Guiberson & Rodriguez (2013) Spanish Nonword Repetition
Guiberson & Rodriguez (2013) Spanish Nonword Repetition
Guiberson & Rodriguez (2013) Spanish Nonword Repetition
Purpose: The purpose of the present study was to (a) describe significantly lower than those of the TD children. Whereas
and compare the nonword repetition (NWR) performance of item-level scoring of NWR items indicated acceptable levels
preschool-age Spanish-speaking children (3- to 5-year-olds) of sensitivity and specificity and suggested positive and
with and without language impairment (LI) across 2 scoring negative likelihood ratios, PPC scoring of NWR items resulted
approaches and (b) to contrast the classification accuracy of a in less than desirable levels of sensitivity and adequate
Spanish NWR task when item-level and percentage of specificity.
phonemes correct (PPC) scoring methods are applied. Conclusion: Item-level scoring of 3- to 5-syllable Spanish
Method: Forty-four Spanish-speaking children participated. NWR items may be useful as part of an assessment battery
Twenty-one children had LI and 23 had typically developing for preschool-age Spanish-speaking children.
(TD) language. Children were administered a Spanish NWR
task and a standardized Spanish language measure.
Results: A developmental pattern in NWR performance was Key Words: nonword repetition, Spanish-speaking,
observed, and the children with LI had NWR scores that were classification accuracy
I
n 2008, Latinos accounted for 22% of the students provide useful information that could assist in identifying LI
enrolled in public schools, and Spanish-speaking in Spanish-speaking children. NWR is a brief process-
children represented the largest portion of young dependent task that is designed to minimize biases associated
English language learners (Fortuny, Hernandez, & Chaudry, with traditional language assessment measures. Nonword
2010). Nationally, 28% of preschool-age children were from repetition has been suggested as a possible phenotypic
households where a language other than English was spoken, marker for some forms of LI (Bishop, North, & Donlan,
and the majority of these children were from Spanish- 1996). Although numerous studies have examined the NWR
speaking backgrounds (U.S. Census Bureau, 2008). In the skills of English-speaking children, the NWR skills of
coming years, this growth trend is expected to continue Spanish-speaking children, particularly those of preschool
(Center for Public Education, 2012). The increasing presence age, need to be examined.
of young Spanish-speaking children in U.S. preschool
classrooms requires language assessment measures to be NWR
developed for this growing segment of the population.
Accurate and reliable Spanish assessment measures are NWR is a task in which children are asked to repeat
needed to differentiate children with language impairment novel phonological forms or nonsense words such as /naõb/
(LI) from children who have typically developing (TD) (Archibald, 2008). NWR mimics an important language
language. A Spanish nonword repetition (NWR) task may learning mechanism: immediate repetition of unfamiliar
words (Archibald, 2008). NWR tasks have been argued to
tap into underlying cognitive skills, including phonological
a
University of Wyoming, Laramie working memory, phonological encoding, phonological
b
The University of New Mexico, Albuquerque awareness or sensitivity, and a general phonological proces-
Correspondence to Mark Guiberson: mguibers@uwyo.edu sing ability (for a review, see Graf Estes, Evans, & Else-
Editor: Marilyn Nippold Quest, 2007). NWR tasks are also thought to be language-
Associate Editor: Gary Troia based processing measures that theoretically deemphasize
Received January 19, 2012 prior language experiences (Dollaghan & Campbell, 1998;
Revision received April 3, 2012 Kohnert, Windsor, & Yim, 2006).
Accepted October 22, 2012 Nonword characteristics. Nonword stimuli have speci-
DOI: 10.1044/0161-1461(2012/12-0009) fic characteristics (Dollaghan & Campbell, 1998; Graf Estes
Language, Speech, and Hearing Services in Schools N Vol. 44 N 121±132 N April 2013 N ß American Speech-Language-Hearing Association 121
122 Language, Speech, and Hearing Services in Schools N Vol. 44 N 121±132 N April 2013
124 Language, Speech, and Hearing Services in Schools N Vol. 44 N 121±132 N April 2013
Kohnert, Windsor, & Yim (2006) 14/28 50 65/71 92 5.92 (2.53±13.85) .55 (.37±.80)
Girbau & Schwartz (2007) 11/11 100 10/11 91 11.00 (1.70±71.28) 0 (N/A)a
Windsor, Kohnert, Lobitz, & Pham (2010) 11/19 58 53/65 82 3.14 (1.66±5.94) .52 (.30±.89)
Gutierrez-Clellen & Simon-Cereijido (2010) 30/49 61 78/95 82 3.42 (2.11±5.56) .47(.33±.68)
For the purpose of comparing studies that have of diagnostic status because either they lack acceptable
reported on the classification accuracy of Spanish NWR classification accuracy or they have precision values that are
tasks, we closely reviewed each study and collected uninformative (Dollaghan & Horner, 2011). Furthermore,
classification accuracy values or raw data reported for there is a need to contrast results obtained through item-level
studies including children with and without LI. We used this and PPC scoring methods. Given the effort required to
information to compute sensitivity, specificity, +LR, 2LR, complete PPC scoring and the relative ease of item-level
and corresponding CIs. The Spanish NWR classification scoring, these findings may have practical applications if
accuracy measures for the four studies are presented in NWR tasks are to be included as a component of
Table 1. The +LR obtained on Spanish NWR tasks ranged comprehensive language assessments.
