歯김영태4.PDF

Transcription

1 Semantic Analyzability of the Intra - and Inter - sentences of the Language - Disordered Children at 2- to 3- year of Language Age* 1 Yo ung Ta e Kim (Ewha Womans University) Young Tae Kim. S e ma ntic a na lyza b ility of the intra- a nd inte r- s e nte nces of the la ngua ge- dis orde re d childre n at 2- to 3-ye a r of la ngua ge a ge. Korea n J ourna l of Com munication Dis orde rs, 19 99, 4, T he purpose of this study was to investigate the semantic analyzability of the intra- and inter- sentences of the language- disordered children. T hirty- five language- disordered children and 35 MLUmatched 2- to 3- year- old normal children served as the participants. Participants ' spontaneous utterances were collected and analyzed semantically. The findings were as follows : (1) the number of grammatical units and the semantic units are significantly correlated in both groups, (2) most frequently used semantic roles and relations were common in the two groups. Only five of the semantic roles were statistically different between the two groups. Although most of the semantic roles and relations occurred more frequently in the language- disordered group, the possessor and communicative devices occurred more frequently in the normal group, (3) among the intersentence relations, the language- disordered group used significantly more juxtaposition and conjoining, while there was no difference in the use of embedding. However, adjective embedding was used more frequently in the normal group, while substantive embedding was used more frequently in the language- disordered group. This study suggested a need for a more systematic semantic analysis of early simple and complex sentences of the language- disordered children. I. INT RODUCT ION It has been noted that young children in the single- word level begin to combine w ords once he has acquired approximately 50 words. If these sentences are analyzed only T his research w as supported by grants from the Ewha Womans University Research Fund

2 in term s of grammatical aspect s, much information would be missed since these early sentences are combined semantically rather than grammatically (Braine, 1963; Bloom, 1970; Brown, 1973; McLean & Snyder - McLean, 1978). Since Bloom ' s (1970) finding that children ' s early sentences are semantically oriented rather than grammatically oriented, many researchers and clinicians have developed sy stematic ways to semantically analyze early sentences. Often semantic roles and relation s are used for analyzing early one- to three- word sentences. It has been agreed that these early semantic roles are univer sal (Bloom, 1970; Schlesinger, 1971; Bowerman, 1980; Owens, 1999). F or example, Stockman and Vaughn - Cooke (1986) reviewed several studies of nonstandard English speaker s and reported similar semantic roles to those of native standard English speaker s. F or single- w ord sentences, Bloom (1973) categorized them into substantive words and function w ords, while Nelson (1973) u sed 4 categories : nominals, action w ords, personal- social words, and function w ords. F or the sentences with more than two words, Brown (1973) reported 8 frequently used 2- word semantic relations in Brown ' s I- II Levels (MLU- m ): agent - action, action - object, agent - object, action - locative, entity - locative, possessor - possession, attribute- entity, and demonstrative- entity. T he agent - action - object, agent - action - locative, and action- object - locative relations were reported as frequently used 3- word semantic relations. McLean and Snyder - McLean (1978) reviewed previou s studies and suggested a more practical and flexible analy sis sy stem. In their sy stem, for example, the demonstrative- entity relation could be substituted by the notice- entity relation and the entity - attribute relation could be substituted by the entity - recurrence/ nonexistence relation. Retherford et al. (1977) reported several less frequently used semantic roles based on their study with 11 children in Brown ' s I- III Levels. T hese less frequently used semantic roles were instrument, comitative, created object, beneficiary/ recipient, and experiencer. Retherford et al. (1981) developed a more detailed system for semantic analysis. T hey suggested (1) 15 main semantic categories (e.g., action, locative, agent ), (2) 4 grammatical categories which have grammatical function rather than semantic function (e.g., entity, negation, attribute, adverbial), (3) conversational devices which have more conversational function (e.g., attention, affirmation ), and (4) routines which are conventional or automatic expressions. T heir sy stem is differentiated from the previou s sy stem s since they attempted

