Morphological Processing in a Second Language:
Behavioral and Event-related Brain Potential Evidence
for Storage and Decomposition
Anja Hahne1, Jutta L. Mueller1, and Harald Clahsen2
D
o
w
n
je
o
un
d
e
d
Abstrait
& This study reports the results of two behavioral and two
event-related brain potential experiments examining the
processing of inflected words in second-language (L2) learners
with Russian as their native language. Two different subsystems
of German inflection were studied, participial inflection and
noun plurals. For participial forms, L2 learners were found to
widely generalize the -t suffixation rule in a nonce-word
elicitation task, and in the event-related brain potential experi-
ment, they showed an anterior negativity followed by a P600—
both results resembling previous findings from native speakers
of German on the same materials. For plural formation, the L2
learners displayed different preference patterns for regular and
irregular forms in an off-line plural judgment task. Regular and
irregular plural forms also differed clearly with regard to their
brain responses. Whereas overapplications of the -s plural rule
produced a P600 component, overapplications of
irregular
patterns elicited an N400. In contrast to native speakers of
German, cependant, the L2 learners did not show an anterior
negativity for -s plural overapplications. Taken together, le
results show clear dissociations between regular and irregular
inflection for both morphological subsystems. We argue that
the two processing routes posited by dual-mechanism models
of inflection (lexical storage and morphological decomposition)
are also employed by L2 learners. &
INTRODUCTION
Previous research indicates that inflectional morphology
is an area of specific difficulty for adult second-language
(L2) learners. L2 learners are said to use inflectional
morphemes in an unsystematic fashion (Meisel, 1991)
and reduced morphological paradigms (Klein, 1986).
Some researchers (Pre´vost & Blanc, 2000; Haznedar &
Schwartz, 1997) speculate that the adult L2 learners’
difficulties with inflectional morphology might be be-
cause of processing reasons. Cependant, the details of L2
morphological processing have not been studied by
these researchers. An interesting hypothesis as to how
L2 processing might differ from native language (L1)
processing has been advanced by Ullman (2001) depuis
the perspective of dual-mechanism models of morphol-
ogy (voir, par exemple., Clahsen, 1999; Pinker, 1999, for reviews).
He argues that processing one’s native language involves
two different brain memory systems, a lexical store of
memorized inflected words that depends upon declar-
ative memory and is rooted in temporal lobe structures,
and a mental grammar which includes combinatorial
rules and is rooted in frontal brain structures. Given
these assumptions, Ullman (2001) claims that L2 pro-
1Max Planck Institute for Human Cognitive and Brain Sciences,
2University of Essex
cessing and representation is largely dependent upon
the lexical memory system and invokes grammatical
computation to a much lesser extent than L1 processing.
He further assumes that reliance on the procedural
memory system should increase with practice (Ullman,
2004). For morphological processing, this means that L2
learners mainly rely on full-form storage of inflected
words, whereas morphological decomposition is under-
used or even absent in L2 processing of inflected words,
and this might perhaps be the reason for why inflec-
tional morphology is hard for L2 learners. Malheureusement,
cependant, there is very little empirical evidence for these
claims, and the details of how adult L2 learners process
inflected words remain largely unknown.
With respect to L1 processing, experimental studies
using a range of different psycholinguistic methods and
techniques have led to a number of consistent and rep-
licable results on how morphologically complex words
are processed in one’s native language, Par exemple,
frequency effects for inflected word forms in lexical
decision tasks and stem-priming effects for regularly
inf lected word forms in different kinds of priming
experiments. The theoretical
interpretation of these
and other results has been the subject of a controversy
between associative single-mechanism models and a
family of dual-mechanism models of inflection. A de-
tailed review of this controversy is beyond the scope of
D 2006 Massachusetts Institute of Technology
Journal des neurosciences cognitives 18:1, pp. 121–134
je
je
/
/
/
/
/
j
F
/
t
t
je
t
.
:
/
/
F
r
o
m
D
o
h
w
t
n
t
p
o
:
un
/
d
e
/
d
m
je
F
r
t
o
p
m
r
c
h
.
p
s
je
je
d
v
je
r
e
e
c
r
t
c
.
m
h
un
je
e
r
d
.
toi
c
o
o
c
m
n
/
j
un
o
r
t
c
je
c
n
e
/
–
un
p
r
d
t
1
je
8
c
je
1
e
1
–
2
p
1
d
1
F
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
oui
7
g
7
toi
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
F
p
e
b
m
oui
b
e
g
r
toi
2
e
0
s
2
t
3
t
/
.
