Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Journal of Speech and Hearing Disorders,Volume 55, 612-620, November 1990

WHO SHALL BE CALLED LANGUAGE DISORDERED? SOME


REFLECTIONS AND ONE PERSPECTIVE
MARGARET LAHEY
Emerson College, Boston

This paper discusses some issues involved in identifying children who have language problems. The perspective taken is that
(a) the goal of identification must be clearly distinguished from other goals of assessment; (b) identification of children with
language disorders is better based on language performance than on inferences about the language knowledge that underlies this
performance; (c) language performance must be sampled in more than one context, including, for purposes of identification,
contexts that stress the language system; (d) the standards of expectations for comparing performance and determining differences
must be explicit; (e) standards used to determine differences are better based on the performance of chronological-age peers than
on the performance of children with similar mental abilities; and (f) children who do not evidence poor language performance but
are considered at risk for language-related problems should be distinguished from children who demonstrate poor language skills.

KEY WORDS: language assessment, language disorders in children, age-equivalence scores, identifying language disorders,
cognition and language

The criteria and procedures for identifying children based on the performance of chronological-age peers than
with language disorders appear to vary widely among on the performance of children with similar mental abil-
research and clinicians. This diversity creates confusion ities; and (f) children who do not evidence poor language
about the meaning of the term language disorder-con- performance but are considered at risk for language-
fusion among clients, students, service-providing agen- related problems should be distinguished from children
cies, and schools, as well as speech-language patholo- who demonstrate poor language skills. These points are
gists. The use of varying criteria could interfere with our further discussed below.
ultimate understanding of language disorders. Further-
more, in these days of increasing health and education GOALS OF ASSESSMENT
costs and decreasing availability of funds, unclear criteria
are more and more likely to be questioned by those called
upon to pay for services to language-disordered children As speech-language pathologists, we are often asked to
(see Frattali, 1990, for a discussion of a similar problem assess children who someone (parent, teacher, doctor,
with defining quality of service). etc.) has hypothesized may be having difficulty learning
Although we do not now all agree on a particular set of language or using language to communicate. One goal of
criteria for defining a language disorder, most of us would such assessment is to confirm or reject this hypothesis.
agree that we need to be explicit about and critically This goal can be viewed as identification.
Identification is followed by careful description of the
evaluate the criteria that are used. Such examination may
allow us to eliminate those criteria that cannot be justified child's communicative behaviors, as we attempt to deter-
mine, for example, what changes could be made, what
or do not prove useful, and perhaps gradually reduce
some of the confusion that exists about who shall be impact these changes might have, and what the reason for
the problem is. Identification of a language disorder is not
called language disordered.
This article evolved from a critical examination of a a statement about etiology or the nature of a problem.
Furthermore, it does not imply that we can help a child
frequently used criterion for identifying children with
nor that we must continue working with a child until the
language problems-a criterion based on a discrepancy
child no longer manifests language problems. Procedures
between language age and mental age. Isolation of that
for identification of a language disorder are different from
one criterion proved impossible without discussion of
those used to determine goals of language intervention;
more general issues related to assessment, language, and
what makes a difference. The perspective taken here is they are different from those used for understanding the
nature of language disorders. Even with these limits,
that (a) the goal of identification must be clearly distin-
however, identification is not easy and is fraught with
guished from other goals of assessment; (b) identification
controversy. Most of the controversy involves issues
of children with a language disorder is better based on
about how we view language and how we determine
language performance than on inferences about the lan-
what a "disorder" is.'
guage knowledge that underlies this performance; (c)
language performance must be sampled in more than one
context, including, for the purpose of identification, con-
texts that stress the language system; (d) the standards of 'Throughout this article I refer to children with language
expectations for comparing performance must be explicit; problems as "language disordered," in line with common usage
(e) standards used to determine differences are better (e.g., Aram & Nation, 1982; Bloom & Lahey, 1978; Carrow-

© 1990, American Speech-Language-Hearing Association 612 0022-4677/905504-061201.00/0

Downloaded From: http://jshd.pubs.asha.org/ by a Universite Laval User on 09/11/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
LAHEY: Language Disorders 613

