Variations in "motherese" pronoun usage

Susan Blackwell, Department of English, University of Birmingham, UK

Abstract

1. Introduction

1.1 A brief overview of the literature on child-directed speech

Chomsky (1965) notoriously made "rather extreme and unsubstantiated claims" (Sokolov & Snow 1994:39) that the speech heard by young children was deformed and degenerate, and therefore an inadequate base on which to build language acquisition unless children were endowed with innate knowledge of Universal Grammar. Attempts to refute these claims led to an explosion in research, both on the language produced by young children and on the language directed at them. Early results showed that far from exhibiting complexity, syntactic errors, false starts and hesitations as Chomsky had claimed, "Baby Talk" (BT) - speech to young children - was typically clear, grammatical and semantically and syntactically simpler than speech to adults (Ferguson 1977).

While controversy continues to rage around the topic of what is now termed Child-Directed Speech [1], there is now a generally accepted list of its features:

Child-directed speech (CDS) tends to consist of short, well-formed utterances, to contain fewer false starts or hesitations, and to include fewer complex sentences and subordinate clauses. It is characteristically higher in pitch, more exaggerated in intonation, and slower in tempo than speech among adults. It is also highly redundant, as reflected in the incidence of part or whole repetitions, much more closely tied to the immediate context, and employs a number of special discourse features which serve to involve the child in interaction and to clarify and upgrade the child's own contributions.

(Pine 1994:15, references omitted)

Some (e.g. Levelt 1975; Moerk 1976, 1983; Cross 1977) have gone so far as to claim that by using CDS mothers are giving their children "language lessons". Such strong views are now generally discredited. It seems more plausible that some aspects of CDS are primarily intended to improve communication with the child while others are used to express affection towards it. Ferguson (1977) classified the features of BT into three kinds of processes: 'simplifying', 'clarifying' and 'expressive and identifying'.

1.2 Pronouns in CDS

Brown (1977:2) included "the use of proper names or kin terms in place of pronouns" in a catalogue of CDS features which he regarded as so transparent that they could "surely be tested against the investigator's intuition and reliably judged to be peculiar to the BT register." The present study nonetheless prefers empirical data to intuition and seeks to question Brown's claims.

Wills (1977) provides the first systematic account of pronouns in CDS, based on observations from five parent-child dyads. She observes that "most BT pronouns are conventional pronouns used grammatically but deviantly in regard to participant role, number, or gender" (Wills 1977:273), and proceeds to classify the deviations by means of eleven rules which derive BT forms from adult-adult usage. The most frequent substitution in her data is reference to the speaker by means of a third-person form such as a relational noun, in place of a first-person singular pronoun:

(M= Mother, F= Father)

The next most frequent is the use of "we" or "us" in place of the first-person singular:

Replacement of second-person forms by the child's name are also extremely common:

An alternative is to replace second-person singular pronouns by first-person plural ones:

(examples 1-5 are taken from Wills (1977))

The cumulative effect of these modifications is "a system which, as a whole, is strikingly different from adult talk" (Wills 1977:276). "Of the basic pronouns, only they is used consistently in its adult-adult sense." (ibid.)

Brown's (1977:2) allusion to "the use of proper names or kin terms in place of pronouns" turns out not to be sufficiently precise or inclusive. It takes no account, for instance, of the use of "we", which Wills observes is "the most ambiguous of all with regard to its exact referent" (Wills 1977:279). Sometimes "we" can be interpreted in context as a genuine 1st-person plural pronoun, including both parent and child; but at other times it is used to denote either the child alone (example 5) or the parent alone (example 3).

Boezewinkel (1995) discusses this phenomenon with reference to Dutch, citing the following examples from Coopmans who apparently considered them helpful to the child in establishing that "we" referred to more than one person:

(Coopmans, personal communication to Boezewinkel, in Boezewinkel 1995:25; my translation)

In (6) the "we" refers to the child; in (7) it denotes the parent. Boezewinkel vigorously contests Coopmans' suggestion that such parental usages help establish the plurality of "we": on the contrary, she states, it causes confusion when the child tries to relate it to normal daily communication.

On other occasions pronouns can be deleted altogether:

The deletion of "it" or a noun phrase also occurred, but less frequently:

Personification (example 10) and Role Substitution (example 11) occurred less frequently in Wills' data; she regards them as related since she considers role substitution as "an extension of personification" (Wills 1977:287):

Wills also encountered instances of ungrammatical case forms such as "you" for "your", which she considered to be infrequent, non-systematic and self-consciously produced.

Why should carers modify or avoid pronouns in CDS? It might well be expected to do more harm than good to the child's language acquisition. Wills herself acknowledges that "The exclusion and objectification accomplished by usage of [third person forms] are ordinarily a threat to people". (Wills 1977:282) Her suggested answer is that replacement of second-person pronouns by third person forms "may help the child to learn his name" (ibid.) However, caregivers are known to make frequent use of the child's name in addressing it (see Clark and Clark 1977:328), and it is unlikely that the replacement of second person pronouns is necessary in addition to this. Moreover, the substituted element may be a noun phrase or a third-person pronoun rather than the child's name.

A more convincing explanation is that parents avoid the use of both first-person and second-person singular pronouns, replacing both of these with a variety of items or even deleting them altogether, because they somehow know or believe that it is hard for young children to grasp the inverse relationship between the first and second persons and their consequent alternation of referent in dyadic discourse: what Jordan (1998) has termed "the speaker principle".