F
o
n
1
8
M.
un
oui
2
0
2
1
the present article. C'est, cependant, important to provide a
brief summary of previous event-related brain potential
(ERP) studies on morphological processing in adult
native speakers.
Replicable ERP effects have been found in studies
of inflectional morphology, in German (Lu¨ck, Hahne,
Friederici, & Clahsen, 2001; Penke et al., 1997; Weyerts,
Penke, Dohrn, Clahsen, & Mu¨nte, 1997), English (Mu¨nte,
Say, Schiltz, Clahsen, & Kutas, 1999; Newman, Izvorski,
Davis, Neville, & Ullman, 1999), Catalan (Rodriguez-
Fornells, Clahsen, Lleo, Zaake, & Mu¨nte, 2001), Italian
(Gross, Say, Kleingers, Mu¨nte, & Clahsen, 1998), et
Spanish (Rodriguez-Fornells, Mu¨nte, & Clahsen, 2002).
As the present study examines German as a target lan-
guage, our focus here is on the three German ERP
violation studies. Two types of violation were tested in
these studies, (un) regularizations, formed by adding a
regular suffix to a verb or noun that requires an irregular
un, (b) irregularizations, in which a verb or noun that
takes the regular default suffix appeared with an incorrect
(irregular) ending. Penke et al. (1997) examined partici-
ple formation in three experiments, Weyerts (1997) et
Lu¨ck et al. (2001) noun plurals. Penke et al. and Weyerts
et autres. presented their stimuli visually, Lu¨ck et al. audi-
torily. In all these experiments, an anterior negativity
entre 300 et 800 msec was found for regulariza-
tion (which was larger over the left than over the
right hemisphere). De plus, Lu¨ck et al. found a centro-
parietally distributed positivity (P600) dans le 800- à
1200-msec time window for regularizations. For irregu-
larizations, both the visual and the auditory studies on
plurals elicited an N400-like negativity compared with
their correct counterparts. These results were inter-
preted as supporting a dual-mechanism account of
morphological processing. From this perspective, regu-
larizations are combinatory violations, c'est, misappli-
cations of the participle -t or the plural -s to (irregular)
verbs or nouns that would normally block these rules, à
produce illegal stem + affix combinations. Par contre,
irregular inflection is based on full-form storage, et
misapplications of irregular inflection produce pseudo-
words, as suggested by the N400 effect for (plural)
irregularizations.
In contrast to the rich experimental literature on L1
morphological processing, very little is known about
how adults process inflected words in an L2. Il y a
three L2 studies employing speeded production or gram-
maticality judgment tasks (Brovetto & Ullman, 2001;
Beck, 1997; Lalleman, van Santen, & van Heuven, 1997).
The findings from these studies are inconsistent and
partly surprising. Although the native speaker controls
showed a consistent response-time advantage for high-
frequency irregulars (but not for high-frequency regu-
lars) in all experiments, most studies failed to replicate
this contrast for the L2 learners. Only Brovetto and
Ullman (2001) obtained the same significant frequency
effect for irregulars in L2 learners as in native controls.
The L2 learners’ results on regulars were even more
inconclusive. Given the inconsistencies across and even
within studies, any general conclusions would appear to
be premature. Clairement, more research on L2 morpholog-
ical processing is needed.
Given the sensitivity of ERPs as a distinctive neuro-
physiological measure of syntactic and lexical processes,
we made use of this technique to further investigate
how L2 learners process regular and irregular inflection.
We investigated two inflectional systems of German,
participle formation and noun plurals, which both con-
tain regular and irregular forms but differ in terms of
their morphological complexity and frequency distribu-
tion. In addition to two ERP experiments, one on par-
ticiple inflection and one on plural inflection, we will
report the results of two behavioral experiments on the
same phenomena, adding off-line performance mea-
sures to the on-line ERP data. D'abord, we will provide a
brief description on how participles are formed in Ger-
man and in our L2 learners’ native language (Russian).
