Clinical Application of the Mean Babbling Level and Syllable Structure Level

LSHSS Clinical Exchange Clinical Application of the Mean Babbling Level and Syllable Structure Level Sherrill R. Morris Northern Illinois University, DeKalb T here is a documented synergy between development of the speech sound system and development of other aspects of language (see Paul, 1998, for an overview). This relationship is particularly true during the toddler years and encompasses several language domains. Stoel-Gammon s (1991) data on normally developing 2-year-olds showed a strong correlation between the number of consonants in a child s phonetic inventory and his or her vocabulary size. That is, children who ABSTRACT: Purpose: This clinical exchange reviews two independent phonological assessment measures: mean babbling level (MBL) and syllable structure level (SSL). Both measures summarize phonetic inventory and syllable shape in a calculated average and have been used in research to describe the phonological abilities of children ages 9 to 36 months. An example of the measures clinical use is presented. Method: Six studies reporting MBL and SSL were reviewed. Combined, the studies included 207 participants. One hundred twenty-eight of those were developing language as expected, and 79 were identified as having speech-language impairment. Conclusion: Overall, the studies indicate that SSL is a reliable measure that increases as children develop. Children with language impairment exhibit lower SSL scores than children with typical language skills. Further, a lower than expected MBL is predictive of a lack of meaningful speech at 24 months of age. Implication: MBL and SSL have clinical application in terms of initial evaluation as well as progress monitoring. Recommendations for standardized corpus inclusion procedures are presented. KEY WORDS: babbling, syllable structure, phonological development, assessment, language impairment produce more words generally have more sounds in their speech sound inventories. Further, typically developing children are selective about the new words they attempt during the first-words stage of development (10 to 14 months; Schwartz & Leonard, 1982). For example, toddlers are more likely to produce a new word if the word shape and the sounds in the word are similar to other words in their spoken vocabulary. In fact, the fewer words that children have in their spoken vocabularies, the more particular they are. This characteristic is known as lexical selectivity. Because the main objective in early intervention is to increase communication, carefully choosing specific lexical targets based on phonological as well as semantic and syntactic characteristics will likely maximize results. Taken together, phonological measures such as phonetic inventory and syllable shape analysis can determine the sounds and syllable shapes that are already present in a child s productive phonology and can aid the speech-language pathologist (SLP) in choosing vocabulary targets that are likely to be attempted by a specific child. Additionally, new words can be presented to provide the child with support in terms of expanding phonetic inventory, syllable shape, or both. Two measures that summarize the information obtained from phonetic inventory and syllable shape analysis are mean babbling level (MBL) and syllable structure level (SSL). These measures have been used in the literature to describe the phonological skills of children under the age of 3 years. MBL provides a metric that summarizes phonetic and syllable shape information for babble into one score that can be used to show progress over time. SSL does the same for word productions. SSL is an extension of MBL; both use hierarchical levels to describe phonetic productions. The difference between the two measures is that MBL restricts tokens to babbling and SSL focuses on productive words. Results of previous studies have indicated that MBL and SSL are viable clinical tools that can be used to assess the LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS Vol. 41 223 230 April 2010 * American Speech-Language-Hearing Association 223

phonetic and syllabic characteristics of babble and early speech in children with delayed language abilities. This exchange describes the history of MBL and SSL and provides an example of their clinical use. History of the Measures In 1987, Stoel-Gammon described instructions for calculating MBL, which is a measure she created to assess phonological diversity in babbling. This measure is appropriate for use with children who are in the babbling stage of language development, which is defined as producing fewer than 10 different words over two separate 1-hr sessions. For children in the babbling stage, MBL is determined by analyzing the phonetic and syllabic properties of the child s vocalizations (approximately 50) recorded during a 30-min parent child play session. Once transcribed, each vocalization is coded according to the levels presented in Table 1. The consonant types in Levels 2 and 3 are considered true consonants that include all consonants with the exception of glides (e.g., w, j) and glottals (e.g., h,?). Average MBL is calculated by counting the number of occurrences of each level and multiplying the number of Level 1 utterances by 1, Level 2 utterances by 2, and Level 3 utterances by 3. The MBL is obtained by dividing the overall sum by the number of babbled utterances. Stoel- Gammonalsocalculatedthepercentage of the sample characterized by each level. For example, a child who produces 9 Level 1 utterances, 34 Level 2 utterances, and 7 Level 3 utterances has an MBL of 1.96. Eighteen percent of the child s utterances are Level 1, 68% are Level 2, and 14% are Level 3. Phonetic productions and corresponding MBL scores for the above example are presented in Table 2. In order for a child s production to be considered in calculating MBL, it must be babble (i.e., nonmeaningful). In cases where it is difficult to determine the beginning and end of a babbled utterance, the vocalization must be bounded at either side by 1 s of silence, a nonspeech-like child production (e.g., scream), or parental speech. Guidelines for corpus inclusion are listed below. & The vocalization was judged by the parent and the examiner/ observer to be non-meaningful; & the vocalization contained, at a minimum, a voiced vocalic element or a voiced syllabic consonant; & the vocalization was produced with an egressive airstream; and Table 1. Initial mean babbling levels (MBLs) created as a measure of phonological diversity in babble. Level Description 1 The utterance is composed of voiced vowel(s), voiced syllabic consonant(s), or CV syllable(s) in which the consonant is a glottal stop, a glide, or /h/. *Note: A glottal stop, /h/, /j/, /w/, and voiced vowel(s) are always classified as Level 1 and never contribute to Level 2 or Level 3 classification. 2 The utterance is composed of CV, VC, or CVC syllable(s) with a single consonantal type. Disregard voicing differences. 3 The utterance is composed of syllables with two or more consonantal types. Disregard voicing differences. Note. From Assessing Prelinguistic and Early Linguistic Behaviors in Developmentally Young Children (p. 122), by L. Olswang, C. Stoel-Gammon, T. Coggins, and R. Carpenter (Eds.), 1987, Seattle, WA: University of Washington Press. Copyright 1987 by L. Olswang. Reprinted with permission. & the vocalization was judged to be speech-like i.e., it could not be a cry, scream, cough or vegetative sound. (Stoel-Gammon, 1987, p. 122) Later, Stoel-Gammon (1989) used MBL to describe phonological changes in a group of 32 children between the ages of 9 and 24 months. Because MBL is calculated only for children within the babbling stage, the number of participants in this category declined throughout the longitudinal study. At the age of 24 months, 2 of the children in the longitudinal study were not communicating as expected. An assessment of the babbling of those children indicated that their vocalization patterns were different from those of the other 30 children, who followed a typical language progression. One of the atypical children had an MBL that was significantly lower than that of her peers between the ages of 9 and 18 months. The second child exhibited a similar MBL as his peers, but his true consonant inventory was limited to velar stops. Thus, Stoel-Gammon suggested that a child s MBL might be predictive of his or her future language impairment. Additional Studies Since Stoel-Gammon s (1987) initial publication of the measure, five teams of researchers have used MBL, or its meaningful speech counterpart, SSL, as a measure of phonological skill. Each of the studies slightly modified the procedures or the specific age range studied in order to address their specific research questions. A brief summary of the design and overall findings of each study is presented in the Appendix. For ease of reporting in this exchange, the tables and appendices refer to all children reported to have impaired or delayed language skills as late talkers even when the authors used other terminology (e.g., specific expressive language impairment, slow expressive language development) in their original work. Similarly, children with age-appropriate language skills are referred to as normally developing. Table 3 provides the MBL data that have been reported in previous studies. Fasolo, Majorano, and D odorico (2008) were interested in determining if the babbling characteristics of monolingual Italian children at 20 months of age were predictive of the children s language status at 24 months of age. Similar to Stoel-Gammon s (1989) study, Fasolo et al. identified participants (n = 12) in a larger longitudinal sample who were not talking at 24 months of age. They analyzed the 20-month language samples of the identified 12 children as well as those of 12 study participants who were communicating as expected at 24 months. They found that the mean MBL of the late talkers at 20 months of age (M = 1.22, SD = 0.19) was significantly lower than the mean MBL of the children in the typical language group (M = 1.50, SD = 0.31). Paul and Jennings (1992) examined the phonological skills of 28 toddlers with slow expressive language development