from 3.14 to 11.00. Three of the four studies reviewed had The purpose of the present study was to (a) describe
suggestive +LR and relatively narrow CIs that were above and compare the NWR performance of preschool-age
the uninformative range. The study with the strongest +LR Spanish-speaking 3-, 4-, and 5-year-olds with and without LI
(Girbau & Schwartz, 2007) was completed with a rather across two scoring approaches and (b) to contrast the
small sample (N = 22) and had extremely broad +LR CIs; classification accuracy of a Spanish NWR task when item-
these characteristics suggest that the tool may lack precision level and PPC scoring methods are applied.
in classifying children (Dollaghan, 2004; Dollaghan &
Horner, 2011). The 2LR obtained on three of the Spanish
NWR tasks ranged from .47 to .55, with narrow corre- Method
sponding CIs. These 2LR values indicated that the Spanish
NWR tasks yielded results that are suggestive but insufficient Participants
to rule out disorder. For the fourth study (Girbau & Forty-four preschool-age children (3;0±5;10) partici-
Schwartz, 2007), 2LR was 0 and corresponding CIs could pated in this study. Families were recruited from two
not be calculated because of the lack of variability observed. regional Head Start and Early Head Start programs in the
As stated earlier, this lack of variability is a cause for concern western United States. Children met the following inclu-
and should be interpreted with caution, at least until these sionary criteria: normal hearing, no known neurological
results have been replicated in multiple carefully conducted impairment, and lack of severe phonological impairment.
large-scale studies with this diagnostic measure (Dollaghan, Only predominantly Spanish-speaking children (who spoke
2007). Based on the classification accuracy measures of the Spanish §80% of the time according to parent report) were
other three studies, Spanish NWR tasks may provide useful included.1 In order to maximize statistical power for group
information when attempting to identify LI in school-age comparisons, we determined at the onset of the study that the
Spanish-speaking children. sample would be approximately split in terms of LI status.
Three sources were obtained to establish LI status: (a)
Current Study identification of LI by a bilingual speech-language patho-
logist (SLP), (b) report of parent concerns about child's
The NWR meta-analysis and studies of preschool-age language development, and (c) expressive language scores on
English-speaking children, combined with the emergent body the SPLS±4 ¦77 (1.5 SDs below the mean). Children were
of NWR research describing Spanish-speaking and Spanish± placed in the TD group if (a) they had not been previously
English bilingual children, indicate the need for additional diagnosed as having LI, (b) their parents did not report
studies describing the Spanish NWR performance of pre- concerns about their language development, and (c) they had
school-age children with and without LI. Such studies will expressive language SPLS±4 scores §85.
help determine the usefulness of Spanish NWR tasks as a
diagnostic marker for LI in preschool-age Spanish-speaking
children. Available measures intended to diagnose LI in 1
The term Spanish-speaking will be used to refer to this sample for the
Spanish-speaking children are only somewhat suggestive remainder of the manuscript.
Table 2. Means, standard deviations, and independent-samples t test for family and child characteristics for the
typically developing (TD) group and the group with language impairment (LI).
Note. SPLS±4 = Preschool Language Scales, Fourth EditionÐSpanish Edition (Zimmerman, Steiner, & Pond, 2002b).
*p , .001.
126 Language, Speech, and Hearing Services in Schools N Vol. 44 N 121±132 N April 2013
Table 3. NWR means and standard deviations for 3-, 4-, and 5-year-old children from the TD and LI groups across tasks
and for the total item-level score.
