Language in Down Syndrome: Delay of Principle A Effect?*

Alexandra Perovic Durham Working Papers in Linguistics Volume 8. 2002. pp. 97-110.

Reproduced with the permission of the author: Alexandra Perovic Department of Phonetics and Linguistics University College London Gower Street London WC1E 6BT

Abstract

1. Introduction

   Research on Down syndrome (DS) has uncovered an unusual disparity between linguistic and cognitive development, with linguistic development significantly lagging behind. Further dissociations have been reported within the linguistic faculty itself, particularly between morphosyntax on the one hand, and lexical knowledge and pragmatics on the other. In an attempt to further elucidate the relationship between different linguistic modules in the selective language impairment in DS, this study of the language of four adolescent girls with DS focuses on Binding Principles. In standard Binding Theory (BT), Binding Principle A governs the distribution and interpretation of reflexives, whereas Binding Principle B is concerned with pronouns. It is well known that typically developing children acquire Principle A early and with few difficulties, whilst their acquisition of Principle B is significantly delayed – the phenomenon known as Delay of Principle B Effect (DPBE). If language development in DS is, as traditionally thought, delayed, but essentially normal, then investigations of the availability of the Binding Principles in DS should demonstrate parallels to normal language development – with Principle A posing few problems but Principle B yielding interpretive difficulties until later stages of language development. The results of this study, however, point to a rather different pattern: in contrast to typically developing children, as compared to data from Chien & Wexler (1990), the subjects violated Principle A but obeyed Principle B.

   In line with the 'delayed' characterisation of language in DS, it may be reasonable to claim that these findings are due to some kind of delay in the acquisition of a particular syntactic principle, thus revealing a 'Delay of Principle A Effect' in this population. However, on the basis of the accounts for the DPBE in typical children, I argue that a satisfactory account of these findings cannot be provided within the framework of standard BT. I argue that the pattern shown in DS is not caused by the unavailability of a Binding Principle but rather a specific deficiency in establishing binding relations. The proposed dissociation between binding, as the expression of referential dependencies, and the ability to establish the syntactic relation of binding in DS, can be accounted for within the framework of Reflexivity of Reinhart & Reuland (1993).

   The comprehension pattern on pronouns, as opposed to reflexives, presented by the four girls with DS in this study has not been evidenced at any stage of typical language development, thus providing further evidence against the claim that language development in DS is severely delayed but essentially non-deviant.

   
   
2. Language in Down syndrome: Delay or deficiency?

   Comparisons of the linguistic development of DS children with that of typical children usually come to the conclusion that the two populations follow the same course of development (Chapman, 1995; Fowler, 1990). It has been argued that DS children acquire vocabulary, use the same range of grammatical morphemes and syntactic structures, induce grammatical rules and impose word order just like typically developing children, albeit with a considerable delay (Fowler, 1990; Rutter & Buckley, 1994; Vicari, Caselli & Tonucci, 2000). However, this 'delayed' characterisation does not seem adequate to describe the course of linguistic development, and particularly the end linguistic achievement in DS. Interestingly, disparities between linguistic and non-linguistic abilities in DS seem to increase with chronological age. DS children up to 3 or 4 years of age have been found to have language skills consistent with their cognitive abilities. As they get older, however, their language skills do not increase at comparable rates to other cognitive skills (Miller, 1988; Chapman, 1995). Fowler, Gelman & Gleitman (1994) report the average mean length of utterance for a group of DS adolescents to be around 3, but their mental age (MA) to be around 6 years.¹ Moreover, Down syndrome is found to be more detrimental to certain aspects of language development than other aetiologies, e.g. autism (Tager-Flusberg & Calkins, 1990). In studies conducted with young adolescents and adults, individuals with DS show consistently poorer performance on linguistic measures to individuals of unknown aetiology of intellectual disorder (Kernan & Sabsey, 1996; Marcell, Ridgeway, Sewell & Whelan, 1995).

   Disparity between language and cognition in DS is further reflected in the interaction of distinct linguistic modules. Inconsistent use and widespread omission of grammatical morphemes such as articles, auxiliaries, copulas, pronouns, conjunctions, prepositions, verbal and nominal inflection have been widely reported (Chapman, Schwartz & Kay-Raining Bird, 1998; Fabretti, Pizzuto, Vicari & Volterra, 1997; Vicari et al 2000), and DS syntax is usually limited to only the simplest structures: constructions involving passives, subject/auxiliary inversion, possessive forms, negation, interrogatives are rarely used by either children or adults (Fowler, 1990; Rondal, 1995). Phonological development in this population is also fraught with difficulties, with processes such as final consonant deletion, consonant cluster reduction, substitution and omission greatly reducing the intelligibility of speech of both children and adults (Dodd, 1976). In contrast, receptive vocabulary and pragmatics ha ve been reported to be relatively less impaired, and, interestingly, rather consistent with general cognitive levels of this population (Chapman, 1995; Rondal & Comblain, 1996). Miller (1988) reports measures of receptive vocabulary as correlating with measures of mental age in children with DS at various stages of development, in line with the argument that mental age measures successfully predict lexical development in both intellectually disabled and typically developing children. In other words, modules of the computational system (e.g. morphosyntax and phonology) appear more severely impaired in DS than more general, multi-modality processing systems (e.g. lexical knowledge and pragmatics).

