Chapter 7: Summary and Conclusions

Chapter 7
Summary and Conclusions
7.1.
Summary of the experimental work
The main objective of the experimental work described in this dissertation
was to uncover linguistic precursors of developmental dyslexia. Specifically,
early signs of dyslexia were sought in the morphosyntactic domain of
language. Most of the present experiments assessed children’s ability to
perceive morphosyntactic dependency relations, although the production of
such dependencies was also tested. The perception and production of the
past participle (which constitutes a morphosyntactic dependency between
the auxiliary hebben and the past participle form of the verb) was highlighted
in this dissertation. Dutch children acquire the past participle early (around
the age of 3:0), which makes this specific morphosyntactic construction ideal
for discovering early deficits in the syntactic competence of children at risk
for developing dyslexia. A secondary goal of this dissertation was to study
the relation between developmental dyslexia and specific language
impairment. Consequently, three groups of children took part in the
experiments, namely children with an increased genetic risk for dyslexia,
children with specific language impairment and normally developing
children.
As mentioned in the introductory chapters of this dissertation, only
around 40% of the children at risk for dyslexia will eventually develop the
syndrome. For this reason, the expectation was that, if linguistic precursors
were present in the at-risk group, they would manifest only in a subgroup of
the at-risk group. In the following paragraphs, the at-risk group is often said
to be different from the control group. The assumption is always, however,
that only a sub-group of the at-risk children are responsible for the different
behaviour of the group as a whole.
In chapter 3, Dutch children’s ability to discriminate between a grammatical
and an ungrammatical morphosyntactic agreement relation was studied.
The main finding of this chapter is that the at-risk children with a genetic
risk for dyslexia (aged 19 months) seem to be less sensitive to a
discontinuous morphosyntactic dependency than their normally developing
peers. When presented with sentences containing either a grammatical
morphosyntactic dependency or an ungrammatical morhosyntactic
dependency, normally developing children prefer to listen to the sentences
containing the grammatical dependency (i.e. sentences containing the
auxiliary heeft and the past participle). At-risk children, on the other hand,
show no such preference. Additionally, it was established that the
154 Summary & Conclusions
perceptual language development of the at-risk children is delayed by at
least six months and that normally developing children’s ability to detect a
discontinuous dependency is limited. Under strenuous conditions, normally
developing children’s perceptual behaviour is similar to that of the at-risk
children.
The experimental data collected in chapter 4 provide evidence that, in
general, children at risk for dyslexia (aged 3;6) are as capable as their
normally developing peers to produce the auxiliary heeft (has) and the past
participle form of the verb. However, at this age, at-risk children generate
significantly fewer instances of the complete past participle construction
than their normally developing peers. This is especially true in situations
where processing resources are taxed. Furthermore, when they fail to use the
past participle form of the verb (where this form is the expected form), the
at-risk children (as a group) use different strategies to complete a sentence
than normally developing children. Whereas normally developing children
generally opt for grammatical verb forms, some at-risk children produce
ungrammatical verb forms. For example, at-risk children use significantly
more infinitives in the complement position of the auxiliary verb heeft than
normally developing children. This suggests that the representation of the
morphosyntactic dependency between heeft and the past participle is less
stable in children at risk for dyslexia than in normally developing children.
Finally, productive performance is affected more in at-risk children than in
normally developing children when processing resources are taxed.
Specifically, in sentences containing a verb with a complex argument
structure, at-risk children omit verbal morphology more frequently and they
produce more ungrammatical utterances than normally developing children.
The effect of verb argument structure complexity on the production of
functional items was further examined in chapter 5 (the study in chapter 4
included only normally developing and at-risk children, while the study in
chapter 5 also included a group children with SLI). Additionally, the effect
of sentence length and of short term memory on the production of functional
items was also investigated (seeing that an increase in argument structure
complexity also causes a sentence to increase in length). The influence of
verb argument structure complexity was assessed with a sentence imitation
task. Statistical analyses of the data gathered from this experiment indicate
that argument structure complexity has an effect on auxiliary omission and
on determiner omission, but not on the omission of ge- (where ge- is a verbal
prefix). The effect of argument structure is most evident in the at-risk group
as these children omit significantly more auxiliaries in sentences containing
an intransitive verb than in sentences containing an intransitive or a
transitive verb. Sentence length has no effect on auxiliary omission or on ge-
Summary & Conclusions
155 omission, but it does have an effect on determiner omission. The length
effect is again most evident in the at-risk group; these children produce
significantly fewer determiners in the longest sentences. In a second
experiment, verbal working memory was assessed by means of a digit span
task. The results of this experiment show that children with SLI and children
at risk for dyslexia have significantly poorer short term memory skills than
normally developing children, a factor that could have influenced their
performance in the sentence imitation task. However, statistical analysis
shows that the capacity of a child to imitate sentences is weakly correlated to
short term memory. Short term memory can therefore not be the sole factor
influencing the omission of functional items.
