The Responses of Bilingual Children in a Phonology Screening Instrument
Yvonne Wren
Sue Roulstone
Speech and Language Therapy Research Unit
The development of a new screening
assessment of children’s speech highlighted the need for an evaluation of the
responses of children assessed on the instrument who are bilingual. The Phoneme
Factory Phonology Screener (Wren, Roulstone and Hughes, 2006) is intended for
use in schools to assist teachers in identifying which children need referral
to speech and language therapy (SLT) due to difficulties with their speech
development. Many of the children who would be assessed on the screener would
be bilingual and therefore it was essential that a sample of such children were
included in the validation. This paper reports on the use of the Screener with
a sample of bilingual children attending
Background
The population of the
There is limited data on phonological
development in children who are bilingual. However, as bilingualism is not a
cause of communication impairment, the rate of referral for children who are
bilingual should be the same as that for the population as a whole. There is
evidence to suggest this is not the case though.
Holm,
What we do know about bilingual speech development is that it is different to monolingual speech development (Watson, 2001). This is because a child who speaks more than one language has to develop phonological systems for each language and ‘cross linguistically’. Watson suggests that they do this in two ways – first by superimposing the unknown system onto the known system and second, by ‘averaging’ the two systems by combining characteristics of the two phonologies. From this, they develop their two separate systems.
Gildersleeve et al (1996) found that bilingual children are less intelligible, make more consonant and vowel errors, distort more sounds and produce more uncommon error patterns than monolingual children. Bilingual children have been shown to demonstrate more phonological processes and have a higher percentage of occurrence of error patterns than monolingual children (Gildersleeve-Neumann and Davis, 1998) while Dodd, So and Li (1996) found that the types of errors evident in bilingual children’s speech would be atypical in a child who is monolingual for either or any of the languages spoken by the child.
Holm and Dodd (1999a) reported the phonological development of two typically developing bilingual Cantonese/English speaking children. This showed that the children had two separate phonological systems for each language. They used different error patterns in each language and language specific phonemes were not used in the ‘wrong’ language. Shared phonemes were often used in one language before another and the same phonemes were simplified differently in each language. Moreover, the errors always obeyed the phonotactic constraints of the appropriate language.
This pattern for surface level errors to be particular to each language was also found by Goldstein and Washington (2001) who observed different error patterns across each language in bilingual Spanish/English speaking children.
Though these patterns have also been observed in children with impaired speech development, it is thought that the underlying deficit in speech development is the same for all languages spoken by a child. Case studies of bilingual children show the same type of impaired speech (i.e., articulatory, phonological delay, consistent phonological disorder or inconsistent phonological disorder) in both languages though both languages may not be equally affected. (Holm et al, 1997; Holm and Dodd, 1999b).
Ethical, legal and professional considerations compel us to provide an equitable service to clients who are bilingual. In the first instance, we need to address access to services and the prime mover for this is the process of referral.
Health
visitors (HV) are the prime referral agents for pre-school children with speech
and language difficulties in the
Once a
child starts school, teachers become the most likely agent of referral. Schools
may be cautious about over-referring when the problem appears to be skill in
the new language rather than language impairment. Advice in some regions is to
allow a child two years to gain conversational fluency of English and only to
refer if the problems remain after that.
Parents
can refer children direct to SLT in the
The family doctor can also refer children to
speech and language therapy. However, they tend to use criteria provided by the
local SLT department. These may consist of checklists or screening tests. Few
of these cater for the specific needs of bilingual children even though we know
their development is different.
Nevertheless
there are many issues to consider in the development of screening assessments
for use with bilingual children. Ideally, they need to be available in a wide
range of languages but also be able to be used by non-speakers of the language
being assessed. They also need to reflect what we know about bilingual speech
development, i.e. that different error patterns may be seen in each language
and that atypical error patterns may be seen in children with typically
developing speech; also that children who are typically developing bilinguals
are likely to have lower intelligibility ratings and PCC scores than
monolingual speakers.
