Effects of age of acquisition (AoA) and proficiency on processing of syntax

in 6- to 8-year-old monolingual and bilingual children: an ERP study

Annika Andersson1, Lisa D Sanders2, Christina Karns3, Helen J Neville3
1Lund University, 2University of Massachusetts, 3University of Oregon

Introduction Participants Monolingual (M) Bilingual (B) Comparisons: M / B

A left lateral negativity between 800-1500 ms A right medial negativity that weakened with time Different distribution of anterior negativity across hemispheric sites
Language proficiency in children is strongly related to A positivity between 300-1500 ms strongest A posterior positivity between 1000-1500 ms between 200-800 ms and across the lateral medial plane between
success in almost all domains, however neurocognitive medially and over occipital sites strongest medially 800-1500 ms.
studies of second language (L2) processing are
Mean difference 200-800 ms Mean difference 800-1500 ms
typically limited to adults with long exposure, who may µV
µV
-2
use general cognitive mechanisms to compensate for -2.5
-2
-1.5
-1
any difficulties in L2 processing. Previous studies of -1.5 -0.5
-1 M 0 M
adult bilinguals have reported especially differences in -0.5
0
B 0.5 B
1
the anterior negativity elicited by syntactic violations 0.5
1
1.5
2
with delays in exposure to English of less than 3 years1 1.5
Left Hemisphere Right Hemisphere
2.5
Lateral Medial
while a precursor to the anterior negativity has been
reported in monolingual children as young as 2.5 years M/B Anterior effect (100-200 ms, S x Ling: F(5,385) = 6.08 , p < .05; S x L x Ling, F(1,77) = 5.79, p < .05; 200-800 ms, S x H x Ling: F(1,77) = 4.20, p < .
05; S x L x Ling: F(1,77) = 4.92, p < .05; S x Ling at F4: F(1,77) = 4.10, p < .05; 800-1500 ms S x L x Ling: F(1,77) = 2.68, p = .106)

of age2. A stronger posterior positivity in M compared to B

  Mean difference 300-500 ms
In the current ERP study, processing of English phrase Mean difference 500-700 ms M/B Posterior effect (300-500 ms: S x
H x Ling: F(1,77) = 6.01, p < .05; at right

structure was explored in 6- to 8-year old monolingual English Proficiency µV

-2
Right hemisphere sites
µV
-2
Medial sites hemisphere regions: S x Ling: F(1,77) = 7.26,
p < .01; 500-700 ms: S x Ling: F(1,77) =
3.23, p = .076, S x L x Ling: F(1,77) = 4.20, p
< .05; at medial sites: S x Ling: F(1,77) =

(M) and bilingual (B) children who acquired English as

-1 -1
M M 4.05, p < .05; 700-1000 ms: S x AP x Ling:
0 0 F(3,385) = 2.55, p = .099; at P and O: S x
B B Ling: F(1,77) = 4.40, p < .05; 1000-1500 ms:
a second language around 4 years of age. 1

2
1

2
S x Ling: F(5,385) = 3.54, p = .064)

3 3

Effects of AoA and English proficiency were explored in 4 4

the following types of analyzes: Mean difference 700-1000 ms Mean difference 1000-1500 ms

•  M compared with B Parietal and occipital sites All electrode sites

µV
-2 -2

•  M and B divided by median split into groups with -1

M
-1
M
0 0

higher (H) and lower (L) proficiency (P) in English Anterior effect (100-200 ms, positivity over T and C: F(1,39) = 3.95, p = .054; 200-800 ms, positivity over right
Fand T: F(1,39) = 6.22, p < .05; 800-1500 ms, S x H x AP:F(5,195) = 2.45, p = .068, S x L x AP: F(5,195) = 4.31, p
Anterior effect (100-200 ms, S: F(1,38) = 4.04, p = .052; S x L: F(1,38) = 3.65, p = .064, S x H x L: F (1,38) =
5.58, p > .05; S at right hemisphere sites: F(1,38) = 4.20, p < .05; S at medial sites: F(1,38) = 4.44, p < .05’ 200-800 1
B
1
B