In chapter 6, school-going children (average age 5; 1) with a genetic risk for
developing dyslexia were assessed on their ability to discriminate wellformed morphosyntactic dependencies from ill-formed morphosyntactic
dependencies. The at-risk children were compared to normally developing
children and to children with SLI. Sensitivity to two structures was tested;
namely to the relation between the temporal auxiliary heeft and the past
participle and to the relation between the modal kan and the infinitive.
Normally developing children (aged 5:0) are sensitive to both dependencies;
these children have no problems discriminating grammatical from
ungrammatical morphosyntactic dependencies. As a group, the at-risk
children are not significantly different from the normally developing
children with regard to their discriminative abilities. Even so, a sub-set of
the at-risk children have striking difficulties in discriminating sentences
containing the temporal auxiliary heeft and the past participle from sentences
containing a bare participle. The group of SLI children performed
significantly more poorly than the other two groups, suggesting a problem
in their grasp of the linguistic demands of the task.
7.2.
Theoretical implications
A secondary aim of this dissertation was to investigate the possible link
between developmental dyslexia and SLI. Therefore, an area of language (i.e.
morphosyntax) that is known to be problematic in children with SLI was
also investigated in children at-risk for dyslexia. Difficulties in the area of
morphosyntax are undisputed in SLI, which is not the case in dyslexia. For
this reason, a linguistic category that is typically problematic in SLI
(auxiliaries) was studied in children at-risk for dyslexia. In other words,
morphosyntactic difficulties that are typical in SLI were used as a yardstick
of possible morphosyntactic difficulties in dyslexia. Consequently, theories
of SLI could be potentially useful in understanding developmental dyslexia
156 Summary & Conclusions
if children with dyslexia are found to have similar difficulties as children
with SLI.
With regard to the perception and production of the past participle,
theories of SLI that presume a deficit in grammatical competence are not
very helpful in explaining the data of the at-risk children in this study. The
at-risk subjects do not seem have a deviant morphosyntactic system, which
would be the prediction of accounts such as the missing feature hypothesis,
the missing agreement hypothesis and the RDDR hypothesis. Theories of SLI
that presume a deficit in processing capacity are more useful in explaining
the behaviour of the at-risk children in this sample. Children with a genetic
risk for dyslexia clearly suffer from a limitation in their ability to process
complex information, a symptom which would be predicted by theories
such as the limited processing account and the surface hypothesis. Some of
the data produced by the at-risk children, notably the data in chapter 6, are
compatible with the predictions of the Extended Optional Infinitive Account
of SLI. It is thus possible that specific developmental stages (such as the
optional infinitive stage) are prolonged in developmental dyslexia. If this is
indeed the case, one can at most say that specific aspects of the language
development of dyslexic children are delayed.
In the end, a limited processing account seems to be the only theory
that can account for all of the findings in this dissertation. It is useful to
summarize how a limited processing capacity theory can explain the various
findings regarding the at-risk children in this study. When the results of the
various studies conducted in this thesis are viewed simultaneously, one
could interpret the findings as follows. Children at-risk for dyslexia suffer
from a limitation in their general processing capacity. Such a limitation
potentially affects (i) the size of the computational region of memory (i.e. the
available work space), (ii) the energy available for computational processes
and (iii) the rate at which information can be processed. In chapter 3, the atrisk children were not able to detect a discontinuous morphosyntactic
dependency when the dependent morphemes were separated by a twosyllable interval. This could be seen as a result of the size of their processing
window. In other words, the at-risk children were not able to process (and
remember) large enough pieces of the input in order to discover that the
input sometimes contained a discontinuous dependency. It is also possible
that the at-risk children’s processing rate was inadequate. When incoming
information is not processed quickly enough, it is susceptible to decay and
to interference from further incoming information. The results of chapters 4
and 5 can be explained in terms of energy. In these studies, the at-risk
children showed evidence of being able to perceive and produce closed class
items; however, this ability decreased when argument structure complexity
increased. In these situations, the task of constructing a sentence is started
but all of the energy available is exhausted before the sentence construction
Summary & Conclusions
157 can be completed. In chapter 6, the at-risk children failed to identify an
ungrammatical morphosyntactic dependency, but only when the
ungrammaticality did not hinder the semantic interpretation of the sentence.