SOME ISSUES RELATED TO been made by many; examples are Sapir's (1921) distinc-
LANGUAGE tion between "langue and parole" and Chomsky's (1972)
distinction between "competence and performance."]
What is observed is the form of communicative behaviors.
No agreed-upon theory of language or language devel-
opment is available to help us in the identification of However, researchers within the mentalistic perspective
language disorders. We cannot wait for such a theory to are concerned with describing development by inferring
resolve some of the problems associated with identifica- changes in the knowledge underlying the observed be-
tion. In fact, an understanding of children who are having haviors (e.g., Bloom, 1970; Bowerman, 1973; Brown,
difficulty learning and using language may well inform 1973).
theories of language and of language acquisition. An Forming taxonomies and counting behaviors continues
understanding of these children is, however, dependent even for those who operate with a mentalistic view of
on identifying them. language. However, the taxonomies formed, the purpose
Early studies of language yielded taxonomic descrip- of forming and counting, and the interpretations of the
tions of the linguistic forms used for expression. Devel- data have changed. At first, the inferred structures were
opmental studies tended to focus on listing the forms that related primarily to the form of language (in particular,
were found at different ages (e.g., Templin, 1957). These rules of syntax), but in the past two decades they have
studies either were atheoretical descriptive studies or also included interpretations of linguistic form and other
were based on a maturational view of development (i.e., behaviors (e.g., gaze, facial expressions, actions on ob-
that the sequence of development is predetermined by jects) so that aspects of language content and use are
genetic code) (Bloom & Lahey, 1978; Kuhn, 1988). The inferred and categorized. Rather than ages at which
data collected in these studies are still used to refer to certain categories of behavior are expected, the new
developmental milestones in language development. taxonomies and their quantification are used for other
Another view of development, behaviorism, was inter- purposes, including the following: inferring the structure
ested only in the observable behaviors and, thus, also underlying the form; inferring the emerging knowledge
focused on the form of language. In this view, develop- of a rule versus the "mastery" of a rule; inferring pro-
ment is a product of environmental influences. Complex cesses related to language learning; examining the influ-
behaviors are considered a constellation of smaller be- ence of context on the production of particular forms; and
haviors that must be isolated and studied in order to comparing description of expression and interpretation
understand the complex behaviors. Research about de- through language with behaviors of different kinds (e.g.,
velopment often involved laboratory studies of learning, Bates, 1976; Bloom, 1970; Bloom & Beckwith, 1989;
in particular paired-associate learning. Learning was Brown, 1973). Although the emphasis on the role of the
viewed as a cumulative process, and the counting of environment in learning is controversial, there is consen-
forms as a measure of development is consistent with this sus that environment and the context of observations
perspective. influence what behaviors are observed and when partic-
In the developmental psycholinguistic literature that ular behaviors appear to emerge.
has predominated since the 1960s, there has been a Like the researcher in normal child development, the
marked trend away from a focus on taxonomies of form, clinician also describes and interprets behavior with the
counting forms, and providing ages associated with the intent of understanding the underlying system and pro-
production of these forms. Language is viewed as knowl- cesses that are not directly observable. Like the re-
edge that underlies the observed behaviors. In contrast to searcher, the clinician has only the data of performance.
a behavioral perspective, it is a mentalistic perspective. But, unlike developmental psycholinguistic researchers,
This perspective has most often considered development clinicians are also asked to identify children who are
in terms of the interaction between the child's genetically having problems learning and using language. One con-
determined endowment and the environment. Differing cern is whether such identification should be based on
views exist concerning the relative role of the child the manifestation of language knowledge (i.e., compari-
versus that of the environment. sons of language performance with some expectation of
What children know about a linguistic system or about performance) or on the inferred underlying knowledge
communicating with language is not directly observable. system (as compared with some expectation for that
They manifest that knowledge through acts of expression system).
and interpretation. [This distinction between language The perspective taken here is that, even within a
knowledge and the manifestation of that knowledge has mentalistic view of language, identification of a language
disorder should be based on differences in language
behaviors (e.g., the production and comprehension of
Woolfolk, 1988; Cole, 1982; Committee on Language Learning units of connected language) and not on inferences about
Disorders, 1989; Holland, 1984; Hubbell, 1981; Lahey, 1988) the nature of the underlying system based on these
and because we in the field of speech-language pathology performance variables. One reason to focus on perform-
usually refer to the various areas of our field as disorders (e.g., ance is the practical consideration that there is no agreed-
voice disorders, phonological disorders, speech disorders). Al-
though some other term (e.g., disability, delay, impairment) may upon way to describe the underlying system and, thus, no
be more appropriate, such discussion is not the purpose of this way that a description of the expected system could be
article. used as a comparison. More important is the fact that we

Downloaded From: http://jshd.pubs.asha.org/ by a Universite Laval User on 09/11/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
614 Journal of Speech and Hearing Disorders 55 612-620 November 990