< .005).
(MHP, MLP, BHP, BLP) to compare proficiency Posterior effect (300-500 ms: S: F(1,39) = 4.29, p < .05; 500-700 ms: S: F(1,39) = 4.84, p < .05; 700-1000 ms:
S: F(1,39) = 7.24, p < .05, S x AP: F(5,195) = 7.19, p < .01; 1000-1500 ms: S: F(1,39) = 28.33, p < .001, S x AP:
ms, F and FT, S x L : F(1,38) = 3.48, p = .07, S x AP: F(1,38) = 5.64, p < .05, S x L x AP: F(1,38) – 5.08, p < .05; S at
F3 and F4: F1,38) = 3.13, p = .085: 800-1500, S x H x AP: F(1,38) = 5.81, p < .05; at F4/8 all ps > .427).
Posterior effect (1000-1500 ms: S: F(1,38) = 5.57, p < .05, S x AP: F(5,190) = 3.01, p = .075; S x L: F(1,38) =
2

3
2

effects within each lingualism group F(5,195) = 7.44, p < .01; S x L: F(1,39) = 7.66, p < .01). 3.06, p = .088).
4 4

Paradigm Difference waves MHP / MLP Difference waves BHP / BLP Comparisons: BHP / BLP
•  Claymation video, featuring “Pingu”, with synchronized stimulus audio
•  Participants seated in comfortable chair and asked to pay close attention
In MHP a left lateral anterior negativity 200-1500 In BHP a right medial negativity 200-1500 ms but An anterior negativity in BHP only
•  5 videos presented as 7-minute segments, each video containing 100 ms but no negativity in MLP no negativity in BLP A stronger posterior positivity in BHP especially over left
sentences (20 syntactical canonical/violation pairs)
E.g. Pingu comes outside of this igloo
In MHP a lateral posterior positivity 300-1500 ms, In BHP a left posterior positivity 700-1500 ms, in hemispheric sites
Pingu comes outside of his this igloo in MLP
VE
a right
HE
temporal positivity 300-500 ms BLP aVE right central
HE
positivity 1000-1500 ms BLP/BHP Anterior (100-200 ms all p’s > .117; 200-800 ms: S x H x AP x BHP/PLP Posterior (700-1000 ms C P O: S x H x AP: F(5,160) =

and a medial posterior positivity 700-1500 ms

Group: F(5,150) = 2.35, p = .087; at F4 and F8 all p’s > .704; 800-1500 ms S x 4.07, p < .05; over left regions: S x Group: F(1,30) = 4.39, p < .05;

EEG recording & conventions Left hemisphere Right hemisphere

H x Group: F(1, 30) = 4.44, p < .05, S x H x AP x Group: F(5,150) = 4.08, p < .
05; at F4 and F8 all p’s > .281
 
1000-1500 ms S x H x Group: F(1,30) = 4.22, p < .05; at subsets of
electrode sites all p’s > .128).

•  29 tin electrodes (Electro-Cap International). FP1/2, Left hemisphere Right hemisphere