The
screening assessment also needs to be based on normative data for bilingual
children. The difficulty though is the large number of languages spoken. The National Literacy Trust (2004) estimates that 300 languages are
used in
To produce
screening assessments for each of these languages is untenable given that the
return on each published assessment is unlikely to match the cost of its
development.
The
Phoneme Factory Phonology Screener, hereon referred to
as the Screener, was
developed primarily as a referral tool for teachers. It was trialled on a
sample of 488
The aim of
the Screener is to provide a recommendation to refer to SLT because of concern
about a child’s speech development. The performance of the 488 children on the
Screener was compared with their performance on a ‘gold standard’ reference
assessment - the Phonology subtest of the DEAP (Diagnostic Evaluation of
Articulation and Phonology, Dodd et al, 2002) - to determine the sensitivity
(71%) and specificity (99%) of the Screener.
The Screener contains 66 pictures for a child to name. The adult administrator listens to the child’s production of one target sound within the stimulus word and then selects from a multiple choice response format which option most accurately reflects the child production of the target sound. The options presented to the administrator reflect those common error patterns used by young children with typically developing speech. The software then uses algorithms based on the number and type of error patterns shown compared with the child’s age to determine whether or not the child should be referred.
The program produces a report that indicates whether or not referral to SLT is recommended and why. It also lists the error patterns recorded and guides the user to relevant activities to use while waiting for SLT (or in partnership with SLT) in an accompanying software program and book.
Method
The 32 bilingual children assessed on the
Screener were in the age range four years to seven years, 11 months (see table
1). Attempts were made to recruit from 6 regions across the
Table 1: Bilingual sample by age
Table 2: Bilingual sample by region
A total of 15 different languages were
spoken by the 32 children in addition to English. These were Arabic, Spanish, Somali, Zambian,
Portuguese, French, Tumbuka, Russian, Chinese, Tamil, Urdu, Punjabi, Pahari,
Creole, and Bengali.
The
question asked in this trial was
how reliable was the
screener in correctly identifying bilingual children who needed referral to
SLT. Children were assessed on both the Screener and the DEAP. A cut-off of 5th
centile or below was used as the criteria to indicate referral on the DEAP and
a score of 1 using the algorithms built into the software was used for the
Screener. The referral decision for the DEAP and the Screener were then
compared.
Results
Both the Screener and the DEAP agreed that 26 of the 32 bilingual children should not be referred for therapy. A further six children received a refer decision from the DEAP but not from the Screener. None of the bilingual children were referred by the screener. That is, the Screener did not identify any children for referral while the DEAP identified 6.
To investigate the discrepancies between the two assessments, the performance of the children recommended for referral on the DEAP assessment was further analysed.
The DEAP manual provides additional data for children who are bilingual for English and one of the Punjabi languages.. Two (referred to below as Child A and Child B) of the six children spoke a Punjabi language in addition to English and so their data was considered.
The means for the 83 bilingual children reported in the DEAP manual were in the low normal range of the monolingual children, equivalent to a standard score of 6 or 7 for the Percent Consonants Correct (PCC) figure. Child A in this trial achieved a PCC score of 63 at age four years 10 months. The mean PCC score for this age range in the DEAP bilingual sample was 82.6 with a standard deviation of 14.1. Taking these figures into consideration, child A was between –1 and –2 standard deviations and would not therefore have been recommended for referral by the DEAP. Child B however, had a PCC score of 79 at age five years, two months. The DEAP Punjabi sample gave a mean PCC of 91.3 with a standard deviation of 5.0 at this age. These figures place this child more than 2 standard deviations below the norm suggesting that he or she would still have been recommended for referral.
A second consideration was given to the case/control status of these six children. As indicated above, of the 32 bilingual children, two were already known to SLT services. Of the two case children, only one had been identified as needing referral by the DEAP. Ironically, this was Child A whose referral status was altered to a no refer decision when the DEAP bilingual data for speakers of Punjabi languages was taken into account (above).
Final thought was given to the opinion of the SLT who carried out the DEAP assessment in this study. She noted her own clinical judgement regarding the appropriateness of referral. Of the 32 children, she identified only one child as in need of referral for his/her speech development. This child was again child A, the case child who was reclassified as no refer using the DEAP bilingual data. This child had the lowest PCC score (63) of all 32 bilingual children assessed on the DEAP.
