PHONOLOGY AND CHILDREN WITH SPECIFIC LANGUAGE IMPAIRMENT: STATUS OF STRUCTURAL CONSTRAINTS IN TWO LANGUAGES UMBERTA BORTOLINI Centro di Fonetica del CNR, Padua, Italy
LAURENCE B. LEONARD Audiology and Speech Sciences, Purdue University, West Lafayette, Indiana
Two studies are reported in which the phonological characteristics of preschool-age children with specific language impairment (SLI) are compared with those seen in younger normally developing children matched for mean length of utterance and consonant inventory size. The productions of both English-speaking and Italian-speaking children with SLI were more likely to deviate from the adult standard than the productions of the younger control children. In Italian, the children with SLI had more difficulty than the younger controls in the use of non-final weak syllables; in English, the children with SLI were poorer than the younger controls in the use of non-final weak syllables, word-final consonants, and word-final consonant clusters. These are the same phonological details that are required for several grammatical inflections and many function words in the two languages. However, the children with SLI were also less consistent than their younger compatriots in using consonants in structurally simple words. These findings provide evidence for the view that for many preschool-age children with SLI, phonological problems go beyond those that might be predicted on the basis of the children’s short utterances and limited consonant inventories. © 2000 by Elsevier Science Inc. Educational Objectives: Readers will be able to (1) describe the phonological differences between children with specific language impairment and younger, typically developing children; and (2) identify the phonological limitations most likely to hinder the use of grammatical morphology. KEY WORDS: Specific language impairment; Phonology; Prosody
Address correspondence to Laurence B. Leonard, Audiology and Speech Sciences, Heavilon Hall, Purdue University, West Lafayette, IN 47907, USA. Tel: (765) 494-3794; Fax: (765) 4940771; E-mail:⬍[email protected]
J. COMMUN. DISORD. 33 (2000), 131–150 © 2000 by Elsevier Science Inc. All rights reserved. 655 Avenue of the Americas, New York, NY 10010
0021-9924/00/$–see front matter PII S0021-9924(99)00028-3
BORTOLINI and LEONARD
INTRODUCTION AND BACKGROUND Children with “specific language impairment” exhibit a significant deficit in language ability, yet they do not show clear evidence of other type of problems that might contribute to their language difficulties. Hearing is within normal limits, there is no evidence of neurological dysfunction, and the children’s relationships with other people and objects are not characteristic of pervasive developmental disorders such as autism. Furthermore, the children receiving this clinical label score at age-appropriate levels on nonverbal tests of intellectual functioning. It is rare that language is the only area in which these children perform below the level of same-age peers. See reviews in Johnston (1988) and Leonard (1998). However, no nonlinguistic factor or combination of factors has yet emerged that is capable of explaining the language disorder. Although children with specific language impairment (SLI) constitute a heterogeneous group, it is clear that phonology is among the areas of language adversely affected in many of these children. For example, in studies aimed at identifying subgroups of children with SLI, children with both marked phonological and morphosyntactic difficulties emerge as the largest subgroup (Rapin & Allen, 1983; 1988). Whether phonological ability is measured in terms of segment accuracy, distinctive features, phonological processes, implication laws, or unusual phonological patterns, children with SLI come up short when compared to typically developing children of the same age. See recent review in Leonard (1998). However, it is also true that children with SLI score below age level on most every developmentally appropriate measure of language ability. Consequently, it is not clear whether phonology constitutes an area of special difficulty for children with SLI or whether problems in this area are simply a natural consequence of a more general limitation in language learning. We address this question in the present study. We approach this question by comparing the phonological characteristics of children with SLI and group of younger normally developing children matched according to mean length of utterance (MLU) and phonetic inventory size. Given that most children with SLI show MLUs that are significantly below age level, any group differences in phonology emerging from this design would reflect problems that exist even after limitations related to output length are taken into account. Consequently, the differences found in phonology might reflect rather serious limitations on the part of the children with SLI. The use of phonetic inventory size as an additional basis for matching permits an examination of difficulties in phonology relative to typical children that go beyond the availability of segments for a well-constructed output. That is, it allows for an inspection of how the children make use of the segments that are available. Two different languages are studied here—English and Italian. Inclusion of these two languages constitutes a stronger test of the status of phonology in
PHONOLOGY AND SLI
children with SLI. For example, English-speaking children often show limitations in the use of both grammatical inflections and function words relative to younger MLU-matched control children. See reviews in Johnston (1988); Leonard (1989, 1998) and Bishop (1992). However, Italian-speaking children with SLI are more likely to differ from MLU controls primarily in the use of function words. Their use of most inflections closely resembles that of their MLU-matched compatriots (Leonard, Bortolini, Caselli, McGregor, & Sabbadini, 1992; Bortolini, Caselli, & Leonard, 1997). If children with SLI are found to differ from MLU controls in phonological ability in both languages despite these crosslinguistic profiles of grammatical morphology, a stronger case could be made for the centrality of phonology in the diagnosis of SLI. The question of special interest in this study is whether children with SLI would lag behind the younger control children in the degree to which their phonologies reflected the prosodic properties of their respective language. Our interest in prosodic structure lies not only in the fact that prosody is a crucial but understudied part of phonology in children with SLI, but also because limitations in prosody could have a deleterious effect on these children’s use of grammatical morphology. For example, some English inflections require codas (e.g., played) or word-final consonant clusters (e.g., pushed). Some function words are weak monosyllables that might constitute unfooted syllables in a prosodic phrase (e.g., the and are in The children are playing). As noted above, English-speaking children with SLI have serious difficulties with many inflections and function words, omitting both types of morphemes. In Italian, very few words have codas. However, as in English, many function words in Italian are weak monosyllables that might be unfooted in a phonological phrase (e.g., la “the” and lo “him” in La ragazza lo spinge “The girl pushes him”). For Italian-speaking children with SLI, few inflections appear to be especially problematic. However, function words such as monosyllabic clitics and articles are omitted more frequently than is seen in MLU-matched controls. It seemed important to determine whether the prosodic structures required of the problematic morphemes are difficult for these children even when they play no morphemic role. Links between prosodic structure and grammatical morpheme use have been uncovered in studies of normally developing children acquiring English. For example, Gerken (1996); as well as Spanish and German (Lleó & Demuth, 1999). If these links hold for children with SLI as well, these children’s documented difficulty with grammatical morphemes may be related in part to limitations in their use of the prosodic structures on which these morphemes depend.
