Transcribing disordered speech: By target or by production?

Abstract

The ability to transcribe disordered speech is a vital tool for speech-language pathologists, as accurate description of a client's speech output is needed for both diagnosis and effective intervention. Clients in the speech clinic often use sounds that are not part of the target sound system and which may, in some cases, be sounds not found in natural language at all. While the IPA provides a wide range of symbols that can be used in clinical transcription, the extended IPA (extIPA) may also be needed to transcribe atypical sounds never or rarely encountered in natural language. When using the IPA and extIPA transcribers aim to show the client's productions, irrespective of the intended target. An alternative tradition of clinical transcription has grown up in the US: the symbols suggested by Shriberg and Kent (SK). In many cases, these symbols are designed to show the intended target with a diacritic illustrating in which way the realization differs from the target. In this article possible confusions that may occur if the SK system is used are discussed, together with problems that may occur when SK and IPA are used together.

Full text

Transcribing disordered speech: By target or by
production?
MARTIN J. BALL
University of Louisiana at Lafayette, Lafayette, LA, USA
(Received 27 January 2008; accepted 20 April 2008)
Abstract
The ability to transcribe disordered speech is a vital tool for speech-language pathologists, as accurate
description of a client’s speech output is needed for both diagnosis and effective intervention. Clients
in the speech clinic often use sounds that are not part of the target sound system and which may, in
some cases, be sounds not found in natural language at all. While the IPA provides a wide range of
symbols that can be used in clinical transcription, the extended IPA (extIPA) may also be needed to
transcribe atypical sounds never or rarely encountered in natural language. When using the IPA and
extIPA transcribers aim to show the client’s productions, irrespective of the intended target. An
alternative tradition of clinical transcription has grown up in the US: the symbols suggested by
Shriberg and Kent (SK). In many cases, these symbols are designed to show the intended target with
a diacritic illustrating in which way the realization differs from the target. In this article possible
confusions that may occur if the SK system is used are discussed, together with problems that may
occur when SK and IPA are used together.
Keywords: Transcription, disordered speech, production vs target
Introduction
It has been stressed on many occasions that a detailed phonetic transcription of disordered
speech is a vital first step in accurate diagnosis and relevant intervention in the clinic (for
example, Carney, 1979; Buckingham and Yule, 1987; Ball, Rahilly, and Tench, 1996;
Howard and Heselwood, 2002). Phoneticians and speech pathologists have long used the
International Phonetic Alphabet (IPA) as their primary resource in transcribing speech—
whether disordered or not. The current Handbook of the IPA (IPA, 1999) describes the
alphabet and its use. However, clients being assessed in the speech clinic may use sounds
that are, in some cases, not found in natural language at all. There have over the years been
various developments of symbol systems to capture these atypical sounds (see the
description in Ball et al., 1996) . Duckworth, Allen, Hardcastle, and Ball (1990) describe
the drawing up of the extended IPA (extIPA) to transcribe atypical speech sounds, and this
system now has the approval of the IPA and was included in the 1999 Handbook.
Correspondence: Dr Martin J. Ball, Department of Communicative Disorders, University of Louisiana at Lafayette, PO Box
43170, Lafayette, LA 70504-3170, USA . E-mail: mjball@louisiana.edu
Clinical Linguistics & Phonetics, October–November 2008; 22(10–11): 864–870
ISSN 0269-9206 print/ISSN 1464-5076 online # 2008 Informa UK Ltd
DOI: 10.1080/02699200802172757
Downloaded By: [Ball, Martin J] At: 15:22 2 July 2009