EXPERIMENT 1: PAST PARTICIPLES
Past participle formation in German involves two end-
ings,
-n appears on all participle forms of so-called
fort (= irregular) verbs and -t appears on participle
forms of all other verbs. Irregular verbs undergo (pho-
nologically unpredictable) stem changes in the simple
past (= preterit) and at times also in the participle (par exemple.,
laufen [infinitive]–gelaufen [participle]–lief [simple
past] ‘‘to walk–walked–walked’’). There are about 160
simplex verbs that fall into the strong (= irregular) class.
Regular (= weak) verbs do not exhibit any stem changes
in the past participle or the simple past (par exemple., tanzen–
getanzt–tanzte ‘‘to dance–danced–danced’’). In addi-
tion, traditional grammars identify a small class of 13
verbs of so-called mixed inflection, which take the
(regular) participle ending -t but also show a stem
change in the simple past and in the participle. Prefix-
ation with ge- (as in ge-tanzt) is required when the
verbal stem is stressed on the first syllable and is
irrelevant for the morphological distinction between
regular and irregular inflection, as it occurs for both
regular (-t) and irregular (-n) participles. The participle
suffix -t is applied under default circumstances in the
sense of Marcus, Brinkmann, Clahsen, Wiese, and Pinker
(1995), c'est, to words for which lexical entries are not
readily available, -n participle forms, by contrast, do not
generalize to nonrhyming nonce words (Clahsen, 1997).
The Russian (perfective passive) participle endings are
-t, -n, and -on, and the selection of one of these suffixes
depends in part upon conjugation class and in part upon
the phonology of the stem. Russian verbs fall into two
conjugations. Verbs that fall into the first conjugation
regular class, c'est, those with stems in -a, always take -n
(par exemple., the participle form of ubra- ‘‘clear up’’ is ubra-n.
Verbs of the second conjugation always take -on (par exemple.,
122
Journal des neurosciences cognitives
Volume 18, Nombre 1
D
o
w
n
je
o
un
d
e
d
je
je
/
/
/
/
/
j
F
/
t
t
je
t
.
:
/
/
F
r
o
m
D
o
h
w
t
n
t
p
o
:
un
/
d
e
/
d
m
je
F
r
t
o
p
m
r
c
h
.
p
s
je
je
d
v
je
r
e
e
c
r
t
c
.
m
h
un
je
e
r
d
.
toi
c
o
o
c
m
n
/
j
un
o
r
t
c
je
c
n
e
/
–
un
p
r
d
t
1
je
8
c
je
1
e
1
–
2
p
1
d
1
F
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
oui
7
g
7
toi
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
F
p
e
b
m
oui
b
e
g
r
toi
2
e
0
s
2
t
3
t
/
.
F
o
n
1
8
M.
un
oui
2
0
2
1
poxvalit (cid:1) poxval-on ‘‘praise–praised’’ participle). Le
effect of the phonology of the stem can be seen within
first conjugation irregular verbs. Par exemple, first conju-
gation stems ending in -o or -u take the participle suffix
-t (pobryznu- (cid:1) pobryznu-t ‘‘wash (cid:1) washed’’). First con-
jugation irregular verbs with consonant-final stems take
-on in the participle (par exemple., unes- uneson ‘‘carry off’’
[stem] (cid:1) ‘‘carried off’’ [participle], and first conjugation
verbs whose present stem ends in /j/ or have a nasal infix
take -t in the participle. The system of participle forma-
tion in Russian is clearly different from the one in German
in that, in Russian, the choice between the three endings
is determined by conjugation class and by phonological
segments at the right edge of verb stems. In German, comment-
jamais, the -t suffix is an overall default, whereas -n participle
formation only applies to the subclass of strong verbs.
Experiment 1A: Elicited Production of Participles
This experiment was designed to determine the gener-
alization properties of the two participle endings -t and
-n in L2 German using a nonce-word production task
that was previously employed with native speakers of
German (Clahsen, 1997). Spécifiquement, we ask whether
the two endings show the same difference in generaliz-
ability in L2 learners as in native speakers of German
(see Methods section). Tableau 1 shows the mean rates
of expected participle suffixation given the simple past
forms the L2 learners supplied.