STUDY 1—ENGLISH Most of the evidence on the phonological abilities of English-speaking children with SLI have come from investigations in which these children have
BORTOLINI and LEONARD
been compared with normally developing children of the same age. Very few studies have compared English-speaking children with SLI and younger typically developing children matched according to MLU or phonetic inventory size. The evidence available to date suggests that children with SLI may be more likely than younger control children to produce open syllables (Schwartz, Leonard, Folger, & Wilcox, 1980; Ingram, 1981) and less likely to use the consonants in their inventory in a consistent manner (Ingram, 1981; Leonard, Schwartz, Swanson, & Loeb, 1987). However, these few studies can provide only a preliminary picture at best. Some of the studies involved a small number of children (Schwartz, Leonard, Folger, & Wilcox, 1980), whereas in others, it is not clear how many of the children exhibited language deficits in addition to their phonological difficulties (Ingram, 1981).
Subjects Eighteen children served as the source of data in this study. These children had participated in an investigation by Leonard, Eyer, Bedore, and Grela (1997). Nine of the children met the criteria for SLI. These children, five boys and four girls, ranged in age from 3;7 to 5;9. Eight of the children with SLI were above age 4;0. Each of these children scored more that one standard deviation below age level on the composite score (based on comprehension and production) of the Test of Language Development-Primary: 2 (TOLD-P:2) (Newcomer & Hammill, 1991). The remaining child, age 3;7, was too young for the TOLD-P:2. This child scored more than one standard deviation (SD) below age level on the Preschool Language Scale - 3 (Zimmerman, Steiner, & Pond, 1992). The MLUs of the children computed in words ranged from 2.8 to 4.8 (M ⫽ 3.75, SD ⫽ 0.54) based on a 100-utterance spontaneous speech sample. MLU was calculated in words rather than morphemes to facilitate comparison with the Italian-speaking children of Study 2. These MLUs were well below age level according to the norms of Templin (1957). All of the children scored above 85 on the Leiter International Performance Scale (Leiter, 1952) or the Performance Scale of the Wechsler Preschool and Primary Scale of Intelligence—Revised (Wechsler, 1989). Scores ranged from 93 to 148. Each of the children passed a hearing screening. Based on observation, medical history, and parental report, the children showed no evidence of neurological dysfunction or disturbed interaction with other people or objects. Because the Leonard et al. (1997) study was concerned with the use of grammatical morphology, each of the children included as a subject had to score above 80% on a picture-naming task requiring them to use word-final /d/, /t/, /z/,and /s/ in monomorphemic contexts (e.g., bed, nose). Given that some children with SLI of similar age cannot meet this criterion, the data in the present study constitute a conservative test for any phonological differences between children with SLI and typical children, particularly with regard to the use of word-final consonants.
PHONOLOGY AND SLI
The consonant inventory sizes of the children were calculated on the basis of their spontaneous speech and their responses to a set of 120 probe items constructed to examine the children’s use of a range of grammatical morphemes, including articles, infinitival to, copula be forms, and noun and verb inflections. The children were credited with a consonant if it appeared in at least two different words in any word or syllable position, and represented the segment expected in the adult system for that context. Using these criteria, the children’s consonant inventory sizes ranged from 15 to 18 (M ⫽ 16.00, SD ⫽ 1.22). The second group of nine children served as controls for MLU and consonant inventory size. These children were developing normally and therefore were necessarily younger than the children with SLI. Ages ranged from 2;5 to 3;3. Five of the children were girls, four were boys. The children scored at age-appropriate levels on all measures of the subject selection test battery. Each child’s MLU fell within .2 words of the MLU of a child in the group with SLI, resulting in MLUs (M ⫽ 3.72, SD ⫽ 0.55) that closely resembled those of the children with SLI. Consonant inventory sizes ranged from 14 to 19 (M ⫽ 16.44, SD ⫽ 1.94).
Procedure Following administration of the subject selection test battery and the spontaneous speech sample, the children were asked to describe sets of pictures through a sentence completion procedure. The pictures and sentence frames (hereafter referred to as the “probes”) were designed to assess the children’s use of grammatical morphology. The 15 items used to assess singular or plural nouns required only single-word responses (e.g., train, hats); the remaining 105 items required responses more than one word in length. These included NP responses such as a dog, Mommy’s car, infinitival phrases such as to play, VP responses such as gets the milk and combed her hair, as well as full sentences such as The balloon is red. Because the children sometimes used words that were unanticipated, there was some variation across children in the lexicon making up the phonological database. However, the overlap was substantial. Some of the words were required for more than one item on the probes (e.g., a cat; The cat is sleeping), permitting us to examine intra-word phonological variability.