   The strikingly low ultimate level of linguistic attainment, comparable to typically developing 2-year-olds (Fowler, Gelman & Gleitman, 1994), and the known disparities in the development of computational vs. conceptual systems, suggest that the language of DS is more than just delayed. Recall that the definition of 'delay' in the language in DS is essentially based on comparisons of linguistic patterns in DS with those found in typical acquisition. However, parallels with typical development are expected if we assume that human language, impaired or not, is constrained by Universal Grammar. The same argument holds for language disorders in general: constraints postulated by UG will restrict the logical number of possible deficiencies whatever the linguistic impairment.

   To determine whether linguistic impairment in DS is merely delayed, or also deficient in important respects, I investigate the end syntactic achievement of four adolescents with DS, in particular, their knowledge of Binding. Binding belongs to the core computational system of the language faculty and constitutes a major part in adult syntactic knowledge. However, recent developments in the theory have suggested an interesting interplay of syntactic and pragmatic factors that influence the process of the acquisition of this module, resulting in apparent delays in acquiring a Binding Principle, thus making it particularly interesting with respect to the delay vs. deficiency argument regarding linguistic development in DS.

   
   
3. Binding Theory and typical language development

   Principles A and B² of Binding Theory of Chomsky (1981, 1986) regulate the distribution and interpretation of nominal expressions—anaphors and pronouns, respectively—within a particular sentence domain, therefore permitting and excluding the constructions in (2) and (3):

   
   
   1. Principle A: an anaphor must be locally bound.³
      Principle B: a pronoun must be locally free.
   2. Mary_i is washing herself_i/*j
   3. Mary_i is washing her_*i/j
   
   

   The theory of Universal Grammar (UG) entails that Binding Principles are innately specified. However, it comes as a surprise that in the process of acquisition children show distinctions between Principles A and B, obeying Principle A very early, from around age 4, but violating Principle B even after the age of 5 or 6. A number of studies have shown that children would accept (4) as grammatical around 50% of the time:

   
   
   4. * Mary_i is washing her_i
   
   

   This phenomenon, often referred to as 'Delay of Principle B Effect', is reported in a variety of languages: English (Jakubowicz, 1984; Chien & Wexler, 1990), Dutch (Philip & Coopmans, 1996), Russian (Avrutin & Wexler, 1992), Icelandic (Sigurjónsdóttir, 1992).⁴ To circumvent the empirical problem that children's apparent violations of Principle B pose to the central claim of UG, namely that knowledge of syntactic principles is innate, researchers have argued that children do have knowledge of this particular syntactic principle but their performance on the reported tasks is masked by other factors. Rather than attributing it to a violation of a syntactic principle, Chien & Wexler (1990), Grodzinsky & Reinhart (1993), argue that children's errors on (4) are due to the immaturity of their pragmatic and/or general processing systems, rather than a failure within the syntax proper.

   Arguing for the idea that binding must be separated from coreference,⁵ the above accounts of the DPBE suggest that children's errors are due to their inability to rule out the coreference reading in illicit contexts such as (4).⁶ The innate knowledge of a syntactic principle, such as Principle B, predicts that children will not allow the bound variable reading. However, coreference is governed by some constraint outside syntax proper, not available to young children. This constraint is pragmatic in nature, and its role is to block the coreference interpretation whenever it is semantically indistinguishable from the bound variable interpretation. Due to the immaturity of children's pragmatic systems (Principle P, Chien & Wexler, 1990)⁷, or limitations on processing systems that hinder coreference computation operations (Rule I, Grodzinsky & Reinhart 1993)⁸, this constraint cannot be implemented in children's grammar, resulting in the application of guesswork to examples like (4) above.

