PHONETICS, PHONOLOGY, AND LANGUAGE DISORDERS

AN ENCYCLOPAEDIA OF LANGUAGE



237



that certain segments will resist full identification. Such modifications are necessary because the IPA symbols (segmental and

diacritic) are devised to deal with the range of sounds possible in the languages of the world as used by adult speakers. The

articulation of both normal and disordered children may (and does) deviate considerably from such ‘normal’ adult speech

sounds. Without specific transcriptional features to capture the idiosyncratic character of the pronunciations of impaired

individuals in particular, there is a considerable risk of data distortion.

As Carney (1979) points out, however, the limitations of standard transcription systems for dealing with disordered speech

are often not acknowledged. The drawbacks are most obvious when a transcription of speech amounting to a phonemic

representation is used. In normal circumstances such a transcription allows the inference of a considerable amount of phonetic

detail, since the range of allophonic variation, for most accents of English, is well-known. Thus (to take one of Carney’s

examples) in RP the transcription of a lateral in different contexts using the same symbol will not mislead: in [klei], [lei] and

[eil], we are able to predict the phonetic variation in clay, lay and ale from the position in which the lateral appears. Following

the voiceless velar stop, it is likely to be devoiced, while pre-vocalically so-called ‘clear’ [l] has what Gimson (1970:201)

describes as a relatively front-vowel resonance, as opposed to the back-vowel resonance of the post-vocalic ‘dark’ l. (For

these differences, see Chapter 2, above.) There is no guarantee however that a child with speech problems will respect the

allophonic variation of the adult language. It is not uncommon for example for such children to produce ‘clear’ l in both prevocalic and post-vocalic positions. A transcription which assumed adult allophonic variation would miss this information

which is potentially valuable for remediation, and so constitute what Carney (op. cit.) would refer to as ‘inappropriate

abstraction’. Careful and detailed transcription by well-trained individuals, using where relevant the recommended symbols of

Figure 2, will overcome most of the problems of too abstract a transcription, and in most instances furnish the speech

therapist with the information needed.

3.2

Instrumental supplementation

It has been argued however that the procedure of phonetic transcription can be unreliable, because the child (normal or

disordered) may be making distinctions, or using articulatory postures that the transcriber cannot hear, however skilled. Since

this information may be relevant to the characterisation and/or remediation of the child’s problem, it may be necessary in

certain areas to supplement an auditory impressionistic transcription with information from instrumental phonetic techniques.

We will consider one example which uses acoustic data from spectrograms, and one from speech production data, using the

electropalatograph.

It has been a general observation of young normal children’s developing speech that the voicing distinction in initial

English stops is neutralised at a certain, quite early stage in their acquisition. An instrumental analysis of the speech development

of normal children (Macken and Barton 1980) revealed that one stage of development, for some children producing their

versions of voiced and voiceless stop targets, involved a consistent but sub-phonemic difference in voice onset time, a crucial

cue for voicing in English and other languages. In distinguishing /p/ and /b/ in English, described respectively as voiceless and

voiced labial stops, the point at which voicing begins, after the release of the stop is crucial. If voicing begins at the time of

release or up to about 30 milliseconds after, then the sound will be interpreted as /b/. But if voice onset is delayed until after

this 30 msec cross-over point, then the sound will be heard as /p/. The VOT range for /b/ (and other voiced plosives) is

referred to as the ‘short lag’ range, and the values for /p/ as the ‘long lag’ range.

The children in the Macken and Barton study, in their early pronunciations (and the age of the children in this longitudinal

study was from about 18 months to 2 years) showed no consistency in their use of short lag and long lag for labial stop targets.

But then for a period before they gave evidence of having controlled adult parameters, they made a consistent VOT

distinction, but within the adult short lag category. This distinction was not one that a transcriber would reliably pick up, and

it required spectrographic analysis to be detected. Similar data for VOT in labial stops (but using pneumotachography as the

instrumental technique) is reported for one of the disordered child subjects considered in detail by Hardcastle and Morgan

(1982). They also considered other aspects of their subjects’ pronunciation instrumentally, with some interesting results. One

technique they used was electropalatography, in which a real-time analysis of tongue dynamics can be made by fitting the

patient with an artificial palate, in which a number of small electrodes are embedded. As the patient speaks, the tongue

contacts he makes are recorded by the electrodes and transmitted to a computer, which records them. Comparisons were made

between contact patterns of the impaired subjects and those of normal children, in the pronunciation of single words. For one

impaired child, for example, it was apparent from the pattern of contacts that for initial alveolar or alveolopalatal sounds such

as the [t] in tent, or [ʃ] in sheep, there was considerable velar contact as well as the more forward contact necessary for the

alveolar obstruents. The velarisation would not have been picked up by a transcriber, but is obviously important for a speech

therapist concerned to have detailed information on articulation available for planning remediation.