A repeated measures analysis of variance (ANOVA)
with suffix (-t vs. -n) as predicted by class (weak, fort,
mixed) as dependent variable revealed a significant main
effect [F(3,48) = 17.47, p < .001]. Pairwise comparisons
between the four conditions in Table 1 (with an adjusted
a level of .008) revealed significant differences for all
comparisons except for ‘‘weak’’ versus ‘‘mixed’’ verbs.
The L2 learners made use of -t participle suffixation, not
only for mixed and weak verbs, but also for nonce verbs,
for which they themselves produced strong simple past
forms. By contrast, the -n participle ending was hardly
ever applied to weak or mixed verbs. These results are
Table 1. Participles of Nonce Verbs
Form
Supplied
in Step 2
Expected
Participle
Ending
Strong past/
nonrhymes
Strong past/
rhymes
Weak past
Mixed past
-n
-n
-t
-t
Mean Rates
of Expected
Participle Endings
in Step 3 (Maximum
Score = 10)
5.3
6.8
9.4
9.1
SD Range
3.5
0–10
2.6
1–10
1.1
1.1
6–10
6–10
similar to those reported in Clahsen (1997) for native
speakers of German; -t was found to generalize widely to
all kinds of nonce verbs, whereas extensions of -n
participle formation were restricted to novel strong
verbs. The present findings suggest that this contrast
also holds for the group of L2 learners under study.
Experiment 1B: Event-related Brain Potentials
to Participles
On-line morphological processing of participles was
assessed by measuring ERPs in response to correctly or
incorrectly inflected regular or irregular participles em-
bedded in visually presented sentence contexts. Two
sets of behavioral data were gathered in this experiment
from the L2 learners: (a) elicited production of participle
forms for the critical experimental items after the elec-
troencephalography (EEG) session and (b) probe veri-
fication during the EEG sessions. Performance on both
tasks was very good (>95% correct responses in the
elicitation and >96% correct responses in the probe
verification task), indicating that the participants actually
read the stimuli to make the required distinction and
that they were familiar with the critical verbs and their
participle forms.
The ERP data of the L2 learners show a regular/
irregular distinction. Regularizations (c'est à dire., the regular -t
suffixed to a strong verb such as laufen ‘‘run’’) elicited
an anterior negativity between 250 et 600 msec as well
as a small parietal positivity between 600 et 1000 msec
compared with their correct counterparts. Par contre,
irregularizations (c'est à dire., -n suffixed to a weak verb such as
tanzen ‘‘dance’’) yielded a centrally distributed negativ-
ity between 450 et 600 msec relative to the correct
participle forms (voir la figure 1).
With respect to the regularization condition, statistical
analyses for the lateral electrodes in the early time
window (250–600 msec) revealed a significant interac-
tion of the factors Condition (cid:2) Site [F(2,34) = 8.75, p <
.01]. Subsequent analyses showed that there was a
reliable effect of condition for the anterior [F(1,17) =
5.71, p < .05], but not for the central and posterior
electrode positions. For the midline positions, there
were no significant effects involving the factor Condi-
tion. Analyses in the late time window (600–1000 msec)
for the lateral electrode positions revealed a reliable
main effect of condition [F(1,17) = 4.95, p < .05] as well
as a significant interaction of Condition (cid:2) Site [F(2,34) =
4.85, p < .05] reflecting the fact that there was a reli-
able positivity over central [F(1,17) = 4.45, p = .05] and
posterior [F(1,17) = 7.53, p < .05], but not over anterior
sites. Analyses for midline positions demonstrated a
main effect for condition [F(1,17) = 4.47, p < .05],
which interacted with electrode [F(2,32) = 3.74, p <
.06]. A significant positivity for regularizations was only
observed for Pz [F(1,17) = 7.11, p < .02] and not for the
other central electrode sites Cz and Fz. These statistical
Hahne, Mueller, and Clahsen
123
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
/
.
f
o
n
1
8
M
a
y
2
0
2
1
Figure 1. Grand average
ERPs (n = 18) for regular
participles, correct -t versus
incorrect -n forms (top),
and for irregular participles,
correct -n versus regularized
(incorrect) -t forms (bottom).
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
.
/
f
o
n
1
8
M
a
y
2
0
2
1
124
Journal of Cognitive Neuroscience
Volume 18, Number 1
analyses confirm the observed anterior negativity be-
tween 250 and 600 msec plus a later parietal positivity
for regularized participle forms.