   The important issue that arises in the accounts of the DPBE presented here is the proposed fractionation of Binding Principles into linguistic and extralinguistic components. Processes involved in the interpretation of anaphoric elements are constrained to syntax proper. In contrast, coreferential interpretation involves processes that relate linguistic expressions to elements outside grammar. This division of labour between syntax and pragmatics in the interpretation of pronominals has interesting implications for our exploration of the linguistic deficit in Down syndrome. If binding is syntactically encoded, and it is assumed that the linguistic deficit in DS is syntactic in nature, the subjects may reveal difficulties with interpreting anaphoric elements whose distribution is constrained by syntactic principles. If the interpretation and distribution of pronouns are regulated by principles that belong to some system outside syntax proper, influenced by the maturation of general processing abilities, the subjects should do better than typically developing children on tasks involving pronouns. However, this prediction runs against the 'delay' characterisation of the language in DS, discussed in previous sections. If children with DS go through identical stages of acquiring syntactic principles to typically developing children, even if these processes get arrested at distinct points in the development, it should be possible to identify at least some stages in the grammar of adults with DS that match the patterns of typical acquisition processes.

   
   
4. Experiment

   4.1. Subjects

   Four adolescent girls with DS, between 17 and 21 years of age, participated in the study. All subjects were students at a Learning Support Unit at an institute of further education in Greater London. Their scores on standardized grammar and vocabulary tests are given in Table 1. Note the disparity between their scores on vocabulary and the test of comprehension of grammar, in line with dissociations between grammar and vocabulary reported in the literature. Scores on British Picture Vocabulary Scales (BPVS), receptive vocabulary, are used as an indicator of verbal mental age for each of the subjects, following the tradition in the literature.

   
   

   Table 1: DS subjects' scores on standardised vocabulary and grammar tests

   Subjects	Age   in years	Receptive Vocabulary   (BPVS) Age Equivalent	Expressive Vocabulary   (RWFVT)9 Age Equivalent	Grammar Comprehension   (TROG)10 Age Equivalent
   LP	17;09	8;01	7;03-05	4;09
   DA	17;02	6;06	5;07-09	5;03
   MK	19;03	7;0	>8.6	4;05
   SL	20;07	5;00	5;07-09	4;0

   4.2. Materials and procedure

   The task used in the study was the Picture truth-value judgement task, adapted from Chien & Wexler (1990),¹¹ eliciting yes-no answers to experimental questions accompanying the picture stimuli which matched or did not match the picture shown.¹² Following Chien & Wexler (1990), four experimental conditions were included: name-reflexive, name-pronoun, quantifier-reflexive and quantifier-pronoun, with eight questions for each condition. In addition to control conditions used in Chien & Wexler (1990), name-name and quantifier-name, two extra control conditions were included: name-name action and attention, with eight¹³ questions for each, bar the attention control condition, which included 16 questions. Examples of questions for each of the conditions are given in the Appendix A. Subjects were presented with a picture showing cartoon characters drying, washing or touching either themselves or other characters. A sentence introduced the characters (e.g. 'This is Cinderella. This is Snow White.') followed by a question (e.g. 'Is Cinderella washing her?'). In the four experimental sessions, each subject was presented with 30 pictures and 30 questions. To ensure that subjects understood the task, four trial questions were used at the beginning of each session. The sessions were conducted in an empty classroom at the girls' college.

   
   
5. Results

   Table 2 presents the percentage of correct responses for the subjects with DS on each of the experimental conditions.¹⁴

   
   

   Table 2: DS subjects' performance on experimental conditions

   	LP	DA	MK	SL
   Name-reflexive
   Match NRM	25.00%	75.00%	62.50%	75.00%
   Mismatch NRX	12.50%	37.50%	75.00%	100%
   Quantifier-reflexive
   Match NRM	12.50%	50.00%	12.50%	25.00%
   Mismatch NRX	12.50%	25.00%	87.50%	62.50%
   Name-pronoun
   Match NRM	100.00%	100.00%	100.00%	100.00%
   Mismatch NRX	100.00%	100.00%	100.00%	100.00%
   Quantifier-pronoun
   Match NRM	100.00%	100.00%	750.00%	100.00%
   Mismatch NRX	100.00%	100.00%	100.00%	100.00%

   
   

   On conditions involving pronouns (NPM, NPX, QPM, QPX), the subjects performed at ceiling: LP, DA and SL correctly rejected locally bound pronouns in the mismatch condition and accepted a referent distinct from the local subject for the pronoun in the match condition 100% of the time. MK's performance was also 100% correct on conditions NPM, NPX and QPX, and with 75%, slightly worse, but still above chance (p= 0.1445)¹⁵ on condition QPM.

