Conversational characteristics of children with semantic-pragmatic disorder.
2: What features lead to a judgement of inappropriacy?

D. V. M. Bishop Department of Psychology, University of Manchester

and C. Adams Centre of Audiology, Education of the Deaf and Speech Pathology, University of Manchester

Key words: children, language disorder, semantics/pragmatics, discourse analysis.

In the field of specific language impairment, a distinction has been drawn between disorders where language is characterised by simplified comprehension and expression of form and meaning, and those where language appears not so much oversimplified as inappropriate. It has been proposed that in the former kind of disorder there may be difficulty in expressing and understanding complex forms and meanings, but language, though limited, is appropriate to the context. In the latter kind of disorder, children may have good mastery of the formal aspects of language, but conversation does not flow smoothly, because they express themselves in an odd way, come out with unexpected utterances and seem unaware of the needs of the conversational partner.

These generalisations are based on descriptive accounts and raise many questions that can only be answered by experimental studies. Is inappropriacy something that can be reliably identified? If so, is it invariably a feature of language disorder or is it found in the course of normal language development? Does inappropriacy characterise one subset of language-impaired children, or is it observed to a greater or lesser extent in all? Can we identify characteristics of utterances that lead people to judge them as inappropriate?

The study of these questions reported here adopts a complementary approach to the accompanying paper (see p. 211). Rather than looking for occurrences of specific categories of conversational behaviour that were defined in advance, this study started from the opposite end by first identifying conversational utterances that were judged to be inappropriate, and then examining how far different raters agreed about such judgements, and what characterised those utterances that were deemed inappropriate. No attempt was made to specify a tight definition of inappropriacy in advance, as the interest was in discovering which aspects of children's conversation led an observer to judge the child's utterances as inappropriate. The subjects, sampling procedures and conversational material used in this study were identical to those in the companion paper.

After preliminary pilot work, 44 conversations were taken from our body of work, selected to include control children (n=20) covering the whole age range (4-12 years) and language-impaired children (n=24) with a variety of types of language disorder. Each transcript was independently analysed by each of the authors, and by a speech therapist who was given 3 hours' training with examples of utterances (from other transcripts) that were agreed by both authors to be inappropriate. The rater was instructed that utterances were not to be regarded as inappropriate just because they were simplified or not grammatically well formed; rather there should be a sense of oddness and disruption of the normal conversational flow. Further, judgements of inappropriacy were to be made on the basis of first impressions of an utterance. Quite often the children would say something which gave an initial impression of oddity, but which the listener could make sense of with hindsight; such cases were to be coded as inappropriate. This third rater had no background information about the age or language status of children. For each transcript, all three raters marked those child utterances that seemed conversationally inappropriate.

For each conversation, the percentage of child utterances identified as inappropriate was computed. Pairwise correlations between the three raters for 44 conversations were 0.874, 0.698 and 0.688 (d.f. =42).

Agreement between raters, whilst not perfect, was high enough to indicate that there was considerable overlap in ratings of inappropriacy made by independent individuals. It therefore seemed worth while to proceed to a full analysis of all the conversations gathered in the study, with the following aims:

All utterances from the first phase of our study were rated as appropriate or inappropriate by one of the authors. In addition, the conversations that had been used by Adams and Bishop (see p.216) to evaluate test-retest reliability were analysed for inappropriacy independently by both authors giving an estimate of inter-rater reliability with a new sample. Test-retest reliability was also computed for those children who had participated in two conversations on different topics, and longer-term stability for those who had taken part in one conversation with us and another with a teacher or therapist 9 months later.

Untransformed inappropriacy data are shown plotted in Figure 1. For statistical analyses, percentages were transformed logarithmically to stabilise variances. Within the control group, there was a significant correlation between age and log percentage of inappropriate utterances (r [65] = -0.785, P <0.01). The age trend was still apparent when analysis was restricted to children between the ages of 8 and 12 years (r [29] =-0.39, P <0.05). The substantial. differences between language-impaired and control children of the same age are evident by inspection, although many language-impaired children obtained scores within normal limits. Inappropriacy scores clearly relate to subtype of language impairment although there are some inconsistencies: a few individuals classified as 'semantic-pragmatic' do not obtain particularly high inappropriacy scores, whilst one child with an extremely high inappropriacy score was not classified as 'semantic-pragmatic'.

For statistical comparisons, control children were pooled into two broad groups: a control group (C) of the same age range as the language-impaired children (8-12 years) and a younger group (Y) of 4 to 6 year olds. Language-impaired children were divided into two subgroups: 14 classified as 'semantic-pragmatic disorder' and 43 'other language impaired'. The

logarithm of the percentage of inappropriate utterances was compared for these four groups using an analysis of variance, giving a significant main effect (F= 39.2; d.f. = 3, l20; P <0.001). Scheffé tests at the 0.05 level of significance were used for specific comparisons between groups. Language-impaired children in the 'semantic-pragmatic' group produced a significantly higher rate of inappropriate utterances than other language-impaired children and age-matched controls, although their mean score did not differ from that of younger control children. Other language-impaired children produced a higher rate of inappropriate utterances than age-matched controls.