Lateral Medial Medial Lateral

Conclusion
Lateral Medial Medial Lateral
F7/8,FT7/8, F3/4, FC5/6, C3/4, C5/6, T3/4, CP5/6, P3/4, T5/6, f7 f3 f4 f8 2μV f7 f3 f4 f8
TO1/2, and O1/2 and three midline sites: Fz, Cz, and Pz. Data
2μV
_ _
Frontal Frontal
from midline sites and frontal pole sites (FP1/2) were not
Comparisons: MHP / MLP
100 500 1000 1500 ms
+
Monolingual children of higher proficiency displayed
100 500 1000 1500 ms
+
included in analyses.
•  Electrodes were placed to below left eye and at canthus of both
relatively mature processing of phrase structure violations eyes to monitor eye movements. ft7 fc5 fc6 ft8 ft7 fc5 fc6 ft8
An anterior negativity in MHP only
•  Online reference: right mastoid. Offline re-referenced to the
as indicated by a left anterior negativity over lateral sites averaged mastoids
Fronto-
temporal
Fronto-
temporal
µV Mean difference 800-1500 ms MHP/MLP Anterior (100-200 ms:
and a posterior positivity. High-proficiency bilingual •  Impedance: facial electrodes < 10kΩ, mastoid- and scalp-
electrodes < 5 kΩ.
-3
-2
F Ft
S x Group: F(1,32) = 3.65, p = .065, S
x L x Group: F(1,32) = 7.97, p < .01; S
x Group over medial sites: F(1,32) =
t3 c5 c6 t4 t3 c5 c6 t4
children tended to display a medial anterior negativity and •  Grass amplifiers (bandpass .01-100 Hz; sampling rate 250 Hz).
Temporal Temporal
-1
0
MHP 4.75, p < .05; 200-800 ms: all ps > .
181; 800-1500 ms: S x AP x Group:
•  ERPs time-locked to onset critical word (canonical or violation). MLP F(5,160) = 3.37, p = .058; F and FT: S
a posterior positivity. •  Artifacts: ‘runica’ routine of EEGLAB software (Delorme &
1
2
x Group: F(1,32) = 3.67, p = .064, S x
H x AP x Group: F(1,32) = 3.54, p = .
Makeig, 2004). Ocular artifacts were identified from scalp 3 069; at subsets of electrode sites all
4 p’s > .072).
topographies and the component time series and removed. ct5 c3 c4 ct6 ct5 c3 c4 ct6
The difference in distribution of the anterior effect across •  Digital low-pass 40 Hz filter and high-pass filter .1 Hz. Central Central
•  Participants with 10 or more artifact free trials in each of the two
groups was explained by AoA. However, English main conditions were included.
proficiency affected distribution and amplitude of the •  Mean amplitude across anterior (frontal, fronto-temporal, and t5 p3 p4 t6 t5 p3 p4 t6 A different distribution of the positivity across groups with a
temporal; 100-200 ms, 200-800 ms, and 800-1500 ms) and
posterior ERP effect and amplitude of the anterior effect across posterior (central, parietal, and occipital; 300-500 ms, Parietal Parietal
stronger effect in MHP over occipital sites
500-700 ms, 700-1000 ms, and 1000-1500 ms).
in response to syntactic violations in both groups. •  4-way within subject repeated measures ANOVA: Syntax (S: µV Mean difference 1000-1500 ms MHP/MLP posterior (300-500 ms: all p’s > .
148; 500-700 ms: S x AP x Group: F(5,160) =
-4
canonical/violation), Hemisphere (H: right/left), Lateral/Medial to1 o1 o2 to2 to1 o1 o2 to2 2.81, p = .081; at P and O : S x H x L x Group:
F(1,32) = 7.32, p < .05; at right lateral P and O: S
position (L: lateral/medial), and AP position (AP: anterior Occipital
-2 x Group: F(1,32) = 4.81, p < .05; 700-1000 ms all
These results suggest that the more automatic syntactic effects: frontal/fronto-temporal/temporal; posterior effects: Occipital
0
p’s > .126; 1000-1500: S x AP x Group: F(5,160)
= 3.45, p = .056: at F, FT, T, and C : S x Group:
central/parietal/occipital).
processing in children (the anterior negativity) is affected
F(3, 96): 4.60, p < .05).
2
•  Between-subjects factor: Lingualism (Ling: M/B), Proficiency MHP anterior (F and FT : S x H x L x AP: F(5,80) = 2.35, p = .069; over F7 and FT7: all p‘s > .287; 800-1500 ms: S x BHP Anterior (100-200: S x H x AP: F(5,75) = 3.08, p =.051; at F4 and F8 all p’s >.21; 200-800 ms: S x H x AP: F(5,75) MHP
= 3.54, p < .05, S x H x L x AP: F(5,75) = 2.19, p = .083; at F4 and F8 all p’s > .386; at F3/4 all p’s > .384; 800-1500 ms: S
by AoA while more controlled, metalinguistic processing (Prof: HP/LP). Greenhouse-Geisser correction was applied to
all measures with more than two levels.
AP:F(5,80) = 15.41, p < .001; S x L x AP: F(5,80) = 3.03, p < .05; at F and FT regions: S x H x AP: F(1,16) = 3.91, p = .
066; over F7 and FT7: S: F(1,16) =5.65, p < .05).
MHP posterior (300-500 ms: S x AP: F(5,80) = 2.17, p = .085; over C, P and O sites: S x H x L x AP: F(2,32) = 4.76, p
x H: F(1,15) = 5.32, p < .05, S x AP: F(5,75) = 5.35, p < .05; at frontal sites: S x H: F(1,15) = 3.57, p = .078; at F4 and F8
all p’s > .417).
4