who were in the meaningful speech stage of language development and those of 25 of their peers who were learning language normally. Because Paul and Jennings were applying MBL to an older age group (18 34 months), they expanded the measure to include words and word-like utterances in addition to babbled syllables. Additionally, they included consonant clusters in Level 3. In order to make it apparent that they were not using the measure with babbling, Paul and Jennings changed the name to SSL (Table 4). The analysis by Paul and Jennings indicated that the difference between children with delayed expressive language and their peers was not the types of syllable structures they produced, but the proportion of higher level structures that included 224 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS Vol. 41 223 230 April 2010

Table 2. MBL calculation example. Transcription MBL Justification esi 2 One true consonant /s/ Si 2 One true consonant /S/ nebo 3 Two true consonants that differ in place and manner /n/ /b/ we 1 One glide /w/ mu 2 One true consonant /m/ kuku 2 One true consonant that is replicated /k/ hu 1 One glottal /h/ jæsas 2 One replicated true consonant /s/; glide /j/ does not increase the level bu 2 One true consonant /b/ dis 3 Two true consonants that differ in manner /d/ /s/ juhi 1 One glide and one glottal /j/ /h/ Giku 2 Two true consonants that differ only in voicing /g/ /k/ hop 2 One true consonant /p/; glottal /h/ does not increase the level ju 1 One glide /j/ lum 3 Two true consonants /l/ /m/ no 2 One true consonant /n/ hə?e 1 Two glottals /h/ /?/ jãp 2 One true consonant /p/; glide /j/ does not increase the level. Ã:k 3 Two true consonants produced in a cluster /:/ /k/ bəbai 2 One repeated true consonant /b/ osiz 2 Two true consonants that differ only in voicing /s/ /z/ di 2 One true consonant /d/ æ 1 One vowel with no consonants zoz 2 One repeated true consonant /z/ sit 3 Two true consonants /s/ /t/ nãn 2 One repeated true consonant /n/ seif 3 Two true consonants that differ in place /s/ /f/ GeI 2 One true consonant /g/ hus 2 One true consonant /s/; glottal /h/ does not increase the level su 2 One true consonant /s/ dei 2 One true consonant /d/ eno 2 One true consonant /n/ deg 3 Two true consonants that differ in place /d/ /g/ Am 2 One true consonant /m/ bebi 2 One repeated true consonant /b/ teid 2 Two true consonants that differ only in voicing /t/ /d/ e 1 One vowel with no consonants mo 2 One true consonant /m/ wa 1 One glide /w/ wi: 2 One true consonant /:/; glide /w/ does not increase the level m 1 A single consonant functioning as a syllable Go 2 One true consonant /g/ in 2 One true consonant /n/ ot 2 One true consonant /t/ weit 2 One true consonant /t/; glide /w/ does not increase the level se 2 One true consonant /s/ fu 2 One true consonant /f/ nænæ 2 One repeated true consonant /n/ Su 2 One true consonant /S/?æpo 2 One true consonant /p/; glottal /?/ does not raise the level Note. Multiply 1 by the number of Level 1 utterances: 1 9 = 09 (18% of sample) Multiply 2 by the number of Level 2 utterances: 2 34 = 68 (68% of sample) Multiply 3 by the number of Level 3 utterances: 3 7 = 21 (14% of sample) Divide the weighted total by the number of utterances to determine the MBL: 98 50 = 1.96 two different true consonants and/or consonant clusters. Table 5 summarizes the SSL that was calculated on all vocalizations (intelligible and unintelligible) of the children between 20 and 24 months of age. Thal, Oroz, and McCaw (1995) conducted a comprehensive study assessing the grammatical and phonological skills of 51 children between 11 and 33 months of age. Children diagnosed as late talkers were separated into two groups: those exhibiting premeaningful speech and those exhibiting meaningful speech. Each of the children with language impairment was matched with a peer of the same chronological age and another peer with the same language abilities. Similar to Paul and Jennings (1992), Thal et al. restricted token inclusion to a single example of each token. The elimination of identical vocalizations may have provided a more robust SSL measure because the resultant score was not inflated or deflated by multiple repetitions of the same vocalization. Thal et al. also calculated SSL Morris: Clinical Application of the Mean Babbling Level 225

Table 3. Summary of reported MBL data. Standard deviation in parentheses when reported. Table 5. Summary of reported SSL averages (standard deviations) when the analysis corpus included all vocalizations. Mean age a LT ND N Average MBL N Average MBL Mean age a LT ND N Average SSL N Average SSL 9 2 1.16 1 32 1.33 (0.24) 1 12 2 1.28 1 32 1.50 (0.27) 1 14 7 1.50 2 15 2 1.48 1 24 1.58 (0.28) 1 16 10 1.50 2 18 2 1.53 1 8 1.65 (0.20) 1 20 12 1.22 b (0.19) 3 12 1.50 b (0.31) 3 21 2 1.47 1 7 1.80 2 22 7 1.50 2 23 10 1.80 2 10 1.90 2 24 2 1.91 1 Note. LT = late talker; ND = normal development. Source indicated by superscript: 1 Stoel-Gammon (1989); 2 Thal, Oroz, & McCaw (1995); 3 Fasolo, Majorano, & D odorico (2008). a Average age in months for all participants in each study. b Fasolo et al. (2008) included only Italian