Analysis of conversations from phase 2 (reliability study) gave the following reliability figures for percentage of inappropriate utterances: inter-rater reliability (20 conversations independently rated by two observers): r=- 0.789; test-retest reliability (two conversations on different topics and different occasions by each or 20 children): r=0.888; test-retest stability over a 9-month interval (two conversations for each of 14 children): r=-0.66l. All these correlations are significant at the 0.01 level.

Although inappropriacy is a subjective concept lacking an operational definition, acceptable levels of inter-rater reliability were obtained for this measure. Furthermore, the high levels of test-retest correlation that were obtained not just over a 2-week period but also over 9 months, indicated that this is a stable property of the individual. There is evidence that the measure of inappropriacy is valid as well as reliable. The percentage of inappropriate utterance showed significant developmental trends in the normal population not only between 4 and 12 years, but also over the range from 8 to 12 years, indicating that it is sensitive to conversational skills that are still being mastered at the primary school age. A major motivation for devising the conversational procedure hod been to devise an objective means of quantifying the types of conversational behaviours that gave a sense of bizarreness and abnormality to the language of some language-impaired children. and this index performed well, with children classified as having semantic-pragmatic disorder typically obtaining exceptionally high scores, even relative to other language-impaired children.

Although there was a very strong relationship between inappropriacy ratings and the classification of children as semantic-pragmatic disordered, this was not perfect. Figure 1 shows one child who was not regarded as having a semantic-pragmatic disorder despite very high levels of inappropriacy, and others who were placed in this group but who looked relatively normal on the inappropriacy measure. The first of these children did have phonological problems that occasionally interfered with intelligibility, and it seems possible that this was the reason he was not identified as a case of semantic-pragmatic disorder, since fluent, articulate speech is usually regarded as a characteristic of this disorder. In all other respects this child appeared to fit the criteria for semantic-pragmatic disorder: he was identified by Adams and Bishop (p. 233) as showing an abnormally high rate of initiations, and a high frequency of interrupts that violated turntaking rules.

Both groups of language-impaired children obtained higher inappropriacy scores than age-matched controls, but they did not differ significantly from younger normal children. This raises the question of whether their inappropriate conversational behaviour should be regarded as a linguistic immaturity. This question is considered further in a final analysis.

The next aim was to explore the nature of those utterances judged to be inappropriate in more detail. The first step was to develop by degrees a classificatory scheme for inappropriate utterances on the basis of discussion between the two authors for all the inappropriate utterances identified in phase 1 of the study. Once a framework had been finalised, all inappropriate utterances from phase 1 of the study were reclassified according to this scheme.

Inappropriate utterances identified in phase 2 were then used to check inter-rater and test-retest reliability. (Table 1). Table 1 shows the categories of inappropriacy that were identified, examples of which will now be given.

This category was used for utterances where the sense of inappropriacy seemed to arise because of unusual syntax and/or semantics. In many cases, the child appeared to have selected a wrong lexical item of the correct form class, closely related to, but not precisely corresponding to, the meaning he or she wished to convey, so that the utterance seemed oddly expressed. Semantic overextensions of this kind were observed for a range of parts of speech, especially function words (connectives, auxiliaries and prepositions). When judging if this category applied, it should be seen whether the utterance could be made more appropriate by substituting a related word or similar construction. More than one type of problem could be identified in a single utterance.

______________________________________________________________________________
Expressive semantics/syntax
Failure to comprehend literal meaning
Pragmatics 1: violation of exchange structure
  Nil response
  ignores initiation
Pragmatics II: failure to use context in comprehension 
Pragmatics III: too little information
  Inappropriate presupposition ('pseudo-ellipsis)
  Unestablished referent
  Logical step omitted 
Pragmatics IV: too much information
  Unnecessary assertion/denial
  Excessive elaboration
  Unnecessary reiteration
  Ellipsis/reference not used
Unusually or socially inappropriate content or style
  Topic drift
  Unmarked topic shift
  Stereotyped/formulaic
  Inappropriate questioning
  Socially inappropriate 
Other
  Lack or knowledge/experience 
Unclassified
_______________________________________________________________________________

Examples of connective errors are:

  1 C: we went on a bus because Lee was sick out of the window

Had the child in (1) used 'and' instead of 'because', the utterance would sound normal.