   Their performance is strikingly different on conditions that involve reflexives. All subjects performed below chance on at least one (match or mismatch) condition, revealing a systematic misinterpretation of these constructions. LP performed below chance on all four conditions involving reflexives, match and mismatch (p<0.05 for NRM, p=0.1445 for NRX, QRM and QRX). DA attained above chance performance on NRM, but scored below chance on the mismatch type of the same condition, NRX (p=0.3632). Her performance on reflexives bound by quantified NPs is much poorer: at chance for QRM and below chance (p=0.1445) for QRX. MK and SK scored above chance on the conditions with reflexives bound by a reflexive antecedent, however, their performance on reflexives bound by a quantified antecedent is poorer: significantly below chance for the match type of this condition (p<0.05 for MK, p=0.1445 for SL), and (slightly) above chance for the mismatch type of that condition. Their performance on control conditions was at ceiling; see Appendix B for exact percentages.¹⁶

   Scores on the same experimental task of typically developing children in various age groups, as reported in Chien & Wexler (1990), are given in Table 3 for comparison.

   
   

   Table 3: Performance of typically developing children on experimental conditions (Chien & Wexler, 1990)

   	Group 1 (n=48)  age: < 4	Group 2 (n=45)  age: 4-5	Group 3 (n=44)  age: 5-6	Group 4 (n=40)  age: 6-7
   Name-reflexive
   Match NRM	79.51%	92.96%	96.97%	96.67%
   Mismatch NRX	30.56%	67.04%	92.80%	99.17%
   Quantifier-reflexive
   Match NRM	77.43%	76.67%	89.39%	94.17%
   Mismatch NRX	29.51%	40.74%	82.95%	84.58%
   Name-pronoun
   Match NRM	91.67%	88.52%	90.15%	94.58%
   Mismatch NRX	30.90%	39.26%	49.24%	76.67%
   Quantifier-pronoun
   Match NRM	88.54%	94.44%	97.93%	98.75%
   Mismatch NRX	46.88%	60.00%	83.71%	86.67%

   
   

   The large disparity between the scores on match and mismatch type on each of the experimental conditions suggests the well known positive bias, thus scores in the mismatch conditions are more likely to be informative. The youngest children (Group 1) performed below chance, or at chance, on all the mismatch conditions. The difference in the performance on pronouns and reflexives when bound by a referential antecedent ('Delay of Principle B Effect') becomes apparent from the age of 4, and persists even at the age 6-7. Group 2 performed better on the condition NRX than on condition NPX. Group 3 had generally better scores on both conditions, but their performance on NPX still points to guesswork. The oldest children achieved highest scores on all conditions, except, again, the condition NPX. Note that children in Groups 2, 3 and 4 performed significantly better on pronouns when bound by a quantified NP (condition QPX), providing support for the argument that children have knowledge of the syntactic principle which rules out local binding of pronouns when the bound variable reading is available (see footnote 6).

   Following the tradition in the literature, three subjects (LP, MK and SL) can be matched to controls in Group 2, on the basis of their scores on the test of the comprehension of grammar (TROG). On the same measure, DA can be matched to controls in Group 3. On the basis of their MA scores, as measured by the test on receptive vocabulary (BPVS), the subjects could be matched to the eldest controls in Group 4. However, regardless of the matching method, none of the groups of controls exhibits the pattern evidenced in the subjects with DS.

   
   
6. Discussion

   The experiment reported here revealed that four girls with DS performed significantly worse on anaphors as opposed to pronouns, revealing a pattern opposite to the well known 'Delay of Principle B Effect' (DPBE) discussed earlier. Such a pattern is not documented at any stage of typical language development, as reported in Chien & Wexler (1990). How can these findings be accounted for?

   One possibility would be to argue for a specific deficiency in the grammar of individuals with DS, which causes a 'delay' in the acquisition of a syntactic principle such as Principle A of standard BT (revealing a "Delay of Principle A Effect"), in line with the 'slow but normal' characterisation of linguistic development in DS discussed earlier. This however appears to be a mere reformulation of the problem. It is not clear why individuals with such a grammar would obey one grammatical principle, and not the other, if both principles underlie the same grammatical knowledge of Binding Theory (BT). Recall that similar arguments were put forward in the accounts of the DPBE in typical language acquisition. Furthermore, if Principle A was unavailable in the grammar of DS, the subjects would be expected to rule in all the sentences violating the Principle, and not to reject any constructions with reflexives as ungrammatical. Our subjects' performance did not show such a pattern.

   The starting point in developing an explanation for the findings reported above will be the same problem faced by researchers accounting for the DPBE in typical language development. Recall that the acquisition data discussed in section 3 exposed the inability of standard BT to account for the distinction between binding and coreference. In this framework, interpretive dependencies crucially rely on structural conditions on coindexing of nominal expressions, with Binding Principles presupposing binding relations. However, this claim cannot hold with respect to constructions such as (5), where pragmatic context makes it possible for the two NPs to corefer without being in a binding relation. Yet in the normal case, such relations must also be blocked by Principle A.