The main competitor to the extIPA symbol system within speech pathology is the system
developed by Shriberg and Kent (most recent edition 2003), from the early 1980s on. This
system (henceforth SK) includes some novel and extremely helpful (to the learner)
features, such as the categories of diacritic placement around the main symbol (e.g. stress,
nasality and lip diacritics above the main symbol; tongue and sound source diacritics below;
and off-glide and stop release diacritics above and to the left).
Transcription philosophies
Apart from some differences in choice of symbols and choice of atypical speech behaviours
to symbolize, the main difference between the approaches of those who utilize IPA and
extIPA and the recommendations for use by Shriberg and Kent (2003) for the SK system
lies in what one might term transcription philosophies. Shriberg and Kent specifically urge
their readers to be aware of the intended target pronunciation before they start transcribing.
They say, ‘when intended targets are unknown, trans cription is almost impossible’, and
further, ‘clinicians must attempt to determine, word-for-word, what the child intended to
say at the time the sample is obtained’ (Shriberg an d Kent, 2003: 138). This approach will
be classed as tran scribing by target. Interestingly, the authors do note that other research has
pointed out the dangers of this approach. Howard and Heselwood (2002: 393) summarize
these, and note, ‘it’s crucial to understand some of the ways in which a native speaker–
hearer’s perceptual system will constrain perception of speech data’. In particular they
point to phonemic false evaluation (see also Buckingham and Yule, 1987), where listeners
place speech sound s into their own internalized phonemic categories, due to the influence
of categorial hearing. Predictions of expected speech sounds through knowin g the intended
target form also have a part to play in producing unreliable transcriptions. Howard and
Heselwood (2002: 395) note, ‘the effects of listener expectations are extremely robust’, and
refer to a study by Ingrisano, Klee, and Binger (1996) which demonstrated that transcribers
who knew the intended target were influenced by this and transcribed what they thought
they heard.
It is common, therefore, for many clinical phoneticians to transcribe by production, and
this author’s experience is that, while admittedly difficult with highly unintelligible clients,
it is not ‘almost impossible’ to transcribe them without knowing the intended target
(indeed, this is a task given to graduate students when the speech samples are spontaneous
with no known targets).
These two different trans criber philosophies have consequences for the symbol systems
themselves. Because in many instances SK requires the use of the phonetic symbol for the
target sound, a range of diacritics are then neede d to show the absence of a particular
articulatory feature, or a change to a particular articulatory feature. Usually, in IPA and
extIPA these characteristics are denoted by the use of a different symbol for the sound in
question. Further, because several target sounds may be altered in disordered speech to the
same realization, this may result under SK conventions in the use of two notably different
transcriptions for the same sound, a practice against the principles of the IPA (see IPA,
1999). Another important result of this approach is that using diacritics to show that some
feature is missing may leave it unclear as to what the actual production of the sound was.
While the initial transcriber may, of course, retain a percept of the realization, no-one else
looking at the transcription will. While it is true that any transcription system is likely to
miss some aspects of a sound’s production, at least the IPA and extIPA attempt to show
that production. See, for example some of the SK diacritics that show a sound is different
Transcribing Disordered Speech 865
Downloaded By: [Ball, Martin J] At: 15:22 2 July 2009

(e.g. lateralization and frictionalized stops below). Fina lly, and least importantly, some SK
diacritics are used with other meanings in the IPA, although often the different usage is
disambiguated through the position that the diacritic is used in. In the following section,
some of the points raised here are illustrated.
Examples of differences between IPA/extIPA and SK
In the following subsections a series of transcriptional differences between the IPA (and
extIPA) approach and the SK approach are examined, and where appropriate any
ambiguities arising from the different transcriptional practices are noted.
Rounded and unrounded vowels
As recent research has demonstrated (see reviews in Ball and Gibbon, 2002), a range of
vowel errors has been identified in disordered speech. Among these are monophthongiza-
tion of diphthongs, and movement towards corner vowels. Also possible are the
unrounding of target rounded vowels and the rounding of target unrounded vowels. If
either of these last two patterns are found, a transcriber of clinical speech samples has to
decide how to denote them. If you wish to emphasize the production (and thus to follow the
IPA route), there are symbols provided by the IPA for front rounded and back unrounded
vowels (i.e. [y, ø,
Œ, œ] and [A, L, ], M]). However, an alternative approach is to emphasize
the target in the transcription. The SK system, therefore, recommends that vowels
produced with lip-rounding the opposite from what would be expected for the intended
target should be transcribed with the symbol for the target together with the IPA diacritics
for more or less rounding. So, a high front rounded vowel for English target /i/ would be
transcribed as [i
c
] and a high back unrounded vowel for English target /u/ would be
transcribed as [u
c
].
A problem arises between IPA and SK usage of these diacritics. As noted, the IPA
defines them as ‘more’ and ‘less’ rounded, whereas the SK usage is to denote opposite
rounding from that of the symbol to which they are attached. Of course, the positioning of
the diacritics differs: IPA mandates their use beneath a vowel symbol, SK as superscripts to
the right of the symbol. Nevertheless, there is clearly potential for confusion here unless it is
clearly stated which usage is intended.
Lateral fricatives
Alveolar lateral fricatives are found in several languages (see, for example, Ladefoged and
Maddieson, 1996), inclu ding Welsh and Zulu. In the clinic, they may also be encountered
for target sibilant fricatives, sometimes described as ‘lateral lisps’. The IPA provides
symbols for voiceless and voiced alveolar lateral fricatives ([
, /]). Again, the SK approach
is to consider the target sound first and then add a diacritic to it. So, for lateral realizati ons
of target /s, z/ one finds [s
L
,z
L
]. Moreover, if, as can also happen, the lateral fricative is being
used for target post-alveolar fricatives, then the lateralization diacritic is added to those
symbols: [s
L
, z
L
]. This is potentially confusing as it is possible to make post-alveolar lateral
fricatives (there are no IPA symbols for them, but some have been used unofficially, see
Ball and Rahilly, 1999). So the SK symbols [s
L
, z
L
] could denote either target post-alveolars
realized as alveolar lateral fricatives, or target post-alveolars realized as post-alveolar lateral
fricatives. Matters are further complicated in that the same diacritic is used by the IPA to
866 M. J. Ball
Downloaded By: [Ball, Martin J] At: 15:22 2 July 2009