are interested not only in children who have deficits in nature of the problem and planning intervention. Thus,
their underlying knowledge of language but also in those despite the fact that most of us operate within a mental-
who might have difficulty in accessing or using that istic perspective and are interested in making inferences
knowledge. about the underlying language system, it seems we
What children can do with the knowledge they have should define language disorders in terms of differences
may be as important for prognosis and intervention as from some expectation about language performance.
what they know about the linguistic system. Transitions
associated with changes in curriculum as well as literacy
demands may result in previously undisclosed difficulties SOME ISSUES RELATED TO BASIS
with language tasks for certain children (Bashir, Wiig, & OF COMPARISON
Abrams, 1988). A few studies have indicated that some
children who score within normal limits on standardized Perhaps the most controversial issues in identification
language tests and who carry on acceptable casual con- have to do with the standards upon which we base our
versations have difficulty learning to read (Liebergott, comparisons and thus judge who shall be called language
Menyuk, & Chesnick, 1988). Many of these children are disordered. The past two decades have provided us with
also different from peers on certain language processing the information necessary for improving our descriptions
tasks. For example, some have difficulty on language of children's performance and with developmental se-
tasks that emphasize speed and accuracy of language quences that are helpful as a standard of comparison for
processing (e.g., speed of naming objects, accuracy of determining the goals of intervention. However, we have
rapid multisyllabic word repetition) or metalinguistic not come to terms with a standard of comparison for the
skills (e.g., phoneme segmentation, doze tasks) (see, e.g., purpose of identification.
Kamhi & Catts, 1986; Liebergott et al., 1988). Further- We always compare a child's performance with some
more, evidence of excessive numbers of dysfluencies standard of reference (implicit or explicit) when identify-
(e.g., pauses, false starts, interjections) in connected ing difference or "disorder." Perhaps the greatest danger
speech have been found predictive of, or negatively comes from the use of intuitive or vague alternatives that
correlated with, academic success (e.g., Damico, Oller, & are not open to examination and lead to confusion as to
Storey, 1983; MacLachlin & Chapman, 1988). Problems what we mean by the term language disorder.
with comprehension versus production skills may also Although the standards of expectations we use for
cause children to differ in performance relative to peers- identifying a language disorder are not always explicit,
some do as well as peers on one type of performance but they are usually based on the communicative behaviors of
more poorly than peers on the other (e.g., Stark & Tallal, other children in the child's culture who either are at the
1981). Although some of these discrepancies may be same level of cognitive development or are of similar
related to differences in the knowledge base underlying chronological age. Even the standard of effective commu-
language, others appear to be related to processing (i.e., nication or conspicuousness (e.g., Van Riper, 1954) is
in accessing or using) that knowledge. It is not always usually determined from the perspective of a child's age
easy, or possible, to differentiate the two. or level of mental ability.
These variations in the manifestation of language One frequently used method of comparison is to com-
knowledge suggest that identification of language disor- pare mental age (MA) and language age (LA) and consider
ders should include descriptions of how children use differences of 6 months to 2 years as evidence of a
language and should not be based solely on inferences disorder (e.g., Stark & Tallal, 1981). The estimate of
about the nature of the knowledge. Performance needs to mental ability encompassed in MA is usually derived
be observed in a number of contexts and cannot be from the accuracy of responses on a test or subtest of
limited to current standardized instruments that describe intelligence, and the estimate of language level is usually
the production and comprehension of certain lexical and derived from the accuracy of responses to tests of lan-
syntactic forms in limited noncommunicative contexts or guage or by comparison of language performance to
to a sample of language in casual conversation. Proce- developmental milestones (e.g., use of a language scale).
dures of identification also need to include conditions Multiple problems exist in using MA-LA discrepancy as
that might stress the language system (such as speeded a criterion. Some of the problems relate to measurement
performance or narration of complex ideas). For such concerns, such as the problems inherent in age-equiva-
procedures to be useful in identifying language-disor- lence scores and in comparing any two measurements.
dered children, we need information on how large num- Other problems have to do with the use of mental
bers of children perform on similar tasks so that we are abilitiesas a comparison, including deciding which men-
aware of the variability in performance among children tal abilities to use and the theoretical relation between
and so that we have a standard of comparison. language and mental abilities.
In the perspective presented here, description of best
performance and of differences between best perform-
ance and performance under stressed conditions are not Measurement Concerns
necessary for identifying children who are having diffi-
culty with language. Such descriptions are more relevant Equivalencescores. The equivalence score is the frame
for other goals of assessment, such as understanding the of reference used when one describes children's perfor-

Downloaded From: http://jshd.pubs.asha.org/ by a Universite Laval User on 09/11/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
LAHEY: Language Disorders 615

mance in terms of language age, mental age, or grade equivalence measures in our journals and other official
level. Such scores generally refer to a measure of central publications, and their use should be discouraged in our
tendency (usually the median or mean) of a segment of clinic and school reports.
the standardization sample. For example, to say that a If we cannot refer to level of development in terms of
child has a language age of 4 years means that the score age or grade equivalence, what are the alternatives?
the child received on a measure of language is the Description of a child's performance relative to the vari-
median, or mean score, received by the 4-year-old chil- ability found in a comparative population is more accept-
dren in some standardization population. This method of able than age equivalence. Percentile ranks,2 standard
describing "normal" development and deviance from scores, or number of standard deviations from the mean
normal is extremely problematic. are indices commonly used in descriptions of language
Equivalence scores usually do not compare children behavior. Such scores are available for most of our re-
with a population to which they belong. For example, if a cently published language tests. Furthermore, some de-
5-year-old child receives a language age score of 4 years, scriptions of language samples can be reported in terms of
this tells us about the child's performance in relation to an number of standard deviations from the mean for children
estimate of the "average" 4-year-old, but does not tell us of different ages (e.g., mean length of utterance as pre-
about the child's performance in relation to other 5-year- sented by Miller & Chapman, 1981; aspects of syntactic
olds. A second problem with equivalence scores is that complexity as computed by Lee, 1974; number of dif-
they do not represent linear units. That is, each month of ferent words using data by Templin, 1957). In the future,
difference in age is not equivalent in terms of number of the use of computerized databases may enable us to
errors, either within a test or between tests, and a lan- provide more varied descriptions in such a manner. In
guage age unit (of months or years) cannot remain con- any case, reporting performance in a way that enables us
stant with varying chronological ages. A 1-year difference to interpret a child's relative standing with peers is
in age score when a child is 3 years old is not equivalent preferable to describing the performance relative to the
to a 1-year difference when the child is 9 (see Anastasi, central tendency of some age group (i.e., an age equiva-
1982, for further discussion). Interpretation of "how much lent).
of a difference makes a difference" is, therefore, difficult. Comparingtwo observations. Even if we refer to stan-
A third problem is that equivalence scores are easily dard deviations or standard scores as an estimate of
misinterpreted. For example, a third grader who receives mental ability and of language, further measurement
a reading grade-level score of sixth grade has done very problems arise when we attempt to determine differences
well on the third-grade test, probably as well as a sixth between the two observations. All observations (be they
grader might have done had he or she taken the third- informal or formal) are subject to multiple sources of
grade test. (Note, however, that often such scores are error. A significant portion of the error that relates to
obtained by extrapolation rather than by actual adminis- contexts and variability of the child could be eliminated if
tration of the test to a sixth-grade group.) It is unlikely that it were possible to observe the child all the time rather
the third-grade child reads as well as a sixth grader or than sample the child's behavior. Because observing the
could score anywhere near the 50th percentile on a entire universe of the child's behaviors is impractical, we
sixth-grade reading test (see Cronbach, 1984, for further need to be concerned with reducing sampling error and
discussion). The skills tested at each grade level are quite estimating its influence. Some of this observational error
disparate, and the use of grade equivalence scores masks is estimated by measures of reliability. In general, the
these differences. larger the sample of behaviors observed, the higher the
These and other problems with equivalence scores reliability. The standard error of measurement (SEM) is a
have been discussed by many professionals in our own way of taking this source of error into account when
literature (e.g., Bloom & Lahey, 1978; Lahey, 1988; Lund interpreting an observation. It can be computed if reli-
& Duchan, 1988; McCauley & Swisher, 1984) and in most ability and standard deviations are available for the ob-
texts on measurement (e.g., Anastasi, 1982; Cronbach, servational technique of interest. Many standardized tests
1984; Salvia & Ysseldyke, 1981), yet they continue to be report SEM, and it should be used to estimate the band
used. A cursory review of articles that appeared in the around which an individual child's true score probably
Journal of Speech and HearingDisordersand theJournal falls.
of Speech and Hearing Research between December In comparing the results of observations of language
1984 and December 1988 revealed that approximately with those of mental abilities, apparent differences or lack
40% of the articles that concerned language learning of differences could be a function of error in the observa-
problems in children (as evidenced by their titles) de- tion or description of each. Observation techniques with
scribed the language-disordered child using only equiv- lower reliability need greater differences, whereas those
alence scores. Another 20% used equivalence scores with higher reliability require less. To determine the
along with standard scores. Furthermore, the American
Speech-Language-Hearing Association (ASHA) sells a
poster suggesting that the parent "Find your child's 2
speech and hearing age," (advertised in Asha, June 1987, Percentile ranks provide a comparison with peers in terms of
relative standing and in that sense are better than equivalence
p. 70). Because of the problems mentioned above, it may scores, but they are less useful than standard scores because the
be that we should call a moratorium on the use of all differences between ranks do not represent equivalent units.

