Relatively easy relatives: Children with syntactic
SLI avoid intervention
Naama Friedmann, Maya Yachini, and Ronit Szterman
Tel Aviv University
naamafr@post.tau.ac.il
We tested whether the relativized relatives approach, which suggests that typically-developing
children encounter difficulties with movement structures in which one lexically-restricted NP
intervenes in the movement of another lexically-restricted NP, also applies to syntactic SLI (SySLI).
We assessed the production of subject- and object-relatives in 175 Hebrew-speaking children and
adolescents with SySLI and 87 controls. The results indicated that the participants with SySLI avoided
intervention, significantly more than the control group, by omitting the relative head or the embedded
NP, or substituting them with a non-lexically restricted NP. Two patterns of SySLI emerged: one
subgroup could produce subject relatives but was sensitive to intervention, and one could not produce
any relative, which had a tree-construction deficit.
1. Introduction
One of Adriana Belletti’s many virtues is her ability to tie strong theoretical analyses with the world,
and harness linguistic theory to improve the quality of people's lives. In this paper we show how
Adriana's theoretical work predicts and explains the difficulties children with syntactic specific
language impairment (SySLI) have in the production of relative clauses.
School-aged children with SySLI show difficulties in the comprehension and production of relative
clauses, and especially in headed object relatives (Cipriani et al., 1998; Delage et al., 2008; Friedmann
& Novogrodsky, 2004, 2007, 2011; Marinellie, 2004; Novogrodsky & Friedmann, 2006; Stavrakaki,
2001; van der Lely, 1997). This difficulty has been demonstrated in their poor comprehension in tasks
of sentence-picture matching and written sentence paraphrases, in relatively few correct relatives
produced in spontaneous speech, as well as in low rates of correctly-produced relatives and high error
rates in elicited production of relatives.
In a recent line of work (Friedmann, Belletti, & Rizzi, 2009; Belletti et al., 2012), Belletti and
colleagues tied the difficulty that children show in object relatives during early stages of language
acquisition to Relativized Minimality and intervention. The main thrust of this approach is that the
intervention of the embedded subject in the dependency between the head of the relative and its basegenerated merge position is what makes headed object relatives hard for children. According to this
approach, object relatives are harder than subject relatives for children acquiring syntax because
whereas Wh-movement of the subject to the head position of subject relatives does not cross the object,
the object, in its movement to the head position of object relative, crosses the embedded subject. This
idea also explains why young typically developing children perform worse in object relatives that
include two full DPs in the moved object and embedded subject positions, compared with object
relatives in which only one of the DPs is lexically restricted (or when the two DPs differ in their arrays
of features which are relevant for movement).

Some of the SLI and control data reported here was collected in the framework of MA and PhD theses of Iris Fattal, Adi
Kesselman, and Revital Guggenheim in Tel Aviv University. We deeply thank them for their generosity in allowing us to
include the data they collected in this analysis. We are also grateful to Beit Ekstein schools for their willful participation.
This research was supported by the Israel Science Foundation (grant no. 1066/14, Friedmann), by the Lieselotte Adler
Laboratory for Research on Child Development, and by the Australian Research Council Centre of Excellence for
Cognition and its Disorders (CE110001021).
Relatively easy relatives
Errors that occur in the attempts to produce object relatives are also consistent with the idea that young
typically developing children in the course of syntactic acquisition have difficulty with intervention.
Friedmann, Belletti, & Rizzi (2009) reported that young Hebrew-speaking children avoided
intervention structures in production by producing subject relatives instead of object relatives. A
similar tendency to produce subject relatives instead of object relatives, as well as other sentence
structures in which intervention is avoided was also found in a study of the acquisition of 16 languages
within the COST A33 action (Friedmann et al., 2012). In a line of theoretical and empirical work on
the production of relative clauses of Italian-speaking children and adults, Belletti (2009; Belletti &
Contemori, 2010; Belletti & Rizzi, 2010; Contemori & Belletti, 2014) proposed that children's
productions at these ages are ways to overcome the intervention of the preverbal lexical subject. For
example, focusing on what she termed "passive object relative", she suggested that these structures are
preferred, because they form a way to satisfy the locality principle of RM, as the derivation of such
sentences involving smuggling (Collins, 2005) does not involve intervention.
The current study examines whether this approach can be extended to account for the difficulties
children with syntactic SLI (SySLI) show in relative clauses. For this aim, we elicited subject- and
object relatives in 175 children with SySLI, and analyzed the rate of grammatical productions of
subject- and object relatives and the types of errors they made when they attempted to produce object
relatives. Specifically, we asked whether their errors indicate avoidance of intervention.