If this is so then it follows that parents of children with particular language disorders can be expected to make more modifications of this kind than the parents of normally-developing children:

Faced with such apparent difficulties with deictic pronouns, adults resort to the use of proper names to aid comprehension and this provides a workable strategy for the child with autism, who only seeks to identify people as individuals rather than as conversational partners.

(Jordan 1998:138)

We may also enquire whether there are differences between languages in CDS pronoun usage or avoidance. Presumably the cognitive problems associated with concepts of deixis are universal, but there may in addition be language-specific complications due, for example, to the wide number of variants (due to gender and case inflections, politeness conventions etc.) which pronouns exhibit in many languages as compared with nouns.

Perhaps most importantly from a practical point of view, linguists may wish to know whether the withholding of pronoun input by adults eases or hinders the child's acquisition of the concept of deixis in general, or the use of specific pronouns in particular. In the case of a child with autism, pronoun avoidance in CDS might lead to a vicious circle, in which the child is perfectly happy with the adult's substitution of unambiguous labels and has no desire to challenge it. Oshima-Takane & Benaroya (1989) claim that the only way to get autistic children to use deictic pronouns correctly is by explicit teaching. We shall consider whether there is any evidence for this.

1.3 Autism

The term "autism" in its modern usage [4] was appropriated in the 1940s by Leo Kanner in the USA (Kanner [1943] 1973) and Hans Asperger in Vienna (Asperger 1944), virtually simultaneously and apparently independently of each other (Wing 1991:98).

All forms of autism probably involve some kind of brain damage [5], albeit caused by a range of agencies and at different ages for different individuals, and notwithstanding the fact that medical researchers have so far failed to locate the precise area(s) of the brain involved. While Kanner insisted that the condition was present from birth, it is now generally accepted that this is not always the case and that autism may arise in a previously healthy child as a result of some kind of infectious agent or injury [6]. As Attwood (1998:143) puts it, "We recognise three potential causes of autism, namely genetic factors, unfavourable obstetric events and infections during pregnancy or early infancy that affect the brain."

Nonetheless there is strong evidence that a constellation of symptoms does co-occur in most autistic patients, justifying the use of terms such as 'autistic syndrome' to describe children such as those in Wing & Gould 's study:

All the children with social impairments had repetitive stereotyped behaviour and almost all had absence or abnormalities of language and symbolic activities.

(Wing & Gould 1978:25)

These impairments of socialization, communication and imagination - commonly known as 'Wing's triad' - have come to be accepted as the standard basis for the diagnosis of autism (Rutter & Schopler 1987).

One of the linguistic features of children with Early Infantile Autism (EIA) is what Kanner termed:

the typical, almost pathognomonic, pronominal reversals which consist of the child's reference to himself as 'you' and to the person spoken to as 'I'.

(Kanner [1946] 1973:46)

Kanner gives as a typical example the language of Donald T., one of his original 13 patients, who was aged 5;1 [7] when Kanner first encountered him:

He always seemed to be parroting what he had heard said to him at one time or another. He used the personal pronouns for the persons he was quoting, even imitating the intonation. When he wanted his mother to pull his shoe off, he said: 'Pull off your shoe.' When he wanted a bath, he said: 'do you want a bath?'

(Kanner [1943] 1973:4)

The feature is so common that it is included in the formal diagnostic criteria for the condition (DSM-III-R, American Psychiatric Association 1987). However, as I have argued in detail elsewhere (Blackwell 2007), the attribution of "pronoun reversal" to autistic children is over-simplistic.

1.4 Down Syndrome

Down Syndrome (DS) is a genetic condition caused by an extra copy of chromosome 21 (trisomy 21), affecting approximately one in every 700 live births (Patterson 1999:75). Its main feature is mental retardation, but some 50% of DS children also have congenital heart disease (Mattheis 1999:41), and it is often associated with a host of disorders involving respiration, digestion, hearing, vision, reduced immunity and accelerated ageing (see Hassold and Patterson (eds.) 1999).

DS children often have significant communication deficits, which may be helped by the use of an augmentive communication device or sign language since their perception is often considerably ahead of their expressive abilities (Cohen 1999:25).

DS children generally show a weakness in verbal short-term memory (Wang 1996): while normal adults have a short-term memory span of 7±2 items, the young adults studied by Fowler (1995) and Rondal (1994, 1995) scored approximately half of this. These memory limitations appear to lie at the root of the linguistic difficulties. One fifth of the parents in one British study reported that their children spoke in sentences of three words or fewer (Buckley and Sacks 1987). Moreover, DS children have problems with their grammar, particularly agreement relations. The linguistic performance of DS children is generally lower than that of comparison groups matched for age and general cognitive function, including children with autism. However, this pattern is broken when pragmatic skills come into play:

Intriguingly, though still open to much interpretation, persons with autism can sometimes master complex grammatical function, far more advanced than typically seen in DS, but at the same time be far less sophisticated than persons with DS in their use of verbs and pronouns that depend on social inferencing skills

(Fowler 1999:167)

Tager-Flusberg's team concluded that there were "similar developmental patterns in the emergence of syntactic and morphological structures in normal, DS and autistic children" (Tager-Flusberg et al. 1990:13). Both sets of language-impaired children seemed to pass through the normal stages but at much later ages and often at much slower rates (ibid:10). The only significant difference was in the proportion of nouns and closed class words used between Stage I (MLU 1.0-2.0) and III (MLU 2.5-3.0) [8]:

Specifically, at the early stages of language development DS children tended to rely more heavily on closed class forms than on specific nouns, whereas the reverse pattern was found for the autistic children.