3 Se ma ntic Ana lyzability of the Intra- a nd Inte r- s e nte nces not to fit all of the children ' s utterances into semantic roles. Rather, they looked at compositional (i.e., grammatical) aspects and automatic speech separately from semantic aspects. T hey suggested that we semantically analyze simple sentences and grammatically analyze the complex sentences. Although some of the researcher s or clinician s semantically analyze only simple sentences, it has been noted that children ' s utterances even in the primitive sentence level include complex sentences. T hese early complex sentences are often incomplete or incorrect. Children initially join two ideas simply by juxtaposing words. Gradually, they begin to conjoin sentences approximately in Stage IV (Miller, 1981). T hese early conjoining sentences tend to be semantically related rather than grammatically related. Research for Korean children 's semantic development has been attempted longitudinally (Zho, 1982) and cross- sectionally (Kim, 1998). Kim (1998) analyzed to 3- year old children and reported 8 frequently used semantic roles (object, entity, agent, action, statement, attribute, locative, negation ). In addition, she reported 5 of the frequently used 2- term sem antic relation s (object - action, entity - st at em ent, background (lo c a t iv e ) - action, agent - action, possessor - possession (e n t ity )). Result s indicated that most frequently used 3- term relations were extensions of the object- action or background- action (i.e., agent- object- action, object - background- action, agent- background- action ). Most frequently u sed 4- term relations were also extensions of the frequent 2- term relations (agent - action, entity - statement ) or 3- term relation s (agent - object - action ). A few of comparative studies between the normal children and the languagedisordered children showed that the language- disordered children used less variable semantic roles. Leonard (1984) and T rauner et al. (1995) reported that the languagedisordered children began combining words later and showed less variable than the normal children did. However, compared with the MLU - matched normal children, the language- disordered children did not show noticeable difference on 2- word semantic relations. For exmaple, Freedman and Carpenter (1976) reported no differences except in the introducer - entity relation between the normal and the language- impaired children at Stage I. Based on these result s, researcher s suggested that the language- disordered children have capacity to develop their semantic aspect of language appropriate to their language age (Morehead & Ingram, 1973; Wells, 1974; Leonard, 1984).

4 Even in the 90' s, semantic relations have been an important goal of language intervention. E specially in the natural or pragmatic language intervention program, semantic relations have been a main target for the language- disordered children at 2- to 3- word sentence level (Kim & Lombardino, 1991; Warren et al., 1994; Kaiser & Hester, 1994). In Korea, many language- disordered children at early sentence level have been trained on the basis of their semantic and pragmatic level of development. However, few studies have been conducted for semantic characteristics of the language- disordered children. In addition, their inter sentence relations have been ignored. T he purpose of the present study was to examine the semantic characteristics of the language- disordered children ' s utterances based on the following aspects : (1) the relation s among the chronological ages, the grammatical unit of the utterance length, and the semantic unit s of the utterance length, (2) the most frequently u sed intrasentence relation s (semantic roles and relation s), and (3) the most frequently u sed intersentence relation s (juxtapositions, conjoining, embedding ). F or this purpose, the MLU - matched normal children ' s data were compared with the language- disordered children ' s data.. MET HOD 1. Pa rtic ipa nts T hirty - five language- disordered children who showed 2- to 3- year of language age served as the participant s of this study. T hey showed more than 2 year s of language delay and have received less than 6 month s of speech therapy. None of the participant s showed physical, sensory, or behavior problems. T o compare the language- disordered children ' s data with the normal data, 35 normal children ' s data were selected from the researcher ' s previou s study (Kim, 1997). T hey w ere identified as norm al by the result s of th e Picture V ocabulary T est (Kim et al., 1995). T h e t w o group s of children w ere matched by their MLU- w (the Mean Length of Utterance in Words), one by one. T he means of MLU- w for the normal and the language- disordered w ere 2.48 ( ) and