   
   
   5. I know what Bill and Mary have in common. Mary adores Bill and Bill adores him too.
   
   

   In the alternative framework of Binding Theory, as developed by Reinhart & Reuland (1993) (subsequently R&R), this problem does not arise: interpretive dependencies do not necessarily coincide with the syntactic relations nominal expressions enter into with their antecedents. Binding Principles, as given in R&R, can have locality effects even if no binding relation is established. A hypothesis I shall explore here is that difficulties with the comprehension of anaphors demonstrated by the subjects in the experimental task in fact reveal a deficit in establishing certain syntactic dependencies, and not with interpretive dependencies. To put it differently, my claim will be that binding, as the expression of referential dependencies, is available in the grammar of individuals with DS, but binding relations are not.¹⁷ The Reflexivity model therefore provides tools to exp lore the apparent dissociation between binding principles and the syntactic relation of binding in DS. Independent evidence for the claim that these subjects may have a deficiency in establishing binding relations comes from their inability to form A-dependencies in passive constructions. Before attempting to analyse the experimental data, it is necessary to briefly outline some of the concepts of R&R's Reflexivity framework relevant to our discussion.

   6.1. Reflexivity (Reinhart & Reuland, 1993)

   The central notion of reflexivity refers to the idea that to be reflexive, a predicate must have two of its arguments covalued. The reinterpreted principles of Binding Theory rely on the assumption that reflexivity must be linguistically licensed ('reflexive-marked'), either in the lexicon (with the head of the predicate being reflexive-marked) or in syntax (with one of its arguments being the complex anaphor self). The standard Binding Principles A and B of Chomsky (1981) are thus replaced with the following conditions on reflexivity:

   
   
   6. Principle A: A reflexive marked syntactic predicate must be reflexive.
      Principle B: A reflexive semantic predicate must be reflexive marked.
   
   

   The important consequence of Binding Conditions, as stated in R&R, is that they account for the distribution of anaphors and (together with the revised chain theory) pronouns, without stating any restrictions on their structural domains.

   In English and Dutch the complex anaphor -self/-zelf appears as an argument of a simple transitive verb that has become reflexivized, with Condition B ruling out both the simplex anaphor (zich in Dutch; no equivalent in English) and the pronoun:

   
   
   7. Johni hates himself_i/*zich_i/*him_i
   8. Jan_i haat zichzelf_i/*zich_i/*hem_i
      John hates himself/*zich/*him
   
   

   Intrinsically reflexive predicates are marked so in the lexicon, as part of verb's functional semantics. In English, they are intransitive,¹⁸ whereas in Dutch and other languages that have a contrast between a complex and a simplex anaphor, inherently reflexive predicates overtly realise their internal Θ-role in the form of a simplex anaphor:

   
   
   9. John_i behaves_i *him.
   10. Jan_i gedraagt zich_i/*zichzelf_i/*hem_i.
       John behaves zich/*himself/*him
   
   

   Since the predicate in (9), (10) is inherently reflexive, there is no need to reflexive- mark it in syntax by the complex anaphor; the conditions on reflexivity are satisfied. The revised chain condition excludes pronouns from the same context.¹⁹ However, in absence of the simplex anaphor, the anaphoric systems of some languages allow pronouns to appear as syntactic arguments of inherently reflexive verbs, in the same way as simplex anaphors do. Frisian displays such a pattern:

   
   
   11. Max_i hâld him_i/*himsels_i²⁰
       Max behaves him/*himself
   
   

   The Reflexivity framework also copes with the long-standing issue of complex anaphors occurring in non- local domains. In the following constructions, the anaphor is not a syntactic argument of the predicate, so Condition A as given in (6) is not violated:

   
   
   12. John_i pulled the blanket over himself_i/him_i.
   13. Physicists like yourself are a godsend.

   6.2. Analysis and predictions

   The hypothesis presented above was that DS subjects will not be able to establish binding relations, but they will show knowledge of conditions on reflexivity as defined in (6). Applying R&R's framework to the present results, it is clear that these subjects showed difficulties understanding constructions involving a transitive predicate reflexive-marked by the complex anaphor:²¹

   
   
   14. Snow Whitei is drying herselfi.
   
   

   The poor performance on the reflexive structures supports the above hypothesis: DS subjects do not seem to be able to use the self anaphor as an argument of a reflexive predicate, because they are not able to bind it in syntax. They may very well be aware of the special function of the anaphor self, namely to impose reflexivization on the predicate, but they cannot establish the syntactic relation between the anaphor and its antecedent. Independent evidence for a specific deficiency in forming A-dependencies comes from reports on the production and comprehension of passives: passive constructions are known to be extremely problematic for the DS population (Fowler, 1990). The results reported here also support this claim: all four subjects failed the task on passives on the test of comprehension of grammar (TROG).

