A LESLLA corpus

1
A LESLLA corpus
Ineke van de Craats i.v.d.craats_at_let.ru.nl Ra
dboud University, Nijmegen Research funded by
NWO (355-70-017)
LESLLA 2010 - Cologne
2
Why LESLLA?

• LESLLA learners differ from highly educated
  learners.
• LESLLA learners may process spoken language
  differently (Petersson et al.
  2000 Reis Castro-Caldas, 1997)
• Language pedagogy to LESLLA learners should take
  into account that
• - abstract words and function words are not
  viewed as words
  (e.g. Kurvers, 2002 )
• - words with little meaning are difficult to
  recall
• (e.g. Tarone et al. 2007)
• - LESLLA-learners have little metalinguistic
  and strategic skills (e.g.
  Kurvers 2002)
• - they may understand feedback /recasts
  differently
• (Tarone et al. 2007)
• - reading and writing skills are lacking or
  restricted.

3
LESLLA learners by accident

• LESLLA corpora avant la lettre
• Longitudinal
• - Heidelberger Forschungsprojekt (Klein
  Dittmar 1979)
• ZISA project (Clahsen et al. 1983)
• ESF project (Klein
  Perdue, 1992 Perdue 1993)
• Cross-sectional
• Lexlern project (Clahsen et al. 1991)
• Is acquisition possible solely on the basis of
  aural input?

4
LESLLA learners at purpose

• Minneapolis Somali literacy study
• (Tarone, Bigelow Hansen 2007 2009)
• What is the impact of literacy on oral L2 use
  and development?
• 3 studies, on corrective feedback,
• elicited imitation
• oral narratives (learners focussing on
  meaning)
• LESLLA-corpus (van de Craats 20092010)
• longitudinal 3 semesters (15-18 months)
• How does the low-literate learner proceed in the
  classroom?

5
LESLLA corpus

• Purpose of the project
• Is the L1 morphosyntax an essential factor of
  stagnation in L2 acquisition
• of these learners?

6
LESLLA corpus

• Research questions
• Is there still impact from a low literacy level
  or from another phonetic script?
• Low-paced development for morphosyntax?
• Is that caused by transfer from L1?

7
LESLLA corpus - design

• Longitudinal study of 3 semesters/cycles (1.5
  years)
• Same tasks administered in each cycle.
• Two language groups L2 Dutch
• L1 Turkish, Moroccan Arabic
• CEF level at start below A1 with a very basic
  vocabulary.
• Elicitation material Reading task
• Receptive tasks
• Productive tasks
• Metalinguistic tasks

8
Turkish participants (at start)
Age Schooling Stay in NL stagnating TU -
Neth. years Zilfi 30 5 15
11 no Hülya 19 5 07 0.5 no Emine 28
5 08 13 no - yes Hilal 19 5 18
2 yes - no Ayfer 37 5 08
18 yes Nazife 31 5 06
1 yes Hatice 45 5 06 26 yes Özlem 31
6 20 5 yes Mean 30 5 10 9.5

9
Moroccan participants (at start)
Age Schooling Staying NL Stagnating MA -
Neth. Mina 23 0 20 4 no Zohra 41
5 07 8 no Soad 34 4 08
12 no Najat 25 4 16 4 yes -
no Hayat 22 5 20 2 yes Nezha 38 0
13 3 yes Fatima 27 7
18 5 yes Mean 30 3.6 13 5.4

10
Reading skills Moroccan participants
Age Schooling Arabic literacy course
Roman script Mina 23
0 yes yes Zohra 41 5 yes no Soad 34
4 yes no Najat 25 4
yes yes Hayat 22 5 yes yes Nezha 38
0 yes yes Fatima 27
52 yes no Mean 30 3.6

11
Literacy level
3 tasks text self paced reading drag and drop
task
12
Reading time for an L2 text (in sec.)
Turkish Moroccan

