Linguistic characteristics of SLI in Afrikaans

Specific language impairment (SLI) has been defined as a significant impairment in the language ability of children in the absence of iden tifiable causal factors or obvious accompanying factors, such as neurological deficits , ognitive delay, hearing disabilities, and emotional or behavioral problems (Leonard 1998:vi; Stark & Tallal 1981). The characteristics of SLI as it presents itself in English and some ot her languages are comparatively wellknown. These characteristics include problems with grammatical morphology and with surface word order. Some morphemes appear to be dis proportionately difficult to master (Dromi, Leonard & Shteiman 1993), and, generally sp eaking, more verb-related than nounrelated errors are made by children with SLI (Hanss o & Nettelbladt 1995; Leonard 1989; Roberts & Rescorla 1995; Rom & Leonard 1990).


Introduction
Specific language impairment (SLI) has been defined as a significant impairment in the language ability of children in the absence of identifiable causal factors or obvious accompanying factors, such as neurological deficits, cognitive delay, hearing disabilities, and emotional or behavioral problems (Leonard 1998:vi;Stark & Tallal 1981).The characteristics of SLI as it presents itself in English and some other languages are comparatively wellknown.These characteristics include problems with grammatical morphology and with surface word order.Some morphemes appear to be disproportionately difficult to master (Dromi, Leonard & Shteiman 1993), and, generally speaking, more verb-related than nounrelated errors are made by children with SLI (Hansson & Nettelbladt 1995;Leonard 1989; 4-year-old Afrikaans-speaking children, eight girls and seven boys.They were 4 years 0 months to 4 years 7 months old (M = 4 years 2.3 months) and had an MLUw ranging from 3.91 to 5.00 (M = 4.56).According to their parents and classroom teachers, the participants in the control groups were typically developing in all respects: Their language, intellectual, and socioemotional development were seen as being age-appropriate, and there was no evidence of any visible neurological deficits.All 30 children exhibited hearing sensitivity within normal limits bilaterally during hearing screening and had no previous referral to, or treatment by, a speech-language therapist.

Experimental tasks
The aim of the experimental tasks was to establish whether or not Afrikaans-speaking children with SLI perform age-appropriately as regards their comprehension and production of various types of grammatical morphemes, specifically those relating to number, person, case and tense.In total, 15 experimental tasks were performed, each of them assessing either the comprehension or the production of the singular/plural distinction, pronouns, possessive se-construction, or tense.An overview of these tasks is presented in Table 1.In general, the tasks were of three kinds: (i) a comprehension task comprising picture selection, where the participant had to select the picture matching an utterance of the first author; (ii) an acceptability judgement task, where the participant had to indicate whether an utterance produced by the author was acceptable in Afrikaans or not; and (iii) a production task entailing sentence completion, where the participant had to complete a sentence initiated by the author.The procedures used in these tasks have previously been used with success to test the comprehension and production of grammatical morphemes by young children of different languages, by researchers such as Hansson & Leonard (2003); Jakubowicz (2003); Loeb & Leonard (1991); and Marchman, Saccuman & Wulfeck (2004).The tasks were all first performed with typically developing Afrikaans-speaking 3-, 4-, 5-, and 6-year-olds, during a pilot study, in order to ensure that test items were appropriate and that the demands placed on the participants were realistic (see Southwood 2005Southwood , 2006)).

Collection of spontaneous language
As was the case for the experimental data, all spontaneous data were collected by the first author.During language sample elicitation, this author and the participant mostly played alone in a quiet room at his/her school, care centre, or home, or in a quiet part of a room in doi: 10.5842/37-0-46 Linguistic characteristics of SLI in Afrikaans 107 which other people were also present.Three of the samples were collected with other children taking part in the conversation: One girl with SLI did not want to participate if her typically developing twin sister could not accompany her to all sessions, and two 4-year-old boys each insisted on having a friend present.
Language sample elicitation took the form of free play with toys that included (i) little figurines with accessories such as radios, hats, mugs, and brooms; (ii) wooden building blocks; and (iii) plastic kitchen furniture.The first author initiated the language sampling interaction by inviting the participant to join her in kitting out the dolls, building a house, and/or assembling the kitchen.If the participant was quiet for extended periods, the author used a variety of techniques to encourage conversation, including parallel play, making statements, and asking questions (both wh-and yes/no-questions).These questions were asked about topics previously found to be suitable for discussion with preschool children, such as their families, pets, and birthday celebrations (see Southwood & Russell 2004).Following Crystal, Fletcher & Garman (1976), the language samples collected in this study were each 30 minutes long.An audio-cassette recording was made of each language sample collection session, using an observable recorder.

Experimental tasks
All responses on the experimental production tasks were recorded on a score sheet.Selfcorrections were allowed; only the final response was scored.