   The hypothesis predicts that DS subjects would, however, show knowledge of the conditions on reflexivity. This is indeed shown in their performance on experimental conditions involving a pronoun as one of the covalued arguments of a transitive predicate that has not been reflexive marked in syntax, in violation of Condition B, as stated in (6). The subjects correctly rejected violations of Condition B nearly 100% of the time:

   
   
   15. *Snow White_i is drying her_i.
   
   

   If this analysis is on the right track, it should be possible to predict how a deficit in establishing syntactic dependencies could affect the general use of pronominals in reflexive and non-reflexive predicates in DS. The following predictions may be made:

   1. Production and comprehension of inherently reflexive verbs in English should pose no difficulty. These verbs are intransitive and no establishing of binding relations is needed (cf. John shaves every day).
   2. Inherently reflexive verbs in languages like Frisian, where a pronoun is used as the argument of the predicate, would be used and comprehended without difficulty.
   3. Transitive predicates reflexive-marked by the complex anaphor self should pose difficulties in languages other than English.
   4. Logophoric use of anaphors, in constructions such as (12), (13) should not be problematic. In these constructions no A-dependency is established.
   5. In languages that use a simplex anaphor with inherently reflexive verbs, such as Dutch, distinct patterns in the use and comprehension of these verbs may be expected. Being unable to establish syntactic dependencies, individuals with DS would show a poor comprehension of reflexive structures, in line with the data reported here. However, when using inherently reflexive predicates, they may resort to whatever is available in their anaphoric system to realise the syntactic argument of the inherently reflexive predicate, and use a locally bound pronoun in place of the anaphor, in line with the pattern evidenced in Frisian.
   
   
7. Conclusions

   The data reported here provide evidence against the claim that language development in DS is severely delayed but essentially non-deviant. The comprehension pattern for anaphors as opposed to pronouns evidenced by the four girls with DS tested in this study is not confirmed in typically developing children at any stage of linguistic development. The DS subjects' performance on experimental conditions involving reflexives and pronouns differ from the performance of typical children in two crucial respects:

   1. Poor performance on experimental conditions involving anaphoric binding points to a syntactic deficit in the grammar of DS, related to the inability to form syntactic dependencies. In contrast, typical children are found to correctly reject locally bound anaphors at early ages of language acquisition.
   2. Faultless performance by the DS subjects on conditions involving pronouns demonstrates knowledge of Binding Principles (as given in the Reflexivity framework of R&R, as well as the pragmatic/general inferential constraint such as Principle P/Rule I (Chien & Wexler, 1990; Grodzinsky & Reinhart, 1993) which rules out illicit coreferential interpretations of pronouns. The latter constraint is argued to be unavailable to young children, forcing them to perform at chance level on constructions involving pronouns bound by referential antecedents.