Because they suffer from a general processing limitation, the at-risk children
‘choose’ to omit information during sentence processing if it is redundant for
comprehending or conveying the intended message. Along these lines, a
theory of limited processing capacity can account for the findings described
in this thesis.
Since the relationship between developmental dyslexia and SLI was
a main focus point in this dissertation and since morphosyntactic problems
are generally accounted for by theories of SLI, it was (to begin with)
considered whether a theory of SLI could also account for the findings
described in this dissertation. In theory, it should also be possible to
interpret the findings within the framework of a current theory of
developmental dyslexia. However, the findings of this dissertation are not
easily explained with any of the traditional theories of dyslexia. The
phonological deficit theory focuses exclusively on the phonological
problems experienced by dyslexics. The link between a pure phonological
deficit and a general processing limitation is definitely not straightforward.
Theories that do mention a processing deficit, like the temporal processing
deficit theory and the magnocellular theory assume dyslexia to be the result
of a very specific processing deficit (namely an inability to perceive and
process short or rapidly varying acoustic events adequately). In contrast to
the predictions of such theories, the at-risk children in this study are capable
of perceiving and processing (relatively) short acoustic events, but they
sometimes fail to do so when their processing resources (in general) are
severely taxed. It might be possible to explain a general processing deficit
with the cerebellar deficit hypothesis. Seeing that the cerebellum is involved
in language processes such as lexical retrieval (Marien et al., 2001) and the
linking of nouns to verbs (Gebhart et al., 2002) one could argue that a
(mildly) dysfunctional cerebellum will cause slow/inefficient language
processing and that this, in turn, will force dyslexics to omit certain aspects
of a syntactic structure during sentence production. However, given the
present results, one would then also have to assume that the dysfunction in
the cerebellum does not affect language production at all times, but only
when complex syntactic structures have to be processed.
7.3.
Clinical implications
The results of this study have several clinical implications. Developmental
dyslexia is commonly described as a phonological deficit. As a result, early
identification of and intervention in dyslexia frequently concentrates on the
field of phonological awareness. Yet, the experiments conducted in this
158 Summary & Conclusions
dissertation clearly show that children at-risk for dyslexia (and by
hypothesis children with dyslexia) suffer from a general processing
limitation, which affects their perception and production of morphosyntactic
structures. Whilst the fact that dyslexics suffer from a phonological deficit is
not disputed here, it is not clear how such a deficit could explain a general
processing limitation. Intervention that concentrates only on phonological
awareness may well strengthen dyslexics’ reading ability, but other complex
tasks (such as the processing of long/complex sentences) will remain
problematic in dyslexic individuals. With regard to the early identification of
dyslexia, clinicians might benefit from a language test that includes a section
assessing syntactic processing.
As mentioned in the general introduction of this dissertation,
researchers have recently questioned the division between developmental
dyslexia and specific language impairment. This dissertation provides
evidence that such a division is essential. The performance of the children
with SLI is noticeably different from that of the children at-risk for dyslexia.
This difference is clearest in the experiment conducted in chapter 6, where
some of the children with SLI experienced noticeable difficulties with the
linguistic demands of the task. It is not clear how researchers or clinicians
would benefit from a situation where dyslexia and SLI are seen as different
manifestations of the same disorder. Such an amalgamation would cause yet
another heterogeneous group of children with language disorders; making it
even more difficult to decide what type of clinical intervention a child needs.
Finally, one has to wonder if intervention can ever cure a dyslexic
suffering from a general processing limitation. It turns out that, in spite of
intense remediation, some dyslexics still struggle with sentence formulation
as adults. A general processing limitation which remains as a life-long
deficit, even after reading skills have been improved, could explain such
problems. A general processing limitation could also explain why the early
(often mild) language problems reported in dyslexic children seem to
resolve in adulthood, as processing capacity is assumed to increase with age
(up to a certain age at least).
7.4.
General conclusion
The work presented in this thesis provides evidence that the
morphosyntactic development of children with a predisposition for dyslexia
is different from that of normally developing children. Although at-risk
children do not, in general, exhibit deviant morphosyntactic development,
their representation of specific morphosyntactic dependencies seems
vulnerable. In taxing situations, for example in sentences containing a verb
with a complex argument structure, these children are more likely (than
their normally developing peers) to generate ungrammatical constructions,
Summary & Conclusions
159 to omit functional items such as auxiliaries and determiners and to omit
verbal morphology. Rather than having a deficient morphosyntactic system,
children at risk for dyslexia have a limited processing capacity that affects
their control over morphosyntactic dependency relations. An inability to
discriminate between well-formed and ill-formed morphosyntactic
dependency relations in taxing situations could thus be described as a
possible linguistic precursor of developmental dyslexia.