Downloaded From: http://jshd.pubs.asha.org/ by a Universite Laval User on 09/11/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
616 Journal of Speech and Hearing Disorders 55 612-620 November 1990

likelihood that disparate findings reported for different Other researchers and theorists have focused on infor-
observations are not due to errors in observation or mation processing components of task performance (e.g.,
description, we can calculate the SEM of the difference Case, 1985; Sternberg, 1985), or mental structures (e.g.,
between the two measures (see, e.g., McCauley & Piaget, 1954) that cut across domains of knowledge and
Swisher, 1984; Powell & Peng, 1989; Reynolds, 1981; are purported to be involved in all intellectual activity. In
Salvia & Ysseldyke, 1981). The SEM of difference should essence these approaches focus on how children solve
be reported and considered in research studies or clinical problems. They suggest that the strategies used by the
reports that identify a language disorder based on com- child are more important than whether or not the solution
parisons of level of language with level of mental ability; was correct. Development is often viewed as a reorgani-
it is not usually referred to in journal articles or clinical zation of strategies or structures that enables more effi-
reports that use mental abilities as the comparison. Esti- cient problem solving. More recently, emphasis has fo-
mates of the standard error of the difference are not cused on the metacognitive understanding of when to
available in test manuals and need to be computed by the apply which strategies (e.g., Kuhn, 1988; Muma, 1986).
clinician or researcher for comparisons of different mea- Information about the individual's real-world adaptation
sures. Although it is essential to determine whether (i.e., the individual's sociocultural context) or about how
differences between two measures are significant, the a person handles novel versus automated experiences is
presence of significant differences does not indicate the considered more important than the accuracy of the
frequency of such differences in the population (although person's responses to particular test items (Sternberg,
this can be estimated given the intercorrelation of the 1985). However, scores on traditional intelligence tests
measures) or whether the differences are of any theoret- are based on the accuracy of a child's response. That is,
ical significance (Reynolds, 1981). Thus, the use of any they are a measure of the product of cognitive activity, not
standard of reference for defining a language disorder that with how a child arrived at the product (Sternberg, 1985).
relies on differences between the two sources of obser- Theoretical relation between language and mental
vation (such as language and mental ability) involves ability. A further problem with the use of scores on
potential compounding of measurement error. In addition intelligence tests as a reference point for identifying
to these measurement problems, there are theoretical children with language learning problems relates to the
reasons why comparing language level with level of theoretical relation between language and mental abili-
mental ability is problematic. ties (for review, see, e.g., Rice & Kemper, 1984; Siegel,
Katsuki, & Potechin, 1985). Tasks on intelligence tests are
not theoretically motivated in terms of language develop-
Concerns About the Use of Mental Abilities as a ment. Measures of responses to intelligence tests (e.g.,
MA or IQ) are often referred to as an index of cognitive
Comparison
development, but they do not represent the view of
cognitive development that has most often been studied
Which mental abilitiesdo we consider? Although some relative to language development. Most studies that have
models and theories of intelligence define intelligence in related language learning to cognitive development have
terms of a unified capacity, others suggest that intelli- focused on how the child represents the world and the
gence is composed of distinct mental abilities (e.g., ways in which the child can manipulate those represen-
Gardner, 1985; Wing, 1981; see Weinberg, 1989, for a tations. They have defined cognition more in terms of
review). In fact, Gardner (1985) has suggested that there Neisser's (1967) definition, that is, as the processes by
are several "types of intelligence," each of which has a which sensory input is transformed, reduced, elaborated,
certain degree of autonomy, as evidenced by prodigies stored, recovered, and used. Thus, developmental psy-
and by the differential effects of brain damage. This view cholinguists have not been concerned with MA, IQ, or
of multiple intelligences leads to the perplexing problem any measure of the accuracy of responses; rather, they
of deciding which particular types of intelligence should have been concerned with how the child arrives at those
form the basis of comparison with language skills. Using responses.
Gardner's types, we might ask whether linguistic intelli- Most of this work in developmental psycholinguistics
gence should be compared with logical-mathematical, has been influenced by the writings of Jean Piaget, who
personal, bodily-kinesthetic, musical, or spatial intelli- was interested in how a child formulates and solves
gence. Few tests measure personal, bodily-kinesthetic, or problems. Many studies have looked at correlations be-
musical intelligence. According to Gardner (1985), most tween aspects of sensorimotor development, as described
of our traditional intelligence tests measure linguistic by Piaget, and early lexical acquisition. For example, a
and/or logical mathematical domains. For the purpose of number have reported that the child's earliest words
identifying a language disorder, is there any reason to appear in sensorimotor Stage 5 (e.g., Bates, 1976). How-
assume that a gap between linguistic and logical-mathe- ever, comparisons of language skills and Piaget's descrip-
matical intelligence is any more important than a gap tions of early cognitive development have revealed
between linguistic and any other type of intelligence? If mixed results. Some studies report positive correlations
autonomy exists among domains of intelligence, is there between object permanence and types of words learned
any value in identifying a problem by comparing one (e.g., McCune-Nicolich, 1981); others have found none
domain with another? (e.g., Corrigan, 1978). These correlations appear to vary