2. Method
2.1. Participants
The participants were 175 monolingual Hebrew-speaking children and adolescents with SySLI aged
7;4-16;6 (M = 11;2, SD = 2;3). All of them were diagnosed with learning disability and deficits in
reading comprehension prior to the study, and had normal IQ. We included children in the SySLI group
only if they failed in at least two tests of comprehension and repetition of sentences with Whmovement (most participants took part in 5 or more syntactic tests, other took part in three).
The control participants were 78 monolingual Hebrew-speaking children and adolescents with typical
language development and without any diagnosis of learning disabilities, aged 7;4-17;0 (M = 9;6, SD
= 2;4, at least 20 participants in each age group, 7;4-8, 8-9, 9-11, 11-17).
2.2. Material and procedure
To elicit right-branching headed relative clauses we used a preference task (developed and reported in
detail in Friedmann & Szterman, 2006; Novogrodsky & Friedmann, 2006). In this task, the
experimenter described two children in two situations, and asked the participant to choose which child
he would prefer to be. The task was constructed in a way that the choice would have to be formed as
a relative clause. To ensure a relative clause response, the experimenter requested the participants to
reply to each question starting with “I would rather be…”.
The order of the two types of relative target sentences was randomized. The questions that elicited
subject relatives described two children (two boys when the participant was a boy, two girls when the
participant was a girl) performing two different actions on the same theme, see example (1), or
performing the same action on two different themes. The questions that elicited object relatives
described two children who are the themes of different actions performed by the same figure, or an
action performed by two different figures0 (2).
(1) Elicitation of a subject relative:
There are two children. One child gives a present and one child receives a present. Which child would
you rather be? Start with “I would rather be…”
Target answer:
(hayiti ma'adif lihiyot) ha-yeled she-mekabel matana
2
Naama Friedmann, Maya Yachini and Ronit Szterman
(would-1sg prefer to-be) the-child that-receives present
(I would rather be) the child who receives a present.
(2) Elicitation of an object relative:
There are two children. The father drives one child and the grandfather drives one child. Which child
would you rather be? Start with “I would rather be…”
Target answer:
(hayiti ma'adif lihiyot) ha-yeled she-aba masia
(would-1sg prefer to-be) the-child that-father drives
(I would rather be) the child who the father drives.
Responses were analyzed for total number on target (subject relative for a subject relative target, object
relative for an object relative target). Furthermore, because we were mainly interested in the types of
responses and the structure of the sentences that children with SySLI produce when they try to produce
object relative clauses, we analyzed the various types of responses to explore whether children with
SySLI avoid intervention. Avoidance of intervention configuration can occur in grammatical sentences
and even in on-target responses. For example, a response in which an object relative is produced with
an arbitrary pro subject or with a demonstrative pronoun head (similar to "the one") is an acceptable,
on-target response, but it indicates avoidance of intervention, because it turns one of the lexical NPs
into non-lexically restricted.
2.3. General method of assessment
Each participant was tested individually in a quiet room. Participants were told that they could have a
break or stop the session whenever they wanted. Each session started with a short casual conversation
with the participant, and then the tests were administered.
No time limit was set, and the experimenter repeated every item as many times as the participant
requested, in order to give the participants full credit for their syntax and reduce the effects of working
memory and attention. No response-contingent feedback was given, only general encouragement, and
a final reward was given after completion of the task.
All the sessions were audio-recorded. Children's responses were fully transcribed during and after the
sessions. Incorrect responses that were later corrected by the child were also transcribed, but only the
final utterance for each item was counted and coded. The transcriptions were double-checked by two
experimenters, and compared with the original transcriptions for reliability.
2.4. Statistical analyses
The comparison at the group level between the performance of a group in two conditions, and between
the SLI and the control groups was done using paired t-test. The comparison of each participant to the
control group was done with Crawford & Howell’s (1998) t-test. An alpha level of 0.05 was used for
all comparisons.
3. Results
The results indicated, first, that the SySLI children had a severe deficit in the production of relative
clauses. Recall that the children who participated in this elicitation task were diagnosed as SySLI on
the basis of their failure in at least 2 of the other 3-5 syntactic tests they participated in, and a diagnosis
of language or learning disability and of written text comprehension difficulties. This relative clause
elicitation task proved very sensitive for detecting syntactic impairment in these participants: 163
(93%) of the children who were diagnosed with SySLI according to these other criteria performed
significantly poorer than the control group on this elicitation task.
3
Relatively easy relatives
Because we were interested in the types of errors that SySLI children make, in order to shed light on
the characterization of their deficit, the next analyses were done on the children whose production of
relative clauses was impaired. As summarized in Table 1, all of them showed impaired production of
object relatives, and their production of object relatives (36.5% correct) was significantly poorer than
their production of subject relatives (76.6% correct), t(162) = 19.1, p <.0001. Their production of
subject relatives, although better than their production of object relatives, was still significantly poorer
than that of the control group, t(248) = 7.6, p < .0001.