With respect to the irregularization condition, statistical
analyses comprising the time window of 450–600 msec
revealed a marginally significant main effect of condition
for the midline electrodes [F(1,17) = 3.57, p < .08], which
indicates a strong tendency toward a centrally distributed
negativity for irregularizations.
These results show that the two types of morpholog-
ical violation elicited different brain responses in the L2
learners. Although the anterior negativity had its maxi-
mum at left frontal sites in the L1 speakers,
it was
observed bilaterally in our L2 group. Note, however,
that topographic variations of anterior negativities in this
time range have been observed in a number of ERP
studies with adult native speakers (see, e.g., Rodriguez-
Fornells et al., 2001, for Catalan; Gross et al., 1998, for
Italian; Weyerts, Penke, Mu¨nte, Heinze, & Clahsen, 2002,
for German, who found either bilateral negativities or
negativities restricted to right anterior sites). Both in
terms of its timing and its distribution, the anterior
negativity we found for the L2 learners falls within the
range of variation that has been observed in studies with
native speakers and can be taken to be an instance of a
‘‘morphosyntactic negativity,’’ an ERP waveform that is
clearly different from the centro-parietal N400, which
has been found to be associated with lexical–semantic
processing (see, e.g., Osterhout, 1997). For inflectional
morphology, anterior negativities for regularizations
have been interpreted as reflecting violations of rule-
based morphological processing (Rodriguez-Fornells
et al., 2001; Penke et al., 1997; Weyerts et al., 1997).
Given this interpretation, our results indicate that L2
learners are indeed employing regular rules of inflection
in on-line morphological processing.
Regularizations also elicited a small parietal positive-
going wave in the L2 learners with a latency of 600 to
1000 msec poststimulus, a waveform that can be identi-
fied as a P600. This effect may not have been seen in
Penke et al.’s (1997) study of native speakers of German,
which could be due a shorter analysis epoch of only
900 msec post stimulus onset (as opposed 1200 msec in
the present study) or insufficient statistical power. In any
case, a biphasic ERP pattern with an early anterior
negativity followed by a P600 has been reported in
a number of sentence processing studies (see, e.g.,
Friederici, 2002, for a review) as well as in studies exam-
ining morphological violations in Catalan (Rodriguez-
Fornells et al., 2001) and German (Lu¨ck et al., 2001).
The P600 has been interpreted to reflect controlled
rather than automatized processing (Friederici, 2002;
Hahne & Friederici, 1999), specifically reanalysis, repair,
or effortful syntactic integration processes at the sen-
tential
level (Friederici, Steinhauer, & Pfeifer, 2002;
Osterhout & Holcomb, 1992). Given that the P600
reflects sentence-level processes, the P600 we obtained
suggests that in L2 learners, regularization errors not
only affect early word-internal morphological processing
(as indicated by the anterior negativity discussed above)
but also later sentence-based processes. In other words,
with respect to the -t regularization, two distinct pro-
cesses could be identified, a relatively early process of
automatic morphological decomposition and a later
process that integrates the participle with the rest of
the sentence.
EXPERIMENT 2: NOUN PLURALS
First, we will give a brief description of plural formation
in German and Russian. German has a zero plural form
and four overt plural suffixes [-e, -er, -(e)n, and -s], some
of which can co-occur with an altered (umlauted) stem
vowel. The use of the different plural allomorphs with
specific nouns is arbitrary to varying degrees, and for
most of them, there are preferred tendencies of plural
formation interacting with the gender system and the
phonological form of the singular form (see Marcus et al.,
1995). In the main ERP experiment on noun plurals, we
compared -s and -n plurals; the following remarks,
therefore, focus on these two forms.
What is common to -s and -n plurals is that in contrast
to the other plural forms, they do not involve any stem
changes (‘‘umlaut’’). In other respects, however, they
are very different from each other. First, among the five
German plural allomorphs, -n is the most common with
a type frequency of 48% and a token frequency of 45% in
the CELEX lexical database (Sonnenstuhl & Huth, 2002),
whereas -s is the least common plural
form (type
frequency, 4%; token frequency, 1.8%). Second, the
form that acts most clearly as the regular default is the
plural -s, despite its low frequency (see the work of
Marcus et al., 1995). The -s plural is not restricted to a
specific phonological environment and serves as the
appropriate plural marking whenever a lexical entry is
not readily available, that is,
it generalizes to novel,
unusual-sounding words and to rootless and headless
nouns derived from other categories. None of these
properties holds for -n plurals. Instead, there are partic-
ular morphophonological properties of nominal stems
that favor -n plurals.