Downloaded From: http://jshd.pubs.asha.org/ by a Universite Laval User on 09/11/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
LAHEY: Language Disorders 617

with the level of language development and the aspect of measurement problems. As a consequence, the problem
sensorimotor development measured (e.g., Bates, 1976; of estimating the compounded effect of measurement
Thal & Bates, 1988). error from two observations is eliminated. The use of CA
The reported correlations have been criticized on a does not, of course, eliminate the need for considering
number of grounds (see, e.g., Bloom, Lifter, & Broughton, observational error in the description of language. An-
1985; Lifter & Bloom, 1988). For example, it is not clear other advantage is that the use of CA as a standard does
that scales of sensorimotor development (e.g., Uzgiris & not imply any unidirectional relation between cognition
Hunt, 1975) are the best estimate of the development and language. Furthermore, the use of CA suggests that
described by Piaget. Most scales, again, tend to focus on all children with poor language skills should be consid-
the product rather than on the means by which the child ered for intervention services even when their language
solves the problem (see Bloom et al., 1985, for a review). skills are similar to those expected, given a measure of
Finally, and most important, traditional intelligence tests mental ability. This perspective was supported in a re-
do not measure this type of cognitive development. No cently reported intervention study in which children with
theoretical basis exists for assuming that the score on an MA equal to LA made as much progress in language
intelligence test (except extremely low scores) should learning as children whose MA was higher than their LA
predict current level of language learning, or that it (Cole, Dale, & Mills, in press). Although many agree that
should be the point of comparison for determining children whose poor language skills are equivalent to
whether a child has a language disorder. 3 their mental abilities should be considered for language
Until we have a better understanding of the relation intervention (e.g., Leonard, 1983; Subcommittee on Lan-
between measures of language and measures of mental guage and Cognition, 1987) others do not provide direct
abilities, we must be cautious when using any description programs for such children (see, e.g., Lyngaas, Nyberg,
of mental ability as the reference point in the identifica- Hoekenga, & Gruenewald, 1983).
tion of language disorders. The confusion may be related One danger in the use of CA is that it too often gets
to methodological problems in measuring both language compared to a language age as, for example, in scales or
and cognition (see Bloom et al., 1985, for further discus- references to developmental milestones. The problems
sion of methodological issues). On the other hand, the with the use of age-equivalence descriptions were dis-
hypothesis that cognitive development sets the limit on cussed above. Considerable variability exists among chil-
language learning may be too vague to be correct or may dren in the age at which different language behaviors are
be incorrect. Certainly language itself ultimately influ- manifested. Comparisons of a child's language perfor-
ences cognition, although when that happens is not clear. mance with that of CA peers need to consider the vari-
The point here is not to argue against the possibility that ability of performance among these peers. To know how
cognitive development leads and limits language devel- a child's language performance compares with that of age
opment at some time in development. That issue is yet to peers requires the use of measures that describe a child's
be resolved. Rather, the argument is that evidence based relative standing (e.g., the number of standard deviations
on current tests of intelligence does not relate to such an from the mean or standard scores). On behaviors for
hypothesis. Given these theoretical concerns about the which there is a wide variability among children, the
relation of measures of language to measures of mental deviation scores will reflect greater differences from the
abilities, and given the concerns about compounded error mean score.
when any two measures are compared, the perspective But taking account of this variability does not tell us
taken here is that language disorders should not currently how far below average (i.e., how many standard devia-
be defined by comparing a level of language ability with tions below the mean) a score must fall to be considered
a measure of mental ability. as evidence of a language disorder. A number of research-
ers and clinicians have settled on somewhere between
one and two standard deviations below the mean (e.g., a
Other Standards of Comparison standard deviation of 1.5) or below the 10th percentile
as a temporary solution (e.g., Lee, 1974; Swisher &
Chronological age as a standard of reference. When Demetras, 1987). To my knowledge, no one has placed
chronological age (CA) is used as a standard of reference, such a criterion in the framework of the standard error of
a child's language performance is compared with that of measurement. If the normative population is made up of
children born at a similar time rather than that of children only "normal" children (as is the case for most of our
with similar scores on measures of mental ability. One language tests), it would seem that the criterion of 1.5
advantage of using CA is that it is rarely subject to standard deviations below the mean should be the high
point within the band of confidence interval-that is,
given the standard error of measurement, the chances are
likely that performance did not fall above -1.5 standard
3
Certainly, the use of intelligence tests has been critiqued by deviations.
many (see, e.g., Baumeister & Muma, 1975). The position here is Until we have evidence of language performance on
not that intelligence tests (or any measures of the product of
intellectual activity) are not useful for some purpose, only that randomly selected children of particular ages (i.e., not
they are not useful as a comparison for expected language just on those who have been determined to be nondisor-
performance. dered), it could even be argued that any performance, no