Interestingly, as can be seen in Table 1, the production of relative clauses in the SySLI group did not
improve with age, neither for the object relatives, r = .02, p = .78, nor for the subject relatives, r = .05, p = .51, and even the participants in ninth and tenth grade in this group still showed a severe
difficulty in the production of both subject and object relatives.
Table 1. The production of relative clauses in the various age groups: Correct responses (grammatical
relatives of the target type) to subject and object relative clauses, in the SySLI and the control groups.
(Average percentage correct and SD)
Group
SySLI
Control
Age group
Number of
participants
Subject
relatives
Object
relatives
Number of
participants
Subject
relatives
Object
relatives
7;4-8
16
8-9
30
9-10
14
10-11
7
11-12
16
12-13
29
13-14
37
14-16;6
14
80.0
(25.3)
33.8
(28.3)
77.7
(24.3)
35.0
(25.3)
75.7
(35.2)
27.9
(21.5)
62.9
(35.5)
27.1
(26.9)
73.1
(35.7)
33.1
(28.5)
80.3
(29.4)
41.0
(28.6)
78.5
(28)
40.8
(30)
71.3
(28.3)
39.3
(29.1)
24
21
20
22
..99
99.0
(3.0)
92.9
(8.7)
97.5
(6.4)
93.5
(9.9)
100
(0)
96.4
(7.3)
(4.1)
94.6
(8.8)
The next analyses were done to examine the "relativized relatives" intervention account, which was
developed for syntactic acquisition in typically developing children, and see whether it also applies for
SySLI.
We therefore analyzed the pattern of responses that the SySLI participants produced when they tried
to produce sentences with object relative clauses. This analysis further supported intervention as the
basis for the syntactic impairment in SySLI. The SySLI participants used various ways to avoid
intervention, the rate of each response type is presented in Table 2. The participants mainly used three
paths in order to avoid the production of object relatives in which the movement to the head position
of the object relative, crosses the embedded subject:
a. Avoid movement: simple sentences or embedded sentences without Wh-movement
b. Avoid intervention configuration: subject relatives
c. Avoid two lexically restricted NPs in the intervention configuration: producing only one NP or
only one lexical NP.
In addition to these three types of avoidance strategies, there were productions in which the participants
attempted to produce object relatives but ended with an ungrammatical sentence.
4
Naama Friedmann, Maya Yachini and Ronit Szterman
Table 2. Percentage of avoidance responses of various types to target object relatives (out of the total
target object relatives) in the SySLI and control groups
Responses
SLI
Control
Avoid movement
Simple sentences
16.9
Grammatical
9.3
Ungrammatical
6.0
Fragments
1.5
Embedded clauses without movement
15.3
Grammatical
3.8
Ungrammatical
11.4
Filled gap –full NP in the gap position
3.8
Avoid intervention configuration: subject relatives
Subject relatives
22.1
Avoid two lexically restricted NPs
A relative clause with NP omission in the embedded
8.3
clause (example 8)
Pronoun relative head (example 9)
3.1
Ungrammatical omissions of the relative head
1.0
(example 10)
Subject relative with a reflexive verb (example 11)
0.5
Subject relative with a passive verb (example 12)
0.4
Arbitrary pro embedded subject (example 13)
2.9
Other
0.9
0.5
0.3
0.3
0.3
0.3
0
0.5
Ungrammatical object relatives
0.1
1.7
2.5
0.7
0.6
0.6
3.6
Let us describe in some more detail these avoidance response types. One strategy that the SySLI
participants used was to avoid relative clauses and movement altogether. Such responses included
simple sentences (example 3), fragments, or embedded clauses without movement (example 4). As
summarized in Table 2, the SySLI group produced such responses 18 times more often than the control
group. A large percentage of the simple sentences and the embedded clauses without movement were
ungrammatical. One could also take the ungrammatical responses that included a filled gap, which
included a resumptive full NP that was identical to the head, different from the head (example 5), or
with a different NP than the head1, as instances of sentences without movement, but this is already a
matter of interpretation.
Examples for avoidance of movement:
(3) Grammatical simple sentence
Target answer:
hayiti ma'adif lihiyot ha-yeled she-aba masi'a
I would rather be the boy that father drives.
Response: hayiti ma'adif et aba
would-1sg prefer ACC father
I would prefer father.
1
An interesting point raised by Rosalind Thornton is that the pronunciation of an exact copy of the head at the trace position
might actually be taken to indicate good chain formation, even if the sentence is ungrammatical. However, such responses
were the minority of the filled-gap responses: only 23 such head-doubling responses occurred out of the 65 gap-filling
responses (40 of the SLI participants produced at least one gap-filling response, only 12 of them produced only head
doublings in their filled gaps).