Other examples of this category of inappropriacy are:

tense

  2 A: why did you have to go the doctor?/
  3 C: I used to have a headache/

verb

  4 A: what were you doing?/
  5 C: er-going in paddling pool/
  6 A: uhuh/
  7 C: and playing out and we are finding to see who won the
       race/

  8 A: what should you do if you see thick smoke coming from the
       window of your neighbour's house?
  9 C: get some water and rinse it/

preposition

  10 A: have you ever had a birthday party?/
  11 C: yeah/
  12 A: can you tell me about it?/
  13 C: cos my birthday's in November/
  14 A: uhuh/
  15 C: cos mine's really on night/

pronoun

  16 A: and what do you think the man will do next? - /
  17 C: they can't go everywhere cos cos cos they need help/

(he means 'anywhere')

adverbial

  18 C: but the driver of the red car only had to bleed here ->
        but he was still all right/

  19 A: so what did your dad do?/
  20 C: well he droved it straight to park and then all of a
        sudden we ate some picnic/

  21 A: and why did you have to go to the doctor?/
  22 C: one day perhaps my foot has got a verucca/

Another category of interest was discourse devices. This term refers to a group of adjuncts and disjuncts which are normally used to establish and link discourse topics, and includes words and phrases such as 'well', 'by the way', 'of course', 'in fact', 'actually', 'anyway'. In normal development, use of these terms is acquired late, so their inclusion in conversation gives an impression of linguistic sophistication. Inappropriacy arose when a child incorporated these terms very freely in conversation, but in inappropriate contexts, e.g.

  23 A: how did you get to your holiday in Spain?/
  24 C: by car/
  25 A: by car/
        who drove you?/
  26 C: and aeroplane/
        by the way, aeroplane/

  27 A: so you usually go to France by boat/
        do you take your car on the boat?/
  28 C: yes/
        of course we went from Dover/
  29 A: it's not of course/
        you don't have to go from Dover/

It is a moot point whether such examples of misuse of discourse devices are best regarded as semantic problems or overextension, with a failure to appreciate the very specific meaning conveyed by a particular word or phrase, or whether they should be more properly regarded as use of stereotyped language, consisting of learned 'formulaic' strings of words which are incorporated in utterances (see below under 'Unusual or socially inappropriate content or style').

The examples considered so far involve instances where the sense of inappropriacy could be reduced by substituting a related word or phase. Another type of error, also described by Fletcher (1986), involves aberrant syntactic formulation. An example from this study is:

  30 A: do you ever go on holiday in your car/
  31 C: oh once I did but not in the aeroplane/

These high level formulation errors appear different from false starts and from what Fletcher, Garman, Johnson, Schelleter & Stodel (1986) have termed 'mazes', i.e. non-fluent lexical and structural repetitions and revisions. A syntactic formulation problem is evidenced by the child having problems, as above with getting negation or a postmodifying clause in the right place or running into difficulties with repeated terms, but the utterance is fluent, e.g.

  32 C: every time we need a car we don't need a car we get a van/

This category was coded when the child gave a response that was not appropriate to the question asked by the adult, but to a related question. This was a common reason for inappropriate responses. The child may apparently mis-interpret the form of A's initiation (usually a question), particularly if this is a complex one. Or he or she may not fully grasp the meanings of particular lexical items or function words in a question. It was not usually possible to establish whether the child was aware of misunderstanding, but felt that in a conversational situation it was necessary to answer and to keep the conversation going, or whether he or she simply had an incompletely specified and woolly idea about what was being asked. (Clearly, If a child does have insight, it would be preferable to request clarification from the adult rather than to respond wrongly, and this might be a useful strategy to teach.)

Examples:

  33 A: where did you go on holiday/
  34 C: in September/

  35 A: what other kind of party could it be?/
  36 C: cakes, drinks/

This category encompasses failures to obey conversational rules about what types of utterance may follow one another to make for coherent communication.

A nil response was coded when there was an interval in which the adult waited for a response but the child produced nothing, not even an 'er - - -' or a non-verbal response. Persistent nil responding alienates the conversational partner and may create an impression of hostility. Even a limited 'don't know' response is greatly preferable to maintain interaction.

Example

  37 A: where did you go on holiday?/
  38 C: Scotland/
  39 A: oh, how did you get there?/
  40 C: and we went to Spain as well/

Note that there are many instances when a child may follow an adult's initiation with another initiation quite legitimately, such as when requesting clarification before answering, or when distracted. Utterances such as the following would not be coded as inappropriate; e.g.

  41 A: what did you eat at your party?/
  42 C: there's a wasp in your hair/
  43 A: aargh!/

In certain cases, the child may show adequate understanding of the literal meaning of an utterance, but miss the adult's intended meaning because the linguistic, environmental or social content is not taken into account. This can lead to an overliteral interpretation, missing the illocutionary force of the adult's utterance e.g.