(Tager-Flusberg et al. 1990:13)

Tager-Flusberg et al. note that Dooley (1976) found a similar preference for pronouns among DS children, and compare these observations for DS subjects with the findings of Bates et al. (1988) and Bloom et al. (1975), which indicate that there are individual stylistic differences among normally-developing children along the nominal/pronominal dimension.

The received wisdom, then, is that autistic children have serious problems with their personal pronouns and may avoid them, while children with Down Syndrome prefer them to nouns. If this is true - or even if it is not but is widely believed by the adults caring for them - we may expect it to be reflected in the use or avoidance of pronouns in parental CDS.

2. The CHILDES database

2.1 Origins and rationale of CHILDES

The idea behind CHILDES originated at the Max-Planck Institut für Psycholinguistik in Nijmegen, the Netherlands, in 1981: its construction began in 1984 (MacWhinney and Snow 1990). The purpose of the project was "to facilitate the sharing of transcript data, increase the reliability of transcriptions, and automate the process of data analysis" (MacWhinney 1995:viii). The resulting CHILDES system consists of three interdependent elements: the database itself, the CHIld Language Data Exchange System; the CHAT transcription conventions (Codes for the Human Analysis of Transcripts); and the CLAN package of analysis programs (Computerized Language ANalysis) (MacWhinney 1995, 2000a, 2000b).

2.2 Corpora selected for this study

The CHILDES handbook (MacWhinney 2000b) was consulted for its description of the various corpora. Potentially suitable subcorpora of the CHILDES corpus, along with the CLAN software suite, were downloaded from the website at Carnegie-Mellon University. [9]

A considerable number of criteria had to be met. The data needed to be dyadic, including full transcriptions of the speech of both mothers and children. In order to exclude as many confounding variables as possible, only the speech of mothers and their male children was examined [10]. Corpora involving multi-party discourse, apart from the occasional contribution by the researcher, were ruled out.

Data were required for a sufficient number of children in both selected languages (English and Dutch), and the files had to be in CHAT format to enable automatic analysis.

In view of evidence that children can use pronouns before the age of 2 (Hyams 1992; Clahsen 1990) and may indeed produce "reversals" at this age (Boezewinkel 1995; van der Geest 1977), it was imperative to examine the pronoun usage of children at this age and younger.

The English corpus eventually selected was the relatively recent Manchester collection (Theakston et al. 2001), a longitudinal study of 6 British boys and 6 British girls between the ages of approximately 2 and 3 years, audio-taped in their homes over a one-year period. An additional advantage of the Manchester corpus is that it includes morphological mark-up and coding for errors, imitation and self-repetition.

The Dutch corpus selected for comparison was the Groningen corpus (Bol 1996). This contains speech from 6 boys and one girl between ages 1;5 and 3;7, talking in an unstructured home setting with a parent and an investigator.

For the language-impaired subjects it was necessary to use data collected in the USA [11]. The Flusberg Corpus (Tager-Flusberg et al. 1990; Tager-Flusberg and Anderson 1991) is a longitudinal study of six autistic subjects and six comparable Down Syndrome children. Video and audio recordings were obtained by observation in the subjects' homes, in line with current psycholinguistic research on normally developing children (Tager-Flusberg et al. 1990:2). All the autistic subjects were male and had been diagnosed as autistic (DSM-III criteria) prior to the age of 30 months. The DS subjects comprised 4 boys and 2 girls who were matched with the EIA subjects on age and Mean Length of Utterance (MLU).

The abbreviations M, G and F will henceforth be used here to denote the Manchester, Groningen and Flusberg corpora respectively, and FA and FD to indicate the Autistic and Down Syndrome sub-groups respectively within the Flusberg corpus. A final selection was made of four male subjects in each group, with four files for each subject. In the case of the Manchester and Groningen corpora, the files were matched as closely as possible for biological age. The boys in the Flusberg corpora were older, but their MLU scores compared surprisingly closely with those of the toddlers in the corpora for normal children. It must be borne in mind nonetheless that "utterance length is clearly determined by factors beyond the domain of language such as memory capacity and the child's ability to efficiently encode and convey complex information" (Theakston et al. 2001:129). It cannot be assumed that MLUs for English are directly comparable with those for Dutch; and the MLU figures for autistic children need to be taken with a generous helping of salt given the propensity of children with EIA to regurgitate entire sentences ("echolalia"). The final selection of subjects and files is shown in Tables 1.1 - 1.4. [12]

2.3 Procedure

Listing possible pronoun forms for automatic retrieval was a non-trivial task:

... if a researcher wants to trace the use of the pronoun 'you', it might be necessary to search not only for 'you', 'ya' and 'yah', but also for all the assimilations of the pronouns with verbs such as 'didya/dicha/didcha' or 'couldya/couldcha/coucha'.