Language sample
The utterances occurring in the first 30 minutes of each language sample were transcribed orthographically.Hereafter, the first 100 complete and fully intelligible utterances were identified.Following Hunt (1970:4), an utterance was considered to be a T-unit, i.e., "one main clause plus whatever subordinate clause and nonclausal expressions are attached to or embedded within it".Accordingly, want 'because', en toe 'and then', and en dan 'and then' were each taken to introduce a new T-unit, as were en 'and' and maar 'but' if these two were followed by a clause containing a verb.Hickey 1991).Also, following the caution stated by Miller & Deevy (2003: 1157-1158), care had to be taken not to create a confound: Morphemes were being examined (in both the experimental task and the language samples); therefore, employing MLU measured in morphemes seemed inappropriate.
Verbs taking the form of noun+verb compounds, such as fietsry 'cycle' (literally 'bicycle+ride'); adjective+verb compounds, such as mooimaak 'beautify' (literally 'pretty+make'); and preposition+verb compounds, such as opklim 'climb up' (literally 'on/up+climb'), were counted as one word, unless the verb part of the compound occurred before the noun, adjective, or preposition, as in Hy klim op 'He is climbing up'.
In Afrikaans, dit 'it' and wat 'which/that' change their form when combined with a preposition:  Correct and erroneous occurrences of grammatical morphemes were not tallied from utterance 101 onwards.However, each utterance which (i) occurred after the hundredth one but before the end of the 30 minutes, and (ii) was in any way deviant (i.e., non-adult-like) was identified and placed in a separate database.

Linguistic characteristics of SLI in Afrikaans revealed by the experimental tasks
In Table 1, an indication is given of whether or not there was a statistically significant difference between (i) the performance of the three groups; (ii) the mean scores of the SLI children and those of the typically developing 4-year-olds; (iii) the mean scores of the SLI children and those of their typically developing same-aged peers; (iv) the mean scores of the two typically developing groups; and (v) the degree of variance occurring in the three groups of participants.The statistical procedure used to ascertain whether or not differences between groups could be assumed was a one-way analysis of variance (ANOVA).Where ANOVA returned a significant outcome, post hoc comparisons were made using Tukey's HSD test to establish between which of the three groups (SLI, TD6, and TD4) the statistically significant differences in performance occurred.Levene's statistic of homogeneity of variance was used to determine whether the intragroup variability in performance differed significantly between the groups, i.e., whether the members of one group showed statistically significantly more doi: 10.5842/37-0-46 variability in their performance than another.Levels of significance were taken to be .05or less throughout.
As can be seen from Table 1, the children with SLI obtained lower scores than their typically developing peers on 14 of the 26 aspects measured by the experimental tasks.For all of these 14 aspects, the children with SLI performed on a par with the typically developing 4-yearolds.In addition, the general pattern was that, where the variability differed between the three groups, the SLI group showed the most intragroup variability.This variance was statistically significant for nine of the 26 aspects measured by the experimental tasks.Because there were no statistically significant differences between the three groups, post hoc analyses were not considered.
c NW=nonsense words d Although there was a statistically significant difference between the three groups, this difference was not strong enough to show up in post hoc testing.
In an attempt to establish whether there was one (or more) general factor(s) responsible for the differentiation among the three groups, factor analysis (principal component, varimax doi: 10.5842/37-0-46 rotation) was performed.Several solutions were considered, including solutions where the many scores obtained on the sentence completion task assessing production of past tense constructions (i.e., the scores on Experimental Task 15) were not taken into consideration.In all the different solutions, the first factor after (varimax) rotation turned out to be stable with consistent high loadings (>.50) for the following seven experimental tasks: (i) the picture selection task assessing comprehension of the singular/plural distinction (Task 1); (ii) the sentence completion task assessing production of regular plural forms of real words (Task 5); (iii) the sentence completion task assessing production of irregular plural forms of real words (Task 6); (iv) the sentence completion task assessing production of plural forms of nonsense words (Task 7); (v) the picture selection task assessing comprehension of pronouns (Task 8); (vi) the judgement task assessing comprehension of pronouns (Task 9); and (vii) the sentence completion task assessing production of pronouns (Task 10).
When considering the composite score on these seven tasks (by adding the z scores), the difference between that of the three groups can be portrayed as in Figure 1: The SLI and TD4 groups appeared to perform similarly and the TD6 group better than the other two groups.
Most variability seemed to occur in the SLI group, with some children in this group performing as well as their typically developing peers and others worse than the 4-year-olds.
However, the difference in intragroup variance of the three groups was not significant (Levene's test; F 2,42 =2.007; p=.147).
doi: 10.5842/37-0-46 Table 2 contains the details of the performance of the three groups on the composite index.A one-way ANOVA returned a significant outcome, which means that a difference between the mean scores of the groups could be assumed (F 2,42 =30.662; p=.000).Post hoc analyses (Tukey's HSD; alpha=.05)revealed that the statistically significant differences were between the SLI and TD4 groups, on the one hand, and the TD6 group, on the other.There was no statistically significant difference between the mean scores of the SLI and TD4 groups.Considering the performance of the individual participants on this composite index, two of the children with SLI obtained markedly lower scores than the rest of their group: One was a boy, participant SLI-6, whose composite score was -13.54; the other was a girl, SLI-5, whose score was -11.00.These scores were noticeably lower than the lowest one in the TD4 group --6.45 -which was obtained by a girl.Another two children with SLI obtained markedly higher scores than their group: Again, one was a boy, SLI-11, with a composite score of 4.64, and the other a girl, SLI-10, with a score of 3.83.These two scores were higher than the lowest four in the TD6 group, illustrating the high degree of variability found in the SLI group.