  44 A: can you tell me about your party?/
  45 C: yes/

  46 A: would you say that the boy looked ill?/
  47 C: the boy looked ill/

In many instances, the child's response would not be regarded as inappropriate if considered only in the light of the immediately preceding initiation; the problem is that it does not take into account the context set by preceding conversational turns, e.g.

  48 A: yes/ 
        how did you got to your holiday In Campomor?/
  49 C: by car/
  50 A: by car/
        who drove you?/
  51 C: and aeroplane/
        of course aeroplane/
  52 A: a plane/
        so it's not in England/
  53 C: no/
  54 A: where is it?/
  55 C: Campomor/
  56 A: are there any other times when you have parties?/
  57 C: no/
  58 A: what about at Christmas/   
  59 C: it snows/

Also coded here were instances of miscomprehension of homonyms, where one would expect the adult's intended meaning to be evident from the context:

(talking about his wearing only a T-shirt in the cold weather. There is another boy in the class with the surname Hardy)

  60 A: I thought you said you were hardy?/
  61 C: no, I wasn't/
        I was Jones/

Good conversationalists obey what Grice (1975) has termed the ^'maxim of quantity', i.e. they make their contributions as informative as is required for the purposes of the exchange, without telling the conversational partner things he or she does not need to know. Children with conversational problems may fail to observe this maxim in one of two ways: by giving their listener too little information so that their meaning remains unclear, or by providing more information than is necessary. Three types of inappropriate utterance were identified in which the maxim of quantity was contravened because too little information was conveyed.

This was coded when the child's response omitted one or more elements, apparently wrongly presupposing that the listener had knowledge of the 'elided' words.

Consider the following example which was taken from a transcript of a 9-year-old language-impaired boy, who had fluent and grammatically complex speech:

  62 A: so what did you do when you were sick?/
  63 C: I Can't remember/
        I DID though when I was run over by a car/

In this example the child has used what appears to be an elliptical form 'DID', where there is no call for it. The adult has not used a verb which could be presupposed by the child's utterance. If the adult had said 'You didn't get hurt though', then it would have been quite appropriate for the child to respond as he did, and elide the presupposed verb 'get hurt'.

This code was only used if it was clear from the rest of the child's transcript that he or she was capable of producing the complete sentence form that should have been used in this context, e.g.

  64 C: I've got a big sister who works in Barclays Bank/
  65 A: mhm/
  66 C: 21/
  67 A: she's 21?/

This was coded when the child introduced a term whose reference had not been sufficiently well established for the listener. The unestablished referent was commonly a pronoun, but sometimes it took the form of a noun phrase which had not been identified earlier in the discourse.

  68 C: over here you can go to the car park and get some more
        petrol but it's not on that car/

  69 C: when you take the ring off it it ones both of them are crowns/

  70 C: and be having bath/
  71 A: who'll be having a bath?/
  72 C: that's what they do in hospitals/
  73 A: what?/
        have baths?/
        do they do anything else?/
        tell me what else they do/
  74 C: they play/
  75 A: mhm/
  76 C: and they play about in the puddle/

Not every non-established referent automatically confers inappropriacy onto the entire utterance: there appears to be a degree of 'non-establishment' which the listener cannot tolerate. Our impression was that in general a non-established referent must represent a key part of the sentence to convey a sense of inappropriacy to the listener.

Where a logical step of the argument or a critical step in the sequence which the child is producing is omitted, the effect is bizarre, and the natural flow of the conversation is interrupted. The listener is left without a crucial piece of information which would link the now inappropriate utterance to those that have gone before.

  77 A: do you ever have a party at school?/
  78 C: yes/
  79 A: when do you have parties at school?/
  80 C: we still have a one in the infant and junior/
  81 A: mm/
  82 C: do my own cake/
  83 A: is that for your birthday?/

  84 A: what will happen if he doesn't get better?/
  85 C: he--get some medicine-and make--and make--/
        my brother was feeling sick on Monday/
  86 A: right/
  87 C: -and I took my trouser off/
  88 A: uhuh/
        Why did you take your trousers off?/
  89 C: he was sick on my trouser/

The maxim of quantity may also be violated by the child providing unnecessary information to the listener.

This was coded when a fact was unnecessarily asserted or denied, where the converse would not normally have been assumed.

  90 C: now the new exhaust wasn't rusty/
  91 A: mhm/
  92 C: and the silencer hadn't dropped off/

  93 A: so where have you been on a boat?/
  94 C: where have I been?/
        - haven't sailed a cruiser you know/

Some children tended to over-elaborate on a topic, saying more in response to a question than was necessary.