Downloaded From: http://jshd.pubs.asha.org/ by a Universite Laval User on 09/11/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
618 Journal of Speech and Hearing Disorders 55 612-620 November 990

matter how poor, that is also observed in non-language- learning or use. (We could call them "at-risk for language-
disordered CA peers cannot be used as evidence of a related problems.") In many ways such a distinction
language disorder. For example, on a test, the lowest would be preferable to the current confusion, which
score obtained by a non-language-disordered child would mixes the two. It also is in line with the categories being
be used as a criterion; only those scores that fell below developed for services to infants and toddlers. However,
the lowest score attained by a nonimpaired child would an "at-risk" category could be a wastebasket and a name-
be used to support the hypothesis of a language disorder. playing game if vague clinical hunches are used without
Eventually, it is possible that our standard of compari- explicit specification of the exact conditions that place
son and the degree of difference from this standard will that child at risk, including a rationale for why the child is
be empirically determined by longitudinal data that pro- at risk. Only when such conditions are made explicit can
vide information about clusters of variables that make a we begin to test these hypotheses.
difference. That is, what deviations on what language Still another potential solution to the problem is to
behaviors coupled with what other information suggest serve all children and not just those identified as lan-
that a child will have trouble in school and in life? (See guage disordered. A number of our colleagues are rede-
also Tomblin, 1983, as cited in Fey, 1986.) Such a solution fining their role as speech-language pathologists within
will not be reached, however, if we continue to ignore the the school environment (see, e.g., Miller, 1989). They are
variability among children by using equivalence scores or acting as consultants who work to facilitate the commu-
if we confound the problem by comparing two sources of nication skills of all children and not just those who have
measurement and assume some theoretical tie between been diagnosed as disordered. This move was not moti-
intelligence scores and language performance. In the vated by the problem of defining language disorders, but
meantime, the use of measures that reflect the wide range it might serve to eliminate the problem where applied.
of language performance among children (in contrast to The model seems more appropriate to the school than the
age equivalences that reflect only central tendencies) clinic, but in any case it will eliminate the problem only
serves constantly to remind us of the variability within if funding for the services of speech-language patholo-
the population of "normal" language learners and makes gists is not tied to the number of children who are called
us cautious in the use of any one observation to label a "language disordered."
child as "disordered."
Although I am suggesting that observations be reported
in a way that takes into account the variability in normal SUMMARY
children, I am not stating that standardized tests are
essential nor that they are the only way to observe Some issues concerned with the identification of chil-
language performance. Descriptions of nonstandardized dren who have difficulty learning or using language have
procedures, such as language samples, can also poten- been presented. The perspective presented distinguishes
tially take into account the variability among children if procedures of identification from procedures related to
data can be obtained on large numbers of children in other goals of assessment, such as understanding the
similar contexts. The problems with standardized tests nature of the problem and planning intervention.
have been well documented (see, e.g., McCauley & The first point made was that identification of children
Swisher, 1984; Muma, 1985). These problems are not, with a language disorder is best based on language
however, automatically solved by the use of nonstandard- performance and not on inferences about the knowledge
ized procedures. All types of observation are open to the that underlies this performance. The reasons given in-
same criticisms as those applied to tests. cluded the fact that children whose major problem is with
Other alternatives. One problem that is often raised accessing or using language knowledge may be in as
about the use of CA as a reference is that some children much need of intervention as those who have deficits in
with average or low-average performance relative to a the knowledge itself. It was suggested that procedures for
national sample of CA peers might be at great risk for identifying children with language disorders include
social and academic failure in a home or community tasks that stress the use of the language system so that
where 80-90% of the children perform well above aver- difficulties in performance would most likely be evident.
age. Should this child be denied help from a speech- Probes to determine a child's best language performance
language pathologist if the child's language performance and to make inferences about the language system under-
can be facilitated by this help? Most speech-language lying that performance are considered more relevant to
pathologists would probably agree that the child should planning language intervention and to understanding the
not. The more controversial question is whether such a nature of the problem than they are to identifying chil-
child should be labelled "language disordered." dren who have problems learning and using language.
One solution is to devise two categories. One category In assessment we often compare descriptions of our
would be for those who evidence poor language skills, at observations with some standard of expectation. The
least when the system is stressed. (They could be called standard for comparison varies with the goal of assess-
"language disordered.") The other category would be for ment. For example, when planning goals of intervention,
those who may demonstrate average or low-average skills the sequence of development of particular behaviors may
but for whom environmental, or other, factors place the be the standard to which we compare our observations
child at risk for future problems related to language (e.g., Bloom & Lahey, 1978; Lahey, 1988; Lee, 1974;

Downloaded From: http://jshd.pubs.asha.org/ by a Universite Laval User on 09/11/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
LAHEY: Language Disorders 619