3.2
The linguistic characteristics of SLI in Afrikaans revealed by the errors in the first 100 utterances of the language samples As stated above, the first 100 complete and fully intelligible utterances in each sample were analysed for errors pertaining to correct and incorrect occurrences of (i) singular and plural forms of nouns, (ii) pronouns, (iii) se-constructions, and (iv) various types of past and present tense constructions.The results of some of these analyses are presented in this section.
However, the whole first 30 minutes of each language sample was examined for errors other than those mentioned above, for instance, for errors pertaining to word order or the inappropriate insertion or omission of a determiner.The results of this examination are given in the next section, together with an indication of the types of errors -specifically those related to word order -which did not occur.
The language samples were first examined for the correct occurrence and the substitution, incorrect insertion, and omission of those aspects assessed by the experimental tasks.Table 3 gives an overview of a selection of those measures which produced statistically significant differences between the groups, specifically (i) the proportion of plural forms which were produced correctly; (ii) the proportion of pronouns produced correctly; (iii) the proportion of present tense constructions produced correctly; and (iv) the number of past tense forms vs.
present tense forms.
In general, the SLI group fared worse than the TD6 group.However, in contrast to the pattern found for the experimental tasks, the SLI group was also, at times, outperformed by the TD4 group.The two typically developing groups fared similarly.Again, the most variability was found in the SLI group, with some children faring as well as the typically developing ones.
doi: 10.5842/37-0-46At times, there was no score for a particular child for a certain measure, simply because the child did not attempt the construction in question.Despite the challenge posed by low frequency of occurrence (or even absence) of some of the measures, it was possible to establish that there were positive correlations between the four measures given in Table 3 in a consistent way.Three out of the six correlations were significant (2-tailed), as can be seen in Table 4.The positive correlations between the four measures of the language sample analysis means that it makes sense to obtain a composite score by summing their z scores.The difference between the three groups in terms of their composite scores on these four measures is portrayed in Figure 2. Unlike the case for the composite score of the experimental tasks, the SLI group seemed to fare worse than both typically developing groups, with the latter two performing similarly.Again, most variability appeared to occur in the SLI group, with some children in this group performing better than the best-performing, and others worse than the worst-performing, typically developing ones.In this case, the difference in variance between the groups was significant (Levene's test; F 2,42 =9.311; p=.000).In Table 5, the details of the performance of the three groups on the composite index for the language sample analysis are given.A one-way ANOVA returned a significant outcome, indicating that a difference between the mean scores of the groups could be assumed (F 2,42 =4.268; p=.021).Post hoc analyses (Tukey's HSD; alpha=.05)revealed that the statistically significant differences were between the SLI group, on the one hand, and the two typically developing ones, on the other.Based on the outcome of a one-way ANOVA, no significant difference between the TD4 and TD6 groups could be assumed.This pattern differs from the one for the composite score of the experimental tasks: There, the children with SLI fared similarly to the 4-year-olds.