(MacWhinney 2000a:40)

Boezewinkel's exhaustive inspection of the van Kampen corpus revealed a plethora of idiosyncratic child forms in Dutch, such as Laura's "iekke" and "ikje" for 1st person singular as well as the standard "ik" and well-known 'kindertaal' (child speech) form "ikke" (Boezewinkel 1995:40-41). However, all the above items were cases of hapax legomena and most of the other idiosyncratic forms were also of low frequency. It was therefore considered unnecessary for the present study to conduct an exhaustive search for possible pronouns: wordlists were generated for each child and scanned for likely pronouns but the context was not examined. Any idiosyncratic items found - such as Dominic's "mine-mine" (M) and Daan's "mije" (G) - were added to a file for that child, along with the child's name and various family terms such as "Mummy".

The "COMBO" concordance program was then run over the selected CHAT files to obtain all pronouns or substitute forms used by either the child or the caregiver in each case. Words were counted as "hits" if they matched a word (or wildcard) in the child's pronouns file; if they began with a capital letter, since these would mostly be proper nouns; and in the case of the Manchester corpus, which contained morphological coding, if they were coded as pronouns, proper names or 'possessive determiners'.

Not everything output by this means would, in fact, be a pronoun or pronoun "substitute", and other cases would be ambiguous and would need to be assigned manually to the appropriate person, number, gender and case category. Consequently the COMBO output files were analysed manually and the resulting figures for each pronoun were compiled into tables. Immediate repetitions of the same word or phrase were only counted once.

2.4 Method of analysis

The possible forms available at each "pronoun choice point" - i.e. a point where an adult would be expected to use either a proper name or a pronoun - were given a code such as "1s/n" for "1st person singular nominative", a sample of which is shown in Table 2.1. The full set of codes used can be found in Table 2.2.

In this hypothetical example, the subject has used "I" 10 times where expected, but has also used "me" (1s/a) 3 times where "I" (1s/n) would be expected. This is, of course, merely a fragment of the full matrix showing all possible forms.

For each child and each mother in the sample, for each file, a coding matrix was completed giving all pronouns and pronoun substitutes used, plotted against the expected form. This was entered into an Excel spreadsheet and totals were generated for each subject. The complete spreadsheets can be found in Tables 2.3, 2.4, 2.5, 2.6, 2.7, 2.8, 2.9, 2.10, 2.11, 2.12, 2.13, 2.14, 2.15, 2.16, 2.17 and 2.18; they can also be downloaded as a ZIP file.

'Canonical' and 'deviant' forms were defined in terms of what one would expect an adult to use to another adult. To ensure consistency, a list of coding criteria was drawn up and adhered to, described in detail in Blackwell (2007). Zeroes, representing null forms, were only used for target categories (i.e. pronouns or proper names deemed superfluous), not for actual forms (pronouns deemed to be missing). This was because any search for omitted words could not be automated, and moreover would inevitably be subjective. As MacWhinney (2000a:38) puts it, "The coding of word omissions is an extremely difficult and unreliable process. Many researchers will prefer not to even open up [sic] this particular can of worms."

3. Findings: comparison of CDS used by the mothers

3.1 Overview

All the children, and all the mothers, produced some "deviant" pronouns. Significantly, and in contrast to the received wisdom on this subject, all but one of the normal children in this study produced some pronoun 'reversals' and some made a considerable number of them. Conversely, some of the autistic children scarcely made any. Moreover, all but one of the Down Syndrome children also produced 'reversals'. This is discussed at length in Blackwell (2007). The present paper is only concerned with the speech of the mothers.

3.2 Mothers' deviant pronouns

Using the totals over the four files for each mother, the most common substitutions were listed and are presented below in Tables 3.1 to 3.4 along with their overall frequencies. All substitutions with a frequency greater than 1 have been included. My findings can be compared with those of Wills (1977), though it must be borne in mind that Wills' categories were coarser than mine. [13]

3.3 Manchester

The Manchester mothers exhibited remarkable homogeneity. Substitution of "we" for subject "you" was ranked as the 1st or 2nd most common feature for all four subjects, and substitution of "we" for "I" was ranked 3rd or above. Use of a proper noun such as "Mummy" for "I" was ranked between 1st and 6th. None of this was surprising since these are among the most frequent features of "motherese" documented by Wills (1977). However, another high-frequency phenomenon was unexpected: the substitution of "he" (3s/nm) for subject "it" (3s/nn). This appeared between the 2nd and 7th ranks for all the Manchester mothers.

This feature is mentioned by Wills under the heading "impersonal -> personal" and ranked 9th in her list, which is substantially below the ranking in my data, especially when one considers that each of Wills' categories covers two or more of mine. Wills describes the phenomenon as follows:

The interchangeability of who and what, he or she and it, and 3P and you (addressing things or beings not generally considered to be endowed with participant potential) produces the familiar BT (and poetic) phenomenon of personification.

(Wills 1977:287)

In alluding to the "interchangeability" of "he or she" on the one hand with "it" on the other, Wills overlooks an important aspect: that of gender. Personification by definition applies to inanimate objects which, to an adult, would not normally be thought of as gendered; or to animals (real, toy or depicted ones) whose sex is indeterminate. What is striking here is the fact that they almost always materialised in masculine form: apart from 2 instances of Dominic's mother replacing subject "it" by "she", the only inappropriate uses of third person feminine pronouns by any of the mothers were when they referred to themselves, i.e. they used "she" to replace "I", not "it".