Analysis The children’s productions were transcribed phonetically from the audiorecordings made during the session. We first computed an overall accuracy measure. Each transcribed word production was compared to the adult form and scored as matching the adult form or as deviating from the adult form in some way. The percentage of word productions that were adult-like was then calculated. For this analysis, each word was counted only once. For those
BORTOLINI and LEONARD
words produced more than once by a child, we scored the most accurate production. Word productions that lacked an obligatory inflection (e.g., Mommy for Mommy’s ) were not scored as errors if the remainder of the word matched the adult form. We then examined the children’s productions in terms of three phonological details that were relevant not only to phonological accuracy but also to the use of many English grammatical morphemes. They are described here as “structural constraints,” as they appear to represent obstacles to the wellformedness of English words in children’s early pronunciations. These are: (1) word-final consonant deletion (hereafter, FCD, e.g., [bo] for boat); (2) word-final consonant cluster reduction (hereafter, FCR, e.g., [faes] for fast), and word-initial weak syllable deletion (hereafter, IWSD, e.g., [naena] for banana). In the case of the latter, consonants or consonantal features from the initial weak syllable might have been preserved (e.g., [bun] for balloon, [zaef] for giraffe); the criterial element was the absence of a weak syllable preceding the strong syllable in the child’s production of the word. For each child, the words produced that were subject to each of these constraints were noted, and the percentage of these productions that included the constraint (e.g., final consonant deletion) was calculated. For words subject to these constraints that were produced by the child more than once, we determined the percentage of the productions that differed in pronunciation. It should be noted that these structural constraints were examined only in monomorphemic contexts. Thus, an attempt at fast was considered for analysis of FCR, but an attempt at passed was not. To determine whether there was a relationship between these structural constraints and the children’s use of grammatical morphology, we selected those grammatical morphemes examined by Leonard et al. (1997) that involved the same prosodic structure, and calculated each child’s percentage of use. These grammatical morpheme productions came from the same task serving as the source of the phonological data. We created a composite that reflected each child’s percentage of use across the consonant inflections, noun plural -s, genitive ’s, third person singular -s, and regular past -ed. In the Leonard et al. (1997) study, these grammatical morphemes were used with lower percentages by the children with SLI than by their younger controls. There was not a sufficient number of items to create a separate composite for consonant inflections that constituted consonant singletons (e.g., keys, played). These items were included with those involving consonant clusters (e.g., dogs, climbed). Correlations were calculated to determine the relationship between the children’s use of consonant inflections and their degree of FCD and FCR. We also calculated correlations between the children’s composite percentages of use of articles and infinitival to and their degree of IWSD, under the assumption that both of these morphemes are weak syllables that typically precede a strong syllable (e.g., the car, to play). These morphemes, too, were
PHONOLOGY AND SLI
used with lower percentages by children with SLI than by younger controls. It should be noted that, for all of the above morphemes, errors were omissions. Thus, percentage of use can be viewed as the percentage of instances in which the morpheme was included. Significant negative correlations would suggest a relationship between grammatical morpheme use and the structural constraints. To assess the two group’s phonological abilities independent of the above structural constraints, we examined each child’s use of word-initial singleton consonants in monosyllabic words (e.g., cup, red) and bisyllabic words with a trochaic pattern (e.g., cookie, rabbit). The percentage of these words produced with the appropriate consonant in initial position was then calculated. Finally, to determine whether any limitations in the use of word-final consonant clusters also applied to non-final word positions, we computed the percentages of non-final clusters that were reduced (e.g., [nek] for snake).
Results The children with SLI were significantly less accurate than the younger controls in their word productions, t (16) ⫽ 5.11, p ⬍ .001. This comparison showed a very large effect size of 2.55, as computed by d (Borenstein & Cohen, 1988; Cohen, 1988). Mean percentages of words correct (and SDs) for the children with SLI and the control children were 43.44 (17.52) and 78.33 (8.29), respectively. The two groups’ productions were than compared in terms of the three structural constraints of particular interest. An analysis of variance (ANOVA) with one between-subjects variable (children with SLI, younger controls) and one within-subjects variable (FCD, FCR, IWSD) was performed. The percentage data were transformed using arc-sine transformations. The ANOVA revealed a main effect for subject group, F (1, 16) ⫽ 72.99, p ⬍ .001, indicating that the productions of the children with SLI were less likely to be free of these constraints than the productions of the younger control children. A difference was also seen among the three constraints, F (2, 32) ⫽ 32.15, p ⬍ .001. Post-hoc testing at the .05 level revealed that FCD occurred with lower percentages than the other two constraints. A significant group by constraint interaction was also seen, F (2, 32) ⫽ 19.06, p ⬍ .001. Post-hoc testing revealed that, for the control children, FCD and IWSD occurred with lower percentages than FCR, with very large effect sizes (d) of 2.80 and 1.36, respectively. For the children with SLI, FCD occurred with lower percentages than both FCR and IWSD. Again, effect sizes (d)—3.95 and 4.62, respectively— were very large. For all these structural constraints, percentages were lower for the control children than for the children with SLI, with very large effect sizes d ranging from 1.72 to 3.01. Figure 1 provides an illustration of the findings.