5
Relatively easy relatives
(4) Grammatical embedded clause without movement
Target answer:
hayiti ma'adif lihiyot ha-yeled she-aba melamed
I would rather be the boy that father teaches.
Response: hayiti ma'adif she-aba yelamed oti
would-1sg prefer that-father teach-future me
I would prefer that father will teach me.
(5) An ungrammatical response with a filled gap
Target answer:
ha-yalda she-safta mecalemet
the girl that grandma photographs.
Response: ha-yalda she-safta mecalemet yalda axeret
the-girl that-grandma photographs girl another
the girl that grandma photographs another girl.
Another way the children used to avoid intervention was to produce relatives without intervention
configuration, by producing subject relatives instead of object relatives. This happened almost 9 times
more often in the responses of the SySLI children than in the control participants' responses.
The subject relatives the SySLI children produced were of several types: subject relatives with a
change of the head where the embedded subject becomes the head of the relative (example 6), reversal
of thematic roles (example 7) and other types that also include avoidance of two lexically restricted
NPs and are presented in the next category.
Examples for avoidance of intervention configuration:
(6) Use of subject relative instead of object relative – change of head
Target:
hayiti ma'adif lihiyot ha-yeled she-saba mecalem
I would rather be the boy that grandpa photographs.
Response: hayiti ma'adif lihiyot ha-saba she-mecalem et ha-yeled
would-1sg prefer to-be the-grandpa that-photographs ACC the-boy
I would rather be the grandpa that photographs the boy.
(7) Use of subject relative instead of object relative – with a reversal thematic roles
Target:
hayiti ma'adif lihiyot ha-yeled she-ha-mora boxenet
I would rather be the boy that the teacher examines.
Response: hayiti ma'adif lihiyot ha-yeled she-boxen et ha-mora
would-1sg prefer to-be the-boy that-examines ACC the-teacher
I would rather be the boy that examines the teacher.
Finally, another important way the SySLI children used to avoid intervention was the production of a
single lexical DP instead of two lexical DPs in an intervention configuration. This was achieved in
various ways: omitting one of the DPs in the embedded clause (example 8), replacing the head of the
relative with a demonstrative pronoun (example 9), an ungrammatical omission of the head of the
relative (example 10), omission of one of the lexical NPs in the embedded phrase using a reflexive
verbs (example 11), and passive voice (example 12), or using an arbitrary pro embedded subject
(example 13). In total, these intervention-avoiding responses that included only one lexical NP
constituted 15.4% of the SLI group's responses to the object relative items. Such avoidance responses
6
Naama Friedmann, Maya Yachini and Ronit Szterman
occurred significantly less often in the control group, who only produce such responses in 4.8% of the
target object relatives, t(248) = 5.8, p < .0001. Whereas the arbitrary pro occurred also in the control
group and hence cannot be taken as an indication of specific difficulty and avoidance, the other types
of responses, both grammatical and ungrammatical, definitely are.
Examples for avoidance of intervention using the production of a single DP:
Target:
(8) Use of subject relative instead of object relative with omission of an NP in the embedded clause
Target:
hayiti ma'adif lihiyot ha-yeled she-saba mecalem
I would rather be the boy that grandpa photographs.
Response: ha-saba she-mecalem
the-grandpa that-photographs
The grandpa that photographs.
(9) Replacing the head of the relative with a pronoun
Target:
hayiti ma'adifa lihiyot ha-yalda she-savta mecalemet
I would rather be the girl that grandma photographs.
Response: hayiti ma'adifa lihiyot zot she-savta mecalemet
would-1sg prefer to-be this-fem that-grandma photographs
I would rather be the one that grandma photographs.
(10) Ungrammatical omission of the head of the relative
Target:
hayiti ma'adifa lihiyot ha-yalda she-savta mecalemet
I would rather be the girl that grandma photographs.
Response: hayiti ma'adifa lihiyot she-savta mecalemet
would-1sg prefer to-be that-grandma photographs
I would rather be that grandma photographs.
(11) Use of subject relative with a reflexive verb instead of object relative
Target:
ha-yeled she-saba mecalem
The boy that grandpa photographs.
Response: ha-yeled she-mictalem
the-boy that-photographs-refl
The boy that photographs himself
(12) Use of subject relative with passive instead of object relative
Target:
ha-yeled she-saba mecalem
The boy that grandpa photographs.
Response: ha-yeled she-meculam
the-boy that-photograph-passive
The boy that is being photographed
7
Relatively easy relatives
(13) Use of arbitrary pro embedded subject
Target:
ha-yeled she-ha-mora melamedet
The boy that the teacher teaches.