denote non-syllabic vowels, though in a mixed IPA and SK transcription ambiguity would
be unl ikely.
Retroflex fricatives
Apart from their use in a range of other languages, retroflex fricatives may be found in some
accents of English. For example, Northern Irish English has retroflex fricatives following
the ret roflex approximant [¡], so ‘papers’ may be realized as [
9
p
h
eepe¡Z
u
]. Retroflex
realizations of alveolar and post-alveolar target fricatives may also be encountered in the
clinic. The IPA approach is to transcribe these in terms of the production, thus [S, Z]. As
was se en in the case of the lateral fricatives, the SK approach is to highlight the target.
Therefore retroflex realizations of target /s, z/ will be transcribed as [s
ß
,z
ß
], whereas retroflex
realizations of target /s, z/ will be transcribed as [s
ß
, z
ß
] (SK give the example of English
‘harsher’ as an instance when a retroflex realization of a target post-alveolar fricative might
occur as a normal variant due to the influence of the r-sounds). While on this occasion there
is no ambiguity as to the nature of the articulation, the end result of this strategy is to have
different transcriptions to denote the same sounds. This is of cours e perfectly possible if
one transcribes via the target rather than via the production, but would normally be
considered something to be avoided if only because the visual representation suggests
something different from the auditory impression.
Frictionalized stops
By this term SK denote what phoneticians would normally call affricated stops. The se
occur naturally in some dialects of English (for example, Liverpool English has affricated
fortis stops in some contexts), and may also occur in speech disorders, characterized by
Shriberg and Kent (2003: 197) as having ‘a gradual, rather than a sudden, movement away
from the closure’. Interestingly, Shribe rg and Kent (2003) do not equate these sounds with
affricates, but the above description seems to correspond with the notion of affricated stops
(rather than full affricates) as described, for example, by Clark and Yallop (1990).
The SK system transcribes these ‘frictiona lized’, or affricated, stops through the use of a
diacritic beneath the relevant stop symbol: [t
6
,k
6
,p
6
]. This usage, however, hides potentially
important information concerning the nature of the fricative release of these stops. So,
whereas affricated release of a velar stop would usually be also velar, affricated release of
bilabial stops could be bilabial or labiodental, and of alveolar stops could be dental,
alveolar, or post-alveolar, as illustrated here with the fortis stops of English: [k
x
,t
s
,t
h
,t
s
,p
f
,
p
w
]. Shriberg and Kent (2003) point out that the use of frictionalized stops in disordered
speech may signal the beginning of the development of fricatives; all the more useful then
for someone looking at a transcription to be able to tell whether the affricated part of the
stop corresponds with the relevant fricative type for that context.
The diac ritic used in the SK system also clashes with the IPA usage (where it is used to
mark mid-centralized vowels). However, the IPA uses the diacritic above vowel symbols
whereas SK uses it beneath stop symbols, so the likelihood of confusion here is slight.
Derhoticization
In this and the subsequent subsection two examples of marking the abse nce of a pho netic
characteristic proposed in the SK system are examined. The most difficult to grasp is the
Transcribing Disordered Speech 867
Downloaded By: [Ball, Martin J] At: 15:22 2 July 2009