The linguistic characteristics of SLI in Afrikaans revealed by the full language samples -other errors
From the above, it appears that the Afrikaans-speaking children with SLI fared on a par with the younger typically developing ones on the experimental tasks, but worse than both groups of typically developing children in terms of correct spontaneous production of the grammatical morphemes related to number, person, case and tense.In this section, some other errors made in the language sample are discussed.The first set of errors is verb-related.A summary of these errors and their frequency of occurrence in the first 30 minutes of the language samples is given in Other verb-related errors (difficult to classify and/or idiosyncratic) a This figure indicates the number of times the error occurred in the 30 minutes of language sample.
b The figure in square brackets indicates how many children in that group made the relevant error.
doi: 10.5842/37-0-46 Errors on infinitives included the incorrect form of het 'have' and wees 'be'; the omission of the infinitival form of the main verb, as in example (1); 2 the omission of a part of the infinitival structure, as in (2), where the infinitival particle te of om te probeer swem 'to try swim' has been omitted; and the inappropriate insertion of a part of the infinitival structure, as in ( 3), where om should not have occurred.
( The intended meanings of ( 9) and ( 10) are not clear.For this reason, only a gloss, and not a target construction, is provided in each case.The second set of errors to be considered here is non-verb-related.These errors are summarised in Table 7.Some illustrative examples of non-verb-related errors are provided below.
As regards omitting the subject, a similar number of children in the SLI and TD4 groups made this error, but the errors in the SLI group were almost three times as many as those in the TD4 group.In the SLI group, one boy -participant SLI-6 -produced 12 of the 29 errors, two girls made three errors each, another five children made two errors each, and one girl omitted the subject once.
Five of the six children with SLI who omitted the object did so only once.The girl who made this error three times -participant SLI-14 -did not omit the subject once.
Prepositions were incorrectly omitted, inserted, and substituted with other prepositions by all three groups of participants, but less so by the TD4 group than by the other two.The two 6-year-old groups had almost the same number of children making this error, but, collectively, the 10 children in the SLI group made this error almost twice as often as did the eight children in the TD6 group.One boy and one girl -participants SLI-11 and SLI-14were responsible for seven and nine of the 36 errors, respectively.Two boys -SLI-1 and SLI-6 -made five errors each, and the rest of the six children with SLI made one or two errors each.
doi: 10.5842/37-0-46 Other non-verb-related error (difficult to classify and/or idiosyncratic) a Due to the nature of the conversation -freeplay with frequent comments on the objects present and the actions being performed with them -children from all three groups at times made use of elliptical utterances, particularly ones from which the subject was omitted.An example would be where a child says Gaan nou hierdie een vat 'Going to take this one now' while he reaches for another wooden block.These subjectless utterances were not included here, not even those of the one boy with SLIparticipant SLI-9 -who had a very strong preference for such subjectless utterances over ones containing a subject.is ons toe toe is ons toe be we closed then be we closed 'Then we were closed' In contrast to the difficult to classify and/or idiosyncratic verb-related errors which were made by only the children with SLI, the non-verb-related errors were made by all three groups of participants.However, the SLI group made far more of them than did the two groups of typically developing children.A total of nine such errors were made collectively by eight children in the TD4 group, whereas nine children with SLI made 33 in total.Of these, 11 were made by one boy, participant SLI-6.A girl and boy -SLI-9 and SLI-12 -made five errors each, another boy -SLI-1 -made three, and the other five children made one or two such errors each.As was noted for the verb-related errors, this category of errors is highly diverse by nature.The examples in ( 18) to ( 21) serve to illustrate the types of errors which were taken to be highly idiosyncratic and/or otherwise difficult to classify.Nie

Word order errors occurring in the language samples
The discussion now turns to the word order errors which occurred in the language samples of the 45 participants.All three groups of participants made word order errors, but not all types of errors were made by all groups.
Only the typically developing 6-year-olds produced utterances in which the subordinate conjunction omdat 'because' was treated as a co-ordinate conjunction, similar to its synonym want 'because'.Three such utterances occurred, illustrated by example ( 22).However, no coordinate conjunctions were followed by a subordinate word order.Main clauses with a Subject-Object-Verb surface word order (the order found in embedded clauses) also occurred -as shown in example ( 25) -but only in the language of two children with SLI.Only one instance of VSO occurred, in the language of participant SLI-6; this utterance is given in ( 26).toe ons daar kom ek het nie eers geskrik toe ons daar kom het ek nie eers geskrik vir hulle vir hulle nie nie when we there come I have not even when we there come have I not even get-a-frightget-a-fright-PAST PART for them not PAST PART for them not 'When we came there, I was not even frightened by them' The children with SLI and the 4-year-olds also made errors in the word order of wh-questions.
Examples are given in ( 32) and ( 33).The wh-element was fronted, but subject-verb inversion did not take place.Utterances with a SwhV or VwhS word order did not occur in the data.
One utterance, from the language sample of a girl with SLI, contained a wh-question in which the subject and verb had the correct surface word order, but in which the adverb occurred in the incorrect position.This utterance is given in (34).
( ons babatjies ons by hier kan kies ons babatjies kan ons by hierdie kies our baby-DIM-PL we by here can choose our baby-DIM-PL can we by there choose 'Our babies we can choose to match these' [= we can choose figurines -ones which match these pieces of toy furniture -to be our babies] The 4-year-olds and the children with SLI made word order errors in utterances containing particle-verbs, i.e., verbs consisting of a verbal stem and a particle belonging to the category noun, preposition, or adverb.Examples ( 37) and ( 38) contain such utterances.
(37) Target: ek sal ry fiets ek sal fietsry I will ride bicycle I will bicycle-ride 'I will cycle' doi: 10.5842/37-0-46 (38) Target: daar val af die een daar val die een af there fall off the one there fall the one off 'There the one falls off' It appears then that a range of word order errors were produced, but that not all three groups produced all types of errors.Table 9 contains a summary of the types of word order errors and the group(s) which made them.As can be seen from this table, a word order error which was unique to the SLI group was that of main clauses with a surface Subject-Object-Verb or Verb-Subject-Object word order.

