Professional Documents
Culture Documents
Simon Williams
Disfluency and Proficiency in Second Language
Speech Production
Simon Williams
Disfluency and
Proficiency in Second
Language Speech
Production
Simon Williams
School of Media, Arts and Humanities
University of Sussex
Falmer, UK
© The Editor(s) (if applicable) and The Author(s), under exclusive licence to Springer Nature
Switzerland AG 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and trans-
mission or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Palgrave Macmillan imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
For Cynthia, Janet and Maureen
Acknowledgements
vii
Transcription Symbols
Below are the main symbols used within the book. Additional symbols
are described in examples in certain chapters.
ix
Contents
1 I ntroduction 1
1.1 Disfluency as Fluency 1
1.2 Introduction to the Six Disfluencies: Formal Descriptions 5
1.3 Effects on Listeners 10
1.4 Main Types of Disfluency Data 12
1.5 Disfluency Types and Proficiency Level 14
1.6 Disfluencies in Public Language Test Descriptors 15
1.7 Comment 16
1.8 Overview 19
References 24
2 S
ilent Pauses 31
2.1 Introduction 31
2.2 Formal Description of Silent Pauses 32
2.3 Effect of Silent Pauses on Listeners 36
2.4 Silent Pauses in Classroom Interaction 47
2.5 Silent Pauses and Proficiency Level 49
2.6 Silent Pauses in Public Language Tests 61
2.7 Comment 64
References 65
xi
xii Contents
3 F
illed Pauses 73
3.1 Introduction 73
3.2 Formal Description of Filled Pauses 75
3.2.1 Location/Position 75
3.2.2 Frequency 76
3.2.3 Length/Duration 77
3.2.4 Pitch 78
3.3 Effect of Filled Pauses on Listeners 78
3.3.1 uh and um Studies 78
3.3.2 Second Language Rater Studies 85
3.4 Speech Environment of Filled Pauses 94
3.5 Filled Pauses and Speaker Proficiency 96
3.6 Filled Pauses in Public Language Tests 105
3.7 Comment 106
References108
4 P
rolongations117
4.1 Introduction 117
4.2 Formal Description of Prolongations 119
4.3 Effect of Prolongations on Listeners 121
4.4 Main Types of Prolongation Data 131
4.4.1 Corpora Studies 132
4.4.2 Natural Language Processing 135
4.4.3 Prolongations in Classroom Interaction 137
4.4.4 Prolongations in Elicited Data 138
4.5 Prolongations and Proficiency Level 139
4.6 Prolongations in Public Language Tests 139
4.7 Comment 141
References141
5 R
epetitions147
5.1 Introduction 147
5.2 Formal Description of Repetitions 149
5.3 Effect of Repetitions on Listeners 157
5.4 Repetitions Outside the Classroom 164
5.5 Repetitions in Elicited Data 165
Contents xiii
6 S
elf-Corrections177
6.1 Introduction 177
6.2 Formal Description of Self-Corrections 180
6.3 Effect of Self-Corrections on Listeners 185
6.4 Main Types of Self-Correction Data 189
6.4.1 Self-Corrections Outside the Classroom 189
6.4.2 Self-Corrections in Classroom Interaction 191
6.4.3 Task Effect on Self-Corrections in Elicited Data 196
6.5 Self-Corrections and Proficiency Level 198
6.6 Self-Corrections in Public Language Tests 202
6.7 Comment 205
References206
7 F
alse Starts213
7.1 Introduction 213
7.2 Formal Description of False Starts 215
7.3 Effect of False Starts on Listeners 217
7.4 Main Types of False Starts Data 222
7.4.1 False Starts Outside the Classroom 223
7.4.2 False Starts in Classroom Interaction 224
7.4.3 Task Effect on False Starts in Elicited Data 231
7.5 False Starts and Proficiency Level 235
7.6 False Starts in Public Language Tests 240
7.7 Comment 242
References243
8 C
onclusion247
8.1 Introduction 247
8.2 Formal Description of Disfluencies 247
8.3 Effects of Disfluencies on Listeners 250
xiv Contents
I ndex269
Acronyms
xv
xvi Acronyms
xvii
List of Tables
xix
xx List of Tables
Table 4.4 Filled pause tokens in four corpora in Clark & Tree (2002) 133
Table 4.5 Key differences in filled pauses vs prolongations in Moniz
et al. (2007) 138
Table 5.1 Three subclasses of disfluent repetition in Plauché &
Shriberg (1999) 155
Table 5.2 Repetitions in post-semester interview home and abroad
participants (extract adapted from Table 8. Comparisons of
repair phenomena for at home and abroad groups, Freed,
1995, p. 141) 159
Table 6.1 CEFR level comparability with OET and Aptis tests 207
Table 7.1 False starts (T-test Mean totals) in four picture narratives
(Tavakoli & Foster, 2011; Foster & Tavakoli, 2009) 234
Table 7.2 Repairs in Van Hest (1996) 237
Table 8.1 Disfluencies and errors in three semi-spontaneous
monologues256
Table 8.2 Measures of speaker speed and breakdown fluency to
achieve consistency (but not statistical significance) across
proficiency levels in the Aptis speaking test 257
1
Introduction
more use of the filled pause than women when they are searching for a
word and want to hold the floor (Shriberg, 2001). The two types of inter-
actional approach, i.e. the language production approach and the social
interaction approach (Faerch & Kasper 1983, 1984; and see Chiang &
Mi, 2011 for a discussion) represent a long-standing, cognitive-social
debate on second-language acquisition (SLA) (cf. Firth & Wagner, 1997,
2007; Gass, 2004).