BORTOLINI and LEONARD
Figure 1. The percentages with which the productions of the English-speaking children with SLI (ESLI) and their younger normally developing controls (END-MLU) reflected word-final consonant deletion (FCD), word-final consonant cluster reduction (FCR), and word-initial weak syllable deletion (IWSD). Values are means with standard errors.
Some of the words subject to the structural constraints were produced more than once by the same child. For these instances, we calculated the percentage of these productions that represented a change in the status of the constraint from one production of the word to the next. For the children with SLI and the younger controls, mean percentages of words showing this variability were 63.00 (SD ⫽ 17.74) and 33.33 (SD ⫽ 21.65), respectively. This difference was statistically significant, t (16) ⫽ 3.00, p ⬍ .01, showing a very large effect size d of 1.50. Much of the variation could be attributed to inconsistency in the use of word-final consonants by children with SLI (e.g., [kaet] and [kae] for cat by the same child). For the children with SLI, there was a correlation between the use of consonant inflections and FCR, r ⫽ ⫺.78, p ⬍ .05. Consonant inflection use was not related to FCD, r ⫽ ⫺.02, p ⬎ .05. The coefficient reflecting the relationship between article and infinitival to use and IWSD did not reach significance, r ⫽ ⫺.41, p ⬎ .05. As can be seen in Figure 1, the control children’s productions revealed only a limited degree of FCD and IWSD. As a result, correlations involving these structural constraints could not be computed. There was a significant relationship between these children’s consonant inflection use and FCR, r ⫽ ⫺.89, p ⬍ .05.
PHONOLOGY AND SLI
We inspected the data to determine whether difficulties with particular segments might have contributed to the findings for the three structural constraints. For example, if a child showed no evidence of [k] in his or her inventory, FCD in book and FCR in box might be expected. The only cases for which this was true were word-final clusters involving liquids. Four children from each group consistently produced [w] in place of  in syllable-initial position (e.g, [wid] for lead) and omitted  from word-final clusters (e.g., [wuf] for wolf, [mak] for milk). Because only two words required these clusters and the same number of children in each group showed this difficulty, segmental factors were not likely to be a major factor in the findings for these structural constraints. On the other hand, there is evidence that the two groups of children differed in their use of consonant segments apart from the structural constraints. A difference favoring the control children was seen in the two groups’ use of word-initial singleton consonants in monosyllabic words (e.g., cup) and bisyllabic words with a trochaic pattern (e.g., cookie), t (16) ⫽ 3.91, p ⬍ .01. This difference showed a very large effect size d of 2.05. For the children with SLI, the percentage of word-initial consonants in error averaged 19.22 (SD ⫽ 9.74); for the younger controls, the mean percentage was 4.44 (SD ⫽ 3.09). There were too few words with the same initial consonant to perform a detailed analysis of the types of errors committed, though place errors seemed more frequent in the productions of the children with SLI. Finally, to determine whether the high degree of word-final cluster reductions by the children with SLI could also be found in other word positions, we compared the two groups of children on the percentage of productions in which non-final clusters were reduced. On this measure, the two groups did not differ, t (16) ⫽ 1.08, p ⬎ .05. Mean percentages of cluster reduction were 48.67 (SD ⫽ 37.31) for the children with SLI and 30.78 (SD ⫽ 28.38) for the typically developing children. As can be seen from the large SDs, high variability characterized each group.
STUDY 2—ITALIAN Italian differs from English in several ways that are likely to influence children’s pattern of phonological acquisition. Word-final consonants and wordfinal consonant clusters are uncommon; they are seen principally in borrowed words (e.g., sport, film). As in English, feet are usually trochaic (e.g., cáne “dog,” gátto “cat”). However, Italian has many words with an initial weak unfooted syllable (e.g., matíta “pencil,” martéllo “hammer”), raising the possibility that IWSD (e.g., [tita] for matíta, [telo] for martéllo) poses an obstacle for Italian-speaking children with SLI. Bortolini et al. (1997) observed that Italian-speaking children with SLI made less use of articles than younger con-
BORTOLINI and LEONARD
trol children. Like unfooted word-initial weak syllables, articles are weak syllables that frequently appear in unfooted contexts preceding nouns. The Italian lexicon also contains a large number of words with an unfooted weak syllable in final position (e.g., álbero “tree,” pécora “sheep”). Strict adherence to the trochaic pattern would result in omission of one of the two weak syllables in such words (e.g., [albo] for álbero, [peka] for pécora). Hereafter this constraint will be referred to as word-medial weak syllable deletion, or MWSD. It should be noted that the only inflection documented as giving Italian-speaking children with SLI special difficulty is the third person plural inflection (Bortolini, Caselli, & Leonard, 1997). This is the only inflection thus far studied that requires an unfooted weak syllable in final position as part of the inflection. Examples are -ono and -ano in dórmono (“they sleep”) and pórtano (“they carry”), respectively. Bortolini et al. found that if the children did not succeed in using the third person plural inflection, they usually produced the third person singular inflection in its place, as in dórme (“he or she sleeps”) and pórta (“he or she carries”). Although such productions certainly constitute grammatical errors, they are also simple trochaic patterns.