Miller, 1981). When identifying children who have lan- dation. The Journal of Special Education, 9, 293-306.
guage learning problems, we also describe a child's BLOOM, L. (1970). Language development: Form andfunction in
emerging grammars. Cambridge, MA: The MIT Press.
language performance in relation to some implicit or
BLOOM, L., & BECKWITH, R. (1989). Talking with feeling: Inte-
explicit expectation. A major controversy, however, lies grating affective and linguistic expression in early language
in agreeing upon the standard to use as the basis of development. Cognition and Emotion, 3, 313-342.
expectations. In this paper I have argued against a stan- BLOOM, L., & LAHEY, M. (1978). Language development and
dard that is based on a measure of mental ability (e.g., the language disorders. New York: Macmillan.
BLOOM, L., LIFTER, K., & BROUGHTON, J. (1985). The conver-
LA-MA discrepancy) and suggested that comparisons of gence of early cognition and language in the second year of
language performance of children with similar CA is a life: Problems in conceptualization and measurement. In M.
better option (although not without problems). Barrett (Ed.), Children's single-word speech (pp. 149-180).
The reasons for preferring CA to MA as a reference New York: Wiley.
included measurement concerns such as problems with BOWERMAN, M. (1973). Early syntactic development: A cross-
linguistic study with special reference to Finnish. London:
the use of equivalence scores and difficulties involved in Cambridge University Press.
determining differences between two scores (each of BROWN, R. (1973). Afirst language, the early stages. Cambridge,
which has measurement error). The reasons also included MA: Harvard University Press.
concerns about the use of mental abilities regardless of CARROW-WOOLFOLK, E. (1988). Theory, assessment and inter-
measurement concerns. For example, it is difficult to vention in language disorders: An integrativeapproach. New
York: Grune & Stratton.
decide what aspect of mental abilities we want as a CASE, R. (1985). Intellectual development: Birth to adulthood.
comparison. Furthermore, claims about the theoretical New York: Academic Press.
relationship between language and cognition have not CHOMSKY, N. (1986). Knowledge of language: Its nature, origin,
been based on intelligence test scores but on measures of and use. New York: Praeger.
COLE, P. (1982). Language disorders in preschool children.
the way children represent and manipulate representa- Englewood Cliffs, NJ: Prentice-Hall.
tions of information; and even here, the relationship is COLE, K. N., DALE, P., & MILLS, P. E. (in press). Defining
controversial. language delay in young children by mental age referencing:
A few alternatives to, or variations on, the use of CA as Are we saying more than we know? Applied Psycholinguistics.
a comparison standard were briefly discussed. As long as COMMITTEE ON LANGUAGE LEARNING DISORDERS (1989,
March). Issues in determining eligibility for language inter-
definitional criteria are unclear, it will be difficult for us to vention. Asha, 31, 113.
further our understanding of the children we serve under CORRIGAN, R. (1978). Language development as related to Stage
this label. None of us has "the perfect solution" to who 6 object permanence development. Journal of Child Lan-
shall be called language disordered. It is unlikely that we guage, 5, 173-198.
CRONBACH, L. (1984). Essentials of psychological testing (4th
will ever approach such a solution or even reach a ed.). New York: Harper & Row.
tentative consensus until we make our current defini- DAMICO, J. S., OLLER, J. W., &STOREY, M. (1983). The diagnosis
tional criteria explicit, subject these criteria to critical of language disorders in bilingual children: Surface-oriented
evaluation, and offer possible alternative criteria for our and pragmatic criteria. Journalof Speech and Hearing Disor-
colleagues to challenge and discuss. If we do not ap- ders, 48, 385-393.
FEY, M. (1986). Language interventionwith young children. San
proach this difficult task, it may be done for us by others Diego: College-Hill Press.
outside the profession. FRATTALI, C. M. (1990, January). Quality assurance today:
Learning the basics. Asha, 32, 39-40.
GARDNER, H. (1985). Frames of mind: The theory of multiple
intelligences. New York: Basic Books.
ACKNOWLEDGMENTS HOLLAND, A. (1984). Language disorders in children. San Di-
ego: College-Hill Press.
A number of colleagues have read earlier versions of this paper HUBBELL, R. D. (1981). Children'slanguage disorders:An inte-
and have made thoughtful contributions. For their time and their grated approach. Englewood Cliffs, NJ: Prentice-Hall.
thoughts, I thank Tony Bashir, Lois Bloom, Phil Connell, Marc KAMHI, A., & CATTS, H. (1986). Toward an understanding of
Fey, Charlie Klim, Pat Launer, Jackie Liebergott, Shelly Lip- developmental language and reading disorders. Journal of
schultz, John Muma, Naomi Schiff-Myers, and Linda Swisher. Speech and HearingDisorders,53, 316-327.
Although I have attempted to take their comments into account, KUHN, D. (1988). Cognitive development. In M. H. Bornstein &
this paper does not necessarily represent their views on the M. E. Lamb (Eds.), Developmental psychology: An advanced
topic. textbook (pp. 205-260). Hillsdale, NJ: Lawrence Erlbaum.
LAHEY, M. (1988). Language disorders and language develop-
ment. New York: Macmillan.
REFERENCES LEE, L. (1974). Developmental sentence analysis. Evanston, IL:
Northwestern University Press.
LEONARD, L. (1983). Defining the boundaries of language disor-
ANASTASI, A. (1982). Psychological testing (5th ed.). New York: ders in children. In J. Miller, D. Yoder, & R. Schiefelbusch
Macmillan. (Eds.), Contemporary issues in language intervention (pp.
ARAM, D., & NATION, J. (1982). Child language disorders. St. 107-112). Rockville, MD: American Speech-Language-Hear-
Louis: Mosby. ing Association.
BASHIR, T., WIIG, E., & ABRAMS, J. C. (1988). Language disor- LIFTER, K., & BLOOM, L. (1988). Object play and the emergence
ders in childhood and adolescence: Implications for learning of language. Unpublished manuscript, Teachers College, Co-
and socialization. PediatricAnnals, 16, 145-156. lumbia University, New York.
BATES, E. (1976). Language in context. New York: Academic LIEBERGOTT, J., MENYUK, P., & CHESNICK, M. (1988). Language
Press. processing abilities in normal and language-impaired chil-
BAUMEISTER, S., & MUMA, J. (1975). On defining mental retar- dren. Miniseminar presented at American Speech-Language-