speakers. using two separate corpora; one on babble and one on intelligible utterances. The different analysis sets produced different results. Specifically, late talkers intelligible SSL was higher than that of their language-matched peers. However, the same late talkers had lower babbled MBL than their peers matched on language ability. Pharr, Ratner, and Rescorla (2000) assessed developmental changes between the ages of 24 and 36 months in children developing language typically and children with language impairment. Their longitudinal study included 20 children with expressive specific language impairment and 15 age-matched peers. Similar to previous studies, Pharr et al. s results of SSL showed significantly lower average scores in the 24-month-old children with impaired language when compared to their peers with typical language skills. Further, SSL was found to increase over time in both groups, with significant differences continuing at 36 months. By 36 months, children with language impairment were performing at similar levels as the 24-month-old population who was developing language as expected. Pharr et al. (2000) made the most significant change in the initial levels of SSL by adding a fourth level to their description (Table 6). 20 9 1.55 (0.41) 1 8 2.22 (0.30) 1 24 20 1.43 (0.30) 2 15 2.18 (0.21) 2 27 19 1.73 (0.35) 1 28 17 17 2.34 (0.17) 1 36 20 2.19 (0.13) 2 15 2.39 (0.10) 2 Note. LT = late talker; ND = normal development. Source indicated by superscript: 1 Paul & Jennings, 1992; 2 Pharr, Ratner, & Rescorla, 2000. a Average age in months for all participants in each study. Their rationale was to more adequately separate consonant cluster productions from the other levels. However, at 24 months, Level 4 utterances were infrequent in both the typical and delayed groups. By 36 months, approximately 8% of the utterances produced by children in the delayed language group were identified as Level 4, and 12% of the utterances produced by the typical language learners were identified as Level 4. Morris (2009) assessed the temporal stability of SSL by analyzing two 20-min speech samples that were collected within a 1-week period for 10 toddlers between 18 and 22 months of age with typical language skills. Similar to Paul and Jennings (1992) and Thal et al. (1995), Morris used Stoel-Gammon s (1987) MBL levels using words instead of vocalizations. In Morris s study, SSL exhibited strong test retest reliability, with the Session 1 SSL average being 2.16 (0.17) and the Session 2 average being 2.21 (0.19) ( p <.001). Table 7 reports SSL data calculated on the intelligible words of 16- to 23-month-old children. Results Although each study used slightly different protocol in data collection and analysis, the described participants in each study were Table 6. Pharr, Ratner, and Rescorla (2000) SSLs, including a fourth level. Table 4. Paul and Jennings (1992) syllable structure levels (SSLs). Level Description Level Description 1 The utterance is composed of a voiced vowel ([A]), voiced syllable consonant ([l]), or CV syllable in which the consonant is a glottal stop ([?o]) or a glide ([ha]), [wi]). 2 The utterance is composed of a VC ([up]) or CVC with a single consonant type ([kek]), or a CV syllable that does not fit the criteria for Level 1. Voicing differences are disregarded. 3 The utterance is composed of syllables with two or more different consonant types, disregarding voicing differences ([pati]). 1 Vocalizations containing a vowel (e.g., [A]), a syllabic consonant (e.g., [S]), or a CV syllable in which the only consonant type is a glottal stop (e.g., [?u]) or glide (e.g., [wi]) 2 Vocalizations containing true consonants with a single consonant (e.g., [ba], [Ãp]) or identical consonants (e.g., [mam], [dædi]) not represented in Level 1 3 Vocalizations containing two or more different consonants, not including consonant clusters (e.g., [dgi], [ kãp], [fini]), and vocalizations containing consonants that differed solely in voicing (e.g., [pãbã]) 4 Vocalizations containing consonant clusters (e.g., [brã], [elifənt]) Note. From Phonological Behavior in Toddlers With Slow Expressive Language Development by R. Paul and P. Jennings, 1992, Journal of Speech and Hearing Research, 35, p. 101. Copyright 1992 by the American Speech-Language-Hearing Association. Reprinted with permission. Note. From Syllable Structure Development of Toddlers With Expressive Specific Language Impairment by A. B. Pharr, N. B. Ratner & L. Rescorla, 2000, Applied Psycholinguistics, 21, p. 436. Copyright 2000 by Applied Psycholinguistics. Reprinted with permission. 226 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS Vol. 41 223 230 April 2010