Method Subjects. The 24 children in this study had participated in the investigation of grammatical morphology by Bortolini et al. (1997). Twelve of the children, nine boys and three girls, had been diagnosed as exhibiting SLI. They ranged in age from 4;1 to 7;0. The children’s MLUs in words based on a 100utterance spontaneous speech sample ranged from 2.88 to 4.89 (M ⫽ 3.76, SD ⫽ 0.66), which were well below age-level expectations. In addition, each of the children scored more than 1.25 standard deviations below the mean for their age on the composite (comprehension and production) score of Esame del Linguaggio (Cossu, 1987). Testing in non-language areas, medical records, and parental report suggested that these children were typically developing in other respects. Each child’s score on the Italian standardization of the Wechsler Preschool and Primary Scale for Children (Wechsler, 1969) was above 85 (M ⫽ 101, SD ⫽ 8.28). Furthermore, hearing was within normal limits and there was no evidence of neurological impairment. Interactions with objects and other people were reportedly age-appropriate. The children’s consonant inventory sizes were calculated in the same manner as for the English-speaking children. The children’s productions came from spontaneous speech and responses to a set of probes designed to elicit 84 non-imitative responses (see Bortolini et al., 1997). The children’s consonant inventory sizes ranged from 15 to 20 (M ⫽ 16.92, SD ⫽ 1.93). The remaining 12 children, seven boys and five girls, were younger, ranging in age from 2;6 to 4;0. These children were attending regular preschools
PHONOLOGY AND SLI
and were reported to be developing normally according to both parents and teachers. The MLU of each child in this group was within .2 words of the MLU of one of the children with SLI. Consequently, the MLUs of these children (M ⫽ 3.73, SD ⫽ 0.66) closely resembled those of the children with SLI in distribution as well as average. Consonant inventory sizes based on the above criteria ranged from 16 to 19 (M ⫽ 17.42, SD ⫽ 0.79). Though the range of inventory sizes was somewhat narrower in this group than in the group with SLI, the two groups were statistically identical on this characteristic, t (22) ⫽ 0.79, p ⬎ .05.
Procedure and Analysis Much of the procedure paralleled that used in the study on English. The children were shown pictures and asked to complete sentences begun by the experimenter. The pictures and sentences (the “probes”) were designed to obligate particular grammatical morphemes in NP (e.g., la bambola “the doll”), VP (e.g., balla “dances”), or full-sentence responses (e.g., La palla è rossa “The ball is red”). Eighty-four responses were obtained from each child. The first analysis concerned the overall accuracy of each word production on the probes. The children’s word productions were scored as in the English study above. Unlike the case for English, all Italian nouns, verbs, and adjectives are inflected. The children’s productions were scored without regard to the grammatical accuracy of the inflection used with the stem, though for most inflections, these children were as accurate as their MLU controls, see Bortolini et al. (1997). In those instances in which an inappropriate inflection was used, we examined the plausibility of a prosodic basis for the substitution. The productions were then examined in terms of the structural constraints, IWSD and MWSD. The procedure used for analyzing the constraints was the same used for English. We first identified each instance in which a constraint was possible. For example, caróta “carrot” requires an initial weak syllable, and thus was potentially subject to IWSD (e.g., [rota]), whereas péttine “comb” requires two weak syllables following the strong syllable, which could make it susceptible to MWSD (e.g., [pete]). Once these words were identified, we noted whether the child’s production reflected the constraint, and a percentage was then calculated. These percentages were used for statistical comparisons between the two groups of children and across the two structural constraints. Instances in which the same word was produced more than once were too few to permit analysis of intra-word variability. We than assessed the relationship between the children’s use of grammatical morphology and the structural constraints. We computed correlations between the children’s use of articles in the Bortolini et al. (1997) study and their degree of IWSD, and between the children’s use of the third person plural inflection and their degree of MWSD.
BORTOLINI and LEONARD
To determine whether the two groups of children differed in their use of phonology apart from the structural constraints, we compared their use of word-initial singleton consonants in bisyllabic words showing the canonical trochaic pattern (e.g., tópo “mouse,” cáne “dog”). Although word-final consonant clusters are rare in Italian, they do appear in other word positions (e.g., scarpe “shoes,” prende “takes”). To facilitate comparisons with our English data, were also examined, for each child, the percentage of non-final clusters that showed reduction.
Results The children with SLI were significantly less accurate in word productions on the probes than the younger controls, t (22) ⫽ 3.13, p ⬍ .01. This difference reflected a very large effect size d of 1.33. The control children averaged 79.08% words produced accurately (SD ⫽ 14.66); the children with SLI averaged only 54.33% (SD ⫽ 21.76). We then compared the two groups of children in terms of the two structural constraints of special interest. The data were analyzed by means of a mixed model ANOVA with subject group as a between-subjects variable and constraint type (IWSD, MWSD) as a within-subjects variable, with arc-sine transformations of the percentage data. A main effect for subject group was observed, F(1,22) ⫽ 10.19, p ⬍ .01, indicating that the children with SLI were more likely than the younger children to exhibit IWSD and MWSD. The effect size f of 0.55 was large [see (Borenstein & Cohen, 1988; Cohen, 1988)]. Neither the main effect for structural constraint type nor the subject group by constraint type was significant. As can be seen in Figure 2, the children with SLI showed a greater tendency toward IWSD (M ⫽ 10.00, SD ⫽ 9.25) and MWSD (M ⫽ 6.42, SD ⫽ 5.55) than the younger control children (for IWSD, M ⫽ 1.67, SD ⫽ 3.26; for MWSD, M ⫽ 1.75, SD ⫽ 3.17). For the children with SLI, there was a significant negative relationship between article use and IWSD, r ⫽ ⫺.85, p ⬍ .05. Similarly, a relationship was seen between use of the third person plural inflection and MWSD, r ⫽ ⫺.79, p ⬍ .05. As can be seen from Figure 2, instances of IWSD and MWSD were infrequent in the control children’s speech; consequently, correlations were not computed for these children. The findings for the structural constraints could not be attributed to special difficulties with particular segments. Nevertheless, as was found for English, when we inspected bisyllabic trochaic words requiring only a singleton consonant in initial position (e.g., rósso “red,” sále “ascends”) the children with SLI were less accurate than the younger control children in their production of initial consonants, t (22) ⫽ 2.30, p ⬍ .05. This difference represented a large effect size d of 0.94. The percentages of word-initial consonant errors for the children with SLI and younger controls were 23.42 (SD ⫽ 22.06) and 7.67
PHONOLOGY AND SLI
Figure 2. The percentages with which the productions of the Italian-speaking children wiht SLI (ISLI) and their younger normally developing controls (IND-MLU) reflected word-initial weak syllable deletion (IWSD) and word-medial weak syllable deletion (MWSD). Values are means with standard errors.