In the remainder of our discussion of phonetics and phonological disability we will for the most part be concerned with data

analyses that rely on auditory impressionistic transcriptions. It should be clear however even from this brief excursus on



238



THE BREAKDOWN OF LANGUAGE



Figure 17 Extracts from suggested transcriptional conventions for disordered speech (reprinted with permission from Grunwell 1987)

PRDS—Recommended additional phonetic symbols

For the representation of segmental aspects of disordered speech

A.



Relating mainly to place of articulation



1.



Bilabial trills



2.



Lingualabials

(tongue tip/blade to upper lip)



3.



Labiodental plosives and nasal

(



4.



is an alternative to the usual ɱ)

Reverse labiodentals

(lower teeth to upper lip)



plosives, nasal fricatives lateral



plosives, nasal fricatives



5.



Interdenta

(using existing IPA convention for advancement)



plosives, nasal



6.



Biodental

(lower teeth to upper teeth)

Voiced palatal fricative

(reserving j for palatal approximant)

Voiced velar lateral

(using existing IPA convention for retraction)

Pharyngeal plosives

(using existing IPA convention for retraction)



fricatives percussive



7.

8.

9.

G.



Relating to inadequacy of data or transcriptional confidence



31.



‘Not sure’

entirely unspecified articulatory segment



Ring doubtful symptoms or cover symbols, thus:



unspecified consonant

unspecified vowel

unspecified stop

unspecified fricative

unspecified approximant

unspecified nasal

unspecified affricate

unspecified lateral

probably platal, unspecified manner (etc.)

probably

probably



but not sure (etc.)

, but not sure (etc.)



Note: A voiced, but otherwise unspecified, fricative may be shown as

and so on.

32.



; similarly, avoiceless, but otherwise unspecified, stop as



;



Speech sound(s) masked by extraneous noise

(( ))

thus

big ((bæd wul))f

or

big ((2sylls))

33. The asterisk. It is recommended that free use be made of asterisks (indexed, if necessary) and footnotes where it is desired to record

some segment or feature for which no symbol is provided.



AN ENCYCLOPAEDIA OF LANGUAGE



239



instrumental analyses that auditory transcriptions will not always be reliable. In particular, explanations of phonological

disability which rely on such transcriptions need to be evaluated carefully. (See Hardcastle et al. 1987. A detailed review of

supplementary instrumental analyses appears in Weismer 1984.)

3.3

Analysis

The introduction of phonological concepts into speech pathology in the 1960s led to a re-interpretation of the data of

‘articulation disorder’ and ‘misarticulations’ (Grunwell 1985a). The initial analyses of available phonetic transcriptions were

within the framework of phonemic theory (e.g. Haas 1963). More recently a variety of generative frameworks has been

applied. The most widely used has been some form of process analysis, particularly in North America (Shriberg and Kwiatowski

1980, Ingram 1981). Some researchers in Britain (e.g. Crystal 1982, Grunwell 1985s have argued for and exemplified a more

eclectic approach to analysis, which combines insights from phonemic theory and process analysis, in an initial description of

a disorder. We will accept this approach in providing illustrations of children’s pronunciation problems in English.

3.4

The phonetic inventory and systems of contrast

The majority of approaches to the assessment of pronunciation problems in English have concentrated on consonants. It used

to be generally accepted that vowels did not present problems to children acquiring the sound system normally (although

Haas 1963 does mention vowel problems in his case study, and Crystal 1982 allows for the analysis of vowels). More recently

problems with vowel acquisition (which, however, probably seem to occur only in a small percentage of cases) have been

reported (Stoel-Gammon and Harrington 1987). However, the data to be reviewed here will refer only to consonants.

Most recent approaches to assessment, following the normal phonological acquisition literature, accept that procedures

need to be sensitive to distributional differences in the availability of phones for contrastive use. The system of contrastive

phones that a child might be able to use in initial position in a monosyllable is usually different from (most commonly, more

extensive than) the system in final position. To provide a full description, then, it is necessary to examine separately phones in

different positions in syllable or word structure. A clear illustration of this appears in Figures 18 and 19.