Table 7. Summary of average SSL values when the analysis corpus included only intelligible words. Standard deviation in parentheses when reported. Mean age a LT ND N Average SSL N Average SSL 16 10 1.5 1 20 10 2.16 (0.17) 2 20 10 2.21 (0.19) 2 23 10 1.8 1 10 1.9 1 Note. LT = late talker; ND = normal development. Source indicated by superscript: 1 Thal et al., 1995; 2 Morris, 2009. a Average age in months for all participants in each study. In total, available research suggests that MBL and SSL have applicability in the clinical evaluation and progress monitoring of phonological skills in children as young as 9 months (Stoel-Gammon, 1989) through the age of 36 months (Pharr et al., 2000). For diagnostic purposes, SSL is a temporally stable measure (Morris, 2009) that distinguishes children with language impairment from those with typical language skills (Paul & Jennings, 1992; Pharr et al., 2000; Thal et al., 1995). Specifically, children with language impairment have lower SSL scores than their age-matched peers. Moreover, Fasolo et al. s (2008) data support Stoel-Gammon s (1989) suggestion that MBL is predictive of language skill at 24 months. Because SSL increases with increased language skills (Paul & Jennings, 1992; Pharr et al., 2000; Thal et al., 1995), SSL can be used for monitoring progress in therapy. similar children between the ages of 9 and 36 months who were typical or delayed in their development of language. An examination of trends in the compiled data in Table 3 indicates that the babble of 9- to 12-month-olds mostly contains vowels with a few true consonants, as shown by MBLs between 1.16 and 1.50. Children continue to add true consonants to their babble and have an average MBL of 1.90 by 23 months, indicating that most of their vocalizations contain a true consonant, with some vocalizations containing two different true consonants. Fasolo et al. (2008) found that the average MBL of the 20-montholds in the group of late talkers (M = 1.22, SD = 0.19) was significantly lower than the average MBL of the 20-month-olds in the typical group (M = 1.50, SD = 0.31). However, Stoel-Gammon s (1989) results remind SLPs that an MBL value without knowledge of the characteristics of the original transcriptions may be misleading. For example, if a child is producing many Level 2 vocalizations but is restricted in his or her manner and place of sound articulation, this child is at risk for being a late talker. Although the Italian study by Fasolo et al. (2008) does follow the same pattern as the English studies in terms of the normally developing group scoring higher than the late talker group, the fact that the Italian-speaking 20-month-old normally developing children scored similarly to 12-month-old English-speaking normally developing children indicates that there may be a language effect. In fact, the syllable structure of Italian does differ from that of English in that word-final consonants and word-final consonant clusters are uncommon (Bortolini & Leonard, 2000). Therefore, Level 3 utterances will probably be a rarity in children who are learning Italian because they will need to produce an initial consonant cluster or a language-atypical syllable structure in order to produce two different true consonants. As can be seen in Table 5, the difference between children with typical language skills and those with delayed language is more dramatic when assessing SSL in English-speaking children between 20 and 28 months of age. The average SSL of the children developing language typically in this age range was >2.0, whereas the average SSL of the children with delayed language was <1.75. Interestingly, Paul and Jennings (1992) did not find a significant difference in average SSL between the typical language learners in the 18- to 24-month age group and those in the 24- to 34-month age group. Additionally, the SSL for 36-month-old normally developing children in the Pharr et al. (2000) sample (M = 2.39, SD = 0.10) does not appear to differ from the SSL for Paul and Jennings 24- to 34-month-olds (M = 2.34, SD = 0.17). DISCUSSION Phonetic transcription of young children s speech is difficult. An appealing aspect of MBL and SSL is that levels are assigned based on the phonetic characteristics (such as manner and place) and not the specific phoneme produced. Therefore, these measures are more reliant on accurate coding of broader classes of sounds and are less reliant on accurate segment transcription, likely increasing their test retest reliability. In order for MBL and SSL to be used most effectively in clinical practice, standardized procedures need to be followed. Although language sample elicitation was almost identical for each reviewed study (i.e., unstructured parent child play sample using age-appropriate toys), there were differences in sample duration and the words that were included in the analysis corpus. Further, clarification is needed for the reliable coding of productions that include vocalic liquids and consonant clusters. Future research will provide definitive guidance concerning which procedures are most appropriate. Meanwhile, practitioners may want to consider following the procedures suggested next. The differences in reported scores between MBL and SSL indicate that there are distinctions in the complexity of premeaningful speech productions and meaningful speech productions. Thus, it is recommended that MBL, as presented in Table 1, be used with children in the premeaningful speech stage of language development and SSL, as presented in Table 4, with children in the meaningful speech stage. Recalling