Russian makes use of a range of case–number mor-
phemes and three declensional patterns for nouns includ-
ing subdeclensions. Moreover, case–plural morphemes
are largely vocalic and often stressed. Timberlake (1993)
distinguishes among declension Ia, which is basically for
masculine nouns; declension Ib, which is almost exclu-
sively neuter; declension II, which is composed primarily
of feminines; and declension III, which is characterized
by the syncretic ending -i in genitive, dative, and loca-
tive, with further subdeclensions. Classes Ia, II, and
IIIa prefer the nominative plural -i, whereas classes Ib
and IIIb have a preference for the nominative plural
-a. The preferred genitive plural is -ov/-ev in class Ia, -ej
Hahne, Mueller, and Clahsen
125
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
/
.
f
o
n
1
8
M
a
y
2
0
2
1
in class IIIa, and -< in the remaining classes. In addi-
tion, there is stem allomorphy in the plural for recog-
nizable groups of nouns along with deviations from the
preferred plural endings. Nationality terms, for example,
use an otherwise unique nominative plural ending -e,
borrowings as well as a substantial number of nouns
have the nominative plural
-a´ implying end stress
throughout the plural (e.g., inspe´ktor–inspektora´).
Clearly, noun plural formation in Russian is very different
from German.
Experiment 2A: Acceptability Judgment of
Noun Plurals
This experiment examines the generalization properties
of the different German plural forms using a nonce-word
plural judgment task that was previously employed with
native speakers of German (Marcus et al., 1995). The L2
learners’ mean ratings for the plural forms of rhyming
and nonrhyming nonce nouns are shown in Table 2.
Table 2 shows that ratings for irregular plurals were
overall higher than for the -s plural and that -s plural
forms were judged as better for nonrhymes than for
rhymes, whereas the opposite trend was seen for irreg-
ular plural forms. ANOVAs revealed a significant main
effect of regularity [F(1,17) = 13,60, p < .05], no effect
of rhyme, but rather a significant interaction of Rhyme (cid:2)
Regularity [F(1,17) = 5.58, p < .05]. The L2 learners’
preference patterns for regulars and irregulars are par-
allel to those reported by Marcus et al. (1995) for native
speakers of German, although the differences between
conditions were smaller in the L2 learners. This might be
because of the fact that the L2 learners have not yet
completely mastered the German plural system, an issue
to which we come back below.
Experiment 2B: Event-related Brain Potentials to
Noun Plurals
To examine on-line morphological processing of noun
plurals, we performed an ERP experiment with correct
and incorrect plural forms embedded in auditorily pre-
sented sentences. Two sets of behavioral data were
gathered in this experiment from the L2 learners, (a)
forms for the critical
elicited productions of plural
Table 2. Ratings of Plural Forms of Nonce Nouns
Highest rated irregular
plural/rhymes
Highest rated irregular
plural/nonrhymes
-s plural/rhymes
-s plural/nonrhymes
Mean Ratings
SD
Range
4.5
4.4
3.5
3.6
0.35
3.8–5
0.37
3.8–5
0.91
0.88
1.3–4.8
1.9–5
experimental
items after the EEG session, (b) probe
verification during the EEG session. Performance on
both tasks was good (>86% correct responses in the
elicitation and >91% correct responses in the probe
verification task), indicating that the participants care-
fully listened to the stimuli and were familiar with the
critical nouns and their plural forms. It should be noted,
cependant, that performance on these two tasks was
worse than in the corresponding tasks for participles.
Visual
inspection of the ERP data reveals a clear
contrast between -s plural regularizations and irregulari-
zations. Recall that the nouns tested in the ‘‘masculine’’
and ‘‘feminine’’ conditions require -n plurals in German
and that in regularizations, the correct -n was replaced
by -s (par exemple., Tube-n vs. *Tube-s). On the other hand,
the nouns used in the ‘‘loan word’’ and the ‘‘name’’
conditions require -s in German, and in irregularizations,
the correct -s was replaced by -n. The ERP data show
que (-s plural) regularizations elicited a late positivity be-
tween approximately 850 et 1250 msec (voir la figure 2).