TD group LI group
3-year-olds 4-year-olds 5-year-olds 3-year-olds 4-year-olds 5-year-olds
Task (N = 6) (N = 8) (N = 9) (N = 7) (N = 8) (N = 6)
1 syllable 3.50 (1.23) 4.00 (.00) 4.00 (.00) 3.57 (1.13) 4.00 (.00) 4.00 (.00)
2 syllable 3.33 (1.63) 3.87 (.35) 4.00 (.00) 3.14 (1.46) 2.50 (1.69) 3.50 (.55)
3 syllable 2.33 (1.97) 3.12 (1.64) 3.89 (.33) .86 (1.22) 2.25 (1.75) 3.50 (.55)
4 syllable 2.00 (1.89) 2.25 (1.49) 2.89 (1.05) .71 (1.50) 1.63 (1.60) 2.00 (1.41)
5 syllable .00 (00) .75 (1.04) 2.00 (1.32) .00 (00) .38 (.74) .17 (.41)
Total item-level score 11.17 (5.91) 14.00 (3.85) 16.78 (2.28) 8.29 (4.27) 11.50 (4.75) 13.50 (2.35)
128 Language, Speech, and Hearing Services in Schools N Vol. 44 N 121±132 N April 2013
TD group LI group
3-year-olds 4-year-olds 5-year-olds 3-year-olds 4-year-olds 5-year-olds
Task (N = 6) (N = 8) (N = 9) (N = 7) (N = 8) (N = 6)
1 syllable 88% (30.61) 100% (.00) 100% (.00) 92% (18.89) 94% (17.67) 100% (.00)
2 syllable 83% (40.82) 98% (4.59) 100% (.00) 84% (37.06) 85% (34.67) 98% (3.61)
3 syllable 71% (40.94) 93% (13.87) 99% (1.66) 46% (32.86) 71% (37.72) 96% (3.43)
4 syllable 67% (40.68) 88% (11.61) 94% (6.61) 43% (38.21) 62% (37.52) 84% (13.89)
5 syllable 43% (38.36) 76% (20.70) 84% (12.27) 23% (23.87) 52% (31.86) 47% (23.78)
Total PPC score 61% (37.06) 86% (12.40) 93% (5.56) 46% (27.81) 66% (32.09) 78% (9.36)
between the children in the two groups. Fifteen of the 21 Group Performance on NWR Across Scoring
children with LI were correctly identified (sensitivity of 71%), Approaches
and 17 of the 23 TD children were correctly identified
(specificity of 74%). The +LR for item-level ±scores was Across both item-level scoring and PPC scoring
2.74 (95% CI = 1.31±5.73), and the 2LR was .39 (95% approaches, a developmental trend was observed with both
CI = .19±.79). groups of children (LI and TD), with older children
For the PPC scoring, the overall Wilks' lambda was outperforming younger children. We also compared the
also significant, l = .83, x2(1, N = 44) = 8.12, p , .01, again item-level scores of an LI and a TD group of Spanish-
indicating that PPC scores and age discriminated between the speaking children and found that the children with LI
children in the two groups. Ten of the 21 children with LI performed significantly more poorly than the TD children
were correctly identified (sensitivity of 48%), and 18 of the 23 did. The significant difference between the LI and TD group
TD children were correctly identified (specificity of 78%). is a promising result, suggesting that NWR item-level scores
The +LR for PPC scores was 2.19 (95% CI = .89±5.36), and tapped into group differences in language skills.
the 2LR was .67 (95% CI = .42±1.06). Our results and earlier Spanish NWR studies indicated
Given that the item-level scores of 3-, 4-, and 5-syllable that children with LI have difficulty with NWR, especially
NWR items resulted in fair discriminant accuracy and with items that are three to five syllables in length (Girbau &
moderately strong +LRs and 2LRs, posttest probability was Schwartz, 2007; Windsor et al., 2010). Longer syllable items
computed (for a review, see Dollaghan, 2007). A positive were also more difficult than shorter syllable items, with the
posttest probability is the probability that a child with LI was 5-syllable items being the most difficult. It was expected that
found positive for LI after the results from the test were longer items would be more difficult, as they likely begin to
obtained. Bayes' theorem provides a method to calculate exhaust children's phonological memory capacity. Two
posttest probabilities using LR values and pretest probability studies reviewed (GutieÂrrez-Clellen & Simon-Cereijido, 2010;
values (Battaglia et al., 2002). For the current study, a Summers et al., 2010) used an NWR task that included items
Bayesian calculator was used that provides pretest prob- that were one to four syllables in length. Neither study made
ability and positive posttest probability coefficients mention of a ceiling effect with NWR items, but in one study,
(University of British Columbia, 2012). The pretest prob- a high percentage of children were able to repeat 4-syllable
ability for the current sample was 48%. The positive posttest nonwords (Summers et al., 2010), and both research teams
probability of NWR item-level scored items was 71%, recommended that future research include longer NWR
indicating that there is a 71% probability that children items. Five-syllable items may be critically important in
with LI will score poorly on item-level scored 3-, 4-, and detecting subtle differences in NWR in Spanish-speaking
5-syllable NWR items. These results are consistent with the children.
discriminant accuracy measures and indicate that item-level
scored NWR tasks may assist in indicating when Spanish- NWR Classification Accuracy
speaking preschoolers have LI.
To contrast scoring methods and evaluate the useful-
ness of a Spanish NWR task in identifying preschool-age
Spanish-speaking children with LI, classification accuracy
Discussion measures were completed as a final step. The two scoring
The aims of this study were to (a) describe and methods we applied yielded very different classification
compare how preschool-age Spanish-speaking children with accuracy values. Item-level scoring yielded adequate sensi-
and without LI perform on NWR across item-level and PPC tivity (71%) and specificity (74%), with moderately +LR
scoring approaches, and (b) to contrast the classification and 2LR and corresponding CI ranges that did not include
accuracy of a Spanish NWR task when item-level and PPC uninformative values. PPC scoring yielded unacceptable
scoring methods are applied. sensitivity (48%) and adequate specificity (78%), with +LR
130 Language, Speech, and Hearing Services in Schools N Vol. 44 N 121±132 N April 2013
132 Language, Speech, and Hearing Services in Schools N Vol. 44 N 121±132 N April 2013