Moreover I would dispute Wills' claim (1977:287) that "these forms are strictly reserved for non-serious situations (e.g. language practice, play, chatting)". This tends to trivialise the consequences of such gendered child-directed speech, which arguably serves a serious educational purpose and can be compared to the kind of interaction that children are likely to have with their teachers on starting school:

What is particularly remarkable in this example is that the mother begins with the correct pronoun "it" but promptly switches to "his" mid-sentence, reinforcing the gender with the "he" in the tag question. Such gendered personification has been commented on anecdotally by feminist linguists since the 1970s (Miller and Swift 1976, Nilson 1973) but to my knowledge has not previously been documented in a corpus-based study of this kind.

Another prominent feature in the CDS of these mothers, ranking 4th or 5th for all of them, was the presence of subject "you" in imperatives, where adult speech would omit it:

The "you" here is not contrastive and is therefore gratuitous. Mothers apparently feel instinctively that a "bare" imperative does not make it sufficiently clear to the child that he is the person expected to carry out the action.

In practice, the "motherese" features tended to fall in clusters: mothers might produce a series of utterances which exhibited no CDS features and then switch into a marked CDS register, with several pronoun substitutions in a single sentence:

In the above utterance "you" is redundant, "Mummy" replaces "me" and "we" replaces "you".

3.4 Groningen

Frequencies for the Dutch data were lower and therefore less reliable, partly because the quantity of data was smaller. There was a further factor: the Manchester mothers tended to use a lot of tag questions in their speech to their sons, e.g.:

Use of tag questions roughly doubles the number of pronouns per sentence, and also permits observation of interesting phenomena such as the switch from "you" to "we" in the example above. Dutch, however, uses invariant tags such as "toch?" and "hè?" which do not include pronouns:

The data sufficed, however, to establish that the Dutch mothers were less homogeneous than their English counterparts. While Abel's mother's CDS pronoun usage was comparable to that of Dominic's mother, the other Dutch mothers made less use of "we" as a substitute for either first or second person singular, preferring to use their own or the child's name respectively:

Nonetheless all the mothers did produce examples of substituting the 1st person plural for both 1st and 2nd persons singular:

In the above example, the child is doing the jigsaw unaided and the "we" must refer to him. In the following example the "we" clearly refers to the mother:

There are some features of the Groningen data which are not shared with the Manchester data. The mothers of Matthijs and Peter both use 3rd person emphatic pronouns in place of 2nd person ones:

In this case both the mother and the researcher refer to the child in the 3rd person.

The "gender issue" described above for the Manchester data does not apply in the same way to the Dutch mothers, because the Dutch language still contains the vestiges of grammatical gender and all nouns are either common gender, with "de" as the definite article, or neuter gender, with article "het". While "de" can be regarded as a gender-neutral determiner, a problem arises when a "de" word has to be referred to by a pronoun: does one use "hij" or "zij"? [14] There is a get-out from this dilemma, in the form of the common gender "die". Notwithstanding considerable use of "die", and the grammatical impossibility of using "hij" or "zij" to point back to a "het" word, the Dutch mothers still produced 2-3 times as many masculine pronouns as gender-neutral ones and, just like their English counterparts, many times more masculine than feminine pronouns.

Use of gratuitous 2nd person nominative in imperatives was present in the speech of two of the mothers:

This appears to be exactly the same phenomenon as that encountered with the Manchester mothers. As in the English data, the redundant "you" often appears when the mother's initial attempts to get the child to do something have not succeeded. Its use, however, was not universal among the Dutch mothers.

3.5 Flusberg Autistic

These mothers resembled their Groningen counterparts in their CDS more closely than the Manchester sample, in that they favoured proper noun substitutions rather than "we". Use of "we" in place of "you" did occur for all mothers, but ranked 5th for Jack's mother and 7th for Rick's. Use of "we" in place of "I" was even less consistent, with Stuart's mother producing no examples at all (and only 1 instance of "us" for "me").

By contrast, gratuitous "you" in imperatives ranked 1st or 2nd for three of the FA mothers, and 6th for the other.

There were, in general, quite high levels of "reversals" between 1st and 2nd person. Most of these are cases of "we", "our", "ours" for "you", "your", "yours" etc.: but not all. There are also numerous cases of mothers prompting their autistic sons by taking their turns for them:

In this example, the mother's prompting has the desired outcome, but this is not always the case. It is likely that the parents of autistic children are well aware of their offspring's propensity for echolalia and try to exploit it as a teaching strategy: they may even be encouraged to do so by the professionals advising them. The problem with this is that the child may have no idea which bit of his mother's speech he is expected to repeat. A good illustration of what can go wrong comes from Brett, who was not among the children included in this study:

Brett obediently repeats his mother's prompt, "will you help me?" but also repeats her response "I will help you". It is doubtful whether autistic children are really learning anything from such rote renditions.

The corpus contains numerous instances of mothers' attempts to prompt their sons backfiring on them. In the following example Mark uses the correct pronoun until his mother confuses him:

The above extract illustrates another common strategy adopted by parents of autistic children, namely writing sentences down as the children utter them. In general this may be helpful, but it can also be unduly constraining. In our next example, Mark's mother has tied herself to using the 3rd person singular and cannot adapt her written version to her son's quite correct use of the 1st person plural in describing what he and his classmates do during fire drills at school:

The efficacy of both the "prompting" and "writing down" strategies is open to question as far as pronoun acquisition is concerned.