5 Se ma ntic Ana lyzability of the Intra- a nd Inte r- s e nte nces 2.45 ( ), respectively. T he means of CA for the normal and the languagedisordered were 2;9 (2;0-3;10) and 6;4 (2;7-13;1), respectively. Result s of t- test indicated that the mean difference of MLU - w for the two groups was not significant (t =.24). 2. P roce dure Language samples of each child were collected during free conver sation in a play setting. T he language samples were audio- recorded for 30 to 40 minutes and transcribed within 1 or 2 days. From each sample spontaneous utterances were transcribed and 70 natural and representative utterances were selected and analyzed. T he utterance w as defined when one of the following criteria was met : (1) when one sentence is finished, (2) when there is a prosodic change or a pause longer than 5 seconds, or (3) when topic is changed (Klee & Fitzerald, 1985; Rondal et al., 1987; Owens, 1999; Kim, 1997). Each utterance was analyzed in terms of (1) utterance lengths measured by the number of words and the number of semantic unit s, (2) intrasentence relation s (semantic roles and relations ), and (3) intersentence relation s (juxtaposition, conjoining, embedding ). Juxtaposition relations were observed in terms of substantive and action/ state roles. Conjoining relations were observed in terms of temporal, concurrent, contrary, conditional/ hypothetical, additional, reason/ cause, purpose/ intention, change, concession, no relation, and background. Embedding relations were observed in terms of subtantive, action/ state, adjective, and background roles. T he semantic roles are coded in T able 1 which is based on Retherford et al.' s system (1981) and supplemented by the researcher of the present study (see Kim, 1998 for detailed coding guidelines). T he intersentence semantic relations are coded in T able 2 which is a composite revision based on several studies (Clancy et al., 1976; Hood & Bloom, 1979; Nam & Ko, 1983). F or ex ample, the utterance of (meaning of "(I am ) full because (I) ate a lot ") is analyzed as follow s: the number of words: 3 the number of semantic unit s : 3 semantic roles : agent, reason, statement, adverbial, action

6 semantic relation s: reason (adverbial- action ) - statement intersentence relation : reason conjoining 3. Data Ana lys is T o examine analyzability of the semantic unit as a measure of utterance length, the Pearson correlation coefficient s w ere calculated among CA, grammatical unit s (MLU - w ), and semantic unit s. T he semantic unit s were further divided into (1) the mean number of semantic unit s (MSU ), (2) the mode number of semantic unit s (MoSU), and (3) the maximum number of semantic unit s (Mx SU ). T o analyze the intrasentence relations, top 10 of the most frequently used semantic roles and relations (2-, 3-, and 4- term s) w ere selected. And the frequency of the each role and relation was analyzed by t- test to compare the tw o groups. T o analyze the inter sentence relations, the mean frequency of each type (juxtaposition, conjoining, embedding ) was analyzed and compared the two groups by t- test. 4. Inte rjudge Re liability Interjudge reliability w as calculated for 25 % of the languge samples. A graduate stu dent in speech pathology serv ed as a secon d ju dg e. Int erjudge agreem ent s w ere 92 % in dividing utterances, 85 % in counting number of words and semantic unit s, and 82 % in coding intra- and inter - sentence roles and relation s.. RESULT S 1. Re latio ns hips amo ng CA, Grammatic a l Units, a nd S e ma ntic Units As seen in T able 3, in the language- disordered group, all of the unit s of the utterance length (MLU- w, MSU, MoSU, Mx SU) were correlated significantly. Especially, MLU- w and MSU showed the highest correlation (r =.91, p <.01). In the normal group, MLU- w