(SD ⫽ 8.71), respectively. The nature of the errors varied; no single feature contributed disproportionately to the group differences seen. The two groups also differed in their degree of non-final consonant cluster reduction, t (22) ⫽ 2.57, p ⬍ .05, showing a very large effect size d of 1.20. Higher percentages of cluster reduction were seen for the children with SLI (M ⫽ 36.33, SD ⫽ 26.81) than for the typically developing children (M ⫽ 13.42, SD ⫽ 3.53).
GENERAL DISCUSSION The results of our study provide strong evidence for the conclusion that children with SLI have serious limitations in phonological ability even when their reduced utterance lengths and consonant inventories are taken into account. In both English and Italian—languages with quite different typologies—children with SLI lagged behind younger MLU- and consonant inventory-matched controls on a general measure of accuracy, a measure of word-initial singleton consonant accuracy, and all measures of structural accuracy except non-final cluster reduction in the case of English.
BORTOLINI and LEONARD
An earlier study by Leonard et al. (1987) suggested that children with SLI are less consistent than younger normally developing preschoolers in using the consonants in their inventory. Some of the findings in the present study appear to confirm this observation. In both English and Italian, children with SLI were less accurate in producing singleton word-initial consonants in prosodically simple words than were younger normally developing children with comparable consonant inventories. Further, in our study on English, we were able to examine the children’s consistency in producing the same word. Again, the children with SLI were less likely than the younger controls to maintain a consistent pronunciation. These findings indicate that we cannot explain the segmental inaccuracies of these children solely on the basis of a limited ability to execute particular articulatory maneuvers. It would seem that the phonological representations underlying these articulatory acts are also relatively unstable. One of the motivations behind this investigation was to determine whether the grammatical morpheme limitations of children with SLI might be exacerbated by restrictions in their prosodic abilities. In both English and Italian, all of the structural constraints that could influence grammatical morpheme use revealed significant group differences between children with SLI and younger control children. In each instance, evidence for the constraint was greater in the speech of the children with SLI. In several of these cases, we found significant correlations between the children’s use of the grammatical morphemes and the corresponding structural constraint. Collectively, the significant correlations involved a large number of grammatical morphemes—noun plural -s, genitive ’s, and regular past -ed for English articles and third person plural inflections for Italian. It is quite possible that additional grammatical morphemes could be adversely affected by some of the structural constraints examined here. For example, Leonard, Sabbadini, Volterra, and Leonard (1988) found that Italianspeaking children with SLI omitted preverb direct object clitics (e.g., lo “him” in Gina lo vede Gina sees him”) much more frequently than English-speaking children with SLI omitted the direct object pronoun him (e.g., Gina sees him) even when pragmatic contexts were similar. One of the differences between the morphemes in the two languages is their typical prosodic position. In Italian, the direct object clitic is a weak syllable that usually precedes a finite verb and follows a weak syllable (as in the example above, Gina lo vede). Its English counterpart, in contrast, is a weak syllable that is often in clause-final position, and more likely to immediately follow a strong syllable (such as sees in the example above). Prosodic restrictions may play a role in the use of grammatical morphemes by children with SLI in other languages as well. For example, Hansson (1997) and Hansson and Nettelbladt (1995) found that Swedish-speaking children were as capable as MLU controls in their use of present tense verb inflections (e.g., -er in klipper “cuts”) but were more likely to omit the present tense copula (är) and auxiliary (har) forms. The present tense inflection is a weak sylla-
PHONOLOGY AND SLI
ble that usually follows the strong syllable of the verb stem. The copula and auxiliary forms, in contrast, are weak monosyllables that must often be produced after a weak syllable. Although it seems likely that structural constraints rendered the children’s use of certain grammatical morphemes more difficult, there is reason to believe that some of the phonological weaknesses seen in these children were independent of their problems with grammatical morphology. First, we found no relationship between article and infinitival to use and IWSD in the speech of the English-speaking children wiht SLI. Thus, althugh the children with SLI omitted infinitival to more frequently than younger controls [Leonard et al., (1997)] and also omitted word-initial weak syllables more frequently than controls (the present study), there was not a close correspondence between the two. In addition, we found no relationship between consonant inflection use and FCD in the speech of the same children. This last finding must be interpreted cautiously; the subject selection criteria in the Leonard et al. (1997) study required that each child produce word-final /d, /t/, /z/, and /s/ with over 80% accuracy. This means that the final consonants giving the children the greatest difficulty were probably not those required for inflection use. A second reason to believe that the children’s phonological and grammatical morpheme limitations were separable to some degree was the finding of group differences on segmental accuracy in contexts that bore little resemblance to those in which grammatical morphemes might be