3.6 Flusberg Down Syndrome

These were the most surprising findings of all. The gendered personifications ("he" for "it") described above for the Manchester data were more frequent for all the FD mothers than the substitution of "we" for "I", and for all but one of the mothers they were also more frequent than use of "we" for "you". "He" in place of "it" ranked 3rd or above for all subjects.

Perhaps the mothers feel that it is more appropriate to talk to their sons in this way because of their low IQs. However, the FA subjects also have lower IQs than normally-developing children, yet the FA mothers do not engage in this behaviour to anything like the same extent: Stuart's mother does not make any substitutions of masculine pronouns for neuter ones. One possibility is that mothers of autistic children avoid personification because this is a part of pretend play, at which children with EIA are notoriously bad (Wing and Gould 1979; Frith 1991). It is difficult enough to persuade an autistic child to treat a toy car as a car instead of merely spinning its wheels: to expect him to talk about it as if it were alive is a completely unrealistic demand.

The familiar replacement of "you" by "we" is also encountered in this data, particularly from Jerry's mother:

The FD children were found to produce almost as many 1st/2nd pronoun "reversals" as the FA ones, and their mothers perceive this as a problem which requires addressing. As with the FA mothers they prompt their sons for the correct pronouns and feed them utterances to repeat, with varying success:

As with the autistic children, too, the reversals can often be attributed to echolalia:

Martin uses "mine" to mean "yours" and "where mine" to mean "here's yours". This is clearly an instance of slightly delayed echolalia and a failure to grasp the requirements of role alternation in turn-taking.

4. Conclusion

The corpus data revealed interesting differences. While the CDS of the Manchester mothers was remarkably homogeneous, the Groningen mothers exhibited more variation in the kinds of modifications they made. There were differences in style between the four groups of mothers: Dutch "motherese" tends to prefer proper names to 1st person plural forms, while the mothers of autistic children are less prone to personify objects as "he" compared to the mothers of normal children and in particular the mothers of children with DS.

Despite the similarities in pronoun usage in the speech produced by the DS and EIA children, the speech used to them by their mothers differs. This may be partly attributable to the advice being given to them by professionals working with different types of disability, but it is likely that at least some of the differences are due to the mothers' instinctive and perhaps unconscious belief that a particular type of CDS is appropriate. It is possible that mothers of children with learning disabilities entailing low IQs instinctively treat them in a more "childish" way with more personifications of animals and inanimate objects, except in the case of autistic children, whose mothers avoid imposing extra burdens of pretend play on them.

Mothers of children with DS and EIA often try to accommodate to their children's propensity for echolalia. The evidence of this study is that this is a risky strategy which is unlikely to benefit the child's long-term language development. Rote repetition of the correct pronoun form is no substitution for a sound grasp of the "speaker principle" which often eludes children with autism and perhaps DS as well.

Notes

[1] The terms "Baby Talk" (BT), "motherese", "parentese", "caretaker language" and "child-directed speech" have all been used to denote a special register used by adults to address babies and/or young children. While "Baby Talk" tends to be a more specific label, relating mainly to infants and emphasising the "affective" elements of the register, the other terms often appear to be interchangeable.

[2] The diminutive ending "-je" is a productive morpheme even in adult Dutch, so it is difficult to decide whether "tandjes" should be translated with English babytalk "toothie" or not.

[3] This example involves the deletion of the copula as well as the subject "it", but Wills does not comment on this.

[4] The term 'autism', derived from the Greek word for 'self', was first used in a clinical context by Bleuler (1908) to describe one aspect of the behaviour of adult patients with schizophrenia, namely their avoidance of social interaction.

[5] Cf. Asperger's description of one of his patients:

"In Hellmuth's case there were clear indications that his autism was due to brain injury at birth. His medical history - asphyxia, fits, endocrine disorder, hyper-salivation, neurologically based apraxia - clearly pointed to an organic cause." (Asperger [1944] 1991:67).

[6] As an example of autism induced by physical injury, see Wing's "case of the late talker", whose problems apparently began "from the age of six months when his head was accidentally bruised" (Wing 1991:103).

[7] I am following the standard practice in the Child Language literature of denoting a child's biological age by the convention years;months,days.

[8] The subdivisions of MLU stages are taken from Brown (1973).

[9] I am indebted to Leonid Spektor of the CHILDES team for his invaluable assistance in modifying some of the CLAN files to enable them to compile successfully.

[10] Speech from fathers was not available for most data types, and in the case of autism, which afflicts boys roughly four times as commonly as girls (Wing 1981), no female children were available.

[11] The CHILDES database currently contains no data from autistic or DS children acquiring Dutch, which prevented my study from being completely balanced in design.

[12] In the case of the Downs and Autistic subjects, only the age at the time of the first sample are given in Tager-Flusberg and Anderson (1991). The ages at the times of the samples used in this study have been inferred on the basis that the children were recorded at 4-month intervals (Tager-Flusberg and Anderson 1991:1126).

[13] For instance she has a single label "S-> we (let's)" which covers all instances of the parent referring to him- or herself in the 2nd person plural, whereas the matrices produced for the current study enable finer distinctions to be made, e.g. between "we" for "I" and "us" for "me". To make direct comparisons with Wills, some of my categories would need to be aggregated.

[14] The answer for adult Dutch appears to be that, in the absence of "natural" gender cues, it is usually "hij" except when the noun denotes an institution, when "zij" is increasingly used.

[15] The word-order in the Dutch is unnatural here.