7 Se ma ntic Ana lyzability of the Intra- a nd Inte r- s e nte nces Table 1. Definitions of Intrasentence Semantic Roles Compositional Role. Substantive Roles. Action / State Roles. Adjective Roles. Communicative Devices Semantic Role Definition agent A performer (animate of inanimate) of an action. experiencer Someone or something that undergoes a given experience or mental state. possessor One who possesses (an) object (s). comitative One who performs or participates together. recipient One who benefitted from an action. object A person or thing that receives the force of an action entity A substantive w ord that is stated without action. created object * A person or thing that has been made or changed. action A perceivable movement or activity engaged by an agent. state mental state* A mental condition experienced by a person or object. entity state* T he stated condition/ labelling of an entity. negative state* T he stated condition that includes negative meaning. attribute A description of size, shape, quality of an object or person. adverbial A description of degree or quality of an action or statement. background* A description of background of action or statement. negation T he negative modifier of an action : nonexistence, rejection, cessation, denial, or disappearance. time T he expression of time of an action or statement. location T he place where an object or action w as located or toward which it moved. instrument Something that an agent uses to carry out an action. reason T he expression of reason, intention, or cause of an action or statement. condition * T he expression of condition of an action or statement. comparison T he expression of comparison of the substantive meaning. recurrence T he expression of additional instance or reappearrence. concession * T he expression of concession or permission of action or statement. attention Use of an individual's name or other expression to gain attention. repetition request Use of terms like "w hat", "huh" etc. to request repetition of speaker ' s utterance. exclamation * Use of terms like "oh", "ah". affirmation Use of affirmative terms to assert that a previous utterance or behave is correct or to indicate compliance with request from previous utterance. emphasis * Use of reduplicated meaning. sounds accompanying Noises or sounds used to replicate sounds of agent or object. greeting Use of greeting like "hi". conjunction * Use of conjuncture in a simple sentence. routine Use of rote counting or recitation of story/ song/ poem. (Based on Retherford et al., 1981; Kim, 1998) *: added or revised semantic role by the present researcher.

8 Table 2. Definitions of Inte rsentence Semantic Relations Intersentence Relations Juxtaposition Conjoining Embedding Description Simply juxtapose (list) several words which have the same semantic roles. T wo phrases / sentences are combined mainly by one of the following conjunctional relations: temporal, concurrent, contrary, conditional/ hypothetical, additional, reason/ cause, purpose/ intention, change, concession, no relation, background Combine two phrases/ sentences by an incorporation of one within the structure of the other. Example,,. [meaning "I like apples, watermelon, grapes."]: agent - object juxtapositions - action,. [meaning "If (you) eat, (I will) give (it)."]: action - conditional conjoining - action [meaning "I told not to do."]: agent - action embedding (negation - action) and M SU also show ed the high est correlation (r =.89, p <.01). How ev er, significant correlation was not found between MLU - w and M osu in the normal group. In the language- disordered group, only low correlation s were found between CA and either unit of the utterance length (grammatical unit and semantic unit ). In the normal group, however, CA show ed significant correlation s with MLU- w (r =.61, p <.01) and with MSU (r =.49, p <.01). 2. Intras e nte nce Re latio ns A. S em antic Role s T he semantic roles which ranked within the 10th were summarized in T able 4. In semantic roles, entity, object, and agent were the most frequently used substantive roles in that order. Most of (all 4 in the language- disordered group, 3 out of 4 in the normal

9 Se ma ntic Ana lyzability of the Intra- a nd Inte r- s e nte nces Table 3. Correlations among CA, Grammatical Unit, and Semantic Units CA Grammatical Unit Semantic Units MSU MoSU MxSU LD N LD N LD N LD N LD N CA * * Grammatical Unit (MLU- w) * *.49 * *.89 * *.39 *.14 N: normal group, LD: language- disordered group, * p <.05, * * p < * *.33 Semantic Units MSU MoSU MxSU.64 * * * *.61 * * * *.63 * *.65 * *.08 group) the action/ state roles ranked within the 10th. Action w as u sed most frequently, follow ed by mental state, entity state, and negative state in the order given. In adjective roles, only the background (location ) and adverbial roles ranked within the 10th. In the case of communicative devices, only the affirmation and attention roles ranked within the 10th. Most of the semantic roles (25 out of 35) were u sed more frequently in the language- disordered group than in the normal group except for the following 10 semantic roles: possessor, comitative, entity, action, entity state, location, negation, condition, recurrence, and communicative devices. However, statistically significant differences were found only in 5 semantic roles: possessor (t = 4.00, p <.01), negative state (t = 2.97, p <.01), reason (t = 2.27, p <.05), repetition request (t = 2.70, p <.01), and exclamation (t = 3.52, p <.01). T hree of them (negative state, reason, and exclamation ) were more frequently u sed in the language- disordered group than in the normal group and possessor and repetition request were u sed more frequently in the normal group. B. S em antic Relation s T he 2- to 4- term semantic relation s which ranked within the 10th are demonstrated

