prosodic boundary.
14
NP
NP
All test sentences in their study were composed of accented words, which have
all of the accented nouns in the left node, or branch of the LB configuration [[XY]Z] in
(3), but not in the RB configuration [X[YZ]] in (4). They claim that this is due to the
15
elements [X] and [Y] in RB structures, and posit the following generalization regarding
The boundary for the IP in (4) above would be the left edge of a syntactic
constituent, which corresponds to the left-branching node containing Yamaguchi no. The
absence of such a boundary in (3) allows the pitch of the accented nouns to be
found similarities between the three languages in terms of prosodic groupings reflecting
syntax above the individual lexical stress and pitch patterns. In the case of Japanese, they
give the following sentence, which can have two syntactic interpretations (p. 299).
(6)
16
b. [[kyonen anda erimaki ga] nusumareta].
Here, the adverb kyonen ‘last year’ can modify the sentence as a whole (a), or just
the relative clause (b). When examining the F0 contours of each of the above
interpretations, it was found that the wider scope interpretation (a) was produced with a
pause coinciding with the phrase boundary that corresponds to the underlying syntactic
structure (1996: 301). They note that either a pause, or a reset in the F0 contour to a
higher pitch can provide the listener clues to the syntactic structure in this type of
investigation of prosody’s role in syntactic acquisition in L1,5 support our study’s claim
by showing that prosody can be a salient clue to syntactic structure in certain sentence
types.
In an earlier study, Venditti (1994) used branching NPs, similar to those that will
(1988) and Pierrehumbert & Beckman (1988) on the nature of syntax-prosody mapping.
5
Their study investigates prosody’s role as a boot-strapping mechanism that aids acquisition of syntactic
structures in L1.
17
(8) [kimidori no [himawari no moyoo]] (RB)
When shown pictures corresponding to the LB configuration (in which only the
sunflowers were green) as in (7), some speakers placed a pause between the genitive case
marker no and the head noun moyoo ‘pattern’ (p. 309). This pause would presumably
provide the listener with a clue to the scope of the modification of the genitive no.
Prosodic breaks and their influence on syntactic parsing were the focus of
research by Clifton (2002), who formulated the Informative Boundary Hypothesis (IBH).
Originally devised for English, it was reorganized by Maynell (2005) for head-final
languages such as Japanese. Paraphrased, the IBH states that for structures [A] [B] [C],
where [A] can be attached to either [B] or [C], prosodic boundaries can encourage either
high or low attachment, depending on location. If the boundary between [A] and [B] is
relatively greater, high attachment should be preferred, but if the break between [B] and
[C] is comparatively larger, low attachment should be favored (p. 6). Although the IBH
originally described prosodic breaks in relative clauses, the occurrence of prosodic breaks
in branching modifiers, such as in (7) and (8) above, will be examined in the current
experiment assuming similar findings will apply. By comparing the relationship between
prosody in oral reading and interpretation, we will investigate the interaction of prosody
18
2.4 Prosody in L2
production ability alone (Lems, 2006; Shibata & Hurtig, 2008). The interaction between
sentence processing mechanisms and prosody, and the usage of prosodic phrasing in
sentence interpretation, have both been well-documented by scholars in L1, but little
quantitative research exists regarding L2 prosody. Lems (2006: 236) attributes this in part
to the fact that groups of L2 learners are often composed of learners from differing
patterns of their native languages are too great a factor in being able to use prosody as an
indicator of sentence parsing decisions. Additionally, other scholars note the particular
difficulty in teaching prosody to L2 learners, or the fact that prosody is often left
Prosody in L1 oral reading was briefly mentioned in the previous chapter. Studies
in L1 reading prosody provided an initial motive for the current study; L1 children who
are still grappling with acquiring adult-like parsing skills can presumably be compared to
1991). That is, errors noted in reading research on prosody of L1 children can
19
Schwanenflugel et al. (2004) conducted research on prosody and reading
comprehension in L1 children. In their study, they tested two assumptions: (1) Prosodic
reading is a by-product of reading comprehension and (2) prosodic reading might also
work conversely, with correct prosodic reading enhancing sentence parsing decisions.6
The reading models which correspond to these two assumptions are (i) reading prosody
prosody model (p. 121). These predictions seem to have some overlap with Clifton’s
on comprehension. This leads to the question that was raised earlier in this study: How
al.’s results suggest that performance on reading comprehension tests and the speed at
which children decode words in L1 contributes to the child’s ability to apply adult-like
parsing ability of a given text. Other studies, such as that by Karlin (1985) on speakers of
comprehension ability. Her study was an early attempt to examine reading prosody in a
specific language population. The subjects in the study all performed well on
comprehension tests, but their prosody deviated from the accepted pattern. Although this
6
Schwanenflugel et al. define reading prosody as the incorporation of prosodic features such as the rise and
fall of pitch and pauses typical of conversational speech into oral reading (2004: 119).
20
is an older study, the following two weaknesses can be pointed out. First, subjects were
all college students at American universities who spoke a dialect of English, but were
specific syntactic structures were focused on for their analysis of the relation between
prosody and structure; subjects were given a document and instructed to read it “with
expression” (p. 10). In order to avoid these shortcomings, in the current study specific
Shibata & Hurtig (2008) conducted one of the few comprehensive studies on
prosody in learners of Japanese. They identify prosody’s vital role in both understanding
when speaking. Their study examines the acquisition of prosody by L1 English learners
motivation for their study is stated that a perceived “foreign accent,” often found in even
advanced learners, hinders the ability of the non-native speaker to function in the society
of the target language (p. 176). At the root of this foreign accent are not only segmental
errors, which are considered relatively easy to identify and correct, but non-segmental
21
prosodic errors, such as pitch accent and sentence-final intonation. They reported that
prosodic factors, such as pitch height, rather than segmental errors had more influence on
an L2 Japanese speaker being perceived as having a foreign accent (p. 178). Shibata &
Hurtig identified and tested several areas of difficulty for English speaking learners of
Japanese in acquiring Japanese prosodic and segmental structure, and tested these with
learners of three proficiency levels, Advanced, Intermediate, and Novice. The areas
length and word accent, ability to produce/perceive prosody for semantically and
intonation.