1. Note, however, that this population displays extreme variability in the levels of linguistic achievement, which cannot be accounted for by comparison with cognitive abilities (Miller, 1988; Rondal, 1995). Cases of nearnormal linguistic complexity have also been reported (Rondal, 1995).
2. Principle C will not be discussed here.
3. a binds b iff a and b are coindexed and a c-commands b. A precise definition of 'locally bound' need not concern us here.
4. See Baauw (2000), McKee (1992) and Varlokosta (2001) for claims that DPBE does not occur in languages with rich inflectional morphology, and especially, clitic doubling: Spanish, Italian and Greek.
5. In standard BT, Binding Principles presuppose binding relations. This is not necessarily the case in (i) below: here the two NPs corefer, despite violating the syntactic restrictions on binding:
   1. I know what Bill and Mary have in common. Mary adores Bill and Bill adores him too.
      Assuming that definite descriptions and proper names also can serve as variable binders, Grodzinsky & Reinhart (1993) claim that ambiguity between the coreferential and bound variable reading is present even in contexts such as (4):
   2. 1. Mary is washing her.
      2. Mary λx (x is washing a)
      3. Mary λx (x is washing x)
6. Support for this view is provided by the acquisition data in Chien & Wexler (1990): in constructions where no ambiguity is available between the coreferential and bound variable interpretation, children make few errors:
   1. Every_i bear washed him_i.
7. Chien and Wexler (1990) propose a non-syntactic principle, Principle P, which rules out the coreferential interpretation of contraindexed NPs unless it is forced by pragmatic context.
8. Rule I: Intransentential Coreference: NP A cannot corefer with NP B if replacing A with C, C a variable Abound by B, yields an indistinguishable interpretation.
9. RWFVT (Renfrew Word Finding Vocabulary Test) assesses expressive vocabulary.
10. TROG (Test for the Reception of Grammar) evaluates comprehension of a range of grammatical constructs, e.g. plural, tense, passives, subordination.
11. I acknowledge the help of Heather van der Lely who kindly provided some of the pictures used in van der Lely & Stollwerck (1997).
12. Mismatch conditions were included to control for a positive bias (opting for 'yes' answers even if unsure, in order to please the (adult) experimenter), well evidenced in the research on typical language acquisition.
13. Chien & Wexler (1990) used six questions per experimental condition.
14. Percentages of correct responses for the control conditions are given in the Appendix B.
15. The p values were calculated using a binomial distribution, on the assumption that subjects were guessing in a random, unbiased way.
16. Two of the subjects, LP and SL showed very poor performance on the control condition involving quantified NPs, mismatch type (QNX). It would be beyond the scope of this paper to explore the possible answers to why these constructions were problematic for our subjects. It is, however, clear that this problem did not hinder their faultless performance on pronouns bound by quantified NPs.
17. It is more than likely that the linguistic impairment in DS is caused by an interplay of a variety of different factors. However, here we begin by assuming that only one particular problem is the cause of the deficiency, and assume that the rest of the grammar is as given by UG.
18. Some English verbs seem ambiguous between being inherently reflexive and purely transitive predicates (e.g. shave, wash). When inherently reflexive, they are intransitive and appear with one argument only (John shaves). When transitive, they can take an object distinct in reference from the subject (John shaved Peter), or become reflexivized just like any other pure transitive predicate, with their internal argument appearing in the form of a complex anaphor (John_i shaves himself_i). Each occurrence of the predicate is listed separately in the lexicon.
    Following the design of Chien & Wexler (1990), wash was one of the verbs used in the experiment. We assume that our subjects treated it as a transitive predicate, as no differences were found in their performance on this verb as opposed to dry and touch, genuine transitive verbs.
19. Arguing that any sequence of coindexed elements meeting the restrictions on government and c-command is a chain, R&R claim that only the head of the chain must be referential, not the tail. Having referential content, pronouns are excluded from the tail position.
20. An explanation for why zich is generally preferred to pronouns in appearing with inherently reflexive predicates has been proposed in terms of economy effects (Reuland, 1997): anaphoric binding is an option cheaper than variable binding. When the cheapest option is not available, the anaphoric system of a language will have to opt for the use of pronouns.
21. It is important to note that some of the subjects did not just provide incorrect answers to the experimental questions involving transitive predicates reflexive-marked by the self anaphor: DA often could not provide any answer (subsequently coded as incorrect), showing a great deal of hesitation and frustration as a result. This seems to reveal the inability to bind the anaphor to the antecedent, with the anaphor remaining uninterpreted.

Appendix A: Example sentences

Appendix B: DS subjects' performance on control conditions

References

Avrutin, S. & Wexler, K. (1992). Development of Principle B in Russian: Coindexation at LF and coreference. Language Acquisition 2. 259-306.

Baauw, S. (2000). Grammatical features and the acquisition of reference: A comparative study of Dutch and Spanish. LOT: Utrecht Institute of Linguistics.

Bol, G. & Kuiken, F. (1990). Grammatical analysis of developmental language disorders: a study of the morphosyntax of children with specific language disorders, with hearing impairment and with Down's syndrome. Clinical Linguistics and Phonetics 4.1. 77- 86.

Chapman, R. (1995). Language development in children and adolescents with Down syndrome. In Fletcher, P. & MacWhinney, B. (eds.), The handbook of child language. Oxford: Blackwell.

Chapman, R., Schwartz, S. & Kay-Raining Bird, E. (1991). Language skills of children and adolescents with Down syndrome: I. Comprehension. Journal of Speech and Hearing Research 34. 1106-1120.

Chapman, R., Seung, H-K., Schwartz S. & Kay-Raining Bird, E. (1998). Language skills of children and adolescents with Down Syndrome: II. Production. Journal of Speech and Hearing Research 41. 861-873.

Chien Y-C., & Wexler, K. (1990). Children's knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics. Language Acquisition 1. 225-295.

Chomsky, N. (1981). Lectures on government and binding. Dordrecht: Foris.

Chomsky, N. (1986). Knowledge of language: Its nature, origin and use. New York: Praeger.

