Transcript Substitution of prepositions as an indicator of SLI in

Substitution and omission of
prepositions as indicators of
SLI in bilingual children
Sharon Armon-Lotem
The Bilingual SLI Project
Bar-Ilan University, Israel
1
Acknowledgements
This work has been done in collaboration
with:
Gabi Danon, Jonathan Fine, Elinor SaieghHaddad, Joel Walters, Bar-Ilan University
Galit Adam, The Open University
With the help of:
Anat Blass, Michal Giladi, Efrat Harel, Audrey
Levant, Ruti Litt, Lyle Lustinger, Sharon Porat,
Efrat Shimon
2
Why Prepositions?
Prepositions are a locus for code interference in
bilingual populations, but are not necessarily
considered an indicator of SLI
 Some SLI children show omissions of
prepositions (Roeper, Ramos, Seymour, and
Abdul-Karim 2001)
 While some prepositions have mainly a
grammatical function, other have a semantic
function, contributing to the meaning of the
sentence.

3
Two Types of Prepositions

Obligatory prepositions (O-prep):
– governed by the verb, e.g., laugh at, dream about,
look at, look for
– Form a meaning unit with the verb, e.g., take out,
turn off

Free prepositions (F-prep) which introduce an
adverbial PP (locatives, temporals), e.g., at
school, near the tree, in the morning, after
lunch
4
Obligatory Prepositions
An obligatory (subcategorized) PP, though a
complement of the verb, is not its argument, i.e.
the theta-role of the verb is assigned to the
nominal complement of P (cf. Neeleman 1997,
Botwinik-Rotem 2004).
 In phrasal verbs the theta-role of the verb is also
assigned to the nominal complement
 Obligatory prepositions are lexically selected by
the verb and only serve a grammatical function,
making them more prone to omission by SLI
children.

5
A Typological Note
In English, a satellite-framed language,
some O-preps (in phrasal or particle
verbs) often contributes to the meaning of
the verb (look for, turn off), adding an
expression of path (Talmy 2000).
 In Hebrew, a verb-framed language, this
meaning is mostly encoded within the
verb (xipes, kiba), and no preposition is
used.

6
Free Prepositions
Free prepositions introduce adverbial
(non-subcategorized) PPs. The PP can be
considered an argument, receiving a
theta-role from the verb, or an adjunct,
and the free preposition contributes to
determining the meaning of the sentence.
 Free prepositions are not selected by the
verb, and their choice is often influenced
by the following NP (at lunchtime, in
the morning)

7
Predictions





Better performance with F-preps since they
contribute to the meaning of the sentence, are
more frequent in spontaneous speech, and are
typologically similar.
Weaker performance on O-prep in both
languages since they mostly serve a
grammatical function.
Weaker performance on O-prep in Hebrew
where they never contribute to the meaning of
the verb.
TD bilingual children are expected to show code
interference (CI) in contrasting environment.
8
SLI children may show omissions (Roeper 2000)
Subjects Selection
5-7 years old sequential bilinguals from Englishspeaking homes who had been exposed to L2
Hebrew in Hebrew-speaking pre-school
programs for more than two years.
 Some of the children attended “language
preschools” due to earlier diagnostics, while
others attended regular preschools.
 All children came from the same neighbourhood
and the same (middle-high) SES.
 Children were tested in both languages using
standardized tests (CELF Preschool for English,
Goralnik for Hebrew).

9
Subjects
8 atypically developing (ATD) bilingual children
(7 girls, 1 boy) – all had a score which was
lower than -1 SD below the norm on the CELF
preschool for English and lower than -1.5 SD
below the norm on the Goralnik for Hebrew. All
these children matched the exclusionary criteria
for SLI.
 11 typically developing (TD) bilingual children (8
girls, 3 boys) - all scored within the norm in both
languages

10
The Elicited Imitation Task
24 simple sentences, in each language, 10
containing free prepositions, and 14
containing obligatory prepositions.
 Sentence length

– In Hebrew: 4-7 words (mean 5.6), 5-10
morphemes (mean 7.5), and word/morpheme
average is 6.5
– In English: 5-8 words/morphemes (mean 6.5).
11
Findings
Percentage of full target responses [N=24]
100%
79%
74%
80%
63%
58%
60%
40%
20%
0%
Hebrew
English
ATD



TD
Significantly more full target responses among TD than ATD (p<0.05
for both languages)
No significant difference between English and Hebrew (even though
Hebrew is the school language)
Quantitative difference between TD and ATD children in the number
12
of errors
Preposition Errors out of all Errors
Ratio of preposition errors and other errors
100%
3%
6%
7%
11%
14%
26%
80%
60%
23%
36%
40%
20%
0%
TD
ATD
TD
Hebrew
English
Other



ATD
Prepositions
Prepositions errors are the source of 10-15% of all errors in Hebrew, but
third of all errors in English.
Though ATD children make more errors, there is no significant difference
in the ratio of preposition errors between TD and ATD, but there is a
significant difference between Hebrew and English for both groups.
No quantitative difference in the number of prepositions errors between
13
TD and ATD
Errors by Preposition Type
Ratio of errors by preposition type
50%
40%
30%
20%
10%
5 4
10 10
6 7
Hebrew
English
Hebrew
19
6
0%
TD
ATD
O-prep



English
F-prep
For TDs there is no significant difference between O-preps and F-preps in
English or Hebrew.
For ATDs there is a significant difference in English between O-preps and
F-preps as well as between English and Hebrew in O-preps (both due to
the relatively large number of errors on O-preps in English)
A significant difference was found between TD and ATD in O-preps errors.
14
Summary - Ratio of Errors
Quantitative difference is found between TD and
ATD children in the number of errors
 No quantitative difference was found in the
number of prepositions errors between TD and
ATD
 A qualitative significant difference was found
between TD and ATD in O-prep errors, showing
that verb governed preposition are a locus of
unique difficulty in the English of the ATD
children.