Among the areas they examined, of interest to our study is the ability of Japanese
was labeled Task 2 in their study. Their experimental Task 2 deals with prosody and its
use in providing cues to mark certain syntactic structures, in this case branching
structures. Phrases containing ambiguous structures, that is, those which can syntactically
and semantically have two interpretations, can be disambiguated by prosody alone (p.
184). This task examined learners’ ability to produce prosodic cues to mark these
In order to establish a baseline prosodic pattern, the test phrase [akai kuruma no
shiito] ‘red car-GEN seat’ was first given to a Japanese native speaker (NS) to establish
an intonation contour. Below are the syntactic trees representing both the RB (9) and LB
(10) configurations of this phrase from Shibata & Hurtig (2008: 184):
22
(9) Right-branching (10) Left-branching
ADJ N P N ADJ N P N
[[akai] ˄ [kuruma no shiito]] [[akai kuruma no] ˄ shiito]
pause at the major phrasal boundary in the RB-structure. In the LB configuration, akai
kuruma no was produced as a unified group, with a slight pause or break occurring before
the final noun shiito ‘seat’. They assume that RB structures in particular are prosodically
The same target was then presented to JFL students, who were instructed to read
the sentence aloud in the frame Kore wa akai kuruma no shiito desu, ‘This is a red car
seat.’ Four other of similarly ambiguous structures were used. Prior to the reading,
learners were allowed to spend a few minutes analyzing the sentences and were also
presented the English translation to clarify meaning. Oral productions were recorded and
later rated for comprehensibility by three NS judges. Shibata & Hurtig conclude that
RB structures was clearly more difficult for learners than that of LB, even at the
23
advanced level. Observed errors include learners’ failure to raise the pitch of the N
kuruma ‘car’ as well as failure to suppress the pitch of the final noun shiito ‘seat’. These
errors by JFL learners in oral productions of the RB structures contributed to the lower
comprehensibility by the NS judges (p. 193). Although Shibata & Hurtig’s study was
providing sentence interpretation, a step in the normal reading process has been
effectively removed, providing little insight on how JFL learners might produce a
2.5 Summary
F0 boost and prosodic breaks have been documented as instances of prosody mapping
onto surface syntax. The question then arises as the nature of this relationship in L2
Japanese.
aspect of Shibata & Hurtig’s study (Task 2) by adding a comprehension task to the oral
reading of branching modifiers. In their production task, subjects were provided with
sentence interpretation, limiting the scope of the task solely to production. By doing so,
little light is shed upon the processing decisions learners make when reading a sentence
24
CHAPTER 3
THE EXPERIMENT
This chapter discusses the design and procedure of the experiment conducted in
this study which examined the relationship between prosody in oral reading and reading
the chapter. Branching modifiers with both unambiguous and ambiguous interpretations,
such as those examined in Shibata & Hurtig (2008) and discussed in Chapter 2, were
and syntax. Similar structures have also been examined by researchers in sentence
processing (e.g. Pierrehumbert & Beckman, 1988; Venditti, 1994; Venditti et al., 1996).
The current experiment was designed to investigate how JFL (Japanese as a Foreign
Language) learners apply prosody during oral reading, and whether or not this prosody
25
1) How do JFL learners apply prosody to sentences containing branching
modifiers?
and ULB sentences will be examined to see if prosody matches syntactic structure and if
this prosody matches sentence interpretation. The AMB sentences, those with two
possible interpretations, will be checked for a match between prosody and interpretation
of JFL learners, comprising the experimental group, will be analyzed and compared to a
3.2 Participants
(experimental group) and 4 native speakers of Japanese (control group). All participants
were either undergraduate or graduate students at The Ohio State University who were
over 18 years old. They were recruited directly by the experimenter through Japanese
language classroom visits to briefly explain the experiment and distribute letters of
solicitation.
The JFL learners were all enrolled in Japanese courses at the time the experiment
was conducted. They were taking one of the following courses: Intensive Track, Level 3,
26
or Level 4 Japanese, which used Japanese: The Spoken Language, Parts 2 & 3 (Jorden &
Noda, 1988, 1990) as the course textbooks. The Intensive Track combines Levels 2 and 3
into one year of study; these participants had all been exposed to nearly the same
materials as Level 3 at the time of the study. However, this group will be analyzed
separately, as the pace of study of this course is much faster than Level 3.7
The experimental group consisted of three subject groups: Group 1 includes seven
participants from Intensive (between approx. 410-450 instruction hours over the duration
of the experiment), Group 2 includes 11 participants from Level 3 (430-450 hours), and
Group 3 contains eight participants from Level 4 (560-600 hours). The experiment was
conducted over the period of about six weeks at the end of the academic year, so contact
hours at the time of participation varied slightly among individuals. All JFL participants
spoke English as their L1, with no formal Japanese study prior to university classes. A
total of eight students had visited Japan, six of who had stayed for less than three months,
with no formal Japanese classes. Two participants from Group 2 had three to six month
stays in Japan, one of whom took Japanese language classes at a university during her
stay.