  95 A: is that a good place to break down?/
  96 C: the answer whether it is a good place to beak down is
        no, because if see if anybody broke down there's no
        telephone, there's no telephone for the breakdown/

Also coded here were cases where the adult expected the child to give some kind of generalisation,. but receive a much more specific answer:

  97 A: so what happens to people who get very ill?/ 
  98 C: they won't be able to go downstairs and watch their
        favourite television programme/ 

In this category were included those utterances where the child attempted to reiterate or to confirm a piece or information that has already been established. Obviously, there are instances where it is appropriate for a person to confirm something just said (for instance if the adult had misheard and the child has picked up a non-verbal cue from the adult to indicate that this was the case), but we noted that particular children tended to overuse this as a conversational device to the extent that it was quite conspicuous.

  99 A: and can you think of any other occasions when we have parties?/
  100 C: sing happy birthday - -/
         all sorts of things/
  101 A: uhuh/
  102 C: sorts of things/

  103 C: you been there?/
  104 A: yes/
  105 C: yeah/
         I been there to have my operation/
  106 A: uhuh/
  107 C: had my operation there/
         jelly and icecream before/

An elliptical form is expected but not used.

  108 A: what's the doctor doing?/
  109 C: the doctor is looking at the boy/

The effect of not using ellipsis where appropriate is to make the conversation seem stilted, and bestows a 'learned' quality to the interchanges. It is possible that in some language-disordered children this may be instilled by speech therapy: a few children invariably produced subject-verb-object conversations when asked to say what was happening in a picture, but soon ceased to speak this way when engaged in more informal conversation.

The difference between this category and other types of semantic and pragmatic problem is that the utterances coded here gave the impression that there was something abnormal about the message the child was trying to convey- not just with the way it was conveyed. Indeed. the child's expression may be very clear, but the utterance seems an inappropriate or even bizarre thing to say in this context.

Five broad classes of conversational behaviour were identified under this heading.

A conversation that never progressed would be very dull indeed, and it is entirely appropriate that speakers should develop a topic and then move on to another. However, some children gave the impression of going off at a tangent in an inappropriate way: they would drift off into talk about something which is in some way connected to the original subject, but not really relevant to the discussion:

(birthday picture)

  110 A: what's going on there?/
  111 C: it's someone's birthday/
         something could be dangerous you know like a fire from the candles/

Some children with circumscribed interests would invariably turn conversations to particular topics:

  112 A: Anyway, I'm glad you enjoyed the fair/
         Let's talk about something else/
         How do *-
  113 C: *Did you ever see the bicentennial state fair?/
  114 A: No, I didn't see that one/
         Hey, how do you like your teacher?/
  115 C: She's really OK/
         She lets me work on my bulletin board/
         She also lets me play with the cars/
  116 A: The cars?/
         What cars are those?/
  117 C: The model cars in the state fair exhibit/
         How much do you really like the state fair?/

Some changes in topic were sufficiently unmarked by the child as to appear non-related to the topic in hand. This stood out as being quite bizarre and abrupt:

  118 A: where might he go?/
  119 C: down to the_____/
         you know I told them about Blue Peter/
  120 A: to the where?/
  121 C: (laughs) well if he was very ill he's go to hospital/

  122 A: are you too old then to have parties?/
  123 C: - I don't know/
         I had a party/
         I got three sisters and one brother and it's me the one brother/

Some children produced utterances that had a stereotyped quality, as if he or she was repeating learnt information or a learnt construction.

  124 A: have you ever been to the doctor?/
  125 C: I had a apple a day/

Sometimes the child invariably produced a stock phrase with a characteristic intonation pattern that was quite inappropriate to the context in which the phrase was uttered. For example, one boy always responded affirmatively by saying 'oh yes' with a pronounced rise-fall on 'yes'.

This was coded when the child asked a question that the adult could not possibly know the answer to, which was not the type of question typically asked about this topic, or to which the child already knew the answer.

Sometimes children gave the impression that they did this because they spent much of their time being asked questions by adults, and therefore thought that asking questions was an appropriate conversational strategy. It is possible that some children have more insight and use a questioning mode as a strategy to avoid being asked more questions which they cannot cope with.

  126 C: do you like candyfloss?/
  127 A: no/
  128 C: do you HATE it?/
  129 A: I think it's all horrible and sticky/
  130 C: why?/
  131 A: you used to like measles?/
  132 C: no, I didn't/
  133 A: I bet you didn't, no/
         it's horrible, isn't it/
  134 C: what colour were they?/
  135 A  what colour were the measles?/
         I don't know/

This was coded when a child made remarks which were over-friendly or over-personal, e.g.