Response: ha-yeled she-melamdim
the-boy that-proarb-teaching-3pl
The boy that is being taught
Attempts to produce relative clauses also led to the production of ungrammatical sentences. These
sentences included incorrect use of embedding markers in the beginning of the embedded clause but
also within it, a totally incorrect order of the words in the sentence, and verb omissions. Another
interesting type of ungrammatical response involved the production of a verb in a verbal template that
reduces the arity of the verb (reflexive, unaccusatives, or passive), but in a way that is inappropriate to
the sentence, usually with the addition of a by phrase. It seems that syntactically, the SySLI participants
who produced these responses could only produce a relative clause with a single NP, but then tried to
also mention the agent, to satisfy the semantic requirements of the message they wanted to convey,
and did it with a by phrase that is grammatically inappropriate. See examples for object relative
responses under (14, where a by-phrase was added to a reflexive verb, or to a sentence with an arbitrary
pro.
(14) Ungrammatical object relatives with verb arity that does not match the number of arguments
a. Target:
hayiti ma'adifa lihiyot ha-yalda she-ha-shxena melamedet
I would rather be the girl that the neighbor teaches.
Response: hayiti ma'adifa lihiyot ha-yalda she-mitlamedet al-yedey ha-shxena
would-1sg prefer to-be the-girl that-learn-refl by the neighbor
I would rather be the girl that is learning-herself by the neighbor.
b. Target:
hayiti ma'adif lihiyot ha-yeled she-aba melamed
I would rather be the boy that father teaches.
Response: hayiti ma'adif lihiot ha-yeled she-nilmad
would-1sg prefer to-be the boy that-taught-passive
I would rather be the boy that taught.
Another important difference between the SySLI group and the TD group related to the production of
resumptive pronouns. Resumptive pronouns are an optional grammatical option for DP object relatives
in Hebrew (15). As summarized in Table 3, whereas the typically developing children produced object
relatives with resumptive pronouns quite often (61.0% of the responses to target DP object relatives,
and 65.0% of the grammatical object relatives), the SySLI group produced much fewer resumptive
pronouns: only 14.4% of their responses to target DP object relatives included a resumptive pronoun,
and only 39.7% of the grammatical object relatives, significantly fewer resumptive pronouns than the
control group, t(248) = 15.20, p < .0001, t(248) = 7.30, p < .0001, in the two analyses, respectively.
(15) Grammatical object relative with a resumptive pronoun
hayiti ma'adif lihiyot ha-yeled she-aba masia oto
would-1sg prefer to-be the-child that-father drives him
I would rather be the child who the father drives him.
8
Naama Friedmann, Maya Yachini and Ronit Szterman
Interestingly, there were differences within the SySLI group in the tendency to produce resumptive
pronouns in the object relatives, which led us to discover two patterns of impairment in relative clauses.
The results seem to suggest one pattern in which children failed to produce object relatives but
produced all their subject relatives correctly. This pattern was probably demonstrated by children who
are able to correctly construct the syntactic tree, including CP, but have difficulties with object relatives
because of the existence of movement across an intervener. Another pattern was shown by children
who had problems even in the production of subject relatives, not only in object relatives. This pattern
indicates a deficit that extends beyond an intervention problem.
These two profiles of impairment gave rise to different use of resumptive pronouns: The 71 SySLI
participants who had problems even in the production of subject relatives (80% correct or below)
produced less than half (average = 1.0 resumptive pronoun, namely 10% of the target object relatives)
of the resumptive pronouns that the 65 children who produced all their subject relatives correctly (2.5
object relatives with resumptive pronoun on the average, 25%), t(134) = 4.53, p < .0001.
A similar difference within the SySLI group was also detected in the tendency to produce avoidintervention responses: the children who probably have a movement-intervention deficit (and who
therefore produce subject relatives well) produced significantly more avoid-intervention responses
than the children who had a deficit that also involved the syntactic structure and could not even produce
the target subject relatives. The first group produced only 0.64 responses with NP omission (6%),
whereas the second group of SLI participants produced 0.88 (9%) such responses. This difference did
not reach significance.
Relatedly, there was a significant difference between the two patterns of SySLI in their productions of
non-relatives instead of object relatives, t(134) = 5.58, p < .0001. The children who failed also in the
subject relatives produced an average of 5.4 (54%) sentences without movement (grammatical and
ungrammatical simple sentences and sentence fragments) in response to the object relative target items,
whereas the participants who produced subject relatives correctly produced only 2.8 (28%) nonrelatives instead of object relatives on average.
Thus, it seems that when the SySLI deficit is characterized by an intervention deficit, and hence the
children can produce structures without intervention such as subject relatives, this also characterizes
the types of responses they produce when they try to produce structures with intervention, such as
object relatives. In this case, they avoid intervention by producing relative clauses without
intervention: subject relatives, use of relative clause sentences with only one lexical NP, etc. When the
SLI impairment involves a deficit beyond intervention, which extends to all kinds of relative clauses,
then the speaker does not avoid intervention in relative clauses but rather avoids Wh-movement
altogether.2
2
There were no differences in the tendency to fill gaps between these patterns, and there were even slightly more filled
gaps in the intervention-deficit group (4.3% filled-gap sentences in the responses to object relatives in the interventiondeficit group, and 2.7% in the relative impaired group). Interestingly, however, most filled-gap responses in which the head
was doubled in the gap position came from the intervention-deficit group.