The control group of four native speakers was recruited through personal contacts
of the experimenter. They had various lengths of stay in the US, but were all enrolled in
either graduate or undergraduate courses at OSU at the time of the experiment. Dialect
variation among this group was not controlled, but three of the four participants were
7
In terms of ACTFL (American Council on the Teaching of Foreign Languages) proficiency levels,
Level 3/Intensive students are about Intermediate; Level 4 is approximately Intermediate High/Advanced
Low.
27
from eastern Japan, and one was from Nagoya.8 All participants were volunteers and
were paid ten dollars as compensation for their participation in the study.
3.3 Materials
stimuli in the pretest session, 16 test stimuli and 16 fillers in the main session, and one
sentence used in the initial instructions. As mentioned above, the test sentences can be
further separated into three types, URB (four sentences), ULB (four sentences) and AMB
(eight sentences). Test stimuli all contained a branching modifier structure, in the
sequence [ADJ] [N1] [N2], while fillers did not employ these structures. All sentences
were presented in Japanese script, with furigana readings above kanji that had not been
covered by the participants. Since the objective of the oral reading task was to attain a
natural reading of the test stimuli, furigana were deemed necessary to facilitate oral
reading fluency.10 Vocabulary items and grammatical structures used in all 38 sentences
were selected from material that all proficiency levels had previously covered. To avoid
any effects of lexical priming, individual vocabulary items were used only once in the
experiment module.
questions were presented in English, to prevent confusion and reduce the length of time
8
Kanto region participants: Chiba (two subjects), Tokyo (one subject).
10
In materials designed for native readers, furigana are normally only used for kanji not included on the list
of Joyo Kanji.
28
to complete the experiment. Below are examples of each of the types of sentences
Aside from the target branching structures, all of the test stimuli were similar in
structure in that they began with a nominative case-marked subject and ended with a verb
in past tense. Within the target branching structures, mora length was not controlled, and
accent was only controlled on N1. For the N1 position, only unaccented nouns, or nouns
with accents on the second mora or later were used, in order to facilitate the data analysis
for presence of prosodic breaks. Below are the sentence types with English translations.11
(1)
Comprehension Question:
c) either a) or b)
d) neither a) or b)
11
Translations of test stimuli were not presented during the experiment.
29
ii) ULB (Unambiguous Left-branching):
(2)
Comprehension Question:
c) either a) or b)
d) neither a) or b)
bracketing, which was disambiguated by the plausibility of word choice, similar to the
test sentences used by Schafer & Jun (2002) in their research on Korean L1 prosody.
They used plausibility of word choice to create a bias toward adjectival modification of
Korean speakers found the structure [stylish model’s designer] to have a clear N1 bias
Similarly, in (1) above, one would not normally assume that ADJ ookii ‘big’
would modify N1 natu ‘summer’ alone. By inserting an implausible word in either the
30
N1 or N2 position, the syntactically ambiguous structures can then be disambiguated. As
demonstrated in the previous chapter, it was hypothesized that prosody would vary
incorrect as a result of the plausibility limiting the answer to only one correct
similar branching structure as the URB and ULB sentences. In (3) below, the ambiguity
arises from the fact that the ADJ can plausibly modify either N1 mati ‘town’ or N2
koosaten ‘intersection.’ These sentences are structurally similar to those used in the
prosody production task in Shibata & Hurtig (2008), although in the present experiment
sentence interpretation was not provided during the initial silent reading phase. The task
of sentence parsing was placed on the participant, thus enabling the later analysis of
prosody-to-interpretation match.
(3)
“Mr. Brown met his friend at the intersection in the dangerous town.”
31
“Mr. Brown met his friend at the dangerous intersection in the town.”
Comprehension Question:
c) either a) or b)
d) neither a) or b)
contained in the experiment, each with a comprehension question of like format to the
test stimuli. Refer to the Appendix A for a complete list of test materials.
(4)
“Mr. Yamanaka became ill and took the day off school.”
Comprehension Question:
32
c) either a ) or b)
d) neither a) or b)
3.4 Procedure
contained sentences of the four types described in the previous section. All participants
saw the same material in an identical, predetermined order. Spoken data was
continuously recorded with Audacity sound-recording software, and sound files were
A short 5-slide introduction, which contained one sentence similar to a filler, was
administered first, to familiarize learners with the experimental procedures and to make
sure their voice was being recorded clearly enough for later analysis. Participants were all
instructed to read in a natural voice, at their usual reading pace. During the introduction
only, English explanations for each step were displayed at the top of the screen along
with the sentence. Verbal instructions were given by the experimenter only if the
participant was unclear of the onscreen directions. The recording was then played back to
participants to verify speech volume and clarity. After the instruction segment concluded
and any procedural questions were answered, the experimenter left the room. A 5-
question pretest session then followed, before the 32-question test session began.
For the experimental task, each sentence had three slides which corresponded to
33
(5) Phases of the experiment:
Japanese, which they were instructed to read silently, while thinking about
the meaning.
b. Oral reading phase: Participants were instructed to read the sentence aloud
into a microphone.
interpretation.
A separate answer sheet was provided for participants to indicate their responses
to the comprehension questions. For the control group, the same procedures were
followed throughout the experiment. On average, the experimental group took between
25-40 minutes to complete the entire experiment, while the control group took 15-20
minutes.
3.5 Results
For the data analysis, one subject from Group 1 was removed because audio data
was missing for half of the test sentences. The remaining 25 subjects in the experimental
group were all included in the analysis below. As mentioned earlier, the relevant prosodic
feature that was analyzed in this study was the location of a prosodic break. In the target
branching modifiers of the sequence [ADJ] [N1] [N2], the presence of a break between
34
between [N1] and [N2] was considered an indicator of an LB structure. In analyzing the
audio data, break locations were judged by the experimenter and a native speaker of
Japanese. Inter-rater agreement on break location was calculated for each proficiency
group as follows: Group 1 (70%), Group 2 (77%), Group 3 (70%) and NS (61%). The
rapid pace of reading by some of the NS subjects made break locations more difficult to
identify, possibly resulting in the low inter-rater agreement. Due to the relatively low
the raters disagreed was evaluated with Praat sound analysis software. In these cases, the
software enabled us to examine the pitch contour and duration of the target structures,
which were used to compare the lengths of the two possible break locations. The longer
of the two breaks was considered to be the location of the break for the target structure.
Break locations were recorded as either matching/not matching the correct pattern in
URB and ULB or matching/not matching the subject’s answer to the comprehension
question in the AMB sentences. This distinction between sentence types was made
because it was assumed that there were two possible prosodic patterns for the AMB
sentences. The figures that follow present data from all groups.
Figure 3.1 below shows the percentage of correct match between prosody and
syntax in the URB, the ULB, and the AMB sentences for JFL learners (Groups 1, 2 & 3)
and native Japanese speakers (NS). Note that both the RB and the LB interpretations are
35
considered correct for the AMB sentences, resulting in a 100% match between prosody
and syntax.
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
JFL NS
The JFL learners’ correct response rate on the URB sentences was markedly
lower than ULB. Their average rate of correct break placement was 61% for URB and
82% for ULB. The difference among the JFL group by sentence type was statistically
match for the ULB sentences in the JFL learners as a group, and indicate a tendency to
place a break after [N1] in the target branching structure. Examining the NS group, an
opposite pattern was found, with better performance on URB than on ULB (94% and
69%, respectively). These types were marginally significant at (F(1,3) = 3.429, p<0.074).
36
of prosody will be presented in Chapter 4. Figure 3.2 below shows a clearer picture of the
performance by proficiency group. Again, note that the AMB sentences were presumed
to have two possible prosodic break locations, which accounts for the 100% prosody to
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Group 1 Group 2 Group 3 NS
Groups 1 & 2, which are both at a comparable level in terms of classroom hours,
performed similarly on placement of breaks in both the URB and the ULB sentences. As
observed in Figure 3.2, their average rate of correct break placement for URB was
relatively low in comparison to ULB, with 63% for Group 1 and 61% for Group 2. For
12
The raters assumed that all sentences had a break in one of the two possible locations when sentences
were evaluated. The break with the longest duration was selected as the break location, resulting in a 100%
prosody-syntax match in AMB sentences.
37
the ULB sentences, the correct placement for Group 1 improved to 75%, as did Group 2
with 75% correct break placement. Turning to Group 3, the average of correct prosody to
syntax match slightly decreased to 59% for the URB sentences. However, Group 3’s
correct response rate to the ULB sentences was the highest of all groups at 97%. At a
glance, from these results we can say that prosody-syntax match was consistently higher
in ULB than in the URB sentences for JFL learners, possibly indicating a tendency
toward an LB reading in this group. Furthermore, the performance on the ULB sentences
seems to improve at the higher proficiency level. However, no significant difference was
found in correct prosody-syntax match when comparing the three groups of JFL learners
(F(2,48) = 1.071, p< 0.345), as all groups showed the same tendency to perform better on
ULB matching.
As observed earlier, the NS control group, which consisted of only four subjects,
performed the best on break placement in the URB sentences, while the ULB sentences
appeared to be less consistent. Besides the small sample size for the NS group, a possible
38
3.5.2 Answers to comprehension questions: All groups
A comprehension question was presented after each test sentence, and subject
performance is shown below in Figure 3.3. Correct responses are indicated by proficiency
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Group 1 Group 2 Group 3 NS
Figure 3.3: Correct response rate to comprehension questions by group and type
The multiple choice comprehension question for each sentence presented subjects
an “either” RB/LB choice and one a “neither” choice. For the URB and ULB sentences,
only one choice was considered correct, whereas for the AMB sentences any answer
13
Responses to Filler-type test stimuli were not included in the main data analysis. Correct response rates
to these sentences were: Group 1 (76%), Group 2 (81%), Group 3 (78%), and NS (80%). The
comparatively lower correct response rates on the Fillers appears to be due to overall low performance on
sentences (2), (14), and (16) in Appendix A. These questions required either a “c” or “d” answer.
39
choice other than “neither” was counted as a correct response. This accounts for the high
performance in all groups, particularly on the AMB sentences. Group 1 had a correct
response rate of 88% (URB), 88% (ULB) and 100% (AMB); Group 2 at 82%, 80% and
90%; Group 3 at 75%, 88% and 98%; and NS with 100%, 94% and 100%, respectively.
When the correct response rates in the URB/ULB were compared, no significant
difference was found by group (F(3,56) = 1.852, p<0.139), sentence type (F(1,24) =
0.032, p<0.845), or interaction by group/type for all groups (F(3,56) = 0.700, p<0.553).