Summary of results
In terms of the comprehension and elicited production of grammatical morphemes related to number, person, case and tense, the Afrikaans-speaking 6-year-olds with SLI fared on a par with the younger typically developing ones, but worse than both groups of typically developing children in terms of correct spontaneous production of such morphemes. 4  Furthermore, as a group, the children with SLI also made more verb-related and non-verbdoi: 10.5842/37-0-46 related errors during spontaneous production of utterances than did either of the typically developing groups.The hypothesis that the children with SLI will experience more problems with the accurate comprehension and production of grammatical morphemes than the two typically developing groups was therefore borne out in part by the elicited data and in full by the spontaneous data.
It was also hypothesised that, unlike those of the two typically developing groups, some of the utterances of Afrikaans-speaking children with SLI will demonstrate an incorrect surface word order.The spontaneous data showed that the SLI groups did indeed produce utterances with ungrammatical word order, but so did the other two groups of participants.However, the number of different types of word order errors produced by the SLI group was larger than that produced by the two typically developing groups.
The third hypothesis was that verb-related and noun-related grammatical morphemes will pose comparable problems for the children with SLI.On the experimental tasks, the children with SLI were indeed outperformed by their typically developing peers in terms of plurals (i.e., noun-related morphemes) and tense (i.e., verb-related morphemes).The spontaneous data also showed that the children with SLI have problems of a comparable degree with the production of both types of morphemes.Although no statistical comparison was made between the two types of morphemes, it appears that Afrikaans verb-related and noun-related morphology both pose sizable problems for children with SLI.