The error patterns identified by the DEAP assessment and the Screener were compared for each of the six children who received a refer decision from the DEAP. With regard to the Screener, only one error pattern, gliding, was shown in just one of the six children. In contrast, the DEAP identified gliding in all six children. This difference in reporting of gliding was also noted in the main trial sample of 488 children. This could reflect the nature of gliding and the fact that SLTs are more likely to carry out the DEAP while the Screener is designed for teachers. It may be that the subtle differences in correct and glided productions of /r/ and /l/ may not be perceived by the non-SLT.
The DEAP also identified other error patterns in some of the six children. The most frequently occurring after gliding was deaffrication and context sensitive voicing that were identified in three children. In addition, two children showed cluster reduction while one showed stopping.
Discussion and conclusion
The sensitivity (71%) and specificity (99%) of the screener relative to the gold standard of the DEAP assessment mean that that there will be high agreement about non-cases, but less agreement about which children are identified as cases. This is reflected in the high numbers of bilingual children who are identified by both assessments as not in need of referral and less agreement about those requiring referral.
The types of errors picked up by the DEAP assessment show some similarity to the differences between the DEAP and the Screener in the main trial of 488 children. In particular, the Screener is less sensitive to gliding than the DEAP and that appears to be true for the bilingual sample. However, other patterns were also observed in the DEAP assessment but not picked up in the Screener. As these varied across individuals, it is difficult to make any general statements about this finding. However it is interesting to note that there were no incidences of ‘non-developmental’ error patterns shown. Given the comments in the literature (Dodd, So and Li, 1996) regarding the presence of error patterns which are atypical in either language in children who are bilingual, one might expect the presence of ‘non-developmental’ patterns in children who speak more than one language.
These findings illustrate the difficulties
of using assessments designed for monolingual English speakers on children who
are bilingual. There are some published assessments for children who speak
English as an additional language (RAMP, Stow and Pert, 1998) but for the vast
majority of languages spoken in the
Given the number of languages that now are
prevalent in
References:
Bowers, R.
and Oakenfull, S. (1996) The role of health visitors
in speech and language therapy. Health Visitor, 69, 319-320.
DfES (2003) Aiming high: raising the achievement of minority ethnic
pupils. Consultation paper.
Dodd, B.,
Hua, Z., Crosbie, S., Holm, A. and Ozanne, A. (2002) DEAP (Diagnostic
Evaluation of Articulation and Phonology)
Dodd, B., So, L. and Li,
W. (1996) Symptoms of disorder without impairment: the written and spoken
errors of bilinguals. In B. Dodd, R. Campbell,
I. Worrall (eds), Evaluating Theories of Language.
Gildersleeve, C. Davis, B., and Stubbe, E.
(1996) When monolingual rules don’t apply:
speech development in a bilingual environment. Paper presented at the
annual convention of the American Speech-Language-Hearing Association,
Gildersleeve-Neumann, C. and
Goldstein, B. and Washington, P. (2001) An initial investigation of phonological patterns in 4-year-old typically developing Spanish-English bilingual children. Language, Speech and Hearing Services in Schools, 32, 153-164.
Holm, A. and Dodd, B. (1999a) A longitudinal study of the phonological development of two Cantonese-English bilingual children. Applied Psycholinguistics, 20, 349-376.
Holm, A. and Dodd, B. (1999b) An intervention case study of a bilingual child with phonological disorder. Child Language Teaching and Therapy, 15, 139-158.
Holm, A., Ozanne, A, and Dodd, B. (1997) Efficacy of intervention for a bilingual child making articulation and phonological errors. International Journal of Bilingualism, 1, 55-69.
National Literacy Trust (2004) www.literacytrust.org.uk/research/lostop3.html. Accessed 15.3.04
Stow, C.
and Dodd, B. (2003) Providing an equitable service to
bilingual children in the
Watson,
Wren, Y., Roulstone, S.
and Hughes, A. (2006) Phoneme Factory Phonology Screener.