Subject answers will be used below to examine the relationship between prosody and
interpretation, that is, whether or not the prosodic break placement matches with the
Figure 3.4 displays the prosody-interpretation match for all groups and sentence
types. For the URB and ULB sentences, even if the response to the comprehension
question was incorrect, if that answer matched the subject’s prosody, it was considered as
40
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Group 1 Group 2 Group 3 NS
For the JFL learners, prosody-interpretation match appears to be greater for the
ULB than the URB across all three proficiency groups. Group 1 had a 54% match for the
URB and a 67% match for the ULB. For Group 2, 55% of the URB sentences matched
and 61% of the ULB sentences matched. Group 3 appeared to have the largest difference
between the two unambiguous sentence types, with a 50% match for the URB and a 91%
match for the ULB. As noted in Figure 3.2, Group 3 also showed the best performance in
the JFL groups on ULB prosody-syntax match. The difference between prosody-
interpretation match in URB and ULB was significant when the four groups were
compared (F(3,28) = 3.021, p<0.031), but no significance was found among the URB and
ULB types (F(1,28) = 2.541, p<0.112) when all groups were included. As expected, the
NS group again showed an opposite trend to that of the JFL groups, with a 94% prosody-
interpretation match on the URB sentences and a 75% match for the ULB. This opposite
41
trend by the JFL and NS groups resulted in a significant interaction between group and
in the JFL groups alone, no statistical difference was found between the groups
(F(1,24)=8.406, p<0.004) among the JFL learners, and the group/type interaction showed
Figure 3.4 above shows the results for prosody-interpretation match for
questions. The URB/ULB sentences were also evaluated by excluding subjects who
interpretation match for subjects who were able to correctly comprehened these sentence
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Group 1 Group 2 Group 3 NS
URB ULB
were analyzed, we see a slight drop in prosody-syntax match, although the overall
performance trends appear unchanged. Group 1’s performance remained the same, while
Group 2 showed a slight dip in percentage of match at 50% for the URB and 59% for the
ULB. Group 3 also had a slightly lower performance with 47% for the URB sentences
and 88% for the ULB sentences. There was a significant difference in ULB performance
among the three JFL learner groups (F(2,72) = 3,632, p<0.030), and a Tukey test
indicated that Groups 2 & 3 were significantly different (p<0.026) on the ULB types.
URB performance showed no significant difference when the three JFL groups were
Focusing on the AMB sentences, which have two possible locations for a
match as proficiency level improves, if Groups 1 & 2 are taken to be of similar level.
Group 1 had a 71% match and Group 2 a 70% match, indicating consistency between the
two groups. Group 3 performed better, with a 80% match, while NS had the highest
prosody-interpretation match at 84%. Table 3.1 gives a further breakdown of the AMB
prosody (the demoninators in the table below are [the number of subjects by group] x 8
x 8 = 32).
43
Group Group 1 Group 2 Group 3 NS
Break type RB LB RB LB RB LB RB LB
Here we see a clear preference for LB prosody—a break placed between [N1] and
[N2]—in all JFL groups. Native speakers tended to have a nearly equal preference for
RB/LB prosodic readings, with a slight tendency toward an LB reading. The NS subjects’
(Figure 3.1), which had a disambiguating noun in the target branching structure.
Figure 3.6 breaks down the AMB sentences by break location—whether the
prosodic pattern.
44
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Group 1 Group 2 Group 3 NS
RB LB
Here we can observe that all groups had a higher prosody-interpretation match
when applying an LB reading. For the NS group, this appears to show that for ambiguous
sentences, they tend towards the application of an LB prosody reading, and when doing
so, the interpretation more frequently matches this prosody. This trend appears similar to
that of the JFL groups in the preference of an LB reading. Individual test stimuli will be
3.6 Summary
In this chapter, the experiment design, procedure, and results were presented. Our
analysis of the JFL groups showed a higher rate of prosody to syntax match in ULB than
in URB. However, NS data of the same sentence types displayed an opposite trend in
45
terms of prosodic break preference, with a higher rate of match in URB. This finding
necessitates closer examination of factors that might have led to these results. In terms of
prosody-interpretation match, a higher percentage of matching was again found for the
ULB over the URB sentences in the JFL groups, while native speaker subjects showed a
For the AMB sentences, match between prosodic break placement and
Interestingly, when the data for the AMB sentences were separated into RB/LB readings,
the NS group also tended to prefer the LB reading slightly over the RB reading.
In the chapter that follows, possible accounts for these results will be presented,
and the answer to our initial research questions will be provided. Additionally, a
46
CHAPTER 4
This chapter will discuss possible accounts for the results of the experiment, as
well as present future directions for research and limitations of the study. The current
study examined how L2 Japanese learners apply prosody in oral reading to sentences
containing branching modifiers, and how this prosodic pattern was related to their
interpretation of the sentence. The research questions posed in Chapter 1 will provide the
The first question proposed for investigation at the outset of this study was 1)
when reading aloud? As little data is available on prosody-syntax mapping in L2, a goal
of this study was to observe the nature of this relationship in L2 Japanese. Two types of
47
further divided into right-branching and left-branching configurations, disambiguated by
lexical plausibility. This type of sentence assumed that subjects were able to utilize the
disambiguating word choice and apply prosody which matched with this syntactic
implausible for modification by the pronominal modifier in the [N1] position and the
ULB used similar word selection for [N2] position. Syntactic representations of the
The JFL learner groups were able to correctly place a prosodic break in 61% of
the URB sentences, and 82% of the ULB sentences. When the JFL learners were
separated by proficiency level, results were similarly low on URB for all groups, with the
only markedly higher figure being the 94% correct match on ULB found in Group 3. 14
configuration. In (2) above, such a reading would mean that [yasui apaato no] ‘the cheap
14
It should be noted that one participant in Group 3 applied an LB prosody to all of the unambiguous
sentences, both URB and ULB, resulting in a 100% prosody-syntax match in ULB for this subject.