Do the errors reveal a possible clinical marker of SLI in Afrikaans?
In order to compare the performance of the three groups of participants across experimental tasks, a different approach was taken, namely that of discriminant analysis.The aim was to ascertain which combination of experimental tasks would result in the most accurate classification of the 45 participants into their three groups (SLI, TD4, or TD6).All experimental tasks were included, using the stepwise procedure to include and exclude the task results in the discriminant analysis (probability F entry .05;removal .10).The combination of the following three tasks was found to be the most successful in correctly placing participants into their respective groups (SLI, TD4, or TD6): doi: 10.5842/37-0-46 (iii) the number of errors after prompting by the researcher, on the sentence completion task assessing production of past tense forms -excluding highly idiosyncratic errors as well as past tense constructions containing het but in which the past participial (ge-) form was replaced by an infinitival one (Task 15j).
Table 10 shows the results in terms of a classification table.This table makes a distinction between the actual group membership and the predicted group membership.This analysis classified almost all TD6 group members as belonging to that group.
Interestingly, the participant misclassified as a typically developing 4-year-old had the second highest MLU of all participants.Problems occurred in differentiating between the members of the TD4 and SLI groups, as could be expected considering the separate analyses of the experimental tasks.Nevertheless, the majority of the TD4 group was classified as such, with only four members of this group misclassified (all four of them as children with SLI).The general pattern observed for the experimental tasks was that the SLI and TD4 groups obtained similar average scores but that the range of scores in the SLI group was larger than that in the TD4 group.This is reflected in the difficulty that this discriminant analysis had with the correct classification of the members of the SLI group: Only seven were classified correctly, five were seen to be typically developing 4-year-olds and another three to be typically developing 6-year-olds.This latter misclassification (i.e. the classification of SLI children as doi: 10.5842/37-0-46 typically developing children) is interesting from a clinical point of view, because these children with SLI are classified as having no language problems.The SLI children classified as belonging to the TD6 group are participants SLI-10, SLI-11, and SLI-12; the composite score of all three of them had a positive value.
Discriminant analysis was also performed with the various measures of the spontaneous language samples -but only with measures related to the grammatical features number, person, case and tense.As in the case of the experimental tasks, the aim was to ascertain what combination of measures would result in the most accurate classification of the 45 participants.Table 11 shows the results of the stepwise discriminant analysis.A combination of the following two measures was selected as the most successful in placing the participants correctly into one of the three groups (SLI, TD4, or TD6): (i) the proportion of correct present tense constructions out of all present tense constructions; and (ii) the number of past tense forms vs. present tense forms.This analysis was slightly less successful at correctly classifying participants in terms of their actual group membership than was the similar analysis involving the experimental tasks.
Based on the selected two measures of the language sample, two thirds (30) of the 45 participants were classified correctly compared to 32 by the similar analysis involving the experimental tasks.
The group most often misclassified was the SLI one: Seven of its members were deemed to be typically developing.This again confirms that most variability occurred in the SLI group: doi: 10.5842/37-0-46Some of its members performed as well as typically developing 6-year-olds.The two participants with SLI who were classified as belonging to the TD6 group were not the same ones as those classified as such by the experimental tasks.This underscores the importance of using a combination of spontaneous and elicited data when diagnosing SLI in Afrikaansspeaking children.
One of the two participants with SLI classified as a TD6 group member by the two measures of the language sample was participant SLI-5, who obtained a composite score of 2.15.Recall that she was one of the two children with SLI who fared worst on the composite score pertaining to the experimental tasks, but that she was one of the two children who made no errors in her language sample except those related to the grammatical features number, person, case and tense.The other child was also a girl -SLI-2 -one who made almost no errors in her spontaneous language sample (the other girl who does not feature at all in Table 8).Her MLU was one of the lower ones (it fell within the bottom third of her group).This could lead one to think that she made use of short utterances in an attempt to avoid problem structures and, by doing so, increased the accuracy of her utterances.As mentioned by Blake, Myszczyszyn & Jokel (2004:31), the fact that children with SLI sometimes differ from controls in terms of correct morphology when comparisons are made based on elicited production but not when based on spontaneous production, could simply be due to avoidance -in their spontaneous language use -of unfamiliar forms.This could be the case for participant SLI-2.However, none of the four children with MLUs lower than hers appeared to use these strategies.The other child who fared poorly on the composite score pertaining to the experimental tasks -participant SLI-6 -also fared worst on the composite pertaining to measures of the spontaneous language sample: He obtained a score of -6.56.Two other children also fared poorly: SLI-7, with a score of -6.13, and SLI-11, with a score of -5.08.
Because the average scores of the TD4 and TD6 groups on measures of the language sample analysis did not differ significantly, it is understandable that some of these groups' members were classified as belonging to the other group.What is of interest is that one typically developing 4-year-old was classified as language-impaired.The MLU of this participant was also the second lowest of all TD4 participants.However, based on the selection of seven experimental tasks, her score was average compared to that of the rest of the TD4 group.
doi: 10.5842/37-0-46 Interestingly, the 6-year-old who had the lowest composite score on the two measures of the language sample, had the second highest MLU of all participants.So, although she made more errors than the rest of her group, she also produced longer utterances than most of her group.
From the discriminant analysis and language sample analysis, it appears that a combination of experimental and spontaneous data differentiates successfully between children with and without SLI, to a great extent.Considering only spontaneous production might lead to underdiagnosis, because it is, at least in theory, possible for children with SLI to avoid certain structures in their spontaneous language use.Elicited production should therefore also be used when diagnosing an Afrikaans-speaking child as SLI.As stated by Blake et al. (2004:38), differences between spontaneous and elicited production tasks make it unlikely that a morphological measure based on spontaneous speech alone will be useful in diagnosing SLI (see also Bedore & Leonard 1998).Whereas elicited production tasks pose their own special difficulties for children with SLI, Blake et al. (2004:39) state that they may also be better at detecting subtle deficits in older children with SLI.
Three of the five measures discussed in this section are related to the production of verbs.It appears then that one should consider the elicited production of past tense forms and the spontaneous production of present and past tense forms in the search for a clinical marker of SLI in Afrikaans.According to Rice, Wexler & Herschberger (1998:1412), such a marker is "a linguistic form, or principle that can be shown to be characteristic of children with specific language impairment".Rice & Wexler (1996) identified finiteness, or tense marking, as a sensitive and specific clinical marker of SLI in English (see also Marchman, Wulfeck & Weismer 1999).
The number of highly idiosyncratic and/or difficult to classify verb-related errors in spontaneous production appears to differentiate very accurately between Afrikaans-speaking children with and without SLI.However, including "difficult to classify / idiosyncratic errors" as part of a clinical marker could be problematic in practice: This category is one of exclusion rather than inclusion -in order to ascertain whether a child made such an error, one would first have to establish what is meant by "classifiable errors" before one will be able to deem any error "difficult to classify".The fact that difficult to classify errors are included when doi: 10.5842/37-0-46 considering a clinical marker of SLI in Afrikaans is not a problem per se -Blake et al. ( 2004) also found such errors, which they called "odd", to be characteristic of the language of their English-speaking participants with SLI.Rather, the practicalities of classifying errors as "difficult to classify" are the problem.Bortolini, Caselli, Deevy & Leonard (2002:90-91) state that the notion 'clinical marker' can be interpreted in two ways.The first is that the marker represents a clear symptom of SLI and also a particular cause for this symptom.The second, a weaker interpretation, is that the clinical marker is representative of the symptom without assuming that the symptom reflects a single cause.Conti-Ramsden & Hesketh (2003:252) argue for a third interpretation, namely that a clinical marker (or risk marker) represents a symptom, but that no assumption is made about whether the marker reflects a single cause or that this symptom alone identifies the disorder."On the contrary, it is assumed that the risk marker is more likely to be used in combination, to complement information available" (Conti-Ramsden & Hesketh 2003:252).
Based on the obtained Afrikaans data, it is recommended that 'clinical marker' should here be given the interpretation of Conti-Ramsden & Hesketh: It may be a useful risk marker when used together with other information that a clinician has on the child, but it does not necessarily reflect a particular cause for the symptom(s) which they represent.