Par contre, misapplications of -n plurals to loan words
yielded a centrally distributed negativity between 300
et 800 msec relative to the correct plural forms (voir
Chiffre 3). Enfin, in the surname condition, correct and
incorrect versions are seen to only differ slightly at
parietal scalp sites, with incorrect items yielding a more
negative waveform than the correct ones (Chiffre 3).
With respect to the masculine condition (voir la figure 2),
statistical analyses for the 850- to 1250-msec time win-
dow revealed a marginally significant main effect of
condition [F(1,17) = 3.37, p < .10] and a reliable Con-
dition (cid:2) Hemisphere interaction [F(1,17) = 4.55, p <
.05]. Subsequent analyses showed that regularized plu-
rals of masculine nouns produced a significantly more
positive waveform over the right hemisphere than the
correct plural forms [left, F(1,17) = 2.13, p < .17; right,
F(1,17) = 4.37, p = .05]. Analyses on midline positions
revealed a reliable main effect of condition [F(1,17) =
4.76, p < .05]. To test for a possible anterior negativity,
which had been observed in native speakers, an addi-
tional analysis was conducted on the data in Figure 2
(upper panel) for the time interval of 300–550 msec,
which, however, did not reveal any significant effects for
the factor Condition.
Regarding the feminine condition (Figure 2,
lower
panel), analyses for the 850- to 1250-msec time window
revealed a significant Condition (cid:2) Site interaction
[F(2,34) = 4.07, p < .05] as well as a reliable three-way
interaction of Condition (cid:2) Hemisphere (cid:2) Site [F(2,34) =
5.22, p < .05]. Regularized feminine nouns elicited sig-
nificantly more positive waveforms than their correct
counterparts over posterior electrode sites [left central,
F(1,17) = 3.61, p < .10; left posterior, F(1,17) = 5.22, p <
.05; right posterior, F(1,17) = 5.03, p < .05]. For midline
positions, there was a significant Condition (cid:2) Electrode
interaction [F(2,34) = 4.11, p < .05]. Subsequent
analyses showed a significant effect of condition at Pz
126
Journal of Cognitive Neuroscience
Volume 18, Number 1
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
/
.
f
o
n
1
8
M
a
y
2
0
2
1
Figure 2. Grand average ERPs
(n = 18) for masculine nouns,
correct -n versus regularized
(incorrect) -s forms (top), and
for feminine nouns, correct
-n versus regularized
(incorrect) -s forms (bottom).
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
/
.
f
o
n
1
8
M
a
y
2
0
2
1
Hahne, Mueller, and Clahsen
127
Figure 3. Grand average ERPs
(n = 18) for loan words:
correct -s versus incorrect
-n forms (top), and for
surnames, correct -s versus
incorrect -n forms (bottom).
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
.
f
/
o
n
1
8
M
a
y
2
0
2
1
128
Journal of Cognitive Neuroscience
Volume 18, Number 1
[F(1,17) = 4.41, p = .05]. Again, as for the masculine
condition, an additional analysis for the time window of
300–550 msec did not reveal any reliable effects of condi-
tion; hence, there were no signs of an anterior negativity.
Concerning irregularizations (see Figure 3), analyses
in the 300- to 800-msec time window revealed a highly
significant main effect of condition for lateral [F(1,17) =
9.41, p < .01] and midline electrodes [F(1,17) = 8.29,
p < .01], which did not interact with any topographical
variables. This confirms the centrally distributed nega-
tivity for irregularizations seen in Figure 3 (upper panel).