48
apartment’ would be read as one grouping, with a relatively longer break being found
Similar results were found by Shibata & Hurtig (2008) regarding the production
disambiguate test sentences prior to oral reading. Recordings of JFL learners’ productions
were played to a group of native speaker judges, who were asked to determine whether a
result found for all JFL levels examined. The percentages, by learner level, of the
sentences correctly identified by the NS judges in their study were: Advanced: RB 74%,
their acoustic analysis showed that in LB sentences subjects used fewer improper
prosodic cues compared to RB, such as misplaced breaks or failure to raise pitch at a
prosodic boundary. Although the methodology of Shibata & Hurtig’s study differs from
the current experiment, their finding that LB prosody was produced more accurately and
was understood more often by the NS raters is consistent with the higher percentage of
prosody-syntax match that was observed in our ULB test stimuli. Note that the current
15
In their study, proficiency level was determined by performance on the Japanese Language Proficiency
Test (JLPT) Level 3 by taking the average of the scores on the grammar and listening sections. Average
scores by group were: ADV 92%, INT 77%, NOV 33%.
49
Shifting the focus now to NS performance on the URB/ULB sentences in our
for the URB/ULB sentences, both the NS and JFL groups had a high percentage of
correct responses (NS: 100% and 94%, Group 1: 88% and 88%, Group 2: 82% and 80%,
Group 3: 75% and 88%; refer to Figure 3.3), indicating that they accurately
comprehended these types of test sentences. Considering this, how can we account for
this divergent trend in the application of prosody between the NS and JFL learners? First,
the observation by Shibata & Hurtig (2008) that RB structures are indicated by some type
the NS’s performance on the URB sentences. By placing a pause between [ADJ] and
[N1], speakers may be vocally indicating the modification relationship in these branching
structures.
Next, the small sample size of the NS group (four participants) meant that each
individual’s performance heavily influenced the data outcome. Three out of four NS
subjects failed to produce the correct prosody for one of the ULB stimuli, [[oboetai
itariago no] zemi] ‘a seminar on the language he wanted to learn, which is Italian,’
resulting in a lower percentage of correct match in ULB. When the constituents of this
target structure were examined more carefully, it was noticed that the noun itariago
‘Italian’ turns the pre-nominal modifier into a non-restrictive relative clause, as can be
noted in the English translation as well. Due to the fact that some speakers may choose to
place a pause between the non-restrictive relative clause and its head in non-restrictive
50
relative clauses, in this particular stimuli the pause between the [ADJ] and [N1] may have
been more prominent than the expected LB prosodic break. Performance on prosody-
interpretation match for the NS group would be predictably higher if this ULB sentence
was removed.
Dialect was also considered as a possible factor influencing prosody, as it was not
performance did not seem to differ from other subjects in terms of preference for a
literature as possible contributors to prosodic phrasing (e.g. Fodor, 2002; Jun, 2003;
Maynell, 2005). Jun (2003) found that ambiguous relative clauses tend to be realized with
such as English tended to produce a prosodic break consistent with a structurally “low”
attachment. In the case of the URB/ULB sentences used in the current study, a high-
attachment preference would correspond to a URB prosody, i.e., a break between [ADJ]
and [N1], while a low attachment reading would correspond to a ULB reading, with a
break between [N1] and [N2]. These prosodic realizations of attachment preferences
seem to correlate with the results found in this study’s data, that is, that the JFL group
performed better on the ULB sentences, while the NS group performed better on URB.
This result necessitates further investigation into L1 transfer as a factor in prosody in oral
reading.
51
In the AMB test stimuli, with two possible prosodic break locations, JFL learners
result was found in all three JFL proficiency groups, and is again consistent with the LB
preference toward LB-consistent prosody as well, although the figures were nearly even
(RB: 14, LB: 18). This result could again be attributed to the sample size, as one subject
in the NS group read six out of eight (75%) of the AMB-type sentences with an LB
prosody, displaying a preference similar to the JFL groups.17 Removing this subject
In this thesis, the question was also raised as to how prosodic phrasing in oral
identifying sources of difficulty for JFL learners, whether they stem from
for URB in all JFL groups, and only Group 3 had a high percentage of match in the ULB
sentences (91%) (refer to Figure 3.4). This low prosody-interpretation match was
16
Refer to Table 3.1.
17
This subject was from the Nagoya area, as noted earlier. Further investigation is necessary to determine if
dialect plays a role in the application of prosody in these sentences.
52
somewhat predictable for these types, as the structure was unambiguous by design,
meaning that comprehension questions also had only one possible correct answer. Even
when subjects produced a prosody which did not match syntactic structure, interpretation
was generally correctly selected, as was observed in the high percentage of correct
responses to comprehension questions (Figure 3.3). This means that JFL learners’ ability
to produce the correct prosodic patterns has not yet been fully developed, although they
may be aware of the syntactic structures. With the NS group, a greater percentage of
prosody-interpretation match in the URB types again suggests that for the unambiguous
questions for the URB/ULB types were correct,18 and that RB is the preferred prosody for
the NS group, this accounts for the greater match in the URB sentences.
It was found that prosody-interpretation match in the AMB sentences was higher
than that for the URB/ULB types in all JFL groups (with the exception of Group 3’s ULB
performance), possibly suggesting an effect of L1 transfer. It was clear that JFL learners
preferred to place a syntactic break in the LB location in the AMB sentences (see Table
3.1), consistent with attachment preferences in English L1 mentioned earlier. When the
noted that the JFL subjects performed better when reading in the LB configuration,
because of their tendency to place a break in the LB position. Considering the fact that
18
Only one incorrect response among four subjects.
53
English native speakers, the JFL group’s preference for an LB interpretation seems
The NS group unexpectedly displayed a result that paralleled the JFL group in the
configuration prosody was applied, the NS group showed a preference for LB breaks, and
a higher percentage of matching was found when they applied prosody as such (89%).