9
Relatively easy relatives
Table 3. The production of object relative clauses with resumptive pronouns in the various age groups:
in the SySLI and the control groups. (Average percentages out of total responses to object relatives,
and out of the grammatical object relatives produced).
Group
SySLI
Age group
% resumptive pronouns
% resumptive pronoun
out of grammatical object
relatives produced
7;4-8
15.6
46.3
8-9
16.0
45.7
Control
% resumptive pronouns
% resumptive pronouns
out of grammatical object
relatives produced
55.8
59.6
66.2
72.0
9-10
7.9
28.2
10-11
11.4
42.1
59.5
64.0
11-12
15.6
47.2
12-13
16.7
40.7
13-14
22.0
54.0
14+
12.0
30.5
63.2
66.0
Finally, unlike in other languages (see for example Belletti & Contemori, 2010, for normal acquisition
of Italian), in Hebrew passive relatives were very scarce in the productions of both groups (example
12). There were only 7 such responses in the SySLI group (0.4% of the responses) and no such response
in the control group. Another type of response that occurred in the SySLI group was somewhat similar
to the passive responses in other languages, in that some sort of smuggling account can be applied to
it: the SLI participants produced 21 sentences with VOS order within the embedded clause (16). This
response was not specific only to the SySLI group, as it occurred in 13 responses in the control group.
(16) hayiti ma'adif lihiyot ha-yeled she-ha-more boxen (oto)
I would rather be the boy that the teacher tests (him).
Response: hayiti ma'adif lihiyot ha-yeled she-boxen oto ha-more
would-1sg prefer to-be the-boy that-tests him the-teacher
I would rather be the boy that tests him the teacher.
4. Discussion
The pattern of production of the 163 children and adolescents with syntactic SLI indicates that the
Friedmann, Belletti, and Rizzi (2009) approach to normal acquisition of syntax may also account for
the pattern of syntactic impairment of children with SySLI.
Friedmann, Belletti, and Rizzi (2009) and Belletti et al. (2012) suggested that the difficulty that
typically developing children show in certain types of headed object relatives results from an
intervention effect in terms of the syntactic principle Relativized Minimality. According to this
approach, when the moved object crosses an intervening subject that shares with it certain features, the
children fail to understand the sentence (see Grillo, 2005, for a similar account for the comprehension
deficit in agrammatic aphasia). This accounts, for example, for the common finding according to
which, in typical acquisition of many languages, subject relatives are easier than object relatives.
Object relatives in which both the moved object and the intervening embedded subject are lexically
restricted include a subject intervener that blocks the chain between the moved relative head and its
trace because its set of features is included in the set of features of the movement object.
The current study asked, through the examination of a large group of children with SySLI, whether
such an approach can also be applied to the difficulty in relative clauses in SySLI. The results of our
current study, as well as results from other studies indicate that it is: The first supporting evidence for
an intervention account for SySLI is the significantly better production of subject relatives compared
with object relatives. This was found in the current study and also in several previous studies on SLI
10
Naama Friedmann, Maya Yachini and Ronit Szterman
(Adams, 1990; Håkansson, & Hansson, 2000; Fattal, Friedmann, & Fattal-Valevski, 2011; Friedmann
& Novogrodsky, 2004, 2007, 2011; Levy & Friedmann, 2009; Marinellie, 2004; Stavrakaki, 2001).
The second step related to whether children with SySLI, when they do produce an object relative, opt
for a relative clause in which one of the NPs is not lexically restricted. Firstly, Novogrodsky and
Friedmann (2006) reported, in a study of 18 children with SySLI, that these children tended to avoid
object relatives, or produced object relatives with a single lexical DP. Additional evidence to this point
comes from a recent comprehension study: Yachini, Szterman, and Friedmann (2013) reported that
children with SySLI understand free object relatives and object relatives with arbitrary embedded pro
subjects significantly better than they understand object relatives with two lexically restricted NPs.
The results of the current study very clearly support the same idea: whereas children their age can
already produce both subject- and object relatives, children with SySLI, even those who are already in
high-school, still find it very difficult to produce these sentences. Instead, they avoid Wh-movement
altogether, by producing sentences that are not relative clauses, avoid intervention configurations by
producing subject relatives instead of object relatives, and, when they produce object relatives with an
intervention configuration, they avoid the production of two lexical NPs by omitting the head of the
relative clause, using a non-lexically restricted pronoun relative head, omitting the NP within the
relative clause, or produce an arbitrary pro as the embedded subject. Whereas the arbitrary pro occurred
also in the control group and hence cannot be taken as an indication of specific difficulty and
avoidance, the other types of avoid-two-lexical-NP responses, both grammatical and ungrammatical,
occurred significantly more often in the SySLI group and are an indication of an attempt to produce
relatively easy relatives, by avoiding intervention.