notion of derhoticization. Shriberg and Kent (2003: 118) define it thus ‘a derhoticized sound
is an /r/ consonant or an r-coloured vowel that is significantly lacking in r-ness (rhotic or
retroflex quality) but does not fall into another phonemic category of English’. There are, in
my opinion, several difficulties with this definition. First, one has to assume that when the
authors use ‘/r/ consonant’ they mean a target /r/, for otherwise if an /r/ lacks r-ness it cannot
be an [r]. Second, the definition invokes phonemic categories although one would assume
that transcription at this level is phonetic rather than phonemic. Finally, the definition is
noted to apply only to English, and so begs the question of whether one could use this
diacritic with languages other than English that a client might use.
As noted above, an advantage of the SK system is that often it allows one to mark in a
transcription the target sound that the client has misarticulated. However, in this specific
case, the disadvantage of marking derhoticization is that the resulta nt transcription does not
show what sound the client used instead. So, one could imagine that a targe t apical variety
of English /r/, made with no tongue tip raising or retroflexion, could result in a [j]
realization; a target bunched /r/ with the tongue pulled slightly too far back could result in a
[
] (see Ball, Lowry, and McInnis, 2006, for an example of just that realization). Shriberg
and Kent (2003) refer to a sound between /r/ and /w/. This could well be a labiodental
approximant wh ich occurs frequently in the clinic, and if so could be transcribed with [u].
Also, the SK diacritic (e.g. [r
K
]) is used in the IPA to denote a short segment (albeit usually
above the symbol), so again confusion might arise in a mixed IPA and SK transcription.
Non-labialized consonant
This category is inclu ded in the SK system among various lip symbols (Shriberg and Kent,
2003: 116). The definition states that the listed diacritic should be used with consonants
that are ‘not articulated with a constriction or narrowing, of the lips’. However, as this
would app ly to a large number of consonants, it becomes clear that the authors mean that it
is only to be used with target labial consonants: ‘If a speaker fails to narrow and protrude
the lips for the /w/ in weed, the non-labialization symbol would be appropriate’ (Shriberg
and Kent, 2003: 116). The IPA approach would be to transcribe such a sound with the IPA
symbol for a velar (i.e. non-labial velar) approximant. So the SK system would give us [
],
while the IPA would use [
].
While there is no ambiguity in this instance of disagreement between SK and IPA norm s,
one might wish that the SK diacritic had a more useful definition. Defined simply as no n-
labialized does not imply necessarily that the sound has lip-spreading (just that it lacks lip
constriction and protrusion). So, while the symbols available to transcribers of either
tradition are suitable to capture sounds such as a non-labial velar approximant without
recourse to a diacritic, SK does not (although extIPA do es) provide the transcriber with the
means to denote a consonant made with lip-spreading. This might, for example, be a case
where a spread-lip mouth shape was used where normal coarticulatory effects would make
one expect a lip-rounded shape (e.g. [t
«
] where one might expect [t
w
] in a word like ‘two’),
or simply to mark that the expected coarticulatory effect of lip-spreading has in fact taken
place in a word like ‘tea’.
1
Trills
This final subsection will consider the SK method of transcribing trills. Best practice in the
narrow transcription of English (e.g. Ball and Mu¨ ller, 2005) suggests that the IPA symbol
868 M. J. Ball
Downloaded By: [Ball, Martin J] At: 15:22 2 July 2009