that the premeaningful speech stage is defined as producing fewer than 10 words during two separate 1-hr sessions, it may be necessary to obtain such a sample, or administer an expressive vocabulary measure such as the MacArthur Communicative Development Inventories (MCDI; Fenson et al., 1993), in order to determine if MBL or SSL is the most appropriate measure for a specific child. The fact that there were no significant differences between the SSL of typical language developers in the 18- to 24-month-old group (M = 2.22, SD = 0.30) and the 24- to 28-month-old group (M = 2.34, SD = 0.17) in Paul and Jennings (1992) study indicates that there may be ceiling effects on the measure. This is a concern because when the majority of scores are at or near the maximum possible for a measure, differences between children cannot be identified. This lack of variability indicates that the measure is not providing the necessary information for diagnosis or progress monitoring. According to the MCDI, the 15 th percentile for 24-month-old boys is 100 words and for girls is 150 words. By 28 months, children at the Morris: Clinical Application of the Mean Babbling Level 227

low end of normal limits are producing approximately 300 words. Because children with larger vocabularies will likely participate in picture-naming tasks that allow for relational instead of independent analyses of phonology, it is recommended that SSL only be used when parent report of a child s vocabulary is less than 150 words. This will help avoid the measure s possible ceiling effect. These possible ceiling effects are likely byproducts of language abilities and are not specific to chronological age. Because SLPs will be assessing children who are delayed in speech-language skills, it is recommended that expressive vocabulary, and not a specific age in months, be used to determine the applicability of SSL. As documented in Pharr et al. (2000), older (36-month-old) children with speech-language impairment will likely function similar to younger (24-month-old) typically developing children, minimizing SSL ceiling effects in the population of children under the age of 3 years with language impairment. With the exception of Pharr et al. (2000), all of the described studies used three levels, which sufficiently showed differences between groups as well as developmental trends. A fourth level is unnecessary due to the limited use of consonant clusters in children at this developmental level. Although consonant clusters do begin to emerge by 24 months of age, many consonant clusters are simplified to include a glide element (Dyson, 1988; Watson & Scukanec, 1997) and thus may not provide evidence that the cluster has the underlying representation of the combination of two true consonants (Chin, 1996). Therefore, when calculating SSL, it is recommended that Appendix B be used, which appends Table 4 to include twoelement consonant clusters that include a glide element in Level 2 and two-element clusters (initial or final position) that include two true consonants in Level 3. An additional clarification is necessary for adequate coding of words that include liquids. Phonetically, because of the fluid nature in which liquids are produced, they are grouped with glides and are categorized as semivowels (Espy-Wilson, 1992). SLPs typically consider liquids as true consonants because liquids are generally later developing and children often substitute a glide for a liquid target. Although not specified in any of the reviewed studies, it is recommended that r-colored vowels (e.g., park, beard, bear, bored) be coded as vowels (not consonants) because they serve as the syllabic nuclei. Developmental trends of the liquid phoneme class indicate that consonantal liquids should be coded as true consonants. However, future research may support classifying liquids similar to glides in children under 3 years of age. SLPs must decide which words to include in their analysis corpus. Obviously, for children in the premeaningful stage of development, all vocalizations should be used because the majority of utterances are unintelligible. For children in the meaningful speech stage of development, the decision is less clear. Two of the studies (Morris, 2009; Thal et al., 1995) restricted the analysis corpus of SSL to intelligible words. The other two studies (Paul & Jennings, 1992; Pharr et al., 2000) used SSL to analyze all of the vocalizations. The justification for limiting the sample to intelligible words was that the phonological system of children with language impairment might be more taxed during word production than during babbling. However, limiting the sample to only intelligible words greatly reduces the number of words that can be analyzed. Even when a child with language impairment is in the meaningful speech stage of development, he or she may only say 11 words in a 30-min sample with simple target structures (e.g., mommy, hi, bye, no, you). Therefore, in order to obtain a more accurate picture of a child s expressive capabilities, the author recommends calculating SSL on a minimum of 50 unique vocalizations (intelligible and