Perhaps because of the high demands of research time and, apart from
test providers and certain government agencies, the paucity of funding
for most spoken language research, questions regarding individual differ-
ences, noted by Engelhardt et al. (2019), and the speech style and stock
of disfluencies produced by the same speakers in different conditions and
in response to social judgements, remain under-researched. The range of
fluencies demanded and codified by cultures was early recognised in first
language studies: the fluency of the DJ is different from that of the dip-
lomat, which is different again from the after-dinner speaker, though all
three would be acknowledged as notably fluent in their professional lives.
Influenced in this way by a mix of speech style and cultural norms, dis-
fluencies as conversational markers become rhetorical strategies, associ-
ated both with speech genres and with personal identity.
The six kinds of disfluency discussed in these chapters have all been
classed as hesitation (Maclay & Osgood, 1959), with silent or unfilled
pauses occurring most frequently.
Perhaps the most recognizable feature associated with the absence of
fluency is pausing. It is present to a greater or lesser extent on its own and
as part of other forms of dysfluency. Many speakers interrupt the flow of
speech, creating a temporary hiatus and a moment of silence. Silent
breaks occur often in conversation and if less than 0.25 milliseconds are
not normally noticed by an interlocutor. Pauses of longer duration, how-
ever, are marked and are likely to cause the interlocutor to search for an
explanation and make a judgement about the quality of the exchange.
For example, speaker pauses to conduct a word search have been shown
to alert the listener to the less ordinary nature of the following word, aid-
ing mutual understanding and contributing to the success of the discourse.
As another form of codified description of (dis)fluency, the level
descriptors of well-known English language tests are useful in illustrating
this point. With first and second language development, disfluencies
1 Introduction 5
never wholly disappear but rather become invisible through cultural stan-
dardisation. Thus, the remarkable mid-juncture silences of the L2 begin-
ner migrate towards end-juncture and become longer though less
noticeable. Filled pauses become lexicalised, e.g. like and yeah, and func-
tion words may be imperceptibly prolonged as the speaker searches for
a low frequency content word. Function words and phrases may also be
repeated in hardly noticeable close succession for the same reason. The L2
learner becomes adept at self-correcting from the previously articulated
function word, so contextualising the repair and minimising the mental
work of the listener. The same principle applies to false starts, with the
added advantage to the discourse that the repair is information more
obviously intended for the benefit of the listener at a here-and-now level
of interaction rather than an attempt to conform to more general social
norms of language accuracy. Eventually, the L2 (dis)fluency patterns are
indistinguishable from those of an L1 speaker, save that wide individual
variation in the use of filled and silent pauses and repetitions, all delaying
tactics, is well-recognised and accepted.
language to make it sound more natural and increase the time available
for processing responses. If any evidence were needed that the status of
disfluencies has been revised, one example is their inclusion in speech
synthesis, e.g. filled pauses (Adell et al., 2010), in order to simulate spon-
taneous rather than read speech.
The interest in second language self-corrections is concentrated par-
ticularly in language teaching practitioners and specialists keen to maxi-
mise the acquisition of the L2 by learners. Evidence exists that at an early
stage, language learners prioritise acquiring the formal rules of the target
language and devote considerable energy to eliciting models from expert
others, gradually internalising them.
Silent pauses (SPs) as stretches of silence within the speech stream
occur naturally at clause and sentence boundaries but are heard as disflu-
encies when they interrupt a word, phrase or clause. Pause duration is less
significant in this regard, though longer boundary clauses are associated
with longer following sentences (Goldman-Eisler, 1972). In this way,
pauses signal an impending information increase, and words following
pauses are found to be less predictable. Words that are not preceded by
pauses are therefore more predictable, and fluency as absence of mid-
clause pausing is associated with predictability or word frequency, provid-
ing some justification for encouraging learners to include formulaic
sequences in their speech as a means to increase fluency. Pauses can be
lengthy at boundaries without sounding marked. Thus, the distribution
of pauses is more significant than their duration.
Filled pauses (FPs) can be defined as voiced hesitation such as er and
erm in English and include lexical fillers such as yeah, like and you know
that carry no additional information. Speakers may produce filled pauses
for at least three reasons, some of which overlap, depending on the situa-
tion. Of interest to psycholinguists is the filled pause as marker for a word
search. Interactionists are concerned with the filled pause as back-
channelling. And discourse analysts will note the turn-taking function of
filled pauses. As with silent pauses, work has centred on the distribution,
duration, and frequency of filled pauses. And like silent pauses, the
unmarked location of filled pauses is at major boundaries, where they
might indicate either a desire to continue speaking or an invitation to the
listener to take the floor, depending on pitch and intonation. Again, like
8 S. Williams
silent pauses, filled pauses are more likely to occur before a low-frequency
content word and before longer clauses; therefore, like silent pauses in
this regard, they indicate syntactic uncertainty. The duration of FPs is
markedly shorter than SPs and produced at lower pitch than the sur-
rounding words.