Conclusion
In order for one to provide a comprehensive theoretical account of SLI as it presents itself in In general, the Afrikaans-speaking children with SLI fared on a par with typically developing 4-year-olds and worse than typically developing 6-year-olds on experimental tasks assessing the comprehension and production of grammatical morphemes related to the features number, doi: 10.5842/37-0-46 person, case and tense.In terms of spontaneous production of morphemes related to these grammatical features, the two typically developing groups fared similarly, with the children with SLI being outperformed by both.A similar pattern was observed for other errors found in the spontaneous language samples.These results indicate that SLI in Afrikaans indeed entails problems with grammatical morphology.However, in contrast to the general trend that children with SLI find verb-related grammatical morphology more problematic than nounrelated morphology, the Afrikaans-speaking children with SLI experienced problems of comparable size with noun-related and verb-related morphemes.In terms of word order errors, it was shown that, although the children with SLI and the two groups of typically developing children made such errors, some types of word order errors were only made by the children with SLI.This indicates that SLI in Afrikaans entails problems with word order, as has been shown to be the case for most of the other languages in which SLI has been studied.
Discriminant analysis and language sample analysis revealed that a combination of five See Table 8 for some of the types of errors made by this girl and the other participants with SLI.

2.
Example material is presented in the following format throughout: 3.This utterance would not have been ungrammatical had the child meant "It's fish we don't have (but all the other animals are here)".However, this utterance was produced in response to a question by the researcher: En het julle visse op die plaas, kinders?
'And do you have fish on the farm, children?'.There was no indication that the child meant to say anything other than "No, we do not have fish".

4.
Even though Rice, Wexler & Redmond (1999) found that children as young as 3 are able to perform acceptability judgements of the type employed in this study, it is important to note that metalinguistic skills are often thought to be not yet developed by the age of 4 years (see, amongst others, Owens 2001:393;Nelson 1998:361).The spontaneous language production did not require metalinguistic skills, whereas the experimental tasks did.It could therefore be that the 4-year-olds in this study were merely too young to show an advantage over the 6-year-olds with SLI in terms of (meta)linguistic knowledge.
these first 100 complete and fully intelligible utterances, the number of occurrences of the following was tallied separately (examples taken from the corpus): (i) each of the various kinds of plural, present tense and past tense constructions produced correctly and produced incorrectly, (ii) use of historic present tense (e.g., Toe sien ek hulle -literally 'Then I see them', a typical Afrikaans rendering of 'Then I saw them'); (iii) passive constructions in the past tense form (e.g., Dit was deur 'n hond gekrap 'It had been scratched by a dog); (iv) each correct and each incorrect occurrence of personal and possessive pronouns, where incorrect occurrences included omissions; and (v) each correct and incorrect occurrence of a se-construction (as in pa se hoed or *ystervarks maag).

Figure 1 .
Figure 1.Box plot of performance per group -Composite score of seven experimental tasks

Figure 2 .
Figure 2. Box plot of performance per group -Composite score of four measures of the language sample analysis luister tog wat het hy gesticker vat he want-to just listen just what have he sticker-PAST PART take Half of the 18 idiosyncratic and/or difficult to classify errors were made by one boy, participant SLI-6.Two other boys made two and three errors each.For the remainder of the SLI participants who made such an error, each made it only once in their 30-minute language sample.doi: 10.5842/37-0-46 was by 'n ou plaas laas jaar was ek op 'n ou plaas last year I be-PAST by a old farm last year be-PAST I on a old farm 'Last year I was on an old farm'(31) Target: the judgement task involving what should have been regular plural forms of real words (Task 2); and Afrikaans (an endeavor which falls outside the scope of this article, but seeSouthwood 2007 in this regard), one needs to know what such an account has to account for.The study discussed in this paper aimed to establish exactly that, by ascertaining what the characteristics of SLI in Afrikaans entail.The general research question was whether SLI in Afrikaans entails problems with word order and with grammatical morphology, as has been shown to be the case for many other languages.

(
elicited and spontaneous) measures would probably differentiate successfully between Afrikaans-speaking children with and without SLI.A composite of these measures could possibly act as a clinical marker, although further investigation is required in this regard.The composite could possibly be simplified, maybe to include only the tense-related measures, in line with the clinical marker proposed for SLI in English.* This material is based on work financially supported by The National Research Foundation.Any opinion, findings, conclusions or recommendations expressed in this material are those of the authors and therefore the NRF does not accept any liability in regard thereto.Notes 1.