Figure 18, adapted from Grunwell 1988, shows the inventory of phones available to two children. Figure 18(a) is a

consonant chart for Simon, aged 4 years 7 months (4;7), while Figure 18(b) shows the range of consonant sounds available to

Graham, aged 9;0. It is clear that Graham has a greater range of sounds available to him, overall. He has the full range of

plosives (p/b, t/d, k/g, plus a glottal stop), fricatives in two places of articulation (f/v, s) and the alveolopalatal affricate /ʧ/.

Simon has no velar sounds, no fricatives, and no glottal stop or affricate sound. Despite Graham’s wider articulatory

repertoire, an analysis of how this repertoire is deployed can reveal limitations on Graham which Simon does not have.

The consonant chart reveals the extent of the child’s articulatory abilities (or limitations). Further analysis is required to

determine how these abilities are employed at different positions in word and syllable structure. Figure 19, again adapted from

Grunwell 1988, reveals how the two children use their articulatory potential in making meaning distinctions, in one position,

syllable final/word final.

If we consider the structure of the word piglet, in terms of its consonant and vowel structure, we can represent it as:

cvccvc

piglet

The word consists of two syllables, and, without considering here exactly where the syllable boundary is, we can safely say that

p is a syllable initial/ word initial sound (SIWI) and t is syllable final/word final (SFWF). These labels are also used for

monosyllabic words: in pet, p and t would still be referred to as SIWI and SFWF respectively.

It has long been an observation in the literature on normal language development that children’s phonological systems are

not monosystemic.

Phonological development is not simply a matter of developing phonemic contrasts which are then

immediately generalisable to all places in word and syllable structure; different systems develop in different positions. In

English it is in general the case that a wider range of contrasts develop earlier in SIWI position than in SFWF. This

generalisation does not however rule out the existence of children who run counter to this tendency or particular contrasts,

e.g. fricatives (Shriberg and Kwiatowski 1980:135), being more readily developed in SFWF.

Since assessment procedures in child language disorders are referenced to normal development, a number of them,

including Grunwell (1985b and Crystal (1982) examine the child’s use of the phonetic inventory at different positions in word

structure. The charts for Simon and Graham in Figure 19 show only SFWF position (from Grunwell’s procedure). Each chart

shows the range of phonetic realisations for target adult phonemes. Thus the top left hand cell of Simon’s chart (Figure 19a)



240



THE BREAKDOWN OF LANGUAGE



Figure 18 Phonetic inventories for Simon (a) and Graham (b) (adapted from Grunwell 1988)

Phonetic inventory (a)



Name: Simon (4; 7)

Labial

Nasal

Plosive

Fricative

Affricate

Approximant

Other

Marginal Phones: ɫ υ



Dental



Alveolar



m

pb



l



Velar



Glottal



Other



?



j



Phonetic inventory (b)



Name: Graham (9; 0)

Labial

Nasal

Plosive

Fricative

Affricate

Approximant

Other

Marginal Phones: ɫ ʏ υ



Palatal



n

td



w



Post-Alveolar



Dental



Alveolar



Post-Alveolar



m

pb

fv



n

td

s



ʧ



w



Palatal



Velar



Glottal



kg



Other



ʔ

h



l



indicates that for all adult target words ending in m, Simon produced m. Graham however (Figure 19b) failed to produce any

realisation at all for a final m target (Ø indicates a zero realisation). A cell by cell comparison shows very obviously that

despite having the more restricted phonetic inventory, Simon has a more extensive range than Graham of potentially

contrastive elements. Pronunciation problems seem to require for their full characterisation not simply an account of phonetic

limitations but also details of the distributional patterning of the segments that are available to the child.

3.5

Process analysis

The phonemic approach embodied within the description of pronunciation problems so far described has either been

supplemented (Grunwell 1985, Crystal 1982) or supplanted by some form of phonological process analysis (Ingram 1981,

Shriberg and Kwiatowski 1980). This is now widely used in assessment procedures, particularly in the United States.

The term ‘phonological process’ derives from Stampe, who sees the phonological system of a language as ‘the residue of

an innate system of phonological processes, revised in certain ways by linguistic experiences’ (Stape 1969:443). The

processes were seen as innate, and acquisition was a matter, in part, of inhibiting those processes not relevant for the language

of the child’s environment. Processes have been commonly observed in sound changes in the world’s languages. A commonly

cited example of such a process is devoicing of word-final obstruents, which synchronically is a feature of German but not

English. Stampe’s account of the English child’s acquisition would require that an innately-present devoicing tendency was

eventually inhibited, to allow for voicing, which is phonemically relevant in English, to occur word-finally; on the route to

mastery we would expect a stage in which all final obstruents were devoiced. The German child on the other hand, will

devoice from the beginning.