10 Table 4. Rank a nd Mean Frequency of the Semantic Roles and Relations Rank Semantic Roles LD N Mean Frequency of Occurrence Semantic Relations 2- term 3- term 4- term LD N LD N LD N 1 action action ent- sta obj- act age- objacacbac- age- obj- age- bac- age- bac- act obj- act (19.2) (22.16) (22.5) (20.6) (16.1) (14.0) (9.3) (14.9) 2 entity entity obj- act ent- sta ent- bacstacadv- age- bac- age- obj- ent- bac- act att- sta (14.6) (17.0) (20.3) (19.0) (14.5) (14.0) (9.3) (14.9) 3 mental state mental state bac- act bac- act obj- bacact ent- bacsta ent- bacadv- sta pos- entbac- sta (9.3) (8.4) (11.7) (12.6) (11.3) (11.2) (7.4) (8.5) 4 object object adv- act bac- sta age- bacacacatt- obj- bac- ent- bac- act- bac- sta bac- act (8.0) (7.1) (8.3) (8.2) (11.0) (8.8) (5.6) (8.5) 5 agent entity age- act age- act age- advacacent- obj- adv- exp- bac- att- ent- state sta adv- sta (5.2) (5.7) (7.9) (7.5) (5.5) (6.4) (5.6) (6.4) 6 entity state agent bac- sta adv- act obj- advact age- advact (4.8) (5.0) (7.3) (6.2) (5.2) (4.4) 7 back back exp- sta pos- ent exp- bacststa pos- ent- (location) (location) (4.4) (4.8) (4.8) (5.7) (3.9) (4.4) 8 adverbial adverbial bac- ent exp- sta ent- advsta ent- advsta (4.4) (3.9) (2.8) (4.0) (2.9) (2.8) 9 affirmation affirmation att- sta pos- sta exp- entsta att- entsta (3.7) (3.6) (2.5) (3.5) (2.9) (2.8) 10 negative state (2.4) attention (3.4) att- ent (2.2) att- sta (3.0) adv- bacact (2.9) exp- bacsta (2.4) * LD: language-disordered group, N: normal group act : action, adv: adverbial, age: agent, att: attribute, bac: background, ent : entity, exp: experiencer, obj: object, pos : possessor, sta: state in Figure 1. In the 2- term relations, entity - state was used most frequently, followed by object - action and background(loc a t ion )- action (see Figure 1a). T he possessor - entity and possessor - state relations ranked within the 10th only in the normal group. T his result seems to be related with the differential u se of possessor between the tw o groups.

11 Se ma ntic Ana lyzability of the Intra- a nd Inte r- s e nte nces Meanwhile, background- entity and attribute- state ranked within the 10th only in the language- disordered group. In the 3- term relations, agent - object - action was u sed most frequently, followed by entity - background- state and object - background- action (see Figure 1b ). Most of the top ten of the 3- term relations (8 out of 10) were composed of the two of the 2- term relation s which ranked within the 10th. F or example, agent - object - action seemed to be extended from agent - action and object - action. T wo of the 3- term relations (possessor - entity - state, attribute- entity - state) ranked within the 10th only in the normal group. Meanwhile, experiencer - entity - state and adverbial- background- action ranked within the 10th only in the language- disordered group. The four- term relations did not appear frequently in both groups (see Figure 1c). The 4- term relations showed more than 4 times in the language- disordered group were agent-background- background- action, agent - object - adverbial- action, and entity -backgroundadverbial- state. In the normal group, agent- background- object - action and entity -backgroundadverbial- state relations appeared most frequently, followed by possessor - entity - backgroundstate and agent - background- background- action. 3. Inte rs e nte nce Re latio ns T he juxtapositions and conjoining intersentence relations were used more frequently in the language- disordered group than in the normal group (t = 2.62, p <.05), the difference being statistically significant. However, the frequency of the embedding intersentence relation s were not significantly different between the two groups. As seen in Figure 2a, entity was the most frequently juxtaposed semantic role, follow ed by object and comitative in the language- disordered group. In the normal group, entity and comitative were the most frequently juxtaposed, while object was not frequently juxtaposed. In conjoining intersentence relations, as seen in Figure 2b, temporal, conditional, and concurrent relations occurred most frequently in both groups. In embedding inter sentence relation s, as seen in Figure 2c, adjective roles were the most frequently embedded in the normal group, while substantive roles were the most frequently embedded in the language- disordered group. Specifically, created object and entity were