• Cycle I II III Mean
• Zilfi 131 108 111 117
• Hülya 86 110 100 99
• Emine 116 133 137 129
• Hilal 143 103 132 126
• Ayfer 122 194 140 152
• Nazife 117 98 99 105
• Hatice 145 183 137 155
• Özlem 182 107 90 126
• Mean 130 129 118 126

• Cycle I II III Mean
• Mina 148 144 129 140
• Zohra 141 - 138 139
• Soad 127 152 119 133
• Najat 172 174 165 170
• Hayat 217 244 174 212
• Nezha 307 264 225 265
• Fatima 120 123 161 135
• Mean 176 183 158 172

13
Self-paced reading task

• Instruction
• Read the sentence aloud and recall the last
  word.
• Push the button.
• Read the next sentence (12 and 16 syllables)
  etc.
• 3 pairs of 2 sentences
• 3 pairs of 3 sentences
• 3 pairs of 4 sentences
• Say the last word of each sentence in the right
  order.

14
Self-paced reading task (16 syllables)
Reading time Turkish Moroccan learners 2
sent.-pair 1 21 sec 30 sec 2 sent.-pair 2 20
sec 36 sec 2 sent.-pair 3 22 sec 41 sec Mean per
sentence 10 sec 18 sec 3 sent. mean 10
sec. 16 sec 4 sent. mean 10 sec. 16 sec (for 10
Turkish and 10 Moroccan learners in Cycle 1)

15
Transfer and reading
L2
L1
Order within the noun phrase
16
Drag-and-drop task

• Purpose What is the influence of the L1?
• Semi-controlled task
• Adapted version of the drag-and-drop task
• The learner has more blocks to drag and drop
  than required for the task.
• This opens the way to investigate
• - pro-drop and/or topic drop
• - choose between an L1 and an L2 structure.

17
Drag-and-drop task
Possessive relationship in the noun
phrase Turkish order possessor possessee
Hassans car (Hasan-in araba-si) Moroccan
Arabic possessee possessor the car
(of) Hassan Dutch 1. Hassans auto Hassan
zn/zijn auto zijn auto 2. de auto van
Hassan de auto van hem

18
Adapted drag-and-drop task

Dat is .
Target Hassan zn auto Not correct auto van
Hassan (correct de auto van Hassan) Number of
moves and reaction time were registrated.
19
Results for 10 possessive noun phrases
TU MA difference Cyc.1 Moves
45,15 40,29 5 RT sec. 286.83 387.18 100
.35 sec Cyc.2 Moves 49.25 40.99 8 RT
sec. 246.7 294.4 47.7 sec Cyc.3
Moves 58.21 40.17 18 RT sec. 293.58 310.58
17 sec Number of moves increases for Turks,
stable for Moroccans. Number of seconds decreases
for Moroccans, not for Turks. Turks start
manipulating the word order, Moroccans not (they
read faster than before).

20
Easiest and most difficult possessive NPs

• Moroccan Turkish
• Easiest NP 2.95 moves 3.16 moves
• jouw kado mevr. Larbi dr man
• your present mrs. Larbis (her) husband
• Most difficult NP 5.5 moves 6.58 moves
• Freeks ouders de opa van Bas
• Freeks parents the grandpa of Bas
• Transfer of the L1 becomes clear in the
  Psor-Psee order.

21
Bas grandpa (de opa van Bas)

• Most frequent variant of Turkish learners
• Bas-van de opa (Turkish genitive)
• Bas-van zn de opa
• (stimulus Dat is . /Bas/ de opa/zn /van)
• De opa van Bas 12,5 correct for the Turks
  (mean6,5) also in 14 / 20 / 22 moves, or in 3 or
  5 moves.
• All Moroccans do it 100 correctly, but they need
  too many moves (mean 4) and too much time (mean
  30 sec.)
• Abstract function words play a crucial role.
• Syntactic development is slow.