It is not necessary to subscribe to Stampe’s views on the innateness of processes to find them useful in characterising

impairment. Processes can be viewed as strategies adopted by the child in the face of the complex task of learning how to

pronounce, and related to structural and physiological aspects of speech production (Shriberg and Kwiatowski 1980:4). We

can illustrate some of these features with examples from V., a Southern English girl of 4;8 with a history of pronunciation

difficulties:

(a) cluster reduction

initial: [t] for



/tr/ in train

/st/ in stamps

/kw/ in queen

/cl/ in clouds



AN ENCYCLOPAEDIA OF LANGUAGE



241



Figure 19 Contrastive possibilities for Simon(a) and Graham (b), SFWF position (adapted from Grunwell 1988)



/kr/ in Christmas



Any consonant cluster target containing a voiceless stop is substituted in V.’s output by a singleton voiceless alveolar stop.

The obvious outcome of this will be considerable homonymy in her vocabulary. The following words, for instance, would all

be pronounced as [teɩ]: tray, clay, stay. Cluster reduction is a widely attested phenomenon in normal and impaired child

phonologies in English and related languages (see for example Magnusson 1983 on Swedish).

(b) assimilation



A commonly reported assimilatory process is consonant harmony, in which for a CVC monosyllable target the child

produces the second consonant at the same place of articulation as the first (there may also be manner assimilation). Examples

from V.:

[ti:d]



lip

cheese

queen



Both stop consonants in V.’s production are alveolar. Of course the relationship between her segments and the targets is quite

complex, showing the simultaneous application of a number of processes, with resultant homonymy. Initially the lateral,

voiceless affricate and cluster are all substituted for by [t]; finally a labial stop, voiceless alveolar fricative, and alveolar nasal,

have [d] substituted. The influence of the [tVd] word-shape on V.’s output at this stage of her development can be gauged by

her production of CV target monosyllables at this point:



242



THE BREAKDOWN OF LANGUAGE



Figure 20 Common natural processes

1.

cluster reduction

2.

final consonant deletion

3.

unstressed syllable deletion

4.

stopping

5.

liquid simplification

6.

palatal fronting

7.

velar fronting

(from Shriberg et al. 1986)



(e.g. tomato—[ˈma:toʊ])

(e.g. lay—[peɪ])

(e.g. shoe—[ðu:])

(e.g. good—[dʊd])



[ti:d]

[tɜ:d]



key

cow



In both cases open syllable targets have a closed syllable representation for V. (We will deal with vowel problems separately

below.) This is an unusual phenomenon in normal developmental terms.

(c) stopping

[ti:d]

[pɔ:]



cheese

four



In these examples from V. fricative targets are represented by stops at the same place of articulation. Like cluster reduction

and consonant harmony, this is well-attested.

The process approach to the description of pronunciation difficulties collapses the separate stages of phonetic inventory and

contrastive analysis seen in the examples above from Grunwell 1985. Its essentially rule-based approach however does make

comparisons across language-disordered children, and between normal and language-disordered, more straightforward.

Shriberg et al. (1986) in a comparison of the characteristics of children with ‘phonological disorders of unknown origin’ and

normals, find that there are eight processes (which they refer to as ‘natural sound changes’) which ‘capture over 90 per cent of

the deletion and substitution errors made by normally developing and speech-delayed children above age 3’ (1986:145).

Three of these processes, cluster reduction, assimilation and stopping, are present in V.’s output, above. All are listed, with

examples where necessary, in Figure 20.

But what is it that distinguishes the ‘phonologically disordered’ children from normal children? In the Shriberg et al. large

sample comparison, it is the different proportion of errors that is most striking. The highest proportion of errors in the normal

group is 15 per cent, while in the disordered group it is 75 per cent. More generally, the current view taken of what constitutes

a phonological disorder is that it consists in the non-elimination by the children involved of processes regularly found in normal

development. There is some data available (reviewed in Leonard 1985) which goes against this generalisation. However, for

most of the current literature it seems reasonable to assert that the child referred to as phonologically disordered is one whose

development of pronunciation skills is delayed or ‘frozen’ relative to other aspects of his language development, particularly

vocabulary size (Ingram 1987).