unintelligible) as well as all of the intelligible words for the entire length of the sample, resulting in two separate scores. Determining unique vocalizations can be difficult in young children, particularly children with language impairment, as they often simplify their productions by creating homonyms (e.g., /gãk/ for truck and stuck). At the same time, young children may phonetically vary productions of the same word. Because those productions may vary considerably in terms of their phonetic characteristics and syllable shape (e.g., [si] and [Xip] for sheep), including the variants may provide a more accurate picture of the child s phonological system and phonetic capabilities. Thus, the author recommends including all unique intelligible word productions in the sample even if they are produced as a homonym with another word. Multiple productions of the same word should be included if a consonant or the syllable shape is changed but not if the productions vary by only a vowel. Finally, because the target of unintelligible words is unknown, only one phonetic representation of each unintelligible word should be included. MBL and SSL promise to be worthy complements to current assessment measures for the under 3 population. Although more research is necessary, SLPs should feel confident that these measures do provide reliable and helpful clinical information. Similar to mean length of utterance in morphemes (MLUm), which was developed to provide a global index of a child s grammatical skills (Brown, 1973), MBL and SSL appear to provide a global index of a child s phonological skills. Although MLUm is not flawless (see Eisenberg, Fersko, & Lundgren, 2001), it does provide summary information concerning a child s overall language abilities. When provided with an MLUm, the experienced SLP will quickly form an impression of a child s language skills. For example, an MLUm of 1.4 is associated with a child who says utterances like want cookie, more, and go byebye, but not with a child who says I want the chocolate cookies. Likewise, an MBL or SSL of 1.4 provides a snapshot of a child whose speech is characterized by vowels and some true consonants (e.g., [ju], [æ], [es]).rarelywouldachildwithanmblorssl of 1.4 produce two true consonants (e.g., [mibim], [kæt]) within an utterance (MBL) or word (SSL). MBL and SSL can be used to present summary information that can document a child s phonological gains throughout the intervention process. For example, a child who begins intervention producing mainly simple consonant vowel syllable shapes with true consonants may have an MBL of 1.9. As the child expands his syllable shapes to include final consonants, his MBL will increase. It will continue to increase as the child adds more sounds and syllable shapes to his inventory. Once the child begins to use words, SSL can be used to monitor his development until the time that relational analyses are more appropriate. ACKNOWLEDGMENT I am grateful for the assistance of Pamela Hadley during all phases of this project. REFERENCES Bortolini, U., & Leonard, L. B. (2000). Phonology and children with specific language impairment: Status of structural constraints in two languages. Journal of Communication Disorders, 33, 131 150. 228 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS Vol. 41 223 230 April 2010

Brown, R. (1973). A first language: The early stages. Cambridge, MA: Harvard University Press. Chin, S. B. (1996). The role of the sonority hierarchy in delayed phonological systems. In T. W. Powell (Ed.), Pathologies of speech and language: Contributions of clinical phonetics and linguistics (pp. 109 117). New Orleans, LA: International Clinical Phonetics and Linguistics Association. Dyson, A. T. (1988). Phonetic inventories of 2- and 3-year-old children. Journal of Speech and Hearing Disorders, 53, 89 93. Eisenberg, S. L., Fersko, T. M., & Lundgren, C. (2001). The use of MLU for identifying language impairment in preschool children: A review. American Journal of Speech-Language Pathology, 10, 323 342. Espy-Wilson, C. Y. (1992). Acoustic measures for linguistic features distinguishing the semivowels /w j r l/ in American English. Journal of the Acoustical Society of America, 92, 736 757. Fasolo, M., Majorano, M., & D odorico, L. (2008). Babbling and first words in children with slow expressive development. Clinical Linguistics and Phonetics, 22, 83 94. Fenson, L., Dale, P., Reznick, J., Bates, E., Thal, D., Hartung, J., & Reilly, J. (1993). MacArthur Communicative Development Inventories. San Diego, CA: Singular. Morris, S. R. (2009). Test retest reliability of independent measures of phonology in the assessment of toddlers speech. Language, Speech, and Hearing Services in Schools, 40, 46 52. Olswang, L., Stoel-Gammon, C., Coggins, T., & Carpenter, R. (Eds.). (1987). Assessing prelinguistic and early linguistic behaviors in developmentally young children. Seattle, WA: University of Washington Press. Paul,R.(Ed.).