Sources

CHILDES = Child Language Data Exchange System, http://childes.psy.cmu.edu

• The CHILDES Database, http://childes.psy.cmu.edu/data/
• CHAT Transcription, http://childes.psy.cmu.edu/manuals/chat.pdf
• CLAN Programs, http://childes.psy.cmu.edu/clan/
• Database Manuals, http://childes.psy.cmu.edu/manuals/

Corpora:

• The Flusberg corpus and other clinical data,
• The Groningen corpus, childes.psy.cmu.edu/data/Dutch/Groningen.zip
• The Manchester corpus, childes.psy.cmu.edu/data/Eng-UK-MOR/Manchester.zip

Max-Planck Institut für Psycholinguistik, http://www.mpi.nl

Short-term memory, in Wikipedia, The Free Encyclopedia, https://en.wikipedia.org/wiki/Short-term_memory

References

American Psychiatric Association. 1987. Diagnostic and statistical manual of mental disorder. 3rd edition, revised (DSM-III-R). Washington, D.C.: American Psychiatric Association.

Asperger, H. 1944. "Die 'Autistischen Psychopathen' im Kindesalter". Archiv für Psychiatrie und Nervenkrankheiten 117: 76-136.

Asperger, H. 1991. "'Autistic psychopathy' in childhood". Translation of Asperger 1944, translated and annotated by Uta Frith. In Frith (ed.), 37-92.

Attwood, T. 1998. Asperger's Syndrome: A guide for parents and professionals. London & Philadelphia: Jessica Kingsley.

Bates, E., I. Bretherson & L. Snyder. 1988. From first words to grammar. New York: Cambridge University Press.

Blackwell, S.A. 2007. Power and Solidarity Revisited: the Acquisition and Use of Personal Pronouns in Modern English and Dutch. Ph.D. dissertation, University of Birmingham.

Bleuler, E. 1908. "Die Prognose der Dementia Praecox - Schziophreniegruppe". Allgemeine Zeitschrift für Psychiatrie 65: 436-464. Translated as "The prognosis of dementia praecox. The group of schizophrenias" in Cutting & Shepherd (eds.), 1987, 59-75.

Bloom, L.M., P. Lightbown & L. Hood. 1975. Structure and Variation in Child Language. (= Society for Research in Child Development Monographs, 40).

Boezewinkel, S. 1995. Van "Mij doen" tot "Haar zusje had twee vriendinnetjes te spelen": Een onderzoek naar de verwerving van pronomina in de Nederlandse kindertaal. "Scriptie" (equivalent to MA dissertation), University of Utrecht.

Bol, G.W. 1996. "Optional subjects in Dutch child language". In Koster & Wijnen (eds.).

Brown, R. 1973. A first language: the early stages. Cambridge, Mass.: Harvard University Press.

Brown, R. 1977. Introduction to Snow & Ferguson (eds.).

Buckley, S. & B. Sacks. 1987. The Adolescent with Down's Syndrome. Portsmouth: Portsmouth Polytechnic.

Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press.

Clahsen, H. 1990. "Constraints on parameter setting. A grammatical analysis on some acquisition stages in German child language". Language Acquisition 1: 361-391.

Clark, H.H. & E.V. Clark. 1977. Psychology and Language: An Introduction to Psycholinguistics. New York: Harcourt Brace Jovanovich.

Cohen, W.I. 1999. "Health Care Guidelines for Individuals with Down Syndrome: Down Syndrome Preventive Medical Checklist". In Hassold & Patterson (eds.), 15-36.

Cross, T.G. 1977. "Mothers' speech adjustments: the contribution of selected child listener variables". In Snow & Ferguson (eds.), 151-188.

Cutting, J. & M. Shepherd, eds. 1987. The clinical roots of the schizophrenia concept. Cambridge: Cambridge University Press.

Dooley, J. 1976. Language acquisition and Down's syndrome: A study of early semantics and syntax. Doctoral thesis, Harvard University, Cambridge, Mass.

Ferguson, C.A. 1977. "Baby Talk as a simplified register". In Snow & Ferguson (eds.), 209-235.

Fowler, A.E. 1995. "Linguistic variability in persons with Down syndrome: research and implications". In Nadel & Rosenthal (eds.), 121-136.

Fowler, A.E. 1999. "The Challenge of Linguistic Mastery in Down Syndrome". In Hassold & Patterson (eds.), 165-184.

Frith, U. 1991. "Asperger and his syndrome". In Frith (ed.), 1-36.

Frith, U., ed. 1991. Autism and Asperger syndrome. Cambridge, New York & Melbourne: Cambridge University Press.

Gallaway, C. & B.J. Richards, eds. 1994. Input and Interaction in Language Acquisition. Cambridge: Cambridge University Press.

van der Geest, T. 1977. "Some Interactional aspects of language acquisition". In Snow & Ferguson (eds.), 89-107.

Hassold, T.J. & D. Patterson, eds. 1999. Down Syndrome: A Promising Future, Together. New York: Wiley-Liss.

Hyams, N. 1992. "The genesis of clausal structure". In Meisel (ed.), 371-400.

Jordan, R.R. 1998. An empirical and theoretical investigation of pronominal development in individuals with autistic spectrum disorders. Ph.D. dissertation, University of Birmingham.

Kanner, L. 1943. "Autistic disturbances of affective contact". Nervous Child 2: 217-50. (Reprinted in Kanner 1973.)