  136 A: right, let's sit over here/
  137 C: you've got purple socks on!/

(8 year old having tried unsuccessfully to explain what his maze is like to an unfamiliar adult)

  138 C: you could come to my house and see what's it's like/

child to speech therapist:

  139 A: who is your best friend?/
  140 C: I haven't got one/
         will you be my best friend?/

This is an instance where some people may disagree with use of the term 'inappropriate', which appears censorious of what may seem rather endearing behaviour. However, these types of comment are simply not usual for a child of this age to offer to a *strange adult. What may seem like loveable naivety in a primary school child starts to look social incompetence in an adolescent, and children who regularly contravene social expectations in this way will attract negative reactions from other people and may be regarded as naughty, rude, stupid or mad

Some children gave responses that seemed simply to reflect the fact that they did not know enough to be able to provide an adequate response. This seemed less of a language problem than limited general knowledge and lack of experience. The following types of interaction were not uncommon in the transcripts of our control 4 year olds:

  141 A: would you say that's a good place to break down?/
  142 C: (shakes head)/
  143 A: why not?/
  144 C: it's not good/

  145 A: he'll have some medicine/
         and where might he go if he was very very ill?/
  146 C: to school/

Sometimes children who are out of their depth give an initial uncertain response which cannot be sustained under further questioning, and appear to revise what they say 'on line', leading to self-contradiction. Again, this seemed more of a problem with general knowledge than with language.

  147 A: would you say that's a good place to break down?/
  148 C: no, not really/
  149 A: why not?/
  150 C: cos it's only a small it's only a small road- but you
         can park - you can push your car out of the way/

Inevitably there were other examples of unusual utterances which did not fall neatly into one of the categories described above, but which were too rare to justify a separate category.

The occurrence of each type of inappropriacy was computed as a percentage of all utterances and logarithmically transformed to maintain homogeneity of variance.

Table 2 shows reliability indices for each subtype of inappropriacy. There was good agreement for the categories of 'violation of exchange structure', 'too much information' and 'unusual content and use', and significant, though less high, correlations for 'use of context in comprehension', 'too little information', and 'immature/inadequate'. Inter-rater agreement failed to reach statistical significance for the categories of 'expressive syntax/semantics', 'literal comprehension and 'other' -

________________________________________________________________________________
Subcategories of Inappropriacy    Inter-rater     Test-retest     Test-retest
                                  agreement       agreement over  agreement  
                                                  2 weeks         over 9 months
                                                                  9 months
                                  (n=20)          (n=20)          (n=14)
________________________________________________________________________________
Expressive semantics/syntax       0.279           0.680+           0.598*
Literal comprehension             0.138           0.177            0.312
Violation of exchange structure   0.929+          0.618+           0.181
Context in comprehension          0.639+          0.464*           0.246
Too little information            0.487*          0.639+           0.367
Too much information              0.924+          0.954+           0.925+
Unusual content/use               0.950+          0.845+           0.248
Immature/Inadequate               0.670+          0.147           -0.067
Unclassified                      0.303           0.471*           0.427
________________________________________________________________________________

With the exception of the 'immature inadequate' category. all those categories that gave significant inter-rater correlations also showed significant test-retest correlations. One unexpected finding was that there was a highly significant test-retest correlation for the category of 'expressive syntax semantics', although inter-rater reliability had been low for this category. This suggests that the two raters were each adopting specific but different criteria for this category, so that although they did not agree with one another, each rater was self-consistent.

Few of the indices were stable over a 9-month interval, although the category of 'too much information' gave impressively high correlations.

Three groups of children were compared: the two language-impaired groups defined above (semantic-pragmatic and other) and control children aged 4-5 years, whose overall inappropriacy scores were similar to those of the language-impaired groups. Older control children were excluded from consideration because they had very low rates of inappropriacy. Results are shown in Figure 2. The error bars show the 95% confidence limits for the 4- to 5-year-old controls. It is apparent from inspection that the two language-impaired groups have different profiles both from one another and from the control children. The semantic-pragmatic group is characterised by high rates of inappropriacy in the categories of expressive semantics/syntax, too little information, too much information and unusual content/style. On each of these indices the mean for the 8- to l2-year-old children with semantic-pragmatic disorder is outside the 95% confidence limits for the 4- to 5-year-old controls. Both language-impaired groups had lower rates of immature/inadequate responses than the controls, with their means being below the 95% confidence limits for controls. The 'other' language-impaired group was also characterised by relatively low rates of comprehension problems, both for literal comprehension and comprehension in context.

Discussion

As can be seen from the data in Table 2, the extent to which raters agreed on categorisation was variable from one category to another, with some showing excellent agreement, others more modest correlations and some non-significant.