With respect to the surname condition, visual inspec-
tion of Figure 3 suggested a more negative waveform
between 500 and 900 msec as well as between 900 and
1500 msec over posterior positions. ANOVAs for the 500-
to 900-msec time window revealed a marginally signifi-
cant interaction of Condition (cid:2) Site [F(2,34) = 3.03, p <
.10] for lateral electrodes and a significant interaction
of Condition (cid:2) Electrode [F(2,34) = 3.90, p < .05] for
midline electrodes. Subsequent analyses showed a
reliable effect of condition in the posterior region
[F(1,17) = 5.82, p < .05] and at Pz [F(1,17) = 5.88, p <
.05]. For the later time window (900–1500 msec), there
was a marginally significant interaction of Condition (cid:2)
Site [F(2,34) = 4.16, p < .06]. However, when analyses
were calculated for each site separately, only the poste-
rior region revealed a marginally significant effect of
condition [F(1,17) = 3.13, p < .10]. There was also a
significant interaction of Condition (cid:2) Electrode for
the midline positions [F(2,34) = 5.72, p < .05], which
gave rise to a reliable condition effect at Pz [F(1,17) =
4.49, p < .05].
These results show that the two kinds of morpholog-
ical violation elicited different brain responses in our L2
learners. Regularizations elicited a late positivity, an ERP
in terms of its timing and distribu-
component that,
tion, can be identified as a P600. Irregularizations, on the
other hand, yielded an N400-like waveform, that is, a cen-
trally distributed negativity between 300 and 800 msec.
This effect was stronger in the loan word condition than
for surnames, and in the surname condition, it occurred
relatively late (after 500 msec) and had a more parietal
distribution. Crucially, however, the negativity found for
irregularizations represents a clearly different ERP com-
ponent from the late positivity (P600) that was seen for
-s plural regularizations of both feminine and masculine
nouns.
The central negativity for irregularizations has also
been found for native speakers of German in both the
visual and the auditory ERP studies (Lu¨ck et al., 2001;
Weyerts et al., 1997) and has been argued to reflect a
pseudoword effect. Indeed, from a linguistic perspec-
tive, misapplications of -n plurals to nouns that take
-s plurals in German produce items that are similar to
pseudowords (e.g., *Waggonen ‘‘wagons’’), whereas
correct regulars (Waggon-s) are existing (morphologi-
cally complex) words. The central negativity we found
for irregularizations corresponds to this difference, and
it resembles the N400 effect seen in many previous
studies for pronounceable pseudowords (Rugg, 1987).
We interpret the late positivity (P600) as resulting
from the L2 learners’ recognition of a morphosyntactic
violation. Hence, it occurs with misapplications of the -s
pluralization rule, but not with misapplications of irreg-
ular patterns. This effect was also seen in Lu¨ck et al.’s
(2001) study of native speakers of German. Note, how-
ever, that an additional left anterior negativity was found
to precede the P600 in the L1 speakers. It is conceivable
that an -s plural regularization error is initially subject to
(possibly automatic) processes of morphological parsing
by which the inflected word is decomposed into its
constituent morphemes; the anterior negativity could
reflect the combinatory violation detected at this level.
At a later stage of processing, the regularized plural form
has to be integrated with the rest of the sentence, and
the P600 effect that Lu¨ck et al. found for L1 German
speakers suggests that regularization errors cause addi-
tional processing at this level, perhaps because partic-
ipants try to repair or reanalyze the regularized plural
form before integrating it with the rest of the sentence.
Given this account of the biphasic ERP pattern in native
speakers, our present findings indicate that the L2
learners do not employ early processes of word-internal
morphological decomposition to parse a plural regular-
ization error, hence, the lack of an anterior negativity,
but employ repair and integration processes at a later
stage of processing, as indicated by the P600.
GENERAL DISCUSSION
The most salient and consistent result of the present
study is that adult L2 learners respond differently to
violations of regular and irregular inflection during on-
line morphological processing. For misapplications of
regular rules of inflection, they showed ERP effects that
have independently been argued to tap morphosyntactic
processing, namely an anterior negativity and/or a P600,
whereas misapplications of irregular inflection revealed
an ERP effect (the N400) that has been claimed to be
characteristic of lexical processing and interpretation.
From the perspective of dual-mechanism models of
inflection, a -t participle affix appearing on an irregular
verb or an -s plural incorrectly affixed to an existing noun
are combinatorial violations that represent misapplica-
tions of a default rule to a verb or a noun that would
normally block the rule. By contrast, irregular inflection
is based on full-form storage, and misapplications of
irregular inflection produce pseudowords. The brain
responses seen in L2 learners for the two kinds of
morphological violation, that is,
‘‘morphosyntactic’’
ERP components for combