Diagnosing Specific Language Impairment

(1)

Diagnosing Specific Language Impairment

A c t a U n i v e r s i t a t i s T a m p e r e n s i s 1113 ACADEMIC DISSERTATION

To be presented, with the permission of the Faculty of Medicine of the University of Tampere, for public discussion in the auditorium of Finn-Medi 2, Biokatu 8, Tampere, on December 9th, 2005, at 12 o’clock.

MARJA ASIKAINEN

(2)

Distribution Bookshop TAJU P.O. Box 617

33014 University of Tampere Finland

Cover design by Juha Siro Layout Aila Helin

Printed dissertation

Acta Universitatis Tamperensis 1113 ISBN 951-44-6454-0

ISSN 1455-1616

Tel. +358 3 3551 6055 Fax +358 3 3551 7685 taju@uta.fi

www.uta.fi/taju http://granum.uta.fi

Electronic dissertation

Acta Electronica Universitatis Tamperensis 480 ISBN 951-44-6455-9

ISSN 1456-954X

ACADEMIC DISSERTATION University of Tampere, Medical School

Tampere University Hospital, Departments of Paediatrics and Otorhinolaryngology

Hospital District of Helsinki and Uusimaa, Departments of Otorhinolaryngology and Phoniatrics Finland

Supervised by Docent Matti Koivikko University of Tampere Docent Erkki Vilkman University of Tampere

Reviewed by

Professor Timo Ahonen University of Jyväskylä Docent Eeva Sala University of Turku

(3)

To Tom-Henrik and Markku.

(4)

(5)

ABBREVIATIONS

ADD Attention Deﬁ cit Disorder

ADHD Attention Deﬁ cit Hyperactivity Disorder

AD(H)D Attention Deﬁ cit Disorder with or without Hyperactivity CG control group

CT computed tomography dB decibel

DSM-IV Diagnostic and Statistical Manual of Mental disorders EEG electroencephalogram

ICD-10 International Classiﬁ cation of Diseases, 10th edition IQ intelligence quotient

ISI inter-stimulus interval LI language impairment LTM long-term memory MLU mean length of utterance MRI magnetic resonance imaging ns. nonsigniﬁ cant

OM otitis media

OME otitis media with eﬀ usion OR Odds Ratio*)

PIQ performance intelligence quotient SD standard deviation

SES social-economic status SLI speciﬁ c language impairment STM short-term memory

WHO World Health Organization VIQ verbal intelligence quotient

WM working memory

*) Odds ratio is a way of comparing whether the probability of a certain event is the same for two groups. An odds ratio of 1 implies that the event is equally likely in both groups. An odds ratio greater than one implies that the event is more likely in the ﬁ rst group.

Some additional abbreviations are explained in the associated text or tables.

(10)

(11)

SUMMARY

A child with speciﬁ c language impairment (SLI) does not acquire normal language ability at the appropriate age despite having adequate hearing and normal nonverbal intelligence, and no neurological or sensorimotor handicap. Current diagnostic criteria include deﬁ cient expressive and/

or receptive language, and a substantial discrepancy between nonverbal performance and language skills (the discrepancy criteria). Common clinical ﬁ ndings are deviant phonological, morphological or syntactic forms used in utterances, omitted words in sentences, and deﬁ cient understanding of words or sentences.

SLI is regarded as a neurobiologic disorder. Th e factors most often suggested to underlie SLI are defi cient discrimination between rapidly changing stimuli, poor short-term memory (STM), and problems in auditory or verbal processing. However, fi ndings in SLI studies are contradictory, and the core problem of the disorder is still under debate. So far diff erential diagnostics between SLI and other disorders in the spectrum of developmental disordes, social-emotional disorders and learning diffi culty are also undetermined.

Th e major objectives of the present work were 1) to carry out a descriptive analysis of fi ndings of a clinical sample of SLI compared to a sample of normally developing children, 2) to evaluate relationships of these fi ndings with fi ndings and conclusions in the literature concerning especially factors underlying SLI and factors related to cognitive development of children with SLI, 3) to fi nd out if the present data provides evidence that defi cits in long- term memory is a factor explaining SLI, and, fi nally, 4) to evaluate merits of the present clinical practice.

(12)

Th e study group is based on clinical material. Th ey are 78 children referred to the Tampere University Hospital on the basis of SLI. Inclusion criteria were nonverbal ability on age level but deﬁ cits in either expressive or receptive language, or in both. Th e control group (101 children whose development had been followed up in a local well-baby clinic) with no known delay in development was gathered from two kindergartens and three well-baby clinics after their parent´s consent. Th e children´s linguistic, verbal and nonverbal performance and motor skills were evaluated clinically and by the ITPA, Reynell, Bo Ege, Boehm, WPPSI, WPPSI-R, WISC-R (or Leiter) tests, and by the motor test of Stott, Moyes and Henderson. Th e study group´s general history was gathered by interviewing parents, whereas the control children´s parents completed a questionnaire.

Th e clinical picture of the study group was in agreement with the fi ndings reported in the literature concerning SLI. Discriminative ability and auditory span diff ered between the LI and control groups, likewise discrepancy between verbal and nonverbal performance. However, in agreement with criticism of the diagnostic criteria for SLI, there were also children in the study group with clear evidence of SLI in their linguistic skills but who did not fulfi l the discrepancy criteria, or the criteria based on standard deviations in linguistic tests. On the other hand, in the control group there were children with age-appropriate linguistic ability but a marked discrepancy between verbal and nonverbal performance. Th us, the present study suggests that the diagnostic criteria should be reconsidered. Further, averaging scores in psychometric tests should be avoided. It seems relevant to include age- appropriate visual reasoning in the diagnostic criterion but the possible eff ects of SLI and underlying factors on psychomotor and social-emotional development should be reconsidered.

Considering the results of the present study and the literature, it will be suggested that deﬁ cient discrimination and memory functions are the factors underlying SLI. Moreover, considering the theoretical background adopted

(13)

in this study, it seems that the environmental support given for children with SLI does not widely enough correspond to the underlying factors.

Finally, it seems that currently SLI disorder is considered far too much as a disorder of mainly speech and language development. Instead, it should be taken as a sign of underlying neural deﬁ cits to be compensated, not only during actual development of speech and language but in later years, too.

In conclusion, the present study suggests that current clinical practice in diagnosing SLI should be reconsidered in many respects.

(14)

(15)

1. INTRODUCTION

1.1. Why identify children with speciﬁ c language impairment?

A narrow deﬁ nition of health focusing primarily on preservation of life is expanding to a much broader conceptualization that incorporates quality of life (Shaywitz et al. 1990).

As acute, life-threatening illnesses diminish as a cause of morbidity in children, physicians increasingly focus on and accept responsibility for more chronic handicapping conditions. For many physicians, the most dramatic shift has been in the increased demand for services related to school problems, particularly learning disabilities (Shaywitz et al. 1990). Further, in 2003, Isabelle Rapin stated that the mystery has now been removed from many acute neurologic issues and from many genetic-metabolic issues. However, issues mostly seen in the clinical practice of child neurology are seizures and headaches, a sprinkling of rarer conditions, and droves of children with delayed language development, hyperactivity and school failure. Yet, as Rapin goes on, many child neurologists consider static encephalopathies like language disorders as lacking in interest and glamour. …“It is time for a wake-up call for child neurologists to embrace this ﬁ eld and bring to it other professionals with specialized knowledge in the clinical neurosciences, so as to develop much more eﬀ ective prevention and management of the developmental disorders.”

Learning disability refers to a child with a discrepancy between his performance and potential ability (Kessler 1980). In the 1990´s in Finland, nearly 18% of all children were given at least extra lessons by a special education teacher, and about 20% of them were placed in a special education classroom because of learning diﬃ culties (Michelsson et al. 2000). In 2004, more than 20% of all children were given extra lessons by a special education

(16)

teacher, and 7% of all children either started their education in a special education class or were moved to a special education class (Tilastokeskus).

Th e percentage of children so transferred increased 8% after 2003. 18% of the children in special education classes had speciﬁ c language impairment, 14% were diagnosed with Asperger syndrome or autism, and 12% were diagnosed with attention deﬁ cit disorder, cerebral palsy or some other neurological disability. Special education given in vocational school settings has increased: in 2003, 6% of pupils in occupational schools were given special education (Tilastokeskus).

Disorders of auditory perception have frequently been suggested as a cause of learning diﬃ culties (see e.g. Cacace and McFarland 1998, Kraus et al. 1996, 1999). Children with speciﬁ c language impairment (SLI) have been shown to exhibit disorders of auditory perception (Bishop 1997). Strong empirical evidence suggests that dyslexia is also a language-based learning disability (e.g. Fazio 1997, Rockett et al. 1987, Share and Silva 1987).

It seems obvious that SLI does not merely have an infl uence on the development of speech but may cause learning diffi culties, too. It is common that children with delayed speech development tend to catch up (World Health Organization 1993). Also, most children with SLI will sooner or later develop communication skills adequate for daily activities. However, both clinical experience and the literature suggest that the disorder does not disappear but the clinical picture changes. Th ese aspects suggest that it is worth identifying children with SLI, not only to help them to communicate but also to prevent learning diffi culties and social-emotional problems.

1.2. Deﬁ nition and criteria for SLI

At least so far there is no clear explanation for SLI: it is diagnosed by exclusion (Cacace and McFarland 1998). It has been deﬁ ned as a failure to learn language despite at least average intelligence, intact peripheral perceptual abilities, no known neurological, physical, emotional, or social problems, and

(17)

an adequate opportunity to learn language (Bishop et al. 1999a, McArthur et al. 2000). In other words, SLI refers to a disorder characterized by failure to acquire normal language at an appropriate age despite adequate hearing and normal nonverbal intelligence (Holopainen et al. 1997).

A child may pass conventional audiometric assessment but nevertheless process sounds abnormally (Bishop et al. 1999a). A possibility exists that children with SLI are not disordered at all but just fall at the lower end of the distribution with respect to language skills (Leonard 1991).

A classic neurological sensorimotor examination of dysphasic (SLI) children is thought to be normal in a vast majority of cases (Tuchman et al.

1991). However, children may have problems with auditory, visual, tactile, phonetic, and dihaptic perception, as well as with motor tasks (Bishop 1990, Johnston et al. 1981, Powell and Bishop 1992). Furthermore, Trauner et al.

(2000) noted that no systematic studies of neurological function have been published. Trauner et al. found abnormalities in as many as 70% of children with language impairment and only 22% of control children. Th e most common abnormalities in the LI group included obligatory synkinesis, fi ne motor impairments, and hyper-refl ex. Th e children with LI with the most abnormal neurological fi ndings had the lowest language scores (Trauner et al.

2000).

Th e spontaneous speech of an English-speaking pre-school aged child with SLI may typically contain a) errors in word order, b) omitted or incorrectly used morphologic endings, c) omitted articles, prepositions, auxiliary verbs or contractions, d) telegraphic speech, or e) incorrectly selected negatives. Furthermore, children may f ) add irrelevant details inappropriately, g) contradict previous statements, h) fail to stay on the topic, or i) fail to respond to the examiner´s question (Dunn et al. 1996). Th e studies referred to in the present work were made almost exclusively for English-speaking children. Th e language that children are to acquire and use naturally aﬀ ects the ﬁ ndings in studies on SLI clinical features and may also play a role in

(18)

studies searching for factors explaining the disorder. In Finnish, features such as rather long words and inﬂ ected forms of words apparently have their eﬀ ect on language development. Features of the Finnish language are presented in Appendix 10.

Th e ICD-10 Research Diagnostic Criteria for Developmental Language Disorders (World Health Organization 1993) include following:

1. Language skills, as assessed on standardised tests, are below the 2 standard deviations limit for the child’s age.

2. Language skills are at least 1 standard deviation below nonverbal IQ as assessed on standardised tests.

3. Th ere are no neurological, sensory, or physical impairments that directly aﬀ ect use of spoken language, nor a pervasive developmental disorder.

4. A distinction is made between receptive language disorder, where comprehension is more than 2 SD below age level, and expressive language disorder, where only expressive language is so severely aﬀ ected, and where understanding is within the 2 standard deviations limit for the child´s age.

Th e diagnostic criteria for “developmental language disorder” in the American Psychiatric Association´s Diagnostic and Statistical Manual (DSM-IV 1994) are very similar to those of the ICD-10, but they do include an additional requirement: language diﬃ culties interfere with academic or occupational achievement or with social communication (Bishop 1997). Another additional requirement that has been commonly applied is Stark and Tallal´s (1981) criterion for performance IQ (PIQ) to be 85 or above.

As Bishop (1997) stated, one has an option of taking a composite index of language ability by averaging language test scores, or adopting the kind of approach favoured by Bishop and Edmundson (1987), who diagnosed SLI if there was a severe deﬁ cit (e.g. 2 SD or more below the mean) on any

(19)

language measure, or moderate levels of deﬁ cit (e.g. 1.5 SD or more below the mean) on two or more measures.

Th e validity of the diagnostic criteria has been criticized (Aram et al. 1992, 1993, Bishop 1994, 1997, Dunn et al. 1996, Plante 1998). Th e diagnostic criteria include defi cient language skills evaluated by standardised tests. Aram et al. (1992) found that approximately 40% to 60% of children considered clinically to be language disordered could not be classifi ed as language disordered using nonverbal IQ-language discrepancy criteria. In addition, over 45% of randomly selected control subjects were identifi ed as language disordered when discrepancy criteria were applied. Application of the stringent exclusionary and discrepancy criteria also resulted in the study by Stark and Tallal (1981) in the identifi cation of only one-third of the 132 clinician-defi ned language-impaired children as meeting criteria for SLI.

Bishop (1997) also stated that some children have all the linguistic characteristics of SLI in their spoken language but still not enough discrepancy between verbal and nonverbal ability. In a twin study, Bishop (1994) reported that it was not uncommon to fi nd identical twins where one met the stringent defi nition of SLI with a large verbal-nonverbal discrepancy, while the other had equally poor language ability, but did not meet the defi nition of SLI because the nonverbal score was somewhat lower. Bishop (1997) suggested that the discrepancy defi nition is over-restrictive.

In 1993 Aram et al. found that no single measure of language, when compared to nonverbal abilities, provided complete agreement with clinical diagnoses in a total of 252 children. Th e greatest overlap between clinical and research defi nitions was achieved when a cut-off below 85 was applied to MLU (mean length of utterance) age scores. Aram et al. stated that it is possible that the professional´s judgments were more sensitive than the psychometric discrepancy formulas. “Although assessment tools are used as a means of objectively operationalizing constructs and defi nitions, these tools address only limited aspects of a behavior in a specifi c situation.”

(20)

Aram et al. (1993) pointed to two main concerns in using a psychometric approach to identify SLI in children: sensitivity (percentage of children clinically diagnosed with SLI who were also identifi ed as such by the specifi ed test; see Dunn et al. 1996) and specifi city (percentage of children clinically diagnosed with normal language who were identifi ed as such by the test).

Aram et al. (1993) pointed out that whereas a psychometric test may be a very reliable and valid measure of an underlying language construct and behavior, there is still the important matter of how useful it may be for a given diagnostic purpose, such as identifying SLI in children. Th ey further noted that basic to the use of nonverbal intelligence measures for deﬁ nition of SLI is one´s theoretical conceptualization of the relationship between language and other aspects of cognitive development (Aram et al. 1993).

Further, according to Aram et al. (1993), no attempt had earlier been made to judge the “correctness” of clinical versus research deﬁ nition of SLI.

Th ey stated that detailed descriptions of both those children meeting criteria and those not doing so would help bridge the gap between clinical and psychometric deﬁ nitions of SLI.

Dunn et al. (1996) suggested that standardized psychometric discrepancy criteria are more restrictive and perhaps less sensitive to language impairment than is clinical judgement based on a child´s language performance in naturalistic contexts. Dunn et al. (1996) found that spontaneous language data indicated that children clinically identiﬁ ed as SLI produced a signiﬁ cantly higher percentage of errors in spontaneous speech than normal children, whether they met psychometric discrepancy criteria or not.

Plante (1998) stated that the 85-IQ cutoﬀ results in somewhat arbitary deselection of potential subjects whose IQs are within normal limits but do not meet the 85-IQ criterion. Plante claimed that in studies that have not restricted IQ to greater than 85 (and have not matched subjects for mental age), the nonverbal IQ scores of children with SLI are commonly lower than

(21)

those for normally developing children, suggesting that low-average IQ is a common component of the SLI proﬁ le.

One problem with standardized tests is that it is conceptually diffi cult to determine how low a cutoff point for test scores should be set (Lahey 1990). Th e cutoff scores can also be empirically derived to maximize the identifi cation accuracy of a given test (Plante 1998). Furthermore, as Plante continued, statistically derived cutoff scores and the resulting identifi cation accuracy may vary considerably from test to test, even within the same sample of children.

An aspect of cognitive development is that older SLI children may obtain lower scores on nonverbal measures because they do not possess the verbal mediation strategies that normally developing children use to improve performance (Dunn et al. 1996). It may be more diffi cult for them to encode the key features of visually presented stimuli (Nelson 1993). Th ese issues may result in lower nonverbal IQ scores for SLI children, making it more diffi cult for a discrepancy to be met (Dunn et al. 1996). Th e available information on nonverbal functioning in SLI may promote a more sophisticated approach to understanding the role of IQ in the defi nition of the disorder (Plante 1998).

To sum up: According to the dominating view, to be classiﬁ ed as a case of speciﬁ cally language impaired, the child has to exhibit a substantial discrepancy between verbal and nonverbal abilities (Bishop et al. 1999a).

However, it is not clear whether children scoring at or below the 10^th percentile on only two language parameters constitute a basically diﬀ erent population from speciﬁ cally language-impaired children. Furthermore, as Rudel (1980) pointed out, tests provide only some basic levels, and one cannot infer from a normal verbal score that the child does not have a language impairment just as, essentially, an average or better than average IQ score does not guarantee normal learning. Rescorla (2002) also stated that a child may perform in tests on age level but may still exhibit, evaluated clinically, clear language impairment. Th e questions concerning selection

(22)

criteria and relationship between language development and intellectual development seem to be crucial in deﬁ ning SLI.

1.3. Prevalence

Precise prevalence estimates of SLI will depend on the criteria adopted (Bishop et al. 1999b). Tallal et al. (1989) took the view that prevalence studies estimate 8%–15% of all preschool children to be aﬄ icted with some form of speech and language disorder.

In the 1980´s, the prevalence rate of SLI was usually estimated to range from 0,5% to 3% (Vilkman and Helminen 1989), whereas about a decade later Tomblin et al. (1997) suggested that as many as 7.4% of pre-school aged children could be speciﬁ cally language impaired. McArthur et al. (2000) raised the possibility of an even higher prevalence rate on the basis of the undetermined relationship between SLI and dyslexia. In 2003, Hartley et al.

referred to the prevalence estimation of 3% to 10% of children.

It is currently an open question whether the incidence of SLI and other developmental disorders has been rising. Wing (1997) speculated the possible rise of incidence for autistic-spectrum disorders and suggested that it may be due to wider diagnostic criteria, greater awareness of these disorders, or a genuine change in incidence. Th e same might be the case for SLI.

(23)

2. REVIEW OF THE LITERATURE

2.1. Neural mechanisms of language and language development

2.1.1. Aspects of higher brain functioning

Th e human brain is a highly complex mechanism organized in 1) a hierarchical structure of many levels and 2) complexly integrated circuits and sub-circuits (Lucas 1980). While there is a unique processing by the separate sensory systems (vision, audition, kinesthesis, and so on), they operate conjointly (Belmont 1980).

Luria (1973) divided the cortical areas of brain concerning higher brain functions into three zones. Primary or projection areas, such as primary visual or auditory areas, respond only to the narrowly specialized properties of stimuli. Secondary cortical areas include many more associative neurons with short axons, enabling incoming excitation to be combined into the necessary functional patterns, and they thus subserve a synthetic function.

Tertiary zones play an essential role in the conversion of concrete perception into abstract thinking.

In the course of the development it is not only the functional structure of higher mental processes that changes, but also their relationship with each other, or, in other words, their “inter-functional organization” (Luria 1973).

Th e participation of the auditory and visual areas of the cortex, essential in the early stages of formation of the activity, is no longer necessary in its later stages, and the activity starts to depend on a diﬀ erent system of concertedly working zones (Luria et al. 1970). In an adult person, with the fully formed higher psychological functions, the higher cortical zones have assumed the dominant role (Luria 1973).

Luria (1973) further stated that proper working of the tertiary zones (abstract thinking) would be impossible without adequate development of

(24)

the secondary cortical zones, which supply the necessary material for the creation of major cognitive syntheses. Rie (1980) wrote about developmental shift from dependence on sensory memories to complex cortical processes. If children continue beyond early elementary years to rely heavily on sensory impressions for one reason or another, then thought, problem solving, generalizing, and so on, become tedious and impossible tasks (Rie 1980).

Luria (1973) referred to high-graded automation and took an example of writing, which during development is converted to a single “kinetic melody”, no longer requiring the memorizing of the visual form of each isolated letter or individual motor impulses. Similar changes are suggested by Luria to take place during the development of other higher psychological processes.

2.1.2. Pre-requirements for language development

According to Bates (1999), linguistic knowledge is not localized in a clear and compact form. Th e infant brain is not, however, a tabula rasa; it is already highly diﬀ erentiated at birth, and certain regions are biased toward modes of information processing that are particularly useful for language, leading, in the absence of local injury, to the standard form of brain organization for language (Bates 1999).

Neurobiological results underscore the extraordinarily plastic and activity-dependent nature of cortical specialization, and support the case for an experience-dependent account of the development of higher cortical functions. Anything that we know, whether it is innate or acquired, must be represented somewhere in the brain (Bates 1999).

Th e range of representations a system can take is much constrained at the architectural level: the right number of units, number of layers, types of connectivity between layers, etc. (Bates 1999). We have sources of input and patterns of output that connect brain regions to the outside world and to one another. Chronotropic constraints refer to innate constraints on the timing of developmental events, such as number of cell divisions that take place in

(25)

neurogenesis, and synaptic growth and pruning, and diﬀ erences in timing between subsystems (Bates 1999).

Bates further argued that representational nativism (direct innate structuring of the mental/ neural representations that underlie and constitute knowledge) would require a huge investment of genetic information, probably more than we have available, as well as a brain with approximately 100,000,000,000,000 synaptic connections. By contrast, architectural and chronotropic nativism both involve much smaller genetic contributions to achieve the same ends, relying heavily on information in the environment.

Evolution is forced to be very conservative, reusing the same genes over and over again, building new functions through minor quantitative tuning (Bates 1999).

Regions of the cortex are specialized for modes of information processing that have a similar eff ect in linguistic and non-linguistic domains. Th e fi nal product emerges from the interaction between these constraints and the specifi c problems that an organism with such a structure encounters in the world (Bates 1999).

Deacon (2000) also suggested that there are no intrinsically pre-speciﬁ ed language circuits in the human brain, only connection patterns that have been biased in unique ways so that they are slightly better suited to the unique demands imposed by language.

Bates referred to Chomsky (1975) and his linguistic theory of Universal Grammar, which suggested language to be an innate property of the human mind, given by virtue of our genetic code. Bates stated that Universal Grammar left little room for learning. According to Bates and Elman (1996), learning is much more powerful than previously believed, and arguments about the innateness of language and other forms of cognition need to take that undeniable fact into account.

Finally, according to Seidenberg (1997), brain organization constrains how language is learned, but the principles that govern the acquisition,

(26)

representation, and use of language are not specifi c to this type of knowledge. Until recently, it has been diffi cult to point to empirical evidence demonstrating that nonverbal mental structures (such as cross-modal matching, recall memory) actually support the growth of language (Meltzoff 1999). Meltzoff (1999) further stated that modern research has discovered that young children know more at earlier ages than has been predicted by classical theory.

Th e key concept for information processing in the brain is population neuronal coding: overlapped populations of neurons with fl exible functional connections within and among the populations (Sakurai 1996). Correlated activity among the neurons constructs the functional connection. Th e number of information items is virtually unlimited. As information processing in the brain is fl exible and changes continuously, it is impossible to maintain that a specifi c structural organization for specifi c information is prepared in advance (Sakurai 1996).

Simultaneous representations are suggested to form representative networks (Edelman 1987, 1992). In Edelman´s (1987, 1992) account, independent signals that co-occur as time-locked inputs activate groups of neurons that are associated with each of these independent signals. Th e crucial result is that these neuronal groups are activated at the same time.

Such time-locked activation results in reciprocal connections among these groups such that the signals become correlated at the neural level into a network of relations. Activation of one group of neurons in the network can result in correlated activation of a reciprocally connected group of neurons that usually respond to some other independent signal, even though that other signal may not actually be present (Edelman 1987, 1992).

Th e processing of co-occurrence relationships among cues that tend to recur together over time may be a mechanism by which cues in various layers of context get selected over developmental time and connected in real time (Roberts 1997). On the basis of past experience with speciﬁ c co-occurrences

(27)

among cues, infants develop expectations about these relations and, thus, anticipate that certain cues will occur in the presence of others (Roberts 1997).

Learning involves gradual changes to the weights on connections between units that determine patterns of activation in the network (Seidenberg 1997).

Although the weights are set on the basis of experience, they can be used to process novel forms. Networks trained on the pronunciations of written words, for example, can generalize to novel forms to which the network has not been exposed.

Th e network is trained through exposure to a large number of examples and gradually discovers diﬀ erent kinds of simple and complex contingencies between diﬀ erent types of information (Seidenberg 1997).

2.1.3. Learning language

According to Seidenberg (1997), learning based on the frequencies and distributions of environmental events is emerging as an essential aspect of cognitive development. From continuous speech, for example, children identify words and their functions. Th ey naturally and automatically encode statistical aspects of caregiver speech without overt guidance or reward (Saﬀ ran et al. 1996).

As children learn words, more and more minimal pairs also draw their attention to the existence of overlapping articulatory gestures that, in turn, point to the underlying presence of the phoneme (Locke 1996).

Accumulation of lexical knowledge normally activates analytical mechanisms that are needed for grammatical development (Locke 1997).

Allen (1997) implemented a network that learns about verbs and their argument structures from naturalistic input. Input variability is not crucial because the model’s performance on any given verb does not solely depend on experience with the verb; the model beneﬁ ts from exposure to other verbs that pattern similarly and diﬀ erently (Allen 1997).

(28)

Th e same mechanisms are involved in acquiring language as in using it (Seidenberg 1997). Learning to use language adequately means that we learn to know what kinds of sentences are acceptable and which are not (Shapiro 1997). For example, is it possible to say, “Th e woman ate soup”, or “Th e soup ate the woman”. Furthermore, we learn to know that some words behave diff erently than others (Shapiro 1997): Nouns can be pluralized (boys, women), verbs cannot (in English). Nouns and verbs can form complex words made up of more than one morpheme, prepositions cannot. Nouns occur in particular and in diff erent parts of a sentence than do verbs; thus, they cannot be substituted for each other. Nouns can be pre-modifi ed by adjectives (very big boy, pretty woman, etc.), verbs cannot (“Very big know”).

Nouns can be quantifi ed and specifi ed (e.g., made defi nite or indefi nite: a boy, the boy), verbs cannot.

Lexical, functional, and phrasal categories are arranged in a hierarchical structure to form clauses and sentences, much as a house is built with a foundation, walls, beams, and a roof (Shapiro 1997). Included in each word’s lexical entry is information about its phonological form, its lexical category (verb, noun, preposition, etc.), semantic information (what it refers to in the real world), and importantly, the legal sentence environments the word is allowed to enter (Shapiro 1997). When a verb, for example, is encountered in a sentence, all of the verb’s argument structure possibilities are activated (Shapiro 1997).

Nouns are easier to learn because they refer to object-reference concepts (e.g., persons, things) that more consistently map onto the perceptual- conceptual structure of the world (Conti-Ramsden and Jones 1997).

Conversely, verbs refer to relational concepts (e.g., activities, changes of state, instruments, causal relations) that show more variability in how they map onto the world. In addition, verbs often represent events that have a limited temporal availability (Conti-Ramsden and Jones 1997).

(29)

2.1.4. Memory processes

Adequate long-term memory (LTM), short-term memory (STM) and working memory (WM) are essential parts of information processing (Luria 1973, Sininger et al. 1989, Gathercole and Baddeley 1990, Lahey and Bloom 1994, Fazio 1997, Gillam et al. 1998, Adams and Gathercole 2000). For example, representations are to be kept in mind until a process is ﬁ nished (Fazio 1997).

Memory span has been shown to be greater for familiar than for unfamiliar words (Hulme et al. 1991), which suggests that phonological (short-term) memory may not be as isolated from long-term knowledge as often presumed (Adams and Gathercole 2000). STM holds that we can keep representations activated (Gathercole and Martin 1996). Keeping phonological representations activated might be related to the phonological loop (see Baddeley 1986) of WM. Most current memory models describe serial memory (remembering the sequence of items) as a consequence of short-term memory processes (Fazio 1997).

According to Fazio, STM is actually often called WM. It is suggested to be the process that allows for ongoing access of a small number of items of information in conscious awareness. In language processing, STM holds incoming linguistic material while the underlying meaning is found to make sense of sequence of words, to solve problems, or to process information into and out of long-term storage (Fazio 1997).

Any action may become automatized (Luria 1973, Fazio 1997). Th ere is then, at least, decreased demand for selective attention (Carte et al. 1996).

WM is needed in motor performances as well. A motor act originates with an idea (Eccles 1977). In every motor performance we monitor our movements and compare the consequences of our actions with the desired goal. Th is deals with the “landscape” or vision of the goals and actions, or executive functions (see Luria´s (1973) basic laws for the cortical functioning, p. 67–79, and Akhutina 1997).

(30)

It has been suggested that neural networks are changing continuously (Sakurai 1996, Bates 1999); neurobiological results underscore the extraordinarily plastic and activity-dependent nature of neural networks.

Neural networks re-organize along with working with representations and representative networks and with external stimuli (see Luria 1973, Rie 1980, Sakurai 1996, Roberts 1997, Seidenberg 1997, and Bates 1999).

2.1.5. Processing a verbal message

When beginning to produce a verbal message we have an idea (Luria 1973).

Th e idea serves as a frame for the action; other parts of the action form the content inside the frame (compare to Shapiro 1997). Following Gillam et al.

(1998), if a representation of an idea fades away on the basis of any internal or external stimuli or just a primary tendency to fade away before we have ﬁ nished the action, the coherency may break down.

Correspondingly, to understand messages spoken by other people, we have to ﬁ nd out the general idea or the scheme running through the expression (Luria 1973). We must discriminate words, identify meanings and functions of the words in sentences, and discover the integrated meanings of the sentences (Luria 1973, Lucas 1980, Sakurai 1996, Bates 1999). Th is may be easy with short and simple sentences but will require more resources when sentences lengthen or include more abstract words (see Fazio 1997).

2.1.6. Language and executive functions

A little child gradually begins to understand speech, and the parents´ control and guidance will be internalised into the child´s own control of actions, ﬁ rst spoken and later inner speech. Words break the unity of sensory-motor ﬁ eld;

they allow the child to free him or herself from slavish submission to the sensory fi eld and to reorganize the fi eld by focusing attention on diff erent fi eld elements (Akhutina 1997). Th is allows the child to build a fi eld of future action, i.e. to make plans and to change them (Akhutina 1997).

(31)

Language is linked to attention through the concept of executive functions. Actions are usually goal-directed. To be able to carry out a plan of action, we need an instrument to detach from concretia and immediate properties of the objects, and submit actioning to a verbally formulated plan, maybe shared with other children (see Luria 1973). Actions controlled by a plan formulated by speech are more resistant to disruptive stimuli, as reviewed by Aro et al. (2001).

2.1.7. Th e role of intelligence in a child’s development

Teele et al. (1990) and Plomin (2001) suggested that intelligence (general cognitive capacity) is an important predictor of a child’s development and cognitive abilities. Kurzweil (1992) found that children’s improvement was signiﬁ cantly related to their parents´ IQ. Chapman et al. (2000) demonstrated that production skills of children with marked language impairment showed considerable variation associated with their mothers´ education.

Rawson (1968) found very favourable adult outcomes for 20 dyslexic boys from a private school. Th e group was unusually intelligent (the average IQ on the Stanford-Binet scale being about 122) and received exceptionally intensive and systematic remedial instruction. Th e majority of the group were on follow-up reported to have careers ordinarily expected to demand very strong reading and writing skills (physicians, lawyers, professors). Despite such a good outcome a number of those studied still claimed that reading and spelling remained diffi cult. Weiner (1980) and Rie (1980) reviewed studies showing that the child’s intellectual level does signifi cantly diff erentiate good and poor outcome groups (for example Menkes et al. 1967, and Koppitz 1971). Weiner stated that it is still possible that the children never lose their feeling that reading is diffi cult.

What is meant by the concept ´intelligence´? Hebb (1949) distinguished between innate intelligence and the level of development of brain functions, the latter being an eﬀ ect of experience as well as innate dispositions. Luria

(32)

(1973) referred to Vygotsky, who in the 1930´s suggested that analysis and generalization form the basis for an intellectual act. Analysis of the components of a given problem, recognition of the most essential features, and creation of a general plan (scheme) for the performance of the task are essential parts of intelligent thinking (Luria 1973). Lashley (1963) referred to the construct of intelligence as a general, integrative function, and stated that the mechanisms of integration (intelligence) lie in the dynamic relations among parts of the nervous system rather than in structure.

Anderson (2001) discussed the concept of g, or general intelligence.

In 1927 Spearman supposed that g refl ects a hypothetical mental power or energy that infuses all intellectual operations. According to Anderson (2001), the modern equivalent of Spearman’s hypothesis is that diff erences in speed of processing are the basis of diff erences in general intelligence.

Detterman (1987, 1992) stated that any complex task requires the operation of a number of basic abilities. In his scheme general intelligence represents an average level of all the independent components that contribute to any and all complex tasks. Bjorklund and Harnishfenger (1990) and Dempster (1991) suggested that it is not only the increase in central capacity of information-processing speed that accounts for the development of intelligence but rather improvement in individuals´ ability to inhibit task- irrelevant information.

Anderson’s theory is framed within a general theory of cognitive architecture proposed by Fodor (1983) that made a distinction between central processes of thought and dedicated processing modules. Anderson stated that intelligence tests measure intelligence through assessing knowledge but that knowledge itself is acquired through the two diﬀ erent routes proposed by Fodor (central processes and modules). Th e observed ability is constrained by the speed of a basic processing mechanism – at slow speed only the simplest kinds of thoughts can be implemented (Anderson 2001).

(33)

Anderson further stated that if we had to “think through” all the perceptual information presented to us in order to construct a three- dimensional view of the world we would be literally lost in thought. Evolution has created special modular devices to allow us to do this automatically. As examples of modules Anderson gave various language acquisition devices and face recognition systems. Th e common features of both the acquired and the innate modules are that they operate automatically and independently of thought and are consequently unconstrained by the speed of the basic processing mechanism (Anderson 2001).

Intellectual disability is deﬁ ned to be caused by either organic brain damage or normal but poor functioning of the brain (Anderson 2001).

Th e developmental view argues that individuals with intellectual disabilities develop more slowly but progress through the same developmental stages as non-impaired individuals. Th e diff erence view suggests that intellectual disability represents a fundamental defi cit usually in a single process and, consequently, development in children with intellectual disabilities diff ers qualitatively compared to development in intellectually normal children (Anderson 2001).

Anderson suggested that not only speed of processing underlies diﬀ erences in IQ but that speed of processing is unchanging through development. Slow speed of processing will be a pervasive feature of the cognitive processes of people with intellectual disabilities. However, the maturation and acquisition of modules represents a developmental dimension that is independent of speed of processing and hence of IQ. Th is dimension would mean that some aspects of development in people with intellectual disabilities would follow a biological and possibly experimental programme that does not deviate from the development of intellectually normal children.

According to Anderson, low-IQ children go through the same kinds of knowledge restructuring but do so more slowly than the average-IQ children. However, children with intellectual disabilities will always suﬀ er

(34)

a fundamental deﬁ cit in intelligent information processing even when compared with mental-age peers. Th at is, a low IQ represents a pervasive and enduring condition, caused by slow speed of processing, that does not improve through cognitive development. Savant syndrome (an isolated cognitive talent in the face of a low IQ) indicates that certain circumstances (probably an obsessive interest in a knowledge domain with a great deal of exposure to relevant material) can lead to the acquisition of complex skills (Anderson 2001).

Intelligence is described as a non-specifi ed mental ability, the defi nition of which is mentioned to have evoked much debate (see glossary by Kanninen et al. 1997). It has earlier been defi ned as an innate and unitary ability but in recent decades discovered as a capacity also involving environmental factors.

Furthermore, intelligence is suggested to consist of several distinct intellectual actions (see Kanninen et al. 1997).

A child with low IQ-level tends to work with less complex language (compare to Luria 1973, Plomin 2001). Th e more a child works and is able to work eﬀ ectively with complex material (also understanding causes and consequences), the more complex representative networks may be formed (see Luria 1973, Lucas 1980, Sakurai 1996, Seidenberg 1997). Fazio (1997) suggested that processing speed improves with practice, along with reorganizing networks.

Luria (1973) stated that higher mental processes are formed and take place on the basis of speech activity. Words are the basic units of language and form multidimensional matrices of diﬀ erent cues and connections (Luria 1973). Luria further suggested that abstract matrices gradually replace the concrete matrices. A single concept can represent a large or even enormous amount of hierarchically structured knowledge (see Kanninen et al. 1997).

Rie (1980) used a term ´integrative-conceptual response´ (in intellectual processes) to describe brain functioning in abstract thinking. Th inking based on concepts leads to more eﬃ cient mental operations and larger

(35)

consciousness (see Kanninen et al. 1997). Very recently, in her Plenary Speech in the X. International Congress of the Study of Child Language in Berlin (2005), Spelke referred to the combinatorial capacity of language: Language is able to combine knowledge in neural subsystems, which further enables the subsystems work in concert. Th erefore, language is a unique faculty aﬀ ecting development of cognitive systems.

2.2. Etiology and brain pathology in SLI

Th e ﬁ ndings of brain imaging studies of SLI children have not referred to brain damage (Plante et al. 1991, Jernigan et al. 1991). Instead, the gross brain structure has been demonstrated to be remarkably normal (Jernigan et al. 1991). Th is is in concert with the notion that gross brain malformations are typically associated with global cognitive impairment, whereas mild brain abnormalities have been suggested to be at the root of developmental disorders aﬀ ecting higher cognitive functions (Bishop 1997). SLI has been shown to be related to abnormal asymmetries of the peri-sylvian regions of the temporal lobes between the left and right hemispheres (Jernigan et al.

1991, Plante et al. 1991). Galaburda and Kemper (1979) studied at autopsy a 20 year-old man with severe reading disability and found neuronal migration anomalies. Trauner et al. (2000) showed that 12 out of 35 children with LI had abnormalities on their MRI scans, while none of the 27 control children had abnormal scans. Abnormal ﬁ ndings included ventricular enlargement (in ﬁ ve), central volume loss (in three), and white matter abnormalities (in four).

Neither did Trauner et al. (2000) postulate that the abnormalities might be an indication of a brain damage but suggested that LI is a more widespread nervous system dysfunction. Furthermore, they noted that there did not appear to be any greater likelihood of neurological abnormalities if the child had an abnormal MRI scan than if the scan was normal.

Lou et al. (1984) found regions with hypo-perfusion; activation studies failed to show the normal increase in the ﬂ ow in relevant cortical regions.

(36)

Hypo-perfusion and low metabolic activity were suggested to be due to subtle morphologic abnormalities not detectable with CT but with important pathogenic implications (Lou et al. 1984). Lashley (1963) suggested that an impaired transmission from one set of neurons to another and from one hierarchical level to another leads to a failure in the mechanisms of integrative (intellectual) functioning on the basis that the transmission does not occur smoothly or eﬃ ciently.

Genes are known to determine brain development (Bishop 1997).

Th ere is now mounting evidence for a genetic contribution in SLI (Bishop and Edmundson 1986, Bishop et al. 1995, 1999b, Tallal et al. 2001). Th e genes are assumed to aﬀ ect neuronal migration, leading to a brain that is not optimally interconnected (Bishop 1997). Other factors known to be capable of disrupting brain development during gestation, delivery or infancy, such as viral infections, lack of oxygen, or trauma, have not been thought to be plausible or at least general factors causing SLI (Bishop 1997). Neither have the extensive reviews of the medical histories identiﬁ ed consistent evidence of toxins or other environmental (chemical) factors (Trauner et al. 2000).

Tallal et al. (2001) showed that in SLI proband families, language impairment occurred in 13% of off spring (excluding proband) with neither parent aff ected, 40% of off spring with one parent aff ected, and 71%

of oﬀ spring in families in which both parents were language impaired.

Impairment rates for fathers and mothers were approximately equal, whereas rates for brothers were signiﬁ cantly higher than for sisters.

In clinical contexts it has often been questioned whether SLI results from an impoverished linguistic environment. Language development has been found to be very sensitive to deprivation (Leung and Kao 1999). However, the general rule seems to be that, after removal from deprivation, the children either improve rapidly or they do not recover the normal use of language at all (Skuse 1993). If recovery is going to occur, substantial achievements will be made within a few months (Skuse 1993).

(37)

Many children with SLI have affl uent and well-educated parents (Bishop 1997). Language diff erences seen between social classes are typically mirrored by similar size eff ects on nonverbal eff ects (Bishop 1997). Whitehurst (1997) studied children growing up in poverty; their language skills were substantially below age level, but the profi le of language impairment was quite diff erent from that typically seen in SLI. Measures of vocabulary and narrative skills were poor, but syntax scores were well within normal limits (Whitehurst 1997). Murray et al. (1996) investigated language development of depressed mothers, who are often relatively unresponsive to their infants´

communicative attempts. Th e eﬀ ect of depression was found to be relatively small and diminished with age.

An average child seems to require a surprisingly small amount of verbal stimulation in order to trigger language development (Bishop 1997). In a study by Sachs et al. (1981), who evaluated the language development of children whose parents are deaf and have limited oral language skills, the majority of the children had no problems in (spoken) language learning, provided that they heard normal speech patterns from other adults for 5 to 10 hours per week.

2.3. Otitis media with eﬀ usion and language development

Th ere has been debate about whether otitis media with eﬀ usion (OME) might cause SLI. Teele et al. (1984) suggested that most children with OME show some degree of conductive hearing loss. Th e average loss (500 to 2000Hz) from otitis has been identiﬁ ed to be 26 to 31 dB (Downs 1985). Downs stated that it is evident that children with such loss will not hear all the speech sounds. Furthermore, Downs suggested that because language is learned by audition, any reduction in the quantity and quality of language input during critical periods of development amounts to auditory deprivation.

(38)

Roberts (1997) referred to a low phonetic substance (salience relative to the surrounding phonetic context) of some words or linguistic elements.

According to Downs (1983), softer speech sounds and voiceless consonants, in particular, may be missed or confused when hearing loss is as little as 10–

20 dB. Roberts (1997) and Petinou et al. (2001) stated that hearing levels that fl uctuate over time because of OM might cause the low-phonetic substance elements to be periodically unavailable to the neural networks, which would have broader implications for the acquisition of language. Th e phenomenon was thought to be amplifi ed in noisy environments. Roberts (1997) futher suggested that variable availability over time may disrupt the processing of important correlations among terms and their meanings. Roberts concluded that it seems reasonable to consider that the eff ects of a history of fl uctuating hearing loss due to OM may be an important factor in SLI.

Klein et al. (1984) found that a large number of days (in their study more than 130 days per year versus fewer than 30 days per year in the control group) with a conductive loss in the early years had caused language-learning delays. In a group of lower SES, however, there were no diﬀ erences between the OME group and the control group. According to Klein’s group, 33 per cent of all children will have three or more episodes of OME by 3 years of age, 24 percent by 2 years, and 17 per cent by 18 months. Klein et al.

concluded that it is no wonder that more and more children are found in learning disabled classes.

Teele et al. (1984) also found that children who had spent prolonged periods of time with middle ear eff usion had signifi cantly lower speech and language scores when compared with those who had spent little time with middle ear eff usion. Th e correlation was strongest in children from higher socio-economic strata. Time spent with middle ear eff usion in the fi rst 6 to 12 months of life was most strongly associated with poor scores (Teele et al. 1984). In a later study by Teele et al. (1990) no associations were found between ear disease occurring during years 4–7 and performance at age 7. In

(39)

fact, as they concluded, had they assessed ear disease during this period only, they “would probably have concluded that no associations existed between ear disease and intellectual and linguistic abilities.”

Luotonen et al. (1996, 1998) evaluated the eﬀ ects of of early recurrent otitis media on linguistic skills and school performance at the age of 9 years.

Early recurrent otitis media (before the age of 3 years) was associated adversely with scores on the reading comprehension test and with mathematical skills, with classroom concentration and oral performance, as rated by the teachers. No association was found between otitis media episodes after the age of 3 and test performance or school achievement. However, Updike et al.