The relatively poor agreement for the category of 'expressive semantics/syntax' probably arose because of difficulties in deciding whether an utterance should be regarded as bizarre or merely immature; if the latter, then inappropriacy would not be coded. There appeared to be two factors that determined whether or not an expressive problem gave a sense of inappropriacy. The first was the degree of mismatch between the child's overall level of expressive development and ability to convey meaning with precision. If a 3 year old speaking in three-word utterances uses the wrong verb or preposition, the listener will usually make allowance for this, because his or her expectations are geared to the child's overall language level. If, however, the same error is made by an older child using long and complex sentences, then this sounds odd. A further factor is whether the child conveys a broader meaning than intended (overextension), or a more specific one. Use of a general rather than a specific term or construction seldom gives rise to a sense of oddness unless the meaning is so imprecise as to make the listener uncertain of what the child intends. However, use of a term with a specific but wrong meaning is much more disruptive for the listener. For instance, many young children use 'and' as a general purpose connective, when more precise terms such as 'because', 'but' and 'so' could be used. This may sound immature, but seldom bizarre. However, the converse, use of 'because' when 'and' is appropriate, can sound very odd indeed (see example 1). Consider also example 2, where the child who is asked why he went to the doctor says 'I used to have a headache'. 'Used to' is a sophisticated syntactic expression referring to a past event occurring regularly. If the child had said instead 'I have headache', this would be detected as grammatically immature (because a past tense is required in this context), but would be less likely to be regarded as inappropriate, unless the child was otherwise syntactically sophisticated.

The low agreement for the category of 'comprehension of semantics/syntax' may well reflect the problem of deciding whether an odd response has arisen because the child has misunderstood and is attempting to answer the wrong question, or whether the adult's initiation has simply been ignored and the lack of relationship between the adult's turn and the child's is better regarded as a violation of exchange structure, e.g. in this case, is the child misunderstanding the question as 'what games do you like?' or merely ignoring the adult?

  151 A: have you ever had a birthday party?/
  152 C: yes/
  153 A: what sorts of things did you do at your party?/
  154 C: play games/
  155 A: what games do you like?/
  156 C: I like eating those cakes with icing on/

Other indices of types of inappropriacy gave much better agreement, however, indicating that subcategorisation of types of inappropriacy is not unrealistic. Furthermore, those indices that characterised the semantic-pragmatic group tended to be those that also gave good inter-rater and test-retest agreement. The most striking case was the category of 'too much information', which seemed to reflect a highly stable conversational characteristic of children with semantic-pragmatic disorder that distinguished them not only from other language-impaired and control children of the same age, but also from younger normal children with similar overall levels of inappropriacy. Previous descriptive studies have reported that young children with semantic-pragmatic disorders may be unable to use ellipsis where it is called for (McTear, 1985; Conti-Ramsden & Gunn, 1986). This study confirms and extends this finding, by showing that such children have several conversational characteristics, that may be subsumed under the heading of providing too much information for the listener.

The classification of subtypes of inappropriacy lends support to the view that children with semantic-pragmatic disorder are not just immature in their conversational behaviour but also show persistent conversational features that are not normal at any age, at least above 4 years old. Adams and Bishop p. 231) noted that although they are described as 'verbose' these children do not differ from others in number of utterances per turn, although they do produce an unusually high rate of initiations. The impression of verbosity seems to reflect the unusual content of their utterances, in which they produce unnecessary repetitions, assert or deny what is already known, or give over-precise and over-elaborate information. It may seem paradoxical that this group of children also scored highly on the category of 'too little information'. High scores on both indices often coexisted in the same child. Clearly the problem is not just one of stating too much or too little, but rather in matching conversation to the needs of the conversational partner.

On some of the inappropriacy measures, children with semantic-pragmatic disorder were abnormal for their age, but did not differ from younger control children. These behaviours, then, could reasonably be regarded as immaturities. Under this heading came violations of exchange structure (failure to respond at all or ignoring of an initiation), and failure to understand the literal or implicit meaning of an adult utterance. This result agrees with clinical accounts (Rapin & Allen, 1983; Bishop & Rosenbloom, 1987; Rapin, 1987) of children with semantic-pragmatic disorders who give tangential answers to questions, or misunderstand the illocutionary force of an utterance.

At the time this categorisation of inappropriacy was being devised, we were unfamiliar with Damico's (1985) clinical discourse analysis. This procedure had similar aims to ours, in that it did not use a predefined set of linguistic categories, but rather set out to classify 'trouble spots' that impeded children's functional discourse. Damico devised his analysis through the study of conversations with 38 individuals aged from 6 to 22 years who were characterised by language impairment, low academic attainment and poor social stills. The main differences between this scheme and Damico's are that he did not include a category of semantic/syntactic errors, and it is unclear where instances of failure to use context in comprehension could be coded. Conversely, our scheme neglects non-fluency, gaps or mazes as sources of discourse failure, and did not include categories for non-verbal behaviours (such as abnormal gaze), message inaccuracy or failure to ask relevant questions. However, in other respects, there is considerable overlap between the two schemes, although the terminology is quite different. Thus Damico includes categories corresponding to failure to understand literal meaning, violation of exchange structure, too much and too little information, and abnormal content. This convergence of two independently devised coding schemes offers some support for the validity of these categories. Damico emphasised that his scheme was descriptive and was not designed for quantification. We would argue, however, that these abnormal behaviours can and should be quantified in order to identify when their occurrence is so frequent as to be abnormal.