It has become common in the speech therapy field to refer to a child who shows up pronunciation problems on one of the

assessment procedures we have described as ‘phonologically disordered’, and we have used this term above. The practice is a

reaction to the historical tendency to assume that all pronunciation problems were a matter of articulatory skills (or lack of

them), and uses the label ‘phonological’ to emphasise the linguistic nature of the disorder. However while modern assessment

procedures all use phonological frameworks (features, phonemes, rule statements, phonotactics, etc.) in order to arrive at a

descriptin of the data, analysts have not rested there. There has been an unfortunate tendency to take such descriptions as

explanations, rather than as a set of phenomena to be explained (Grunwell 1987, Locke 1983). The assessment procedures we

have used as examples are neutral concerning the sources of the problem they represent (in the terms of Hewlett 1985, they

are data-oriented rather than speaker-oriented). One example will perhaps suffice to emphasise the need to search beyond the

descriptions for the basis of a problem. The instance cited comes from normal development.

Smith (1973) in his detailed longitudinal account of the sound system development of an individual child notes a point at

which cluster reduction in SFWF positions occurs. Final nasal+stop clusters (e.g. nt, nd) are reduced to singletons ([t], [n]

respectively). In Smith’s analysis, this difference is dealt with solely in terms of realisation rules—i.e. in production terms.

The child’s lexical representation is assumed to be the same as the adult from the beginning—i.e. the child’s phonological

perception is perfect; errors in pronunciation and their elimination depend upon production rules and developmental changes



AN ENCYCLOPAEDIA OF LANGUAGE



243



in them. However, as Braine (1976) pointed out, this particular type of cluster reduction may have an auditory basis. Vocalic

elements in English (including vowels and nasals) are appreciably longer before voiced stops than before voiceless stops. The

representations chosen by the child in this instance for nt and nd clusters may then reflect the relative salience of the nasal

consonants auditorily in these sequences. The assessment procedures that we have considered can only consider production

data. The example from Smith shows that at least some instances of cluster reduction in the production data may have a

perceptual basis. Other data which indicates a perceptual basis for some pronunciation difficulties is reviewed by Locke

(1980).

The descriptive approaches to so-called phonological disorder have provided useful guidelines for remediation by speech

therapists, as well as a wide range of information on error types among children with pronunciation problems. It is clear

however that a new phase of research has begun in which the descriptive status of current procedures is recognised, and the

search for explanations of the patterns recognised is on (see Grunwell 1987, Menn 1987, Ingram 1987).



4.

GRAMMAR AND LANGUAGE DISORDERS

While the study of disorders of pronunciation has the longest history, the influence of transformational generative grammar

brought applications in the field of language disorders from the early 1970s, and from that time a variety of grammatical

approaches have been developed, particularly in the English-speaking world. The recognition that a child’s grammatical

system could be impaired—or at least that a language disorder could be characterised in terms of deficits in the grammatical

system—has generated research which has yielded useful information. This phase of research has been (properly) descriptive,

and has provided (a) useful frameworks for assessment and thus therapeutic application and (b) identification, usually in

surface terms, of the grammatical system deficits (most often with respect to English) that children classed as SSLD tend to

have. In this phase of research whatever repertoire of deficits emerged were labelled as grammatical problems, and it is only

now that a new research thrust is developing, which looks beyond the level at which the problems are identified, and asks

what the basis for these disorders is, and how they can be explained. The remainder of this section reviews the historical

development of research into grammatical disorders, and considers current trends.

4.1

Rule-based analyses

The most careful early investigation of the grammatical differences between normal and SSLD children is reported in

Morehead and Ingram (1973). They attempted to resolve the issue of whether SSLD children had qualitatively-distinct linguistic

systems by taking considerable care in subject matching for their samples. In addition to the usual IQ and socio-economic

criteria that had been applied, they also used a measure of sentence length (mean utterance length in morphemes) as a more

reliable indicator of relative linguistic development than age.

Morehead and Ingram compared the grammatical abilities of the two groups of children, using the components of a

standard theory model transformational grammar. They found that the phrase structure rules needed to account for the

children’s utterances in the samples they collected were ‘nearly identical’ for the two groups. A comparison of

transformations used by the two groups showed that, with certain exceptions, the majority of transformations used were

similar across the two groups. Where then does the difference between the two groups lie?

The most obvious difference between the groups is the age at which particular linguistic milestones are achieved. The normal

and language-impaired groups were matched for mean length of utterance, as we have pointed out. The age-ranges of the two

groups are however dramatically different. The span for the normal group was 1 year, 7 months to 3 years, 1 month. The

impaired group ranged in age from 3 years, 6 months to 9 years, 6 months. So for any particular linguistic level, the impaired

group show a marked chronological delay:

‘Given that normal children initiate and acquire base syntax between approximately 18 and 40 months, it appears that

deviant children take on the average three times as long to initiate and to acquire base syntax.’ (Morehead and Ingram

1973:216).