for a post-alveolar centr al approximant, [r], should be used to transcribe the normal
realization of English /r/, and that [r] be used only for trill realizations in English or in other
languages that a client might use. (In fact, in Ball and Mu¨ ller, 2005, it is recommended that
/r/ be used for English at the phonemic transcription level exactly to avoid the
complications encountered with the SK proposals.) Because SK retain [r] at the phonetic
as well as the phonemic levels for the English approximant-r, they are forced to provide a
diacritic for trills (though not for taps, where the normal IPA tap symbols are used).
So, whereas in IPA a trilled-r is [r], in SK it is [r
3
]. Shriberg and Kent (2005: 119) also
suggest that this diacritic be used when transcribing languages other than English that use
trills. However, if a trill is the default value for a rhotic cons onant, it would seem more
sense to use [r] (if apical) as the trill symbol, and [r] for any approximant realiza tion that
might be found in normal or disordered speech.
Conclusion
Transcribing by production has long been the tradition within clinical phonetics as practiced
by phoneticians. It is probably true, however, that where possible transcription by target may
well be the choice of the clinician confronted by unintelligible speech. However, it is not
always possible to know the target of each and every word, so to some extent transcription by
production has to be undertaken unless large amounts of useful speech data are to be ignored.
The IPA and extIPA approach to transcription can be understood to be production-based.
The SK system, on the other hand, is overtly target-based and the authors of the system feel
that this is the only realistic way to proceed. The result of this, however, as has been shown
above, is that in some instances the SK system produces unclear transcriptions. While the two
approaches may well be able to co-exist, some thought might be given as to how the insights of
SK (such as diacritic placement categories) might be used to improve the IPA/extIPA
approach, the avoidance of target symbols with diacritics might be adopted by SK to avoid the
uncertain transcriptions that have been highlighted. Such a move would not mean that SK
users would have to change to transcribing via the production (much as that might be
welcomed by many phoneticians), but that, for example, a lateral fricative is transcribed as a
lateral fricative, not as a fricative with added lateralization.
Note
1. It is true that SK does mark some features not covered by IPA/extIPA. These include ‘inverted lip’, and
different diacritics to mark fronted and retracted in consonants as opposed to vowels. The first is rare according
to the authors, while the second would appear unnecessary.
Declaration of intere st : The author reports no conflicts of interest. The author alone is
responsible for the content and writing of the paper.
References
Ball, M. J., & Gibbon, F. (Eds) (2002). Vowel disorders. Woburn, MA: Butterworth Heinemann.
Ball, M. J., & Mu¨ ller, N. (2005). Phonetics for communication disorders. Mahwah, NJ: Erlbaum.
Ball, M. J., & Rahilly, J. (1999). Phonetics. The science of speech. London: Edward Arnold.
Ball, M. J., Lowry, O., & McInnis, L. (2006). Distributional and stylistic variation in /r/-misarticulations: a case
study. Clinical Linguistics and Phonetics, 20, 119–124.
Ball, M. J., Rahilly, J., & Tench, P. (1996). The phonetic transcription of disordered speech. San Diego: Singular
Publishing.
Transcribing Disordered Speech 869
Downloaded By: [Ball, Martin J] At: 15:22 2 July 2009

Buckingham, H., & Yule, G. (1987). Phonemic false evaluation: theoretical and clinical aspects. Clinical Linguistics
and Phonetics, 1, 113–125.
Carney, E. (1979). Inappropriate abstraction in speech assessment procedures. British Journal of Disorders of
Communication, 14, 123–135.
Clark, J., & Yallop, C. (1990). An introduction to phonetics and phonology. Oxford: Blackwell.
Duckworth, M., Allen, G., Hardcastle, W., & Ball, M. J. (1990). Extensions to the International Phonetic
Alphabet for the transcription of atypical speech. Clinical Linguistics and Phonetics, 4, 273–280.
Howard, S. J., & Heselwood, B. C. (2002). Learning and teaching phonetic transcription for clinical purposes.
Clinical Linguistics and Phonetics, 16, 371–401.
Ingrisano, D., Klee, T., & Binger, C. (1996). Linguistic context effects on transcription. In T. W. Powell (Ed.),
Pathologies of speech and language: Contributions of clinical phonetics and linguistics (pp. 45–46). New Orleans:
ICPLA.
IPA (1999). Handbook of the International Phonetic Association. Cambridge: Cambridge University Press.
Ladefoged, P., & Maddieson, I. (1996). The sounds of the world’s languages. Oxford: Blackwell.
Shriberg, L., & Kent, R. (2003). Clinical phonetics, 3rd ed. Boston: Allyn & Bacon.
870 Transcribing Disordered Speech
Downloaded By: [Ball, Martin J] At: 15:22 2 July 2009