Prolongation is the continuation of the immediately prior phoneme
and is classed as disfluent when it continues for longer than usual relative
to similar phones produced by the same speaker. The prolonged sound
and the usual length of production can be measured digitally for com-
parison. Prolongations are one way of the speaker’s achieving extra time
for planning and information processing without resorting to a more
noticeable repetition, filled pause, or silent pause. In this sense, prolonga-
tions may be the first option for speakers who need extra time. Three-
quarters of disfluent prolongations occur in function words rather than
content words. Speakers’ preferred location for prolongations is the long
vowel in the nuclei of syllables but the location is capable of migrating,
e.g. to a sonorant in the coda, when options are reduced.
Repetitions are contiguous repeats of the same sound, syllable, word,
or phrase that convey no additional meaning for the listener and are more
likely to be heard as disfluent than discontinuous repeats, i.e. repeats
separated by some intervening language. Like the location of prolonga-
tions, repetitions are more likely to be function words that occur before
content words, presumably allowing the speaker time to search for the
lexical item. Repetitions are most often of single words, occasionally
more, and rarely less. Like filled pauses, repetitions are often found near
to, or at the start of, an utterance (Schegloff, 1987), possibly to repair
speaker overlaps. By analysing intonation and pause patterns surround-
ing repetitions, Plauché and Shriberg (1999) offer a model of three rep-
etition types: canonical, covert self-repair, and stalling, illustrated with
the definite article representing a function word. As the term suggests,
canonical repetition is the most numerous of the three subtypes, and the
repetition, retrospective in nature, provides some continuity for the lis-
tener. Covert repetition is rather prospective, its rising intonation expres-
sive of the speaker’s planning uncertainty. The function of stalling
repetitions may be to keep the floor during planning problems and they
seem to be simultaneously prospective and retrospective.
1 Introduction 9
Together with SPs, FPs were responsible for 59% of listener judgements
of temporal disfluency in Rossiter (2009); but Cucchiarini et al. (2002)
conclude from their findings that FPs are a poor indicator of fluency and
may instead be a strategy reflecting sentence planning. Perhaps FPs rather
reflect individual differences or broader cultural norms.
Prolongations in the form of longer syllables have been found to occur
before the vast majority of syntactic junctures and are also typical of gram-
matical junctures, where length of syllable is generally longer, as it is before
rather than after pauses (Martin, 1970). Listeners identify prolongations
as disfluencies yet tolerate them well. In rater studies, listeners prefer pro-
longations to compound disfluencies (Betz et al., 2015) and, consistent
with their effect at grammatical junctures, rate prolongations that occur
within conjunctions highly. Prolongations are thus a means of preserv-
ing fluent speech while planning more complex language (Moniz et al.,
2007). Disfluent prolongations, i.e. certain extended marked forms, espe-
cially in function words such as thee for the, have been shown to prime
listeners for a following referent that is new or unfamiliar (Arnold et al.,
2003). Listeners’ response seems to be inferentially learned rather than
associative because when prompts are given by an L2-accented speaker,
the attenuation effect disappears, i.e. listeners show no preference for old
or new referents following a prolongation. It may be that the wide varia-
tion in length of prolongations is what makes them notably acceptable to
listeners. Furthermore, because speakers seem to plan ahead during longer
prolongations, overall speed is not reduced (Betz et al., 2017).
Like FPs, repetitions may serve as planning devices for speakers
(Lennon, 1990). Similarly, when speakers repeat a function word such as
an article, they may be conducting a word search for a following noun
(Temple, 1992). The fact that repetitions have been reported to increase
in number in study abroad groups, presumably as a result of pressure on
speakers to keep pace in interaction with L1 speakers (Freed, 1995),
reveals the consequences of the additional real-time planning necessary in
the new speech environment. Although listeners are often oblivious to
repetitions in interaction and repetitions impose no additional burden on
listener comprehension, their Event Related Potential (ERP) effect,
recorded by Electrical Geodesics (EEG), is similar to prolongations and
consistent with semantically incongruous input (McAllister et al., 2001).
12 S. Williams
The incongruity may explain why listeners in rater studies see them as a
key indicator of disfluency (Rossiter, 2009).
In general, self-corrections seem not to influence listeners’ judgements
of speaker fluency: although listeners are sensitive to repairs, their judge-
ment is barely affected (Bosker et al., 2013). An exception is reported in
McRobie (1993), whose raters awarded higher fluency scores to non-
corrected vs corrected speech. Expert L1 and L2 raters’ written comments
as they audited L2 speech showed that they consistently reacted more
negatively to self-corrections than false starts (Rossiter, 2009). To sum
up, where self-correction is a named fluency measure in listener rating
studies, it hardly figures in raters’ perceptions.
False starts are another disfluency that holds meaning for listeners, in
this case by recycling material to create ‘discourse coherence’ (Levelt,
1983, p. 42). ERP data shows that false starts also involve a processing
cost (McAllister et al., 2001) and slow monitoring times (Hindle, 1983),
which becomes evident from the more numerous false starts of beginner
and intermediate-level speakers compared to advanced-level speakers.
Lower level speaker FSs also tend to be ill-formed (van Hest, 1996), pos-
sibly reflecting working memory capacity. Although advanced learners
produce fewer false starts, for that reason their false starts may be more
distracting. Comprehension is also retarded by the occurrence of false
starts in the middle of an utterance, and near the start but introduced by
a conjunction, rather than without any marker at the beginning. In gen-
eral, however, rater studies reveal that although listeners notice false starts,
they are one of the disfluencies least likely to elicit a negative reaction
(Rossiter, 2009).
from the central section, it may feature within one of the other sections.
For example, no separate section exists for elicited data in the chapter on
silent pauses. Yet elicited data is central to the rater studies reported
within the ‘effect on listeners’ section in the same chapter.
In fact, all six disfluencies feature in the core group of elicited rater
studies, most of which follow a similar design. Listeners rate speech
recordings of L2 learners for fluency. The recordings are subsequently
analysed for acoustic measures such as speed and pauses and the results
compared with the ratings. Inferences are made about the features that
listeners responded to in making their judgements on fluency. These
core studies are given a separate section, Effects of [the disfluency] on lis-
teners, in central chapters. Data from classroom interaction is included
in Chap. 2: Silent pauses for a number of reasons. Walsh and Li (2013)
illustrate the problem of assigning between-turn pauses in interaction
to one or other speaker and discuss the implications when one of the
speakers is a teacher eliciting responses to questions and the other a
student who is socially reticent or requires time to think or both. The
shorter pauses observable in learner pair work contribute to fluency in
dialogic speech (Tavakoli, 2016). Longitudinal study of learners shows
a steady decline in pause time with proficiency gains, though quantity
and pause distribution mid-clause still convey disfluency (Mora &
Valls-Ferrer, 2012).
Apart from appearing in other sections such as effects on listeners and
proficiency, filled pause data in Chap. 3 is restricted to the uh and um
studies in the elicited data section. Prolongations are the only disfluency
whose study data is largely based on corpora and natural language pro-
cessing: they do not enjoy autonomous sections in any other chapter.
Although Chap. 4: Prolongations includes short sections on classroom
and elicited data, prolongations feature only in transcripts of classroom
interaction and are rarely the subject of analysis; they also figure in elic-
ited data collected in a classroom and gathered into a corpus in Moniz
et al. (2007). There is no classroom data section in Chap. 3: Filled pauses
or Chap. 5: Repetitions. Corpora and natural language processing sec-
tions are included only in Chap. 4. Natural language data collected out-
side the classroom is confined to Chaps. 5, 6 and 7 on repetitions,
self-corrections, and false starts.
14 S. Williams
Pauses are cited in at least one band level descriptor for all five tests but
none except OET, which refers to ‘fillers’, specifies whether SP or
FP. Tavakoli et al. (2017) recommends that Aptis rating scales and training
materials should specify pauses as silent, filled, or mid-clause as necessary.
Prolongations are not referred to in any of the descriptors. Like pauses,
repetitions are included in one or more descriptors from all five scales.
(Self-)corrections are referred to in the CEFR descriptors, the IELTS and
OET tests, and as the more generic ‘reformulations’ in the Aptis test.
False starts are named in the CEFR descriptors, and those for the Aptis,
PTE Academic and TOEFL iBT rating scales. Anomalously, Aptis there-
fore refers to both reformulations and false starts.
The disfluencies that are mentioned most often are pauses and repeti-
tions. Presumably, these forms are most salient to listeners and exam-
iner raters.
1.7 Comment
The preceding outline has been largely descriptive, reflecting the early
state of the disfluency field. It seems that writers share a common pur-
pose: to better understand the causes, workings, and consequences of
disfluencies in spoken language and, in the case of L2 investigators, the
special considerations that might apply to speakers of a second language.
Yet, secondary interests, including language acquisition, task-based per-
formance, human-machine interfacing, test standardisation, and listener
responses, to name just a few, mean that differences in working defini-
tions, methodologies, and conclusions, have resulted in a centrifugal
push of knowledge concentration into a number of different specialisms
rather than a centripetal pull towards a common understanding.
The constructs of fluency and disfluency can best be understood as
interactive, socially-produced, and jointly-constructed phenomena (cf.
Jaspers, 2016; Segalowitz, 2016) that have significance for the iterative
reproduction of social relations. In this respect, Segalowitz’s (2016)
framework offers more global insights into fluency and therefore disflu-
ency, and is helpful in foregrounding the crucial role of the social context
and its intimate relation with speaker motivation. The essential
1 Introduction 17
L2 speech
production
Cognitive-
perceptual Motivation
processing
Fig. 1.1 Framework for L2 (dis)fluency and proficiency (based on Segalowitz, 2016)
18 S. Williams
When children learn to use the socially approved variety of spoken lan-
guage in school, it is not from what their teachers explicitly teach in class,
but rather from adjusting their speech to match the speech of other chil-
dren, in halls, on the playground, and outside of school, and thus gain their
approval.
(Fasold & Connor-Linton, 2014, p. 10).
Although both self-repairs are noticeably disfluent, the first with silent
pauses, filled pauses, prolongations, and repetitions; the second with
filled pauses, Inez accomplishes her communicative objective. Which
intervention strategies speakers adopt to achieve a temporary break in
fluent speech will thus vary as to their self-ascribed identity, the values
they align with, and the interactant. In fact, the disfluencies speakers
produce in the service of repair bear only an indirect relationship to
metadiscoursal social judgements; it is rather the speech style adopted by
the speaker as a consequence of their social judgement that shapes the
speech process and defines its points of vulnerability. That style, whether
impetuous or measured, familiar or formal, confident or hesitant—hav-
ing regard to topic, social context, size of audience, and other environ-
mental factors—determines the speaker’s pre-planning, their lexical
range, syntactic complexity and so on, and predicts where breakdowns
and disfluencies are likely to occur.
1.8 Overview
Apart from introduction and conclusion, the inner six chapters (Chaps.
2–7) are each devoted to a recognised disfluency. Most of the chapters are
organised in a similar way, with some variation reflecting the sort of
methodology and data characteristic of work in that disfluency area.
20 S. Williams
gain processing time. Precisely for this reason, prolongations are of inter-
est to automatic speech recognition researchers, whose concern is to
establish the maximum tolerable length of prolongation for the listener.
Prolongations have long been recognised and reproduced in transcrip-
tions by conversation analysts, conventionally by a colon following the
extended syllable. The second section offers a formal description, noting
the tendency of prolongations to fall on function words such as and, the
and so. The largest section reports the effects of prolongations on listen-
ers, including extended forms of function words such as a and the occur-
ring before uh and um, the FPs discussed in the previous chapter. Apart
from studies in human-machine interaction, the majority of prolonga-
tion data is found in corpora, e.g. the corpus in Fox Tree and Clark
(1997), which examines the incidence of thi:y as an extension of the
strong form thee (ordinarily thuh). Three more sources of data exempli-
fied in this section are natural language processing, interaction, elicited
data. Prolongations and proficiency level are discussed only in the context
of language test descriptors and then by inference from mention of hesi-
tations. The functional association of prolongations with filled pauses is
noted, along with their appearance in contrasting environments.
Chapter 5 distinguishes immediate repeats from language repeated
after some delay and suggests that immediate or ‘concatenated’ repetition
(Maclay & Osgood, 1959, p. 148) is heard as disfluent. The second sec-
tion offers a formal description of repetitions and includes the repetition
often found in repairs and a description of Plauché and Shriberg’s (1999)
taxonomy of repetitions by intonation. The following section examines
the effect of repetitions on listeners and reports several of the rater studies
with reference to findings on repetitions. After noting L2 study outside
the classroom, and in a set of empirical studies, and the effect of task on
L2 learners’ repetitions, a section on repetition and proficiency level
revisits some of the rater studies for conclusions drawn on the association
of repetitions with fluency. The penultimate section surveys references to
repetitions in the public descriptors of English language speaking tests.
The chapter ends by noting the joint occurrence of repetitions with
22 S. Williams
pauses, after which the speaker restarts the utterance with a function
word preceding the lexical word causing trouble, and the fact that listen-
ers often fail to notice speakers’ repetitions of this kind, which are most
likely sites of speakers’ covert repair.
Chapter 6 concerns self-corrections, one of the two compound refor-
mulation disfluencies. Self-corrections are distinct from false starts, the
other kind of reformulation, for a number of reasons. In a self-correction,
the speaker attempts to repair the utterance towards a form compliant
with a social norm. For the same reason, because the repair is usually
towards a standard in pronunciation, grammar, or lexis, the trouble
source is marked as non-standard and the whole repair heard as a disflu-
ency. A formal description of self-corrections in the second section refers
to the seminal paper of Schegloff, Jefferson and Sacks (1977) and their
work on L1 repairs. Levelt (1983), also working with L1 (Dutch) data,
defines five categories of repair, one of which is self-correction, and three
of the others forms of false start. It is Levelt that L2 researchers such as
Kormos build on to describe features of L2 learner self-corrections and
other repairs. The next section reports studies on the effects of self-
corrections on listeners. The following three sections discuss studies of
self-corrections whose data has been collected outside the classroom,
studies of self-corrections in classroom interaction, and the effect of task
on self-corrections in elicited data. Van Hest (1996) and Williams and
Korko (2019) report that self-corrections decline with proficiency gains;
and Tavakoli et al. (2017) note that lower-intermediate learners make
abundant self-corrections as they actively experiment with achieving tar-
get language forms. Only two English language speaking tests, IELTS
and the Occupations English Test (OET), refer to self-correction in their
public descriptors, but the references may be to the more generic ‘repair’,
i.e. including false starts.
Chapter 7, the last chapter on disfluency types, describes false starts,
as distinct from L2 self-corrections, the other kind of repair. Of more
interest than self-corrections as a linguistic opportunity for intake and
acquisition, false starts represent the speaker’s revision of a standard
form that would not necessarily be recognised as a trouble source by the
1 Introduction 23
References
Adell, J., Bonafonte, A., & Escudero-Mancebo, D. (2010). Modelling filled
pauses prosody to synthesise disfluent speech. In Speech Prosody 2010-Fifth
International Conference.
Ahmadian, M. J., Abdolrezapour, P., & Ketabi, S. (2012). Task difficulty and
self-repair behavior in second language oral production. International Journal
of Applied Linguistics, 22(3), 310–330.
Arnold, J. E., Fagnano, M., & Tanenhaus, M. K. (2003). Disfluencies signal
theee, um, new information. Journal of Psycholinguistic Research, 32(1), 25–36.
Betz, S., Wagner, P., & Schlangen, D. (2015). Modular synthesis of disfluencies
for conversational speech systems. Studientexte zur Sprachkommunikation:
Elektronische Sprachsignalverarbeitung, 2015, 128–134.
Betz, S., Zarrieß, S., & Wagner, P. (2017). Synthesized lengthening of function
words - The fuzzy boundary between fluency and disfluency. In Proceedings of
the International Conference Fluency and Disfluency.
Bortfeld, H., Leon, S. D., Bloom, J. E., Schober, M. F., & Brennan, S. E. (2001).
Disfluency rates in spontaneous speech: Effects of age, relationship, topic,
role, and gender. Language and Speech, 44, 123–147.
Bosker, H. R., Pinget, A. F., Quené, H., Sanders, T., & De Jong, N. H. (2013).
What makes speech sound fluent? The contributions of pauses, speed and
repairs. Language Testing, 30(2), 159–175.
Bosker, H. R., Quené, H., Sanders, T., & De Jong, N. H. (2014). The percep-
tion of fluency in native and nonnative speech. Language Learning,
64(3), 579–614.
Brennan, S. E., & Schober, M. F. (2001). How listeners compensate for disflu-
encies in spontaneous speech. Journal of Memory and Language,
44(2), 274–296.
Cenoz, J. (1998). Pauses and communication strategies in second language speech.
Chiang, S. Y., & Mi, H. F. (2011). Reformulation: A verbal display of interlan-
guage awareness in instructional interactions. Language Awareness,
20(2), 135–149.
Clark, H. H. (1994). Managing problems in speaking. Speech Communication,
15(3–4), 243–250.
Council of Europe. (2001). Common European framework of reference for lan-
guages: Learning, teaching, assessment. Cambridge University Press.
1 Introduction 25
Tavakoli, P., Nakatsuhara, F., & Hunter, A. M. (2017). Scoring validity of the
Aptis Speaking Test: Investigating fluency across tasks and levels of proficiency.
ARAGs Research Reports Online. British Council.
Temple, L. (1992). Disfluencies in learner speech. Australian Review of Applied
Linguistics, 15(2), 29–44.
Van Hest, E. (1996). Self-repair in L1 and L2 production. Tilburg University Press.
Van Os, M., De Jong, N. H., & Bosker, H. R. (2020). Fluency in dialogue:
Turn-taking behavior shapes perceived fluency in native and nonnative
speech. Language Learning, 70(4), 1183–1217.
Verhoeven, L. T. (1989). Monitoring in children’s second language speech.
Interlanguage Studies Bulletin (Utrecht), 5(2), 141–155.
Walsh, S., & Li, L. (2013). Conversations as space for learning. International
Journal of Applied Linguistics, 23(2), 247–266.
Watanabe, M., Hirose, K., Den, Y., & Minematsu, N. (2008). FPs as cues to the
complexity of forming phrases for native and non-native listeners. Speech
Communication, 50(2), 81–94.
Wenger, E. (1998). Communities of practice: Learning as a social system. Systems
thinker, 9(5), 2–3.
Williams, S. A., & Korko, M. (2019). Pause behavior within reformulations and
the proficiency level of second language learners of English. Applied
PsychoLinguistics, 40(3), 723–742.
Zuniga, M., & Simard, D. (2019). Factors influencing L2 self-repair behavior:
The role of L2 proficiency, attentional control and L1 self-repair behavior.
Journal of Psycholinguistic Research, 48(1), 43–59.
Another random document with
no related content on Scribd:
There are six of these strange compositions, upon the stories of
David and Goliath, of David and Saul, of Jacob and Leah, and
others. Some years later they undoubtedly suggested to Sebastian
Bach the delicate little capriccio which he wrote upon the departure
of his brother for the wars. Apart from this they are of slight
importance except as indications of the experimental frame of mind
of their composer. Indeed, beyond imitation and to a small extent
description, neither harpsichord nor pianoforte music has been able
to make much progress in the direction of program music.
The various movements lack definite form and balance. The first is in
rather heavy chord style, the chords being supported by a dignified
counterpoint in eighth notes. This leads without pause into a fugue
on a figure of lively sixteenth notes. The key is B-flat major. There
follows a short adagio in E-flat major, modulating to end in C minor,
in which key the last movement, a short allegro in triple time, is taken
up. The whole is rounded off by a return to the opening movement,
signified by the sign Da Capo.
On the other hand, by comparison with the vocal style, the organ
style is free. Where the composer of masses was restricted by the
limited ability of the human voice to sing wide intervals accurately,
the organist was limited only by the span of the hand. Where
Palestrina could count only upon the ear of his singers to assure
accurate intonation, the organist wrote for a keyboard which,
supposing the organ to be in tune, was a mechanism that of itself
could not go wrong. Given, as it were, a physical guarantee of
accuracy as a basis for experiment, the organist was free to devise
effects of sheer speed or velocity of which voices would be utterly
incapable. He had a huge gamut of sounds equally at his command,
a power that could be mechanically bridled or let loose. His
instrument could not be fatigued while boys could be hired to pump
the bellows. So long as his finger held down a key, or his foot a
pedal, so long would the answering note resound, diminishing,
increasing, increasing, diminishing, according to his desire, never
exhausted.
During the first half of the seventeenth century the harpsichord was
but the echo of the organ. Even the collections of early English
virginal music, which in some ways seem to offer a brilliant
exception, are the work of men who as instrumentalists were
primarily organists. In so far as they achieved an instrumental style
at all it was usually a style fitting to a small organ. The few cases
where John Bull’s cleverness displayed itself in almost a true
virtuoso style are exceptions which prove the rule. Not until the time
of Chambonnières and Froberger do we enter upon a second stage.
He learned little of it from what had been written for the organ, but
much from music for the lute, which, quite as late as the middle of
the century, was interchangeable with the harpsichord in
accompaniments, and was held to be equal if not superior as a solo
instrument. It was vastly more difficult to play, and largely for this
reason fell into disuse. The harpsichord is by nature far nearer akin
to it than to the organ. The free style which lutenists were driven to
invent by the almost insuperable difficulties of their instrument, is
nearly as suitable to the harpsichord as it is to the lute. Without
doubt the little pieces of Denis Gaultier were played upon the
harpsichord by many an amateur who had not been able to master
the lute. The skilled lutenist would find little to give him pause in the
harpsichord music of Chambonnières. The quality of tone of both
instruments is very similar. For neither is the strict polyphony of
organ music appropriate; for the lute it is impossible. Therefore it fell
to the lutenists first to invent the peculiar instrumental style in which
lie the germs of the pianoforte style; and to point to their cousins,
players of the harpsichord, the way towards independence from
organ music.
The freedom from polyphonic restraint, inherited from the lute, and
the profusion of graces which have sprouted from the nature of the
harpsichord, mark the diversion between music for the harpsichord
and music for the organ. In other respects they are still much the
same; that is to say, the texture of harpsichord music is still close—
restricted by the span of the hand. This is not necessarily a sign of
dependence on the organ, but points rather to the young condition of
the art. It is not to be expected that the full possibilities of an
instrument will be revealed to the first composers who write for it
expressly. They lie hidden along the way which time has to travel.
But Chambonnières, in France, and Froberger, in Germany, opened
up the special road for harpsichord music, took the first step which
others had but to follow.
[2] At the head of Sebastian Bach’s Musikalisches Opfer stands the Latin
superscription: Regis Iussu Cantio et Reliqua Canonica Arte Resoluta. The initial
letters form the word ricercar.
[5] There was a form of suite akin to the variation form. In this the same melody or
theme served for the various dance movements, being treated in the style of the
allemande, courante, or other dances chosen. Cf. Peurl’s Pavan, Intrada, Dantz,
and Gaillarde (1611); and Schein’s Pavan, Gailliarde, Courante, Allemande, and
Tripla (1617). This variation suite is rare in harpsichord music. Froberger’s suite on
the old air, Die Mayerin, is a conspicuous exception.
[6] 'Denn warum sollte man auf dem Clavier nicht eben wie auf anderen
Instrumenten dergleichen Sachen tractieren können?’ he writes in his preface to
the ‘Seven New Partien,’ 1692.
[7] So they were called in France, which until the time of Beethoven set the model
for harpsichord style. In Germany they were called Manieren.
[8] D’Anglebert published in 1689 a set of pieces, for the harpsichord, containing
twenty variations on a melody known as Folies d’Espagne, later immortalized by
Corelli.
In round figures the years between 1700 and 1750 are the Golden
Age of harpsichord music. In that half century not only did the
technique, both of writing for and performing on the harpsichord,
expand to its uttermost possibilities, but there was written for it music
of such beauty and such emotional warmth as to challenge the best
efforts of the modern pianist and to call forth the finest and deepest
qualities of the modern pianoforte.
Each of the three men whose work is the chief subject of this chapter
is conspicuous in the history of music by a particular feature.
Domenico Scarlatti is first and foremost a great virtuoso, Couperin
an artist unequalled in a very special refinement of style, Sebastian
Bach the instrument of profound emotion. In these features they
stand sharply differentiated one from the other. These are the
essential marks of their genius. None, of course, can be
comprehended in such a simple characterization. Many of Scarlatti’s
short pieces have the warmth of genuine emotion, and Couperin’s
little works are almost invariably the repository of tender and naïve
sentiment. Bach is perhaps the supreme master in music and should
not be characterized at all except to remind that his vast skill is but
the tool of his deeply-feeling poetic soul.
I
It will be noticed that each of these great men speaks of a different
race. We may consider Scarlatti first as spokesman in harpsichord
music of the Italians, who at that time had made their mark so deep
upon music that even now it has not been effaced, nor is likely to be.
His father, Alessandro, was the most famous and the most gifted
musician in Europe. From Naples he set the standard for the opera
of the world, and in Naples his son Domenico was born on October
26, 1685, a few months only after the birth of Sebastian Bach in
Eisenach. Domenico lived with his father and under his father’s
guidance until 1705, when he set forth to try his fame. He lived a few
years in Venice and there met Handel in 1708, with whom he came
back to Rome. Here in Rome, at the residence of Corelli’s patron,
Cardinal Ottoboni, took place the famous contest on organ and
harpsichord between him and Handel. For Handel he ever professed
a warm friendship and the most profound admiration.
Scarlatti wrote many operas in the style of his father, and these were
frequently performed, with success, in Italy, England, Spain, and
elsewhere. During his years at St. Peter’s he also wrote sacred
music; but his fame now rests wholly upon his compositions for the
harpsichord and upon the memory of the extraordinary skill with
which he played them.
We have dwelt thus briefly upon a few events of his life to show how
widely he had travelled and in how many places his skill as a player
must have been admired. That in the matter of virtuosity he was
unexcelled can hardly be doubted. It is true that in the famous
contest with Handel he came off the loser on the organ, and even his
harpsichord playing was doubted to excel that of his Saxon friend.
But these contests were a test of wits more than of fingers, a trial of
extempore skill in improvising fugues and double fugues, not of
virtuosity in playing.
Scarlatti wrote between three and four hundred pieces for the
harpsichord. The Abbé Santini[12] possessed three hundred and
forty-nine. Scarlatti himself published in his lifetime only one set of
thirty pieces. These he called exercises (esercizii) for the
harpsichord. The title is significant. Before 1733 two volumes, Pièces
pour le clavecin, were published in Paris; and some time between
1730 and 1737 forty-two ‘Suites of Lessons’ were published in
London under the supervision of Roseingrave. More were printed in
London in 1752. Then came Czerny’s edition, which includes two
hundred pieces; and throughout the nineteenth century various
selections and arrangements have appeared from time to time, von
Bülow having arranged several pieces in the order of suites, Tausig
having elaborated several in accordance with the modern pianoforte.
A complete and authoritative edition has at last been prepared by
Sig. Alessandro Longo and has been printed in Italy by Ricordi and
Company.
By far the greater part of these many pieces are independent of each
other. Except in a few cases where Scarlatti, probably in his youth,
followed the model of his father’s toccatas, he keeps quite clear of
the suite cycle. The pieces have been called sonatas, but they are
not for the most part in the form called the sonata form. This form
(which is the form in which one piece or movement may be cast and
is not to be confused with the sequence or arrangement of
movements in the classical sonata) is, as we shall later have ample
opportunity to observe, a tri-partite or ternary form; whereas the so-
called sonatas of Scarlatti are in the two-part or binary form, which
is, as we have seen, the form of the separate dance movements in
the suite. Each ‘sonata’ is, like the dance movements, divided into
two sections, usually of about equal length, both of which are to be
repeated in their turn. In general, too, the harmonic plan is the same
or nearly the same as that which underlies the suite movement, the
first section modulating from tonic to dominant, the second back from
dominant to tonic. But within these limits Scarlatti allows himself
great freedom of modulation. It is, in fact, this harmonic expansion
within the binary form which makes one pause to give Scarlatti an
important place in the development of the sonata form proper.
We have said that often he varies his key when thus repeating
himself, and that such variety saves from monotony. But it must be
added that even where there is no change of key he escapes being
tedious to the listener. The reason must be sought in the sprightly
nature of the figures he chooses, and in the extremely rapid speed at
which they are intended to fly before our ears. He is oftenest a
dazzling virtuoso whose music appeals to our bump of wonder, and,
when well played, leaves us breathless and excited.
The pieces are for the most part extremely difficult; and this, together
with his ever-present reiteration of special harpsichord figures, may
well incline us to look upon them as fledgling études. The thirty
which Scarlatti himself chose to publish he called esercizii, or
exercises. We may not take the title too literally, bearing in mind that
Bach’s ‘Well-Tempered Clavichord’ was intended for practice, as
were many of Kuhnau’s suites. But that Scarlatti’s sonatas are
almost invariably built up upon a few striking, difficult and oft-
repeated figures, makes their possible use as technical practice