Another interesting result related to the difference in the production of resumptive pronouns between
the SySLI and the control groups. Resumptive pronouns in object relatives are optional in Hebrew.
Interestingly, the control participants used this option in more than half of their object relatives,
whereas the SySLI participants as a group used significantly fewer resumptive pronouns. This is
reminiscent of the findings from young children acquiring relative clauses in Hebrew: at a stage in
which they still have a problem with movement of one lexically restricted NP across another, they do
not benefit in comprehension from the addition of a resumptive pronoun. Moreover, within the SySLI
group - the ones who had difficulties even in the production of subject relatives produced less than half
of the resumptive pronouns produced by the SLI children who produced all their subject relatives
correctly. This might be interpreted as follows: when the deficit in SLI is related to the construction of
the syntactic tree (possibly of CP), resumptive pronouns are not produced, because the whole syntactic
structure of the relative is not constructed correctly, whereas when the deficit is more closely related
to movement and intervention, resumptive pronouns may have a position in the structure.
Relatedly, we observed differences within the SySLI group also with respect to avoid-intervention
attempts: the children who had a deficit that seemed to also involve syntactic structure building and
could not even produce the target subject relatives mainly produced non-relatives instead of object
relatives, whereas the SySLI children with a probable movement-intervention deficit (and who
therefore produce subject relatives well) produced significantly more avoid-intervention responses, of
relative clauses in which one of the NPs was omitted or not lexically restricted.
Some additional insights can be gained from this large-scale assessment of relative clauses in SySLI.
The first relates to the effect of age, or actually, the lack thereof. Many studies discuss syntactic
problems of young children, aged 4-8, but very rarely test older individuals with SySLI. The situation
is similar in the education system. In Israel, for example, children in the kindergarten are identified
and treated for language problems, but once they are in the school system, they are diagnosed with
"learning disability" and receive less specific treatment. Our results show that relative clauses are
acquired at around the age of 6 in typically developing children, and that individuals with SySLI suffer
very similar difficulties when they are 7 and when they are 16. Persistent linguistic difficulties in
adolescence were also reported by Delage and Tuller (2007) for French-speaking adolescents (14-15
11
Relatively easy relatives
year olds) with SySLI. Hence, these results suggest that the syntactic difficulties persist and do not
dissolve with time, and that syntax should be assessed and treated even in high-school.
A second point is methodological: we selected participants for this study based on their poor
performance in two out of 3-5 syntactic tests of sentence comprehension and repetition, as well as
general reports of learning disability and written test comprehension difficulties. The relative clause
elicitation task proved very sensitive for detecting syntactic impairment in SySLI, as 93% of the
participants who were diagnosed with SySLI according to these other multiple criteria performed
significantly poorer than the control group on this elicitation task. These results suggest that, at least
in Hebrew, this preference task can be used as a sensitive screening task to explore whether schoolaged children and adolescents have syntactic deficits.
To conclude, this study, encompassing the detailed examination of 175 children and adolescents with
syntactic SLI, indicated that there are some relatives that are relatively easy for individuals with
syntactic SLI: sentences that do not include two lexical NPs in an intervention configuration.
5. References
Adams, Catherine. 1990. Syntactic comprehension in children with expressive language impairment. British
Journal of Disorders of Communication, 25, 149-171.
Belletti, Adriana. 2009. Notes on passive object relatives. In P. Svenonius (Ed.), Functional structure from top
to toe. Oxford: Oxford University Press.
Belletti, Adriana, Friedmann, Naama, Brunato, Dominique & Rizzi, Luigi. 2012. Does gender make a
difference? Comparing the effect of gender on children's comprehension of relative clauses in Hebrew
and Italian. Lingua, 122(10), 1053–1069. doi: 10.1016/j.lingua. 2012.02.007
Belletti, Adriana & Contemori, Carla. 2010. Intervention and attraction. On the production of subject and object
relatives by Italian young children and adults. In J. Costa, A. Castro, M. Lobo, & F. Pratas (Eds.),
Language Acquisition and Development, 3. Proceedings of Gala 2009 (pp. 39−52). Cambridge, UK:
Cambridge Scholars Publishing.
Belletti, Adriana & Rizzi, Luigi. 2010. Ways of avoiding intervention: Object relatives, passive and control. In
R. Berwick & M. Piattelli Palmarini (Eds.), Rich grammars from a poor input. Oxford: Oxford University
Press.