(1998). Exploring the speech-language connection. Baltimore, MD: Brookes. Paul, R., & Jennings, P. (1992). Phonological behavior in toddlers with slow expressive language development. Journal of Speech and Hearing Research, 35, 99 107. Pharr, A. B., Ratner, N. B., & Rescorla, L. (2000). Syllable structure development of toddlers with expressive specific language impairment. Applied Psycholinguistics, 21, 429 449. Schwartz, R., & Leonard, L. (1982). Do children pick and choose? An examination of phonological selection and avoidance in early lexical acquisition. Journal of Child Language, 9, 319 336. Stoel-Gammon, C. (1987). Language Production Scale. In L. Olswang, C. Stoel-Gammon, T. Coggins, & R. Carpenter (Eds.), Assessing prelinguistic and early linguistic behaviors in developmentally young children (pp. 120 150). Seattle, WA: University of Washington Press. Stoel-Gammon, C. (1989). Pre-speech and early speech development of two late talkers. First Language, 9, 207 223. Stoel-Gammon, C. (1991). Normal and disordered phonology in two-yearolds. Topics in Language Disorders, 11, 21 32. Thal, D., Oroz, M., & McCaw, V. (1995). Phonological and lexical development in normal and late-talking toddlers. Applied Psycholinguistics, 16, 407 424. Watson, M. M., & Scukanec, G. P. (1997). Profiling the phonological abilities of 2-year-olds: A longitudinal investigation. Child Language Teaching and Therapy, 13, 3 14. Received July 8, 2008 Accepted January 25, 2009 DOI: 10.1044/0161-1461(2009/08-0076) Contact author: Sherrill R. Morris, School of Allied Health and Communicative Disorders, Northern Illinois University, DeKalb, IL 60115. E-mail: srmorris@niu.edu. Morris: Clinical Application of the Mean Babbling Level 229

APPENDIX A. STUDIES REPORTING MEAN BABBLING LEVEL (MBL) AND SYLLABLE STRUCTURE LEVEL (SSL) Source n Child s language status Age range in months Vocalizations included Sample length and conversation partner Summary of results Stoel-Gammon (1989) Fasolo et al. (2008) Paul & Jennings (1992) Thal et al. (1995) Pharr et al. (2000) 32 ND 9 24 Babble only 50 consecutive utterances; 20 25 min with parent 12 ND 20 Babble only (Italian 12 LT speakers) 8 9 ND LT 18 24 Mix of words and babble 17 ND 24 34 19 LT 10 LT 11 28; 10 L match meaningful speech 10 CA match 7 LT 12 33; 7 L match premeaningful 7 CA match speech 15 ND 24 36; 20 LT longitudinal sample Intelligible words were analyzed separately from babble Mix of words and babble A low MBL or restrictive phonetic inventory is predictive of impaired language skill at 24 months. 30 min with mother MBL was significantly lower for the LT group than for the ND group. 50 different utterances; 10 min with mother 30 min (10 min with parent, 20 min with experimenter) Morris (2009) 10 ND 18 22 Intelligible words Two 20-min sessions with mother The LT group had fewer advanced syllable types (VCs, clusters, disyllables); SSL in the LTs did not change significantly between 18 34 months; phonology in LTs is delayed, not deviant. LTs intelligible SSL is higher than that of the L match group but lower than that of the CA match group. LTs MBL is lower than that of the L match and CA match groups. 10 min with mother SSL increases with age. SSL of the LT group is delayed in comparison to that of the ND group; SSL may be a predictor of later language outcomes. SSL is a temporally reliable measure. Note. ND = language skills developing as expected; LT = late talker; L match = language match; CA match = chronological-age match. APPENDIX B. RECOMMENDED SYLLABLE STRUCTURE LEVELS Level Description 1 The utterance is composed of a voiced vowel ([A]), voiced syllable consonant ([m]), or CV syllable in which the consonant is a glottal stop ([?o]) or a glide ([ha], [wi]). 2 The utterance is composed of a VC ([up]), CV ([pu]), or CVC with a single consonant type ([kek]) that does not fit the criteria for Level 1, or a two-element consonant cluster that includes a glide element ([twi]). Voicing differences are disregarded. 3 The utterance is composed of syllables with two or more different consonant types, disregarding voicing differences ([pati]), or two-element clusters (initial or final position) that include two true consonants ([blu], [best]). Recommended SSL calculation conventions 1. Count all r-colored vowels and diphthongs as vowels, not consonants. 2. Include phonetic variants of the same word if the variation includes a consonant or syllable structure change. 3. Include only one instance of each unintelligible word. 4. Calculate two SSL values. a. Fifty unique vocalizations (intelligible and unintelligible) b. All intelligible words for the duration of the sample (minimum 20 min) 230 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS Vol. 41 223 230 April 2010

Clinical Application of the Mean Babbling Level and Syllable Structure Level Sherrill R. Morris Lang Speech Hear Serv Sch 2010;41;223-230; originally published online Sep 15, 2009; DOI: 10.1044/0161-1461(2009/08-0076) The references for this article include 5 HighWire-hosted articles which you can access for free at: http://lshss.asha.org/cgi/content/full/41/2/223#bibl This information is current as of January 16, 2013 This article, along with updated information and services, is located on the World Wide Web at: http://lshss.asha.org/cgi/content/full/41/2/223