12 the most frequently embedded in both groups. In addition, attribute w as the much more frequently embedded role than adverbial in the language- disordered group, while no difference was found in the normal group. Figure 1a. Most Frequently Used 2- Te rm Semantic Relations Figure 1b. Most Frequently Used 3- Te rm Semantic Relations

13 Se ma ntic Ana lyzability of the Intra- a nd Inte r- s e nte nces Figure 1c. Most Frequently Used 4- Te rm Semantic Relations Figure 2a. J uxtapos ing Inte rsentence Relations Figure 2b. Conjoining Inte rsentence Relations ADJ ECTIVE EMBEDDING SUBSTANTIVE EMBEDDING Figure 2c. Conjoining Inte rsentence Relations

14 . DISCUSSION High correlations between the number of grammatical unit s and the semantic unit s in both groups indicate that the number of semantic unit s, especially the mean number of semantic unit s (MSU), can be used as an important measure of expres sive language development of young children. However, the correlation s between the number of unit s of the utterance length and chronological age were not significant in the language- disordered group. T his result seems to be due to the participants ' wide range of chronological ages in the language- disordered group. Many researchers reported low correlation s between MLU and CA beyond age of 4. Poor correlations betw een the number of semantic unit s and CA in the language- disordered group indicate that the mean number of semantic unit s, like MSU, is not a sensitive measure of language development beyond certain age. More detailed research could be conducted to investigate the relationships between MSU and CA with the normal children beyond age of 4. Result s of the analy ses of frequently used semantic roles and relations suggest that there are no quantitative differences betw een the language- disordered children and their MLU- matched normal children. However, some qualitative differences were found in this study. F or ex ample, some of the semantic roles (e.g., posses sor and repetition request ) w ere used significantly more in the normal children, while some (e.g., negative state, reason, exclamation ) w ere used significantly more in the language- disordered children. T hese differences seem to be related to the language- disordered children ' s limited use of communicative devices and confirmative expres sion s. T hese result s support the previou s assumption that the language- disordered children have language- age- appropriate semantic capacity to develop language. Most of the frequent 3- term relations were composed of the frequent 2- term relations which ranked within the 10th. T hese results support Slobin ' s suggestion (1973) that acquisition of a new languge structure is often coordinated with the existing language structures. Although some of the researcher s semantically analyze only simple sentences and grammatically analyze complex sentences of the young children ' s language, both the simple

15 Se ma ntic Ana lyzability of the Intra- a nd Inte r- s e nte nces and complex sentences w ere analyzed semantically in this study. Among the intersentence relations, the language- disordered group used significantly more juxtaposition and conjoining, while there was no difference found in u se of embedding. However, adjective embedding was used more frequently in the normal group, while substantive embedding was used more frequently in the language- disordered group. T he language- disordered children showed even more delay in embedding (especially in adjective embedding ) than the normal children at the same language level. T he result s indicate that it is more difficult for the languagedisordered children to acquire embedding inter sentence relation s. Based on the result s of this study and the present researcher ' s clinical experiences, the previous dichotic analy sis sy stem seems to miss a lot of information and to lead to an underestimated assessment. Further research could be conducted to develop and verify the semantic as ses sment and intervention of young children ' s complex sentences. REFERENCES (1997) , 2, (1998). 2-3 :,,., 3, (1995).. :. (1982). :. :. (1983).. :. Bloom, L. (1970). Lang uag e developm ent: Form and f unction of em erg ing grammars. Cambridge: The MIT Press. Bloom, L. (1973). One word at a tim e: The use of single- word utterances bef ore syntax. T he Hague: Mouton. Bowerman, M. (1980). Language development. In H. C. T riandis & A. Heron (Eds.). Handbook of Cross- Cultural Psychology (Vol. 4). Boston: Allyn and Bacon. Braine, M. (1963). The ontogeny of English phrase structure: The first phrase. Lang uag e, 39, Brown, R. (1973). A f irst lang uag e. Cambridge, MA : Harvard University Press. Clancy, P., Jacobson, T. & Silva, M. (1976). The acquisition of conjunction: A cross- linguistic