(
Transcript number) Transcript of utterance Target utterance in Afrikaans Literal English translation of actual utterance Literal English translation of target Grammatical/Idiomatic English version of utterance doi: 10.5842/37-0-46 Linguistic characteristics of SLI in Afrikaans 139

Table 1 .
Overview of the difference in results between the three groups of participants on the 15 experimental tasks

Table 2 .
Summary of performance per group -Composite index consisting of a selection of seven experimental tasks (Tasks 1, 5-10)

Table 3 .
Overview of the measures of the language sample analysis which produced statistically significant differences

Table 5 .
Summary of performance per group -Composite index consisting of four measures of the language sample analysis

Table 6 .
Table 6, with illustrative examples following the table.Frequency of verb-related errors in the language samples of the three groups of participants Errors on infinitives were made by all three groups of children.However, the number of errors made by the SLI group was more than double that made by the TD4 group, although only two more children in the SLI than in the TD4 group made errors of this nature.Of the 10 children in the TD4 group who made this type of error, six made it only once and the remaining four made it twice each.By contrast, of the 12 children with SLI who made this type of error, only four made it only once; the others made it two to three times each, with two children -participants SLI-6 and SLI-14 -each making five such errors.In example (4), the verb gaan 'go' is omitted, and, in (5), reën 'rain' Only the children with SLI made verb-related errors which were highly idiosyncratic and/or difficult to classify.By nature, this category of errors is a particularly diverse one.Examples (8) to (10) serve to illustrate this diversity.In (8), the target construction could be either a passive one -ek word deur 'n volstruis daar op my hand gepik 'I am pecked there by an ostrich on my hand' -or an active one -'n volstruis het my daar op my hand gepik 'an ostrich pecked me there on my hand'.
I have everything do-PAST PART infinitive-I have everything do-PAST PART infinitivecomplementiser try swim complementiser to try swim 'I did everything to try and swim' (3) Target: mens hoef nie om te betaal nie mens hoef nie te betaal nie one have-to not infinitive-complementiser one have-to not to pay not to pay not 'One does not have to pay' The second type of error concerned the inappropriate insertion (or, from the view point of copy theory, the inappropriate repetition) or the inappropriate omission of a main verb in doi: 10.5842/37-0-46 finite constructions.'Now they are getting wet in the rain' As was the case for errors on infinitives, a comparable number of children in the SLI and TD4 groups inserted or omitted a main verb, but the number of errors made by the SLI group was more than double that made by the TD4 group.The same pattern emerged as before: Of the five children in the TD4 group, four made the error once only.Three of the six children with SLI made the error only once and one made it twice.However, one boy -participant SLI-6made it four times and one girl -SLI-12 -seven times.Het 'have' as a main verb was also omitted and inserted inappropriately, but only by the SLI group and only twice: once each by two boys.The utterance in (6) serves as an example of the inappropriate omission of het as a group.Of the six children from this group who made this error, most made it only once, but one child each made the error twice, three times, and four times.doi: 10.5842/37-0-46 (8) ek word 'n volstruis het daar op my hand gepik I be-PASS-PRESENT a ostrich have there on my hand peck-PAST PART /PASS PART

Table 7 .
Frequency of non-verb-related errors in the language samples of the three groups As in the case of prepositions, determiners were omitted and inserted inappropriately by all three groups of participants, but more children in the SLI than in the other two groups made More children with SLI than typically developing ones made this type of error, although six of the 10 children with SLI who did make this error did so only once.A boy and girl -participants SLI-6 and SLI-14 -made three b This figure indicates the number of times the error occurred in the 30 minute language sample.cThefigure in square brackets indicates how many children in that group made the relevant error.An example of the substitution of one preposition with another is Ek slaan hom *deur die kop 'I hit him through the head' instead of Ek slaan hom oor die kop 'I hit him over the head'.An where we to vacation hold-PAST PART have where we vacation hold-PAST PART have 'where we went on vacation' doi: 10.5842/37-0-46 'Because he plays with other dogs'Although the children who omitted or inserted the negation element nie inappropriately mostly did so only once, far more children with SLI than typically developing ones made this error.An example of the omission of nie is given is (13), 3 and one of inappropriate insertion is given in(14).What is termed "other omission" in Table7entailed the omission of lexical items or phrases in which no clear pattern could be detected; diverse errors involving omission were grouped doi: 10.5842/37-0-46 together under this rather uninformative label.
Table8contains a summary of the errors made by the SLI group, other than those related to the grammatical features person, number, case or tense.In total, 13 of the 15 members of this group made the types of errors found in Table8.The two girls who did not -participants SLI- doi: 10.5842/37-0-46 doi: 10.5842/37-0-46

Table 8 .
Summary of the frequency of errors (excluding those related to the grammatical features number, person, case and tense) made by the SLI children en hulle meet om hulle op die lorrie te gaan en hulle meet hulle om op die lorrie te gaan and they measure infinitive-complementiser and they measure them infinitive-complementiser

Table 9 .
Summary of word order errors made per group

Table 10 .
Results of classification of participants into three groups based on a selection of

Table 11 .
Results of classification of participants into three groups based on a selection of two measures from the language samples