used (e.g., wordinitial consonant accuracy in words such as cup and cookie). Such differences emerged in both English and Italian. If some grammatical morpheme difficulties of children with SLI do not covary with a seemingly related structural constraint and some of these children’s phonological difficulties surface in contexts quite removed from those involved in grammatical morpheme use, it seems reasonable to conclude that phonological and grammatical morpheme deficits can probably co-occur without the former affecting the latter. Indeed, several findings in the literature suggest that the two problems need not be viewed as part of the same cluster, their frequent co-occurrence notwithstanding. For example, if preschoolers are identified on the basis of problems with phonology, a sizable minority of these children will show no problems with morphosyntax. For example, Paul and Shriberg (1982); Shriberg and Kwiatkowski (1994). By six years of age, the comorbidity of these problems is even lower (Shriberg & Tomblin, 1999). Recent work in the genetic study of speech and language impairments suggests that phonological and morphosyntactic deficits have different genetic sources (Tomblin, Shriberg, Nishimura, Zhang, & Murray, 1999). These findings, coupled with those of the present study suggest that while phonological difficulties may compound grammatical difficulties, each type of problem can exist on its own and warrants careful attention. We have cast the structural constraints examined in this study in rather traditional terms such as “final consonant deletion” and “initial weak syllable de-
BORTOLINI and LEONARD
letion.” It will be of interest to determine whether attempts to apply current theories such as optimality theory (Prince & Smolensky, in press ) to phonological development [e.g. (Bernhardt & Stemberger, 1998; Demuth, 1996, 1997)] will lead to deeper insights into the workings of these constraints. For example, within applications of optimality theory, it is assumed that constraints are ranked and that the “optimal” form is a form that violates only those constraints with relatively low ranking. This might suggest that for the English-speaking children with SLI in this study, a constraint such as Not Complex (Coda) (consonant clusters should not appear in final position) held a high ranking, for it was rarely violated. Similarly, the constraints Prominent (Foot, Left) (feet must be organized such that the syllable on the left is prominent) and Binary (Foot, σ) (a foot should have no more than two syllables) must be ranked quite high, given the high degree of IWSD in words such as banana. However, perhaps the greatest contribution of applications of optimality theory will come from future attempts to explain the variability associated with structural constraints such as FCR, IWSD, and the like. These constraints were not operative in all productions, and it seems possible that this variability might be explained by the presence or absence of other properties of the word that are associated with constraints with a higher ranking than these structural constraints. In the meantime, the present investigation seems to provide an affirmative answer to the more basic question of whether the phonological limitations of children with SLI are in some way extraordinary, lagging behind other language abilities which themselves are developing slowly. It is clear that the domain of phonology stands out as an especially serious area of deficiency in many children with SLI.
REFERENCES Bernhardt, B., & Stemberger, J. (1998). Nonlinear phonology and phonological development: A constraint-based aproach. New York NY: Academic Press. Bishop, D. (1992). The underlying nature of specific language impairment. Journal of Child Psychology and Psychiatry, 33, 3–66. Borenstein, M., & Cohen, J. (1988). Statistical power analysis: A computer program. Hillsdale, NJ: Erlbaum. Bortolini, U., Caselli, M.C., & Leonard, L. (1997). Grammatical deficits in Italian-speaking children with specific language impairment. Journal of Speech and Hearing Research, 40, 809–820. Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Hillsdale, NJ: Erlbaum. Cossu, G. (1987). Esame del Linguaggio. Parma, Italy: University of Parma.
PHONOLOGY AND SLI
Demuth, K. (1996). Alignment, stress, and parsing in early phonological words. In B. Bernhardt, J. Gilbert, & D. Ingram (Eds.), Proceedings of the UBC international conference on phonological acquistion (pp. 113–125). Somerville, MA: Cascadilla Press. Demuth, K. (1997). Multiple optimal outputs in acquisition. University of Maryland Working Papers in Linguistics, 5. Gerken, L. (1996). Prosodic structure in young children’s language production. Language, 72, 683–712. Hansson, K. (1997). Patterns of verb usage in Swedish children with SLI: An application of recent theories. First Language, 17, 195–217. Hansson, K., & Nettelbladt, U. (1995). Grammatical characteristics of Swedish children with SLI. Journal of Speech and Hearing Research, 38, 589– 598. Ingram, D. (1981). Procedures for the phonological analysis of children’s language. Baltimore, MD: University Park Press. Johnston, J. (1988). Specific language disorders in the child. In N. Lass, L. McReynolds, J. Northern, & D. Yoder (Eds.), Handbook of speech-language pathology and audiology (pp. 685–715). Toronto, ON, Canada: Decker. Leiter, R. Leiter International Performance Scale. (1952). Chicago, IL: Stoelting. Leonard, L. (1989). Language learnability and specific language impairment in children. Applied Psycholinguistics, 10, 179–202. Leonard, L. (1998). Children with specific language impairment. Cambridge, MA: MIT Press. Leonard, L., Bortolini, U., Caselli, M.C., McGregor, K., & Sabbadini, L. (1992). Morphological deficits in children with specific language impairment: The status of features in the underlying grammar. Language Acquisition, 2, 151–179. Leonard, L., Eyer, J., Bedore, L., & Grela, B. (1997). Three accounts of the grammatical morpheme difficulties of English-speaking children with specific language impairment. Journal of Speech and Hearing Research, 40, 741–753. Leonard, L., Sabbadini, L., Volterra, V., & Leonard, J. (1988). Some influences on the grammar of English- and Italian-speaking children with specific language impairment. Applied Psycholinguistics, 9, 39–57. Leonard, L., Schwartz, R., Swanson, L, & Loeb, D. (1987). Some conditions