6
MLP

(P600) may be related to language proficiency. •  Receptive Language (CELF3), Receptive Vocabulary (PPVT4),
Fluid Reasoning nonverbal IQ (SB-55). Right-handed (Oldfield,
1971), normal or corrected-to-normal vision, normal hearing,
< .05; S at right hemisphere C,P, and O sites: F(1,16) = 4.78, p < .05; 500-700 ms: S x AP: F(5,80) = 3.63, p = .059; at T,
C, P, and O sites: S: F(1,16) = 6.70, p < .05, S x H x L x AP: F(3,48) = 3.45, p < .05; 700-1000 ms: S x AP: F(5,80) = 11.67,
p < .005; at T, C, P, and O sites: S: F(3,48) = 5.23, p < .05, S x AP: F(3,48) = 9.65, p < .005; 1000-1500 ms: S: F(1,16) =
7.39, p < .05, S x AP: F(5,80) = 13.97, p < .005, S x L x AP: F(5,80) = 3.16, p < .05).
BHP Posterior (700-1000 ms: S x H: F(1,15) = 5.50, p < .05; S at left hemisphere sites: F(1,15) = 7.78, p < .05;
1000-1500 ms: S x H: F(1,15) = 4.40, p = .053, S x AP: F(5,75) = 6.15, p < .05; at T, C, P, and O: S: F(1,15) = 6.68, p < .
05, S x H: F(1,15) = 4.67, p < .05).
8
BLP Anterior (100-200 ms: S x H x L: F(1,15) = 10.07, p < .01, S x L x AP: F(5,75) = 3.67, p < .05; at F, FT, T, and C: 10

References
S: F(3,45) = 5.17, p < .05, S x H x L: F(1,15) = 4.34, p = .052; 200-800 ms all p’s > .261; 800-1500: S x H x AP: F(5,75) = F FT T C P O
were not known to have behavioral or neurological problems MLP posterior (300-500: S x H x AP: F(5,80) = 2.86, p = .053; at right hemisphere T: S: F(1,16) = 5.32, p < .05; 2.28, p = .093)
1Weber-Fox & Neville, 1996, 2Oberecker, Friedrich, & Friederici, 2005, 3Clinical Evaluation of Language Fundamentals; Semel, Wiig, & 700-1000 ms: S: F(1,16) = 4.80, p < .05; 1000-1500 ms: S: F(1,16) = 11.86, p < .01, S x L: F(1,16) = 4.92, p < .05).
BLP Posterior (1000-1500 ms at right hemisphere C and P: S: F(1,15) = 5.33, p < .05).
Secord, 1995, 4Peabody Picture Vocabulary Test; Dunn & Dunn, 1997, 5Stanford Binet Intelligence Scale; Roid, 2003

Contact: Annika Andersson, Lund University Humanities Lab, annika.andersson@humlab.lu.se

Acknowledgements We gratefully acknowledge the assistance of Olivia, Bagdade, Ted Bell, Jeff Currin, Jessica Fanning, Petya Ilcheva,
Gloria Larson, Brittni Lauinger, Nicole Makarenco, Brittany Olsen, David Paulsen, and Laura Sabourin who all spent many hours testing
the children in this study.