Downloaded From: http://jshd.pubs.asha.org/ by a Universite Laval User on 09/11/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
620 Journalof Speech and Hearing Disorders 55 612-620 November 990

Hearing Association National Convention, Boston, MA. dial education. Boston: Houghton Mifflin.
LUND, N., & DUCHAN, J. (1988). Assessing children's languagein SAPIR, E. (1921). Language. New York: Harcourt Brace & World.
naturalisticcontexts. Englewood Cliffs, NJ: Prentice-Hall. SIEGEL, G., KATSUKI, J., & POTECHIN, G. (1985). Response to
LYNGAAS, K., NYBERG, B., HOEKENGA, R., & GRUENEWALD, L. "Contemporary accounts of the cognition/language relation-
(1983). Language intervention in the multiple contexts of the ship" (Rice, 1983). Journal of Speech and Hearing Disorders,
public school setting. In J. Miller, D. Yoder, & R. Schiefel- 50, 302-303.
busch (Eds.), Contemporary issues in language intervention STARK, R., & TALLAL, P. (1981). Selection of children with
(pp. 239-252). Rockville, MD: American Speech-Language- specific language deficits. Journal of Speech and Hearing
Hearing Association. Disorders,46, 114-122.
MACLACHLIN, B., & CHAPMAN, R. (1988). Communication STERNBERG, R. (1985). Beyond IQ: A triarchictheory of human
breakdowns in normal and language learning-disabled chil- intelligence. New York: Cambridge University Press.
dren's conversation and narration. Journal of Speech and SUBCOMMITTEE ON LANGUAGE AND COGNITION. (1987, June).
Hearing Disorders, 53, 2-7. The role of speech-language pathologists in the habilitation
MCCAULEY, R., & SWISHER, L. (1984). Use and misuse of and rehabilitation of cognitively impaired individuals. Asha,
norm-referenced tests in clinical assessment: A hypothetical 29, 53-55.
case. Journal of Speech and Hearing Disorders, 49, 338-348. SWISHER, L., & DEMETRAS, M. J. (1987, June). Test-generated
MCCUNE-NICOLICH, L. (1981). The cognitive bases of relational heterogeneity in children with specific language impairment.
words in the single word period. Journal of Child Language,8, Paper presented at the International Conference on Language
15-34. Acquisition and Language Impairment in Children, Parma,
MILLER, J. (1981). Assessing language production in children: Italy.
Experimentalprocedures. Baltimore: University Park Press. TEMPLIN, M. (1957). Certain language skills in children. Min-
MILLER, J., & CHAPMAN, R. (1981). The relation between age neapolis: University of Minnesota Press.
and mean length of utterance in morphemes. Journal of THAL, D., & BATES, E. (1988). Language and gesture in late
Speech and Hearing Research, 24, 154-162. talkers. Journalof Speech and Hearing Research, 31, 115-123.
MILLER, L. (1989). Classroom-based language intervention. Lan- TOMBLIN, J. B. (1983). An examination of the concept of disorder
guage, Speech, and Hearing Services in Schools, 20, 153-169. in the study of language variation. Proceedings from the
MUMA, J. (1985). No news is bad news: A response to McCauley FourthWisconsin Symposium on Research in Child Language
and Swisher (1984). Journalof Speech and HearingDisorders, Disorders. Madison: University of Wisconsin.
50, 290-293. UZGIRIS, I. C., & HUNT, J. M. (1975). Assessment in infancy.
MUMA, J. (1986). Language acquisition: A functionalistic per- Urbana: University of Illinois Press.
spective. Austin, TX: Pro-Ed. VAN RIPER, C. (1954). Speech correction: Principlesand meth-
NEISSER, U. (1967). Cognitive psychology. New York: Appleton- ods. Englewood Cliffs, NJ: Prentice-Hall.
Century-Crofts. WEINBERG, R. A. (1989). Intelligence and IQ: Landmark issues
PIAGET, J. (1954). The construction of reality in the child. New and great debates. American Psychologist, 44, 98-104.
York: Basic Books. WING, L. (1981). Language, social, and cognitive impairments in
POWELL, T. W., & PENG, C. J. (1989). A profile analysis approach autism and severe mental retardation. Journal of Autism and
to interpretation of The Carrow Auditory-Visual Abilities Test. Developmental Disorders, 11, 31-44.
Journal of Speech and Hearing Disorders, 54, 214-225.
REYNOLDS, C. R. (1981). The fallacy of "two years below grade Received April 20, 1989
level for age" as a diagnostic criterion for reading disorders. Accepted February 9, 1990
Journal of School Psychology, 19, 350-358.
RICE, M., & KEMPER, S. (1984). Child language and cognition. Requests for reprints should be sent to Margaret Lahey, Ed.D.,
Baltimore: University Park Press. Division of Communication Disorders, Emerson College, 168
SALVIA, J., &YSSELDYKE, J. (1981). Assessment in special reme- Beacon Street, Boston, MA 02116.

Downloaded From: http://jshd.pubs.asha.org/ by a Universite Laval User on 09/11/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

You might also like