BORTOLINI and LEONARD
that promote unusual phonological behavior in children. Clinical Linguistics and Phonetics, 1, 23–34. Lleó, C., & Demuth, K. (1999). Prosodic constraints on the emergence of grammatical morphemes: Crosslinguistic evidence from Germanic and Romance languages. In A. Greenhill, H. Littlefield, & C. Tano (Eds.), Proceedings of the 23rd annual Boston University Conference on Language Development. Volume 2 (pp. 407–418). Somerville, MA: Cascadilla Press. Newcomer, P., & Hammill, D. (1991). Test of Language Development—Primary: 2. Austin, TX: Pro-Ed. Paul, R., & Shriberg, L. (1982). Assocations between phonology and syntax in speech-delayed children. Journal of Speech and Hearing Research, 25, 536–547. Prince, A., & Smolensky, P. (in press). Optimality theory: Constraint interaction in generative grammar. Cambridge, MA: MIT Press. Rapin, I., & Allen, D. (1983). Developmental language disorders: Nosologic considerations. In U. Kirk (Ed.), Neuropsychology of language, reading and spelling (pp. 155–184). New York, NY: Academic Press. Rapin, I., & Allen, D. (1988). Syndromes in developmental dysphasia and adult aphasia. In F. Plum (Ed.), Language, communication, and the brain (pp. 57–75). New York, NY: Raven Press. Schwartz, R., Leonard, L., Folger, M.K., & Wilcox, M.J. (1980). Early phonological behavior in normal-speaking and language disordered children: Evidence for a synergistic view of linguistic disorders. Journal of Speech and Hearing Disorders, 45, 357–377. Shriberg, L., & Kwiatkowski, J. (1994). Developmental phonological disorders I: A clinical profile. Journal of Speech and Hearing Research, 37, 1100–1126. Shriberg, L., & Tomblin, J.B. (1999). Prevalence of speech delay and comorbidity with language impairment. Poster paper presented at the Third International Symposium of AFASIC, York, U.K. Templin, M. (1957). Certain language skills in children. Minneapolis, MN: University of Minnesota Press. Tomblin, J.B., Shriberg, L., Nishimura, C., Zhang, X., & Murray, J. (1999). Association of developmental language impairment with Loci at 7q3. Poster paper presented at the Third International Symposium of AFASIC, York, UK. Wechsler, D. (1969). Scala Wechsler a Livello Prescolare e di Scuola Elementare. Firenze, Italy: Organizzazioni Speciali.
PHONOLOGY AND SLI
Wechsler, D. (1989). Wechsler Preschool and Primary Scale of Intelligence— Revised (1989). New York, NY: Psychological Corporation. Zimmerman, I., Steiner, V., & Pond, R. (1992). Preschool Language Scale— 3. San Antonio, TX: Psychological Corporation.
CONTINUING EDUCATION Phonology and Children with Specific Language Impairment: Status of Structural Constraints in Two Languages QUESTIONS 1. The use of grammatical morphemes might be related to prosodic structure because: a. these morphemes usually involve late-developing speech sounds b. prosody allows the children to focus on grammar to a greater extent c. these morphemes often require use of final consonants, consonant clusters and non-final weak syllables d. utterances produced in a monotone promote omission of grammatical morphemes e. all the above 2. In Italian, the prosodic structures most relevant to grammatical morpheme use are: a. those that require weak syllables in unfooted positions, such as wordinitial position and word-final position in multi-syllabic words b. word-final consonant clusters c. the same as required in English d. any structure that contains final /s/ or /z/ e. none of the above 3. The rationale for matching children with SLI and younger normally developing children according to mean length of utterance and consonant inventory size is: a. to determine how many years behind the children with SLI are in their phonological ability b. to determine if phonological problems go beyond those that might be expected with short utterances and a limited consonant repertoire c. to ensure that children with SLI are not put in an unfavorable light by being compared to same-age peers d. to ensure that comparisons can be made using developmentally simple materials e. all of the above
BORTOLINI and LEONARD
4. Comparisons between the children with SLI and younger normally developing children revealed that: a. the speech of the children with SLI was influenced by structural constraints to a greater extent than the comparison children, but consonant accuracy was comparable in the two groups b. the speech of the children with SLI differed from the comparison children not only in vulnerability to structural constraints but in consonant accuracy as well c. the childern with SLI used consonants and prosodic structures in a manner that matched that of the younger normally developing children d. the children with SLI were more limited in their use of grammatical morphology, but not in their use of prosodic structures when these were used in monomorphemic contexts e. c and d 5. The most appropriate conclusions to draw from the findings are that: a. the grammatical morpheme problems of children with SLI are actually the result of phonological limitations pertaining to prosody b. the use of some grammatical morphemes by children with SLI can be adversely affected by phonological limitations, but not all problematic grammatical morphemes appear to have a prosodic basis c. children with SLI can be divided roughly into those with phonological problems only and those with grammatical morpheme problems only d. findings from Italian indicate that when prosodic constraints are few in number, children with SLI do not show grammatical morpheme difficulties e. a and d