As with phonological disorder, at least some of the problem for children whose language impairment is revealed by

grammatical analysis is that they are not able, for whatever reason, to make the transition from one stage to another of the

acquisition process. This chronological mismatch, which goes far beyond the variability found in large-sample studies of

normal development (e.g. Wells 1986), has been a consistent finding of subsequent studies.



244



THE BREAKDOWN OF LANGUAGE



While the Morehead and Ingram study indicates that, so far as syntactic structure is concerned, SSLD children do not

develop ‘bizarre linguistic systems’, there was one point of difference that did emerge when the analysis went beyond a

comparison of phrase structure and transformational rules. A significant difference was found in the number of major lexical

categories (N, V, A) used by the two groups. This is interpreted as a restriction on the variety of construction types available

to the language-impaired group, where these types are seen simply as strings of category labels.

4.2

Surface structure approaches

Since the Morehead and Ingram study the points of contact between language impairment and generative theory, as it has

developed, have been slight. There are several reasons why this should be so. One of them is the finding that, viewed from the

perspective of syntactic rules, language-impaired children have ‘normal’ systems, even though these may take an inordinately

long time to develop, or, for some of the children, may be only partial systems ‘frozen’ at a particular stage of acquisition.

Another reason concerns the tendency of studies in non-normal language development to follow trends in the study of normal

language acquisition. In the 1970s this research was much less dependent on current trends in linguistic theory than was earlier

the case; it is only recently, in a period of relative stability in linguistic theory that acquisition researchers have once again

turned to the dominant linguistic model for insights (e.g. Atkinson 1986, Goodluck 1986). Research on language disorders

similarly departed from a reliance on transformational generative theory, but has still to reconnect so far as grammar is

concerned. Clinical approaches to grammatical problems since the late 1970s have tended to rely on surface structure

descriptive frameworks, and it is to one of these that we now turn.

4.3

LARSP

A grammatical profiling procedure widely used in the UK and elsewhere is the Language Assessment, Remediation, and

Screening Procedure (LARSP; for details see Crystal et al. 1989). This has been applied to child disorders and to adult

disordered data (e.g. Penn and Behrmann 1986), and versions of it have been applied to data from other languages (Feilberg

1987—Norwegian; Bol and Kuiken 1987—Dutch). Figure 21 shows a part of the profile which will illustrate the general

principles on which it is based.

In the figure, a section across the profile shows what is referred to as Stage III of grammatical development. The framework

used for the categories represented in the profile was based on that used in Quirk et al. 1972, for their grammar of

contemporary English. This is emphatically a surface structure approach, in which clause structure is represented at three

separate levels:

1. clause: major clause constituents are labelled as S, V, C, O, or A.

2. phrase: here category labels such as Det, N, V, Aux, Adj, Prep are used to label members of noun, verb or prepositional

phrases.

3. word: at this level of analysis morphological processes are recognised.

Stage III of the LARSP profiling approach reflects the clause/phrase/word division, with cols. 2, 3 and 6 representing clauses

(commands, questions, and statements respectively), while col. 7 lists noun and prepositional phrase types, together with

pronouns and auxiliary premodification. The final column shows some morphological features (past participle, third person

plural, and possessive). This approach, carried through the developmental range from the beginnings of syntax at about

eighteen months to the point at about five years of age when the child has acquired most of the major syntactic structures,

provides a template against which those structures actually occurring in a child’s language sample can be checked. The

LARSP profile, systematically applied to individual cases or groups, can provide a picture of the strengths and weaknesses of

the grammatical repertoire. The profile can be used either as the basis for further enquiry, or, for the clinician, as the startingpoint for remediation.

As a format for preliminary data-organisation then, the profiling approach has advantages. It is perhaps best viewed as a

(partial) analogue, at the grammatical level, of the transcription that is the initial representation of the data for phonological

analysis (Garman 1989). It is not based on a coherent theoretical framework, and does not of itself admit any conclusions

about the status of the child’s grammatical system. It will however identify categories or construction types which are

consistently lacking in a child’s repertoire, and thus direct more detailed and principled enquiry into the precise nature and

bases of these problems. The areas of difficulty identified for the SSLD child tend to be consistent, though they differ for

younger and older groups of children. In younger SSLD children it is possible to locate obvious areas of structural deficit,



AN ENCYCLOPAEDIA OF LANGUAGE



245



Figure 21 Stage III of the LARSP syntactic profile



whereas in the older children the identification of such gaps is not so straightforward. In the remainder of this section we will

consider some of the deficits that have been identified in SSLD children and the possible bases for them.