16 study. Child Lang uag e D evelop m ent, 12, Freedman, P. P. & Carpenter, R. L. (1976). Semantic relations used to normal and language- impaired children at Stage. Journal of Sp eech and H earing R esearch, 19, Hood & Bloom (1979). What, when and how about why : A longitudinal study of early expressions of causality. M onograp hs of the Society for R esearch in Child D evelopm ent, 6 (Serial No. 181). Kaiser, A. P. & Hester, P. P (1994). Generalization effects of enhanced milieu teaching. Journal of Sp eech and H earing R esearch, 37, Kim, Y. T. & Lombardino, L. (1991). T he efficacy of script contexts in language comprehension intervention with children who have mental retardation. Journal of Sp eech and H earing R esearch, 34, Klee, T. & Fitzerald, M. (1985). T he relation between grammatical development and mean length of utterance in morphemes. J ournal of Child Lang uag e, 12, Leonard, L. B. (1984). Semantic considerations in early language training, In K. Ruder & M. Smith (Eds.), D evelopm ental lang uag e intervention, Baltimore: University Park Press. McLean, J. E. & Snyder- McLean, L. K. (1978). A transactional app roach to early lang uag e training. Columbus: Charles E. Merill. Miller, P. J. (1981). A ssessing lang uag e p roduction in children. Baltimore: University Park Press. Morehead, D. M. & Ingram, D. (1973). T he developmental of base syntax in normal and linguistically deviant children. Journal of Sp eech and H earing R esearch, 16, Nelson, K. (1973). Structure and strategy in learning to talk. M onographs of the Society f or R esearch in Child D evelop m ent, 38. No Owens, R. E. (1999). Lang uag e disorders (3rd Ed.). Needham Heights, MA : Simon & Schuster. Retherford, K., Schwartz, B. & Chapman, R. (1977). T he changing relationship between semantic relation in mother and child speech. Paper presented at the Second Annual Boston University Conference on Language Acquisition. Retherford, K., Schwartz, B. & Chapman, R. (1981). Semantic roles in mother and child speech: Who tunes into whom? J ournal of Child Lang uag e, 8, Rondal, J., Ghiotto, M., Bredart, S. & Bachelet, J. (1987). Age- relation, reliability, and grammatical validity of measures of utterance length. Journal of Child Lang uag e, 14, Schlesinger, I. (1971). Production of utterances and language acquisition. In D. Slobin (Ed.). The ontog enesis of grammar. New York: Academic Press. Slobin, D. (1973). Cognitive prerequisites for the acquisition of grammar. In C. Ferguson & D. Slobin (Eds.). S tudies of child lang uag e developm ent. New York: Holt, Rinehart &

17 Se ma ntic Ana lyzability of the Intra- a nd Inte r- s e nte nces Winston. Stockman, I. J. & Vaughn- Cooke, F. (1986). Implications of semantic category research for the language assessment of nonstandard speakers. Top ics in Lang uag e D isorders, 6(4), T rauner, D., Wulfeck, B., T allal, P. & Hesselink, J. (1995). Neurologic and MRI profiles of language impaired children. T echnical Report CND- 9513, Center for Research in Language, University of California at San Diego. Warren, S. F., Gazdag, G. E., Bambara, L. M. & Jones, H. A. (1994). Changes in the generativity and use of semantic relationships concurrent with milieu language intervention. J ournal of Sp eech and H earing R esearch, 37, Wells, G. (1974). Learning to code experience through language. Journal of Child Lang uag e, 1, *2( ) 2-3, 70. (1),,, (2), (3).,., 2-3, ,,,. 2-4 * e- mail: youngtae@mm.ewha.ac.kr

18 .,,. 2-3.,.,,.