For these AMB sentences, it was expected that NS would prefer an RB prosody, along
with an interpretation that matched this pattern as well. Examination of individuals in the
NS group found no clear pattern of error on any particular test stimuli, although one
individual could be singled out as having a lower prosody-interpretation match rate (63%
for all AMB types). Another factor possibly contributing to the results on this type is the
Selection of this answer would indicate that the subject recognized the ambiguity in the
branching structure, and eight out of 32 (or 25%) of the AMB-type comprehension
questions were answered with this choice by the NS group. Furthermore, because “either”
choices were considered as matching for data analysis of AMB, this could have increased
19
This choice was available in the comprehension questions to the URB/ULB types as well.
54
4.2 Limitations of the Study
Before concluding, a few limitations of this study must be pointed out. For future
research, a larger sample group would be desirable, particularly for the control group.
were difficult, and a larger sample size may yield different results. The word selection in
the target branching structures used in the test sentences must also be re-examined.
while all other sentences were restrictive relative clauses. In our discussion earlier it was
noted that some speakers may place a pause in non-restrictive relative clauses, regardless
of the branching structure. Regarding the analysis of prosodic break locations, this task
was performed by two people, the experimenter and a native speaker volunteer. Inter-
rater agreement on break location was often quite low for certain groups (see Chapter 3),
and a larger group of raters would have been desirable. Although disagreements on break
location were resolved with sound-analysis software, a larger group of raters would
reduce the need to rely on this step, as one of the goals of this study was to identify
4.3 Conclusion
This thesis presented a case study investigating prosody in oral reading by JFL
learners. The experiment employed branching modifiers and was designed to examine
a) how JFL learners apply prosody to sentences containing branching modifiers when
55
reading aloud and b) if this overt prosody matched interpretation to a comprehension
the experiment. Results indicate that the JFL learners had a higher rate of success in
matching prosody to syntax in left-branching (ULB) sentences than when reading right-
branching (URB) sentences. It was also found that in syntactically ambiguous (AMB)
sentences, all JFL groups preferred an LB prosody when reading, and had a greater rate
of prosody-interpretation match when reading aloud with this prosody. On the contrary,
the NS group had an opposite tendency when reading aloud, with the right-branching
sentences being read with a higher percentage of correct prosody-syntax match when the
test stimuli, the native speakers tended to prefer an LB prosody when reading AMB
sentences aloud, though again the small sample size of this group made drawing
conclusions from this result difficult. These results appear to correspond to cross-
and for JFL learners, it indicates that there was L1 transfer in prosody. Additionally, the
results of this study suggest that the JFL learners’ sentence comprehension and
production of prosody are separate, meaning that production errors are not necessarily
An initial motivation for this study was to investigate the relationship between
prosodic cues in oral reading and syntactic parsing. In terms of overt prosody being
utilized by the language teacher to determine how a learner parses a sentence, the current
results are inconclusive but suggestive. It could be said that L1 transfer is seen in prosody
56
as well, meaning that RB or high attachment readings may be difficult for L1 English
speakers to produce. Shibata & Hurtig’s (2008) experiment found that Japanese learners
made more prosodic errors when attempting to produce RB prosody. As little research
pedagogy. The examination of additional structures, such as relative clauses, along with a
phonetic account of the prosody applied in oral reading, would serve as a starting point
57
APPENDIX A
TEST SENTENCES
58
Instruction Sentence:
Practice Sentences:
59
Unambiguous Sentences -- Right-Branching (URB):
Left-Branching (ULB):
61
‘Mr. Deguchi married the daughter of a famous police officer.’
‘Mr. Deguchi married the famous daughter of a police officer.’
(6) Nishida-san ga [genki na butyoo no hisyo] to syuttyoo simasita.
Mrs. Nishida-NOM energetic section chief-GEN secretary-with business trip do-PAST
‘Mrs. Nishida went on a business trip with the secretary of the energetic section
chief.’
‘Mrs. Nishida went on a business trip with the energetic secretary of the section
chief.’
(7) Wilson-san ga [benri na konpyuutaa no keesu] o takusan urimasita.
Mr. Wilson-NOM convenient computer-GEN case-ACC many sell-PAST
‘Mr. Wilson sold a lot of cases for convenient computers.’
‘Mr. Wilson sold a lot of convenient cases for computers.’
(8) Jones-san ga [hen na sima no hito] o oozee tasukemasita.
Mrs. Jones-NOM strange island-GEN person-ACC many help-PAST
‘Mrs. Jones helped many people from a strange island.’
‘Mrs. Jones helped many strange people from an island.’
Filler Sentences:
62
Mr. Shirota-NOM condition-NOM bad become-PAST because medicine-ACC
drink-PAST
‘Mr. Shirota felt bad, so he took the medicine.’
(4) Bailey-san ga tenki no ii hi ni umi ni dekakemasita.
Mr. Bailey-NOM weather-NOM good day ocean-to go out-PAST
‘Mr. Bailey went to the ocean on a day when the weather was good.’
(5) Oda-san ga tatami no heya ni tomarimasita.
Mr. Oda-NOM tatami-GEN room-LOC stay-PAST
‘Mr. Oda stayed in a room with tatami mats.’
(6) Barker-san ga uketuke no hito ni nedan o kikimasita.
Mrs. Barker-NOM reception-GEN person-DAT price-ACC ask-PAST
‘Mrs. Barker asked the receptionist the price of a room.’
(7) Taguti-san ga Sinzyuku de tikatetu ni notte, sigoto ni ikimasita.
Mr. Taguchi-NOM Shinjuku-LOC subway ride work-to go-PAST
‘Mr. Taguchi rode the subway at Shinjuku and went to work.’