4.4

The English auxiliary

It is a general clinical finding, well-supported in the research literature, that language-impaired children have problems with

verb-forms in English, particularly auxiliaries (Fletcher and Peters 1984, Ingram 1972, Johnson and Kamhi 1984, Johnson and



246



THE BREAKDOWN OF LANGUAGE



Schery 1976, Steckol and Leonard 1979). By comparison with normal children of the same age, SSLD children will have a

less developed auxiliary system, and when normal and SSLD children are matched in terms of mean length of utterance

(MLU), the SSLD children are found to have a more restricted range of auxiliaries and to use them less frequently. To identify

a consistent gap in a grammatical profile is however merely to identify a problem which requires an explanation. Why should

the language-impaired child find the English auxiliary system problematical? There are a range of possible factors that are

relevant, which may operate either singly or in conjunction.

For some areas of the auxiliary system, which in normal development takes a considerable time to be acquired, cognitive

explanations have been preferred. One example of an attempt at this type of explanation was the claim by Cromer (1974) that

the late acquisition of the present perfect (as in I’ve finished, have you seen her) relative to past tense, in the children studied

by Brown and his associates at Harvard, was because of its relative cognitive difficulty. The present perfect form, in its

relation of a past action to the time of speaking (the so-called ‘current relevance’ meaning), was argued by Crome to present a

cognitive complexity over and above the simple past (which locates an action at some time prior to the moment of speaking).

Subsequent research has made it clear that Cromer’s explanation for the late acquisition of the present perfect (at up to five

years of age, in his sample) was a consequence of his studying American children. One of the dialectal differences between

American and British English is that the former has a much lower frequency of occurrence of present perfect, and when we

examine British children’s acquisition, we find that they develop present perfect much earlier than the American children (see

Wells 1979, Fletcher 1981, 1985:33ff; and particularly Gathercole 1986). The alternative (and simpler) environmental

explanation rules out the semantic-cognitive one in this particular case.

This does not mean that in certain areas of normal language development generally, and verb-form use in particular, we

will not want to appeal to the child’s conceptual abilities to account for the nature of change, or lack of it. As Johnston points

out,

‘if children learn language by analysing context, and mapping form to meaning, then their conceptual and factual

resources should constrain the acquisition process’ (1985:48).

One of the pieces of linguistic evidence claimed by Johnston in support of this view is the acquisition of members of some

form classes in a predictable order over a protracted course despite their formal similarity. In the area of verb-forms, the

obvious example of this in English is the lengthy development of modal auxiliaries, which begins in the third year and

continues into the primary school (Stephany 1986). If it is appropriate to ground aspects of the course of normal language

development in the child’s conceptual abilities, then there is every reason to consider the possible involvement of these

abilities in the same aspects of the linguistic behaviour of some language-impaired children.

That said, grammar, as Bruner (1983:169) reminds us, still ‘constitutes its own problem space’. Each child has to determine

which meanings are coded, and how, in the particular language he is trying to learn. Auxiliaries in English present the

following features which may affect the child’s establishment of a linguistic representation:

(i) Two auxiliaries are in construction with non-continuous morphemes (have+past participle, be+ing). All other auxiliaries

are not.

(ii) Have and be have a range of variant forms, contractible and non-contractible; for contracted third person, the have and be

form is the same (’s).

(iii) All auxiliaries (including a number of modals) tend to be phonetically brief, reduced forms in input language, particularly

in initial position. This might present a particular problem to language-impaired children. (Fletcher 1983; see also

Leonard et al. 1987).

So either the constituent structure of the verb-form (i), or the paradigms associated with particular auxiliaries (ii), or their

particular realisations (iii) could affect, independently or interactively, the (normal or non-normal) child’s learning.

A structural deficit for auxiliaries, identified on the grammatical profile of a language-impaired child, may therefore have a

variety of possible sources. It may have a cognitive basis, it may be at root a grammatical problem, or it may have its basis in

the problems in interpretation of input, or some combination of these factors. Future research will require the testing of

hypotheses which relate to the basis of the problem.

Surface-structure approaches of the LARSP type will isolate broad areas of difficulty such as the auxiliary system, or

complex sentences, another widely recognised problem area for SSLD children (Crystal 1984). There may be deficits

however which relate to the grammatical system, construed more broadly, but which a profile of the LARSP type will not

locate. We will look briefly at two of these: adverbials, which while structurally simple have functional differentiation; and

processing constraints.