Although the concept of inappropriacy is ill defined, independent raters overlap substantially when asked to assess the extent to which a conversation is inappropriate, and these judgements differentiate normal from language-impaired children, and distinguish between children identified as having semantic-pragmatic disorders and those with other types of language difficulty.

There was no single explanation for utterances that were judged inappropriate; peculiarities of semantics, syntax and pragmatics could all lead to a sense of oddness.

A question that frequently crops up when discussing inappropriate language is how far the problem is linguistic and how far it is cognitive. Although this distinction is very hard to make in practice, it is not empty. Johnston (1985) raised this point when discussing the language of a child whose use of reference was abnormal, so definite terms were used when the referent was unclear to the listener. She noted two possible explanations. Either the child understood the relationship between definiteness and old information, but miscalculated his listener's need to know (i.e. a problem of social cognition rather than language), or he may have recognised his partner's need but lacked the linguistic knowledge of discourse categories that is necessary to select definite vs. indefinite forms appropriately (i.e. a linguistic, pragmatic problem).

The same sorts of arguments can be applied to many of our categories of inappropriacy. Consider example 1, where a child says 'We went on a bus because Lee was sick out of the window'. One possibility is that the child has a distorted concept of causality, i.e. the language reflects an underlying cognitive disorder. Alternatively. this child may be perfectly well aware that Lee's sickness has not caused the bus journey, but he does not appreciate the difference in meaning between 'because' and 'and', a problem of semantic overextension. We cannot know which explanation is correct simply from observing such utterances, but it makes sense to explore the child's appreciation of linguistic distinctions in such cases. Professionals who are not familiar with language disorders often interpret children's utterances at their face value, particularly if the child uses fluent, complex language, and we find that examples such as example 1 can be treated as ostensive evidence that the child is thought disordered or psychotic. A less extreme interpretation in terms of semantic overextension is compatible with at least some instances of 'inappropriate language'.

However, some types of inappropriacy do not seem explicable in linguistic terms. Under the heading of unusual content or style were categorised utterances where the abnormal language appeared to be the vehicle in which disturbances of cognition were made manifest. The abnormality was not in how the message was conveyed, but in the message itself. All the behaviours listed under this heading have been described in autistic children, and our impression

was that children who produced utterances of this kind tended to have other autistic features.

Johnston (1985) noted that the literature on conversational skills in language-impaired children had not yielded impressive results. Adams and Bishop (p. 237) proposed that one reason for this was that many studies did not take into account heterogeneity of language-impaired children. This analysis suggests that a further reason may be that, in relying on analytical procedures derived from linguistic analysis of normal conversation, investigators may be missing those behaviours that can make children's conversation seem inappropriate. Most research to date has restricted consideration to topic shift, turntaking and repairs, which are features that accounted for only a small proportion of conversational abnormalities observed in our study.

Clearly, more work is needed to specify and define the variety of conversational difficulties that lead to a sense of inappropriacy. The subcategorisation reported here represents a preliminary attempt at this exercise that can undoubtedly be improved upon, but which does demonstrate the feasibility of this approach. Although some subcategories could not be identified reliably by independent raters, others could, and these included some categories that characterised the utterances of children with semantic-pragmatic disorders and were not found with any frequency in the transcripts of younger normal children.

For clinical use, we have produced a pilot version of an analysis based on the experimental procedures described in these two papers, entitled Analysis of Language Impaired Children's Conversation (ALICC). This can be used in two ways. First, it can provide a measure of conversational exchange structure, cohesion, turntaking, repairs and appropriacy which can be interpreted relative to data from normal children of different ages. This approach can be useful in research studies where it might be necessary to have a quantitative index of conversational ability in order to, for example, compare subgroups of children, look at correlates of conversational disability or consider how conversational competence varies across different settings. Secondly, the procedure provides a basis for a detailed analysis of the nature of inappropriacy which may help the speech therapist identify specific areas of difficulty to work on in remediation. It will undoubtedly be necessary to revise and improve this analysis, but it is hoped that in its current form it will provide a useful first step for others interested in analysing conversational problems in more detail and that it may stimulate therapists to develop approaches to the sorts of impairments seen in children with semantic-pragmatic disorders.

Copyright Holders Note:
Permission has been granted to put a copy of this article on this World Wide Web page by Whurr Publishers Limited on behalf of the copyright holders: The Royal College of Speech and Language Therapists. It is not for reproduction without permission.

Back to NAS Surrey Branch Welcome Page