Kanner, L. 1946. "Irrelevant and metaphorical language in early infantile autism". American Journal of Psychiatry 103: 242-246. (Reprinted in Kanner 1973.)

Kanner, L. 1973. Childhood Psychosis: Initial studies and new insights. Washington, D.C.: V.H. Winston & Sons.

Koster, C. & F. Wijnen, eds. 1996. Proceedings of the Groningen Assembly on Language Acquisition. Groningen: Centre for Language and Cognition.

Levelt, W.J.M. 1975. What became of LAD? Lisse, the Netherlands: Peter de Ridder.

MacWhinney, B. 1995. The CHILDES Project: Tools for Analyzing Talk. 2nd edition. Hillsdale, N.J. & Hove, UK: Lawrence Erlbaum Associates.

MacWhinney, B. 2000a. The CHILDES Project: Tools for Analyzing Talk. Volume I: Transcription Format and Programs. 3rd edition. Mahwah, N.J. & London: Lawrence Erlbaum Associates.

MacWhinney, B. 2000b. The CHILDES Project: Tools for Analyzing Talk. Volume II: The Database. 3rd edition. Mahwah, N.J. & London: Lawrence Erlbaum Associates.

MacWhinney, B. & C. Snow. 1990. "The Child Language Data Exchange System: an Update". Journal of Child Language 17: 457-472.

Mattheis, P. 1999. "What New Parents need to Know: Informing Families about their Babies with Down Syndrome". In Hassold & Patterson (eds.), 37-44.

Meisel, J.M., ed. 1992. The Acquisition of verb placement, functional categories and V2 phenomena in language acquisition. Dordrecht: Kluwer.

Miller, C. & K. Swift. 1976. Words and Women: New Language in New Times. Harmondsworth: Penguin.

Moerk, E.L. 1976. "Processes of language teaching and training in the interaction of mother-child dyads". Child Development 47: 1064-78.

Moerk, E.L. 1983. The mother of Eve - as a first language teacher. Norwood, N.J.: Ablex.

Nadel, L. & D. Rosenthal, eds. 1995. Down Syndrome: Living and Learning in the Community. New York: Wiley-Liss.

Nilson, A.P. 1973. Grammatical Gender and its Relationship to the Equal Treatment of Males and Females in Children's Books. Ph.D. thesis, University of Iowa.

Oshima-Takane, Y & S. Benaroya. 1989. "An alternative view of pronominal errors in autistic children". Journal of Autism and Developmental Disorders 19(1): 73-85.

Patterson, D. 1999. "Understanding the Importance of Genes on Chromosome 21". In Hassold & Patterson (eds.), 75-78.

Pine, J.M. 1994. "The language of primary caregivers". In Gallaway & Richards (eds.), 15-37.

Rondal, J. 1994. "Exceptional cases of language development in mental retardation: the relative autonomy of language as a cognitive system". In Tager-Flusberg (ed.), 155-174.

Rondal, J.A. 1995. Exceptional Language Development in Down Syndrome. A Case Study and Its Implications for Cognition-Language and Other Language-modularity Issues. New York: Cambridge University Press.

Rutter, M. & E. Schopler. 1987. "Autism and pervasive developmental disorders: concepts and diagnostic issues". Journal of Autism and Developmental Disorders 17: 159-86.

Snow, C.E. & C.A. Ferguson, eds. 1977. Talking to Children: Language Input and Acquisition. Cambridge: Cambridge University Press.

Sokolov, J. & C. Snow, eds. 1994. Handbook of research in language development using CHILDES. Hillsdale, N.J.: Erlbaum Associates.

Tager-Flusberg, H., ed. 1994. Constraints on Language Development: Studies of Atypical Children. Hillsdale, N.J.: Erlbaum Associates.

Tager-Flusberg, H. & M. Anderson. 1991. "The Development of Contingent Discourse Ability in Autistic Childen". Journal of Child Psychology 2 (77): 1123-1134.

Tager-Flusberg, H., S. Calkins, T. Nolin, T. Bamberger, M. Anderson & A. Chandwick-Dias. 1990. "A longitudinal study of language acquisition in autistic and Down Syndrome children". Journal of Autism and Developmental Disorders 20: 1-21.

Theakston, A.L., E.V.M. Lieven, J.M. Pine & C.F. Rowland. 2001. "The role of performance limitations in the acquisition of verb-argument structure: an alternative account". Journal of Child Language 28: 127-152.

Wang, P.P. 1996. "A Neuropsychological profile of Down syndrome: cognitive skills and brain morphology". Mental Retardation and Developmental Disabilities Research Review 2: 102-108.

Wills, D.D. 1977. "Participant deixis in English and baby talk". In Snow & Ferguson (eds.), 271-298.

Wing, L. 1981. "Sex ratios in early childhood autism and related conditions". Psychiatry Research 5(2): 129-37.

Wing, L. 1991. "The relationship between Asperger's syndrome and Kanner's autism". In Frith (ed.), 93-121.

Wing, L. & J. Gould. 1978. "Systematic recording of behaviours and skills of retarded and psychotic children". Journal of Autism and Childhood Schizophrenia 8: 79-97.

Wing, L. & J. Gould. 1979. "Severe impairments of social interaction and associated abnormalities in children: epidemiology and classification". Journal of Autism and Childhood Schizophrenia 9: 11-29.