AN ENCYCLOPAEDIA OF LANGUAGE



247



4.5

Adverbials

Fletcher and Garman (1988) report on a comparison between a group of older language-impaired children in residential

schools in the UK, and matched normals. On a variety of syntactic measures (including noun and prepositional phrase

structure, verb phrase structure, and availability of complex sentences) the groups are not significantly different. When adverbials

— words, phrases, or clauses that serve as adjuncts in clause structure—are examined in terms of their function rather than

their structural properties, differences emerge, particularly in the use of time adverbials. Time adverbials are involved in a

complex system of specification of tense and aspect choices on the main verb and reflect in their lexical realisations both

calendric (day, month, hour, etc.) and non-calendric (seasons, festivals, holidays) temporal reference points. To use these

adverbials successfully, the child has to learn the relationship between particular tense/aspect choices and the forms, the

semantics of time, and when it is appropriate to use temporal adverbials in discourse. Fletcher and Garman (1988) report that

the SSLD children used fewer time adverbials than their age-matched normal peers, and that they had a very limited grasp of

calendric time, as revealed both in their infrequent use of the relevant lexical items, and errors such as:

I do go home on the weekend on Fridays and Wednesdays.

They also showed uncertainty, as a group, over the appropriate provision of temporal specification. Speakers provide an overt

time reference for their hearers unless one can be construed from the context. Fletcher and Garman (1988) examined the use of

past tense by normal and SSLD children, and identified instances where the reference time could not be determined from

context, and asked whether an appropriate time adverbial was provided by the speaker. Normal five and seven-year-olds

provide the time adverbial on the majority of occasions, but the SSLD group supply fewer than a third of the necessary time

references.

It might be argued that these rather subtle limitations do not merit inclusion under the heading of grammatical disorders,

but rather represent the two aspects of what has been referred to as a ‘semantic-pragmatic’ disorder (Rapin 1982). The

semantic aspect of the disability could be said to be revealed in the problems found by the SSLD group in structuring the

lexical information that relates to conventional time organisation, while the pragmatics of temporal deixis and the

requirements of conversational behaviour are apparently violated by the non-provision of temporal adverbials. It is true that in

using the ‘semantic-pragmatic’ label writers have tended to highlight problems of word-finding (semantic) and problems of

conversational structure and inference (pragmatic) (Rapin 1982, McTear 1985, Conti-Ramsden and Gunn 1986), but the term

is so vague that it is not easy to see how the temporal problems identified can be excluded. The general point is, of course, that

once we take a broader interpretation of grammar than possible structural types and/or the categories represented in them, and

consider the functions of adverbials (and determiners, pronouns, modal auxiliaries, etc.) then we have a much wider range of

deficit phenomena which could be included under the heading of ‘grammatical disorders’. The availability of a single coherent

theory within which the phenomena of disability could be interpreted would avoid current problems of demarcation. In the

absence of such a theory, it is perhaps sensible to avoid labels, and to adopt a heuristic strategy which combines syntactic

profiles and a functional perspective, to develop a typology of linguistic deficits.

Implicit in this strategy is the assumption that even if we do not yet know the full range of possibilities for grammatical

impairment, broadly construed, sub-types with consistent linguistic profiles will emerge. Some recent work by Chiat and

Hirson (1987) suggests that in certain cases a quite heterogeneous grammatical profile will be found, which may not be

explicable as, at base, a grammatical disorder at all.

4.6

Processing constraints: children

In their analysis in detail of a girl of 10, Ruth, Chiat and Hirson explore the data from a perspective that considers the

processing of individual utterances, and ask whether the perturbations found in this individual’s output can be explained in

terms of processing limitations—consistent and specifiable constraints on the real-time production of utterances. The analysis

works within a psycholinguistic framework that conceives of the production of an utterance beginning with a ‘conceptual

intention’, and then proceeding through various stages of ‘mapping’, including a final phonological stage which structures the

output that we hear and record.

The account of Ruth’s problems considers first of all the range of conceptual intentions that she wishes to communicate.

Chiat and Hirson are satisfied that her intended meanings, identifiable in the structures she produces in relation to their context,

cover a wide variety of the types of conceptual relations expressed by language. These include basic propositional structure

(verb+arguments), embedded propositions as arguments, adverbials of various structural types as modifiers of propositions,

negation, and interrogation. Despite this relative richness of conceptual intentions, Ruth’s output shows many deviations from