15
Type of Errors





Substitution with Code Interference (ci-sub):
The baby laughed on the clown.
Substitution no Code Interference (sub):
The baby laughed to the clown.
Omission with Code Interference (ci-om):
The elephant pulled *(down) the zebra's pants.
Omission no Code Interference (om):
The baby laughed *(at) the clown
Code Switching (cs):
The baby laughed al ha-clown
16
Type of Errors – by Groups
Frequency of errors by type of error
*
20
15
*
10
5
0
CI-om
CI-sub
om
ATD
sub
TD
A significant difference was found between TD and ATD
in errors which are not due to code interference. This is
true both for non-CI omissions and non-CI substitutions.
17
Type of Errors by Preposition Type
Frequency of errors by type
*
*
20
15
10
5
0
O-prep F-prep O-prep F-prep O-prep F-prep O-prep F-prep
Hebrew
English
Hebrew
TD
English
ATD
ci-sub
sub
ci-om
om
18
Summary - Type of Errors

In Hebrew, no significant difference was found in
the total number of errors, number of
substitutions or number of omissions.
Nonetheless, a significant difference was found
between TD and ATD in non-CI omissions of
obligatory prepositions.

In English, no significant difference was found in
the total number of errors, number of
substitutions or number of omissions.
Nonetheless, a significant difference was found
in the use of obligatory prepositions in the total
number of errors, due to a significant difference
in the use of non-CI substitutions
19
Major Finding 1
Substitutions and omissions due to code
interference are not a unique feature of
the language of ATD children, and neither
are unsystematic substitution errors, while
unsystematic omission errors are unique
to ATD children.
20
Discussion - Omissions
Though some omission errors can be explained
by code interference, we would argue that the
restriction of omission errors which cannot be
explained by code interference to the ATD
population resembles those errors reported by
Roeper et al (2001) and might indicate
optionallity in the linguistic representation.
 The restriction of omission errors to O-preps is
explained by their limited contribution to the
semantics of the sentence (Clahsen, Bartke and
Göllner 1997, Tsimpli 2001).

21
Major Finding 2
Unsystematic errors (mostly substitutions)
are significantly more prevalent among
ATD children, and are due to significant
difference in their performance with Opreps.
22
Discussion – Unsystematic
Substitutions
Non CI substitution errors by TD children mostly
reflect the use of a synonym, and are therefore
not really unsystematic, but rather semantically
based.
 Unsystematic substitutions of free prepositions
by ATD children show that they know that a
preposition is needed. The semantics of the
adverbial helps the child choose the correct
preposition, and thus there are fewer errors.

23
Discussion – Obligatory
Prepositions

Unsystematic substitutions of obligatory prepositions by
ATD children can suggest that they know the theta-grid
for each verb and therefore know that a preposition is
needed (Botwinik-Rotem 2004), but do not know which
preposition it is. In the absence of semantic basis for the
choice, they pick up any preposition, with preference for
in and on, the semantics of which is less restricted.

The significant difference in unsystematic substitutions
for obligatory prepositions can be explained by their
limited contribution to the semantics of the sentence
(Clahsen, Bartke and Göllner 1997, Tsimpli 2001).
24
Conclusion

Unsystematic omission errors are unique to ATD
children and can be indicative of SLI. These
errors might stem from optionallity in the
linguistic representation.

Unsystematic errors (mostly substitutions) in the
use of obligatory prepositions are significantly
more prevalent among ATD children, and can
serve as a secondary indication for SLI. These
errors can be attributed to the limited
contribution of these prepositions to the
semantics of the sentence, but also to
processing limitations.
25
Partial Reference
Botwinik-Rotem, I. 2004. The category P: Features, projections,
interpretation, Ph.D. thesis, TAU.
Clahsen H, Bartke S and Göllner S. 1997. Formal features in impaired
grammars: a comparison of English and German SLI children.
Journal of Neurolinguistics 10: 151-171
Neeleman, A. 1997. "PP-Complements", NLLT, 15:89-137.
Talmi, L. 2000. Toward a Cognitive Semantics. Cambridge: MIT Press
Tsimpli I (2001) ‘Interpretability and language development: A study of
verbal and nominal features in normally developing and SLI Greek
children’, Brain and Language 77: 432-52.
Ullman, M.T. & Pierpont, E.I. 2005. Specific Language Impairment is
not Specific to Language: The Procedural Deficit Hypothesis. Cortex
41, 399-433.
26
Thank you
‫תודה‬
27