22
Transfer and reading
L2
V finite
L1
Order within the sentence
23
The finite verb

• 1. In the drag and drop task
• Turkish SOVfinite
• Moroccan Arabic SVfinite O
• Dutch SVfinite O
• Make a sentence
• ..
• get Freek a fine
  gets

24
The finite verb

• 1. In the drag and drop task
• Turkish Cycle I II III
• Freek een bon krijgt/en. 50 37 37
• Moves (mean) 3,8 4,4 3,8
• RT (mean) in sec. 17 19,9 18,5
• Moroccan
• Freek een bon krijgt/en. 0 14 0
• Moves (mean) 3,1 3,6 3,8
• RT (mean) in sec. 35 27,9 20,7

25
Narratives

• Development of morphosyntaxis
• in a relatively free task
• film retelling and picture telling story.

26
Bare verbs

• Do these literacy learners produce abstract,
  grammatical (semantically redundant) morphemes,
  such as
• - inflectional endings (3sg)
• and
• - grammatical free morphemes (copulas,
  modals, auxiliaries)?
• that are difficult to process, or mainly bare
  verbs?

27
Turkish learners
28
Turkish and Moroccan learners compared
29
Conclusions

• The Turkish group produced more bare (long) forms
  than the Moroccan learners, although they were
  literate learners.
• The picture is opposite for the Moroccan group,
  although they were the moderate/advanced literacy
  learners with more short forms.
• Cause??
• Not literacy, but interplay between L1 and L2 is
  at issue.
• The short forms are mainly default forms

30
Default forms

• Beginning learners use default verb forms.
• Moroccan learners of Dutch prefer other default
  verb forms than Turkish learners. The L1-L2
  interplay is the motivation for this preference.
• Turkish learners take long forms (infinitives),
  Moroccans short defaults (finite).
• Turks have to acquire movement of the verb.
• Moroccans have to learn what an infinitive is.

31
A finite default form

• kan niet VP gaat fiets Najat
• can not go.3sg bike
• een vriendin is zij VP zegt kom Soad
• a girlfriend is she say.3sg come
• ik ga buiten VP speelt Fatima
• I go-1sg outside play.3sg
• For Moroccan learners the default form is
• short and finite.

32
The finite verb in picture telling task

dummy auxiliary

Zilfi (Turkish) Vader niet komen Cycl. 1
Vader is niet komen father is not
come.INF Cycl. 2 Vader is niet kom
come.1SG/STEM Cycl. 3 (pro) kom niet die vader

(v.d.Craats 2005, 2009)
33
The finite verb in picture telling task
another dummy auxiliary

• Moroccans
• Sneeuwman ga kijken tableaus
• snowman go look paintings
• Dan ga loop naar de raam
• Then go walk to the window

Number of ga-patterns 7 Turks 10 7
Moroccans 313
34
Dummy auxiliaries

• Dummy auxiliaries may
• Emerge when (bound) verb morphology has not fully
  been acquired yet
• Realise one of more grammatical features normally
  part of a lexical verb.
• Disappear after a specific developmental stage.
• Typically (??) emerge in the speech of vulnerable
  learners such as LESLLA learners, SLI children.

35
Conclusions

• Decoding fluency may influence results of written
  tasks with low-educated learners, but will have
  disappeared at A2-level (CEF).
• Low-paced development of these LESLLA-learners
  shows the small steps such as insertion of dummy
  auxiliaries to overcome composite morphosyntactic
  changes (L1 influence is persistent).
• Difficulty of grammatical morphemes has more to
  do with salience in an L1-L2 interplay. Lack of
  meaning is one aspect of salience (unstressed is
  another).
• Literate learners are also intended to skip
  grammatical morphemes in sentence imitation.
• The morpheme learning steps lexical item free
  functional item bound functional item are also
  observed in literate learners (e.g. VanPatten
  1995).
• More comparative research between literates and
  non/low-literates is necessary., e.g. about
  insertion of dummy auxilairies.

36
See you next LESLLA