Dodd, B. (1976). A comparison of the phonological system of mental age matched normal, severely subnormal and Down's syndrome children. British Journal of Disorders of Communication 11. 27-42.

Fabretti, D., Pizzuto, E., Vicari, S. & Volterra, V. (1997). A story description task in children with Down's syndrome: lexical and morphosyntactic abilities. Journal of Intellectual Disability Research 41. 165-179.

Facon, B., Grubar, J-C. & Gardez, C. (1998). Chronological age and receptive vocabulary of persons with Down syndrome. Psychological Reports 82. 723-726.

Fowler, A. (1988). Determinants of rate of language growth in children with Down syndrome. In Nadel, L. (ed.), The psychobiology of Down syndrome. Cambridge, MA: MIT Press.

Fowler, A. (1990). Language abilities in children with Down syndrome: evidence for a specific delay. In Cicchetti, D. & Beeghley, M. (eds.), Children with Down syndrome: A developmental perspective. Cambridge: Cambridge University Press.

Fowler, A., Gelman, R. & Gleitman, L. (1994). The course of language learning in children with Down syndrome. In Tager-Flusberg, H. (ed.), Constraints on language acquisition: Studies of atypical children. Hillsdale, NJ: LEA.

Jakubowicz, C. (1984). On markedness and binding principles. In Jones, C. & Sells, P. (eds.), NELS 14. Amherst: University of Massachusetts.

Kernan, K. & Sabsay, S. (1996). Linguistic and cognitive ability of adults with Down syndrome and mental retardation of unknown etiology. Journal of Communication Disorders 29.5. 401-422.

Marcel, M., Ridgeway, M., Sewell, D. & Whelan, M. (1995). Sentence imitation by adolescents and young adults with Down's syndrome and other intellectual disabilities. Journal of Intellectual Disability Research 39.3. 215-232.

McKee, C. (1992). A comparison of pronouns and anaphors in Italian and English acquisition. Language Acquisition 2. 21-54.

Miller, J. (1987). Language and communication characteristics of children with Down syndrome. In Pueschel, S., Tingey, C., Rynders, J., Crocker, A. & Crutcher, D. (eds.), New perspectives on Down syndrome. Baltimore: Paul Brookes.

Miller, J. (1988). The developmental asynchrony of language development in children with Down syndrome. In Nadel, L. (ed.), The psychobiology of Down syndrome. Cambridge, MA: MIT Press.

Philip, W. & P. Coopmans (1996). The double Dutch Delay of Principle B Effect. In Stringfellow, A., Cahana-Amitay, D., Hughes, E. & Zukowski, A. (eds.), Proceedings of the Annual Boston University Conference on Language Development 20, Cascadilla Press, Somerville, Mass.

Reinhart, T. (1983). Anaphora and semantic interpretation. London: Croom Helm.

Reinhart, T. (1986). Center and perip hery in the grammar of anaphora. In Lust, B. (ed.), Studies in the acquisition of anaphora. (Vol. 1). Dordrecht: D. Reidel.

Reuland, E. (1997) Primitives of Binding. OTS Utrecht Institute of Linguistics.

Rondal, J. (1993). Down's syndrome. In Bishop, D. & Mogford. (eds.), Language development in exceptional circumstances. Hillsdale, NJ: LEA.

Rondal, J. (1995). Exceptional language development in Down syndrome. Cambridge: Cambridge University Press.

Rondal, J. & A. Comblain (1996). Language in adults with Down syndrome. Down Syndrome Research and Practice 4.1.

Rutter, T. & S. Buckley (1994). The acquisition of grammatical morphemes in children with Down's syndrome. Down Syndrome Research and Practice 2.2. 76-82.

Sigurjónsdóttir, S. (1992). Binding in Icelandic: Evidence from language acquisition. PhD dissertation, University of California. van der Lely, H.K.J. & Stollwerck, L. (1997). Binding Theory and specifically language impaired children. Cognition 62. 245-290.

Varlokosta, S. (2001). The acquisition of pronominal reference in Greek. Talk given at the UCL Department of Phonetics and Linguistics, London.

Vicari, S., Caselli, M. & Tonucci, F. (2000). Asynchrony of lexical and morphosyntactic development in children with Down syndrome. Neuropsychologia 38. 634-644.

Tager-Flusberg, H. & Calkins, S. (1990). Does imitation facilitate the acquisition of grammar? Evidence from a study of autistic, Down's syndrome and normal children. Journal of Child Language 17.3. 591-606.

Wexler, K. & Y-C. Chien (1985). The development of lexical anaphors and pronouns. Papers and Reports on Child Language Development 24. 138-149.