(8) Clark-san ga intaanetto de nihon no hoteru o sirabemasita.
Mrs. Clark-NOM internet-INST Japan-GEN hotel-ACC look up-PAST
‘Mrs. Clark looked up hotels in Japan on the internet.’
(9) Perry-san ga Tyuugoku no ohasi o umaku tukaimasita.
Mr. Perry-NOM China-GEN chopsticks-ACC well use-PAST
‘Mr. Perry used the Chinese chopsticks well.’
(10) Takayama-san ga ginkoo no madoguti de okane o moraimasita.
Mrs. Takayama-NOM bank-GEN window-LOC money-ACC receive-PAST
‘Mrs. Takayama received the money at the bank window.’
(11) Lee-san ga Nagoya no zimusyo ni basu de kayotte imasu.
Mr. Lee-NOM Nagoya-GEN office-LOC bus-INST commuting
63
‘Mr. Lee takes the bus to the Nagoya office.’
(12) Yamaguti-san ga kookoo no toki ni eego o benkyoo simasita.
Mrs. Yamaguchi-NOM high school-GEN time English-ACC study-PAST
‘Mrs. Yamaguchi studied English when she was in high school.’
(13) Chen-san ga ie ni kaetta ato de, gomi o sutemasita.
Mr. Chen-NOM house-to return-PAST after trash-ACC throw away-PAST
‘Mr. Chen threw the trash away after he returned home.”
(14) Yamamura-san ga nimotu o orosite kara, takusii o yobimasita.
Mrs. Yamamura-NOM luggage-ACC drop off taxi-ACC call-PAST
‘Mrs. Yamamura dropped off her luggage, then called a taxi.’
(15) Norman-san ga karui hako o haha ni okurimasita.
Mr. Norman-NOM light package-ACC mother-to send-PAST
‘Mr. Norman sent the light package to his mother.’
(16) Kitamura-san ga hidoi ame no naka de go-kiro mo unten simasita.
Mrs. Kitamura-NOM terrible rain-GEN in five kilometers drive-PAST
‘Mrs. Kitamura drove 5 kilometers in the heavy rain.’
64
APPENDIX B
SENTENCE PRESENTATION
65
Instruction Sentence:
Practice Sentences:
Test Sentences:
67
Q: What did Mrs. Grey eat?
a) four of her favorite hamburgers
b) more than one hamburger
c) either a) or b)
d) neither a) or b)
68
d) neither a) or b)
69
c) either a) or b)
d) neither a) or b)
70
b) at a small inn in a village
c) either a) or b)
d) neither a) or b)
71
a) the time of the TV program that she wanted to watch
b) the TV program that she wanted to watch
c) either a) or b)
d) neither a) or b)
72
Q: Where does Mr. Lee commute to?
a) to his office
b) to Nagoya
c) either a) or b)
d) neither a) or b)
73
d) neither a) or b)
74
c) either a) or b)
d) neither a) or b)
76
REFERENCES
---. 1991. “Speaking of Prosody: Fluency's Unattended Bedfellow.” Theory Into Practice
30.3: 165-175.
Eagan, R. 1975. “An Investigation into the Relationship of the Pausing Phenomena in
Oral Reading and Reading Comprehension.” The Alberta Journal of Educational
Research 21: 278-288.
Fodor, J.D. 2002. “Prosodic Disambiguation in Silent Reading.” NELS 32: 113–132.
77
Jorden, Eleanor H., and Noda, M. 1988. Japanese: The Spoken Language, Part 2.
New Haven, CT: Yale University Press.
---. 1990. Japanese: The Spoken Language, Part 3. New Haven, CT: Yale University
Press.
Kang, S., Speer, S. and Nakayama, M. 2006. “Effects of Intonational Phrase Boundaries
on Ambiguous Syntactic Clause Boundaries in Japanese.” Japanese/Korean
Linguistics 14: 77-88.
Kowal, S., O'Connell, D., O'Brien, E., and Bryant, E. 1975. “Temporal Aspects of
Reading Aloud and Speaking: Three Experiments.” American Journal of
Psychology 88: 549-569.
78
Lovric, N., Bradley, D, and Fodor, J.D. 2001. “Overt Prosody and Ambiguity Resolution
in Silent Reading.” Poster presented at the 7th Annual Conference on
Architectures and Mechanisms in Language Processing (AMLaP). Saarbrücken,
Germany.
Mathson, D., Allington, R, and Solic, K. 2006. “Hijacking Fluency and Instructionally
Informative Assessments” 106-119. Fluency Instruction. Eds. Rasinski, T.,
Blanchowicz, C., and Lems, K. New York: Guliford.
Schwanenflugel, P.J., Hamilton, A.M., Kuhn, M.R., Wisenbaker, J.M. and Stahl, S.A.
2004. “Becoming a Fluent Reader: Reading Skill and Prosodic Features in the
Oral Reading of Young Readers.” Journal of Educational Psychology 96.1: 119-
129.
79
Speer, S. and Blodgett, A. 2006. “Prosody” 505-537. Handbook of Psycholinguistics, 2nd
Edition. Eds. Traxler, M.J. and Gernsbacher, M.A. NY: Elsevier Inc.
Venditti, J.J. 2006. “Prosody in Sentence Processing” 208-217. The Handbook of East
Asian Psycholinguistics, Vol. 2. Eds. Nakayama, M., Mazuka, R. and Shirai, Y.
Cambridge: Cambridge University Press.
---. 1995. “Japanese ToBI Labeling Guidelines.” Unpublished Manuscript. The Ohio
State University.
80