Cipriani, Paola, Bottari, Piero, Chilosi, Anna Maria & Pfanner, Lucia. 1998. A longitudinal perspective on the
study of specific language impairment: The long term follow-up of an Italian child. International Journal
of Language and Communication Disorders, 33, 245–280.
Collins, Chris. 2005. A smuggling approach to the passive in English. Syntax, 8(2), 81-120.
Contemori, Carla & Belletti, Adriana. 2014. Relatives and passive object relatives in Italian-speaking children
and adults: Intervention in production and comprehension. Applied Psycholinguistics, 35(6), 1021-1053.
doi:10.1017/S0142716412000689
Crawford, John R. & Howell, David C. 1998. Regression equations in clinical neuropsychology: An evaluation
of statistical methods for comparing predicted and observed scores. Journal of Clinical and Experimental
Neuropsychology, 20, 755-762. doi: 10.1076/jcen.20.5.755.1132.
Delage, Helene, Monjauze, Cecile, Hamann, Cornelia & Tuller, Laurie. 2008. Relative clauses in atypical
acquisition of French. In: A. Gavarró and J. Freitas (eds): Language Acquisition and Development:
Proceedings of GALA 2007, Barcelona. Cambridge Scholars Publishing, Newcastle, 166–176.
Delage, Helene, & Tuller, Laurie. 2007. Language Development and Mild-to-Moderate Hearing Loss: Does
Language Normalize with Age? Journal of speech, language and hearing research, 50, 1–14.
Fattal, Iris, Friedmann, Naama & Fattal-Valevski, Aviva. 2011. The crucial role of thiamine in the development
of syntax and lexical retrieval: a study of infantile thiamine deficiency. Brain 134, 1720–1739. doi:
10.1093/brain/awr068.
Friedmann, Naama, Belletti, Adriana, Tuller, Laurie, Costa, João, Guasti, Maria Teresa, Haman, Ewa…
Varlokosta, Spyridoula. 2012. The acquisition of relative clause production in sixteen languages.
Unpublished ms. COST A33
Friedmann, Naama, Belletti, Adriana & Rizzi, Luigi. 2009. Relativized relatives: Types of intervention in the
acquisition of A-bar dependencies. Lingua, 119, 67–88. doi: 10.1016/j.lingua.2008.09.002
12
Naama Friedmann, Maya Yachini and Ronit Szterman
Friedmann, Naama & Novogrodsky, Rama. 2004. The acquisition of relative clause comprehension in Hebrew:
A study of SLI and normal development. Journal of Child Language, 31, 661–681.
Friedmann, Naama & Novogrodsky, Rama. 2007. Is the movement deficit in syntactic SLI related to traces or
to thematic role transfer? Brain and Language, 101, 50–63.
Friedmann, Naama, & Novogrodsky, Rama. 2011. Which questions are most difficult to understand? The
comprehension of Wh questions in three subtypes of SLI. Lingua, 121, 367–382.
Friedmann, Naama, & Szterman, Ronit. 2006. Syntactic movement in orally-trained children with hearing
impairment. Journal of Deaf Studies and Deaf Education 11(1), 56-75. doi: 10.1093/deafed/enj002.
Håkansson, Gisela & Hansson, Kristina. 2000. Comprehension and production of relative clauses: A
comparison between Swedish impaired and unimpaired children. Journal of Child Language, 27, 313–
333.
Grillo, Nino. 2005. Minimality effects in agrammatic comprehension. In Proceedings of ConSOLE XIII,
Blaho, S., Schoorlemmer, E., and Vicente, L. (eds.), 107-120.
Levy, Hagar & Friedmann, Naama. 2009. Treatment of syntactic movement in syntactic SLI: A case study.
First Language, 29, 15–50. doi:10.1177/0142723708097815
Marinellie, Sally A. 2004. Complex syntax used by school-age children with specific language impairment
(SLI) in child-adult conversation. Journal of Communication Disorders, 37, 517–533.
Novogrodsky, Rama & Friedmann, Naama. 2006. The production of relative clauses in syntactic SLI: A window
to the nature of the impairment. Advances in Speech-Language Pathology, 8(4), 364–375.
Yachini, Maya, Szterman, Ronit & Friedmann, Naama. 2013, July. The effect of intervention on comprehension
and production in developmental syntactic impairment. Presented in the 12th IMAM conference,
Language and Brain Lab, Tel Aviv University.
Stavrakaki, Stavroula. 2001. Comprehension of reversible relative clauses in specifically language impaired and
normally developing Greek children. Brain and Language, 77, 419–431.
van der Lely, Heather K. J. 1997. Language and cognitive development in a grammatical SLI boy: Modularity
and innateness. Journal of Neurolinguistics, 10, 75–107.
13