(1992) found signiﬁ cant diﬀ erences in performance on all tests of auditory processing ability and reading ability, although the children studied were already 6 to 7 years old.

Opposite results have also been found (Hubbard et al. 1985, Roberts et al. 1986, Wright et al. 1988). Roberts et al. (1986) concluded that there is no association between ear disease and verbal and academic performance in later life. Bishop (1997) stated that it would be premature to rule out any role of OME in causing language impairments. In those cases where the condition is chronic, sometimes leading to perforation of the eardrum, and associated with conductive hearing loss lasting months, there is evidence of more persistent deﬁ cits. However, middle ear disease is not an adequate general explanation for the cause of SLI (Bishop 1997).

2.4. Factors underlying SLI

2.4.1. General aspects

Orton (1937), one of the fi rst researchers to study LI, suggested that the root cause of aphasic children’s language inadequacies is a diffi culty with temporal ordering. He asked whether LI children, having disability in reproducing sequentially ordered auditory stimuli such as sentences, series of numbers, or series of sounds, might have a “diffi culty in sequence building”. In 1980

(40)

Weiner stated that the hypothesis of temporal ordering had received more attention from investigators of developmental language disorders than had any other.

Despite years of research, there is little general agreement about the most critical mechanisms underlying SLI (Gillam et al. 1998, Rapin and Allen 1998, Bishop 2003, Hartley et al. 2003). A wide variety of linguistic and cognitive defi cits have been proposed to explain the diffi culties (Bishop 1997). Competing models place the underlying defi cit within limits on, for example, the child’s information processing abilities, symbolic ability, or memory capacity (Rescorla and Ratner 1996). It has also been suggested that expressive and/or receptive delay may have its basis in limitations on speech motor production (Whitehurst et al. 1991, Stark and Heinz 1996). Some researchers claim that SLI arises from impairment in an innate grammatical module (see Gopnik and Crago 1991, Van der Lely 1994).

2.4.2. Auditory defect

In 1965 Lowe and Campbell hypothesized that there is a perceptual defi cit at the root of LI. Eisenson (1968) proposed that children with LI have diffi culty in identifi cation and discrimination of auditory stimuli, including speech sounds. Tallal and Piercy (1974) concluded that it is the brief duration of formant transitions that results in dysphasic children’s inability to discriminate consonant stimuli, and this defi cit may be suffi cient to explain the speech disorder of these children. On all tasks studied, dysphasics´ discrimination of consonant stimuli was signifi cantly inferior both to their discrimination of vowel stimuli and to their discrimination of nonverbal auditory stimuli of the same duration. However, there was no marked diff erence in discriminating when consonants were artifi cially stretched, or vowels artifi cially shortened (Tallal and Piercy 1975). It has subsequently been found that temporal processing of language can be improved by training, by artifi cially stretching brief elements of speech (Merzenich et al. 1996).

(41)

Tallal and Piercy (1973) trained children to associate each of two complex tones easy to discriminate, presented in isolation, with a speciﬁ c key-press response. Once this association had been learned, an Auditory Repetition Test (ART) was given: sequences of two tones were presented, and the child had to press the corresponding keys in the correct sequence. Children with SLI performed accurately when the tones were separated by an inter-stimulus interval (ISI) of over 300ms, but their performance deteriorated when shorter ISI was used. In contrast, control children maintained a high level of performance at shorter ISI, and performed above chance levels with ISI as brief as 8ms. Th e same pattern of results was obtained when the child was required to indicate if two tones in a sequence were the same or diﬀ erent (Bishop et al. 1999a).

A matter of controversy is whether diffi culties in segmenting, discriminating, and identifying speech sounds have their basis in a more fundamental auditory perceptual defi cit that aff ects the processing of all sounds, not just speech (Bishop et al. 1999b). According to a current version of the theory, the impairment is not seen as specifi c to the auditory modality (Bishop et al. 1999b). However, as Bishop’s group suggested, this multi- modal rapid processing defi cit is thought to have an especially severe impact on language development, which is crucially dependent on the ability to distinguish and identify brief and rapid auditory events.

Not all studies support the hypothesis that impaired auditory temporal processing is a major cause of SLI (Bishop et al. 1999a, Reynolds and Fucci 1998, Rosen 2003). Under certain experimental conditions, language- impaired children seem to show adequate discrimination of brief or rapidly changing auditory stimuli (Tomblin et al. 1995, Helzer et al. 1996, Sussman, 1993).

In 1999 Bishop et al. (1999a) found no link between auditory deﬁ cit and SLI. Th ey speculated that it is possible that they included a disproportionate number of cases whose diﬃ culties had a non-auditory basis.

(42)

Bishop et al. (1999b) demonstrated that normally developing children may also do poorly on the Auditory Repetition Test of Tallal and Piercy (1973). Bishop et al. (1999b) argued that the fi nding of children with normal language level but poor auditory processing poses diffi culties for any theory that regards auditory defi cits as a necessary and suffi cient cause of LI.

Furthermore, the hypothesis that SLI is a single disorder, having its basic pathological-genetic deﬁ cit in rapid auditory processing, for example, has been unacceptable to neurologists and others (Rapin and Allen 1998).

According to Rosen (2003), auditory deﬁ cits appear not to be causally related to language disorders, but only occur in association with them.

Rosen reviewed the available literature and concluded that typically only a minority of SLI listeners exhibit any auditory defi cits, and there is little or no relationship between the severity of the auditory and language defi cits in SLI groups. Furthermore, control groups sometimes exhibit stronger relationships of this type. Rosen (2003) argued that the claim that the defi cit is specifi c to rapid temporal processing is almost certainly wrong.

Nevertheless, (Bishop et al. 1999a) stated that it would be dangerous to conclude that auditory defi cit plays no role in the etiology of LI. Bishop’s team found a weak but statistically signifi cant association between LI and auditory impairment. Th ey concluded that such a small eff ect, which would be hard to detect in small samples with low power, could arise if impaired auditory temporal processing were a moderating variable, which is neither necessary nor suffi cient to cause LI, but which only exerts an eff ect on language development in children who are already at genetic risk. Th is could account for their observation that some children showed auditory defi cit but normal language development, because it predicts that auditory defi cit should assume importance only in those at genetic risk (Bishop et al. 1999a).

Finally, as Bishop et al. (1999a) suggested, although tasks such as Tallal´s ART diﬀ erentiate children with SLI from matched controls, it is not the case that all children with SLI are impaired on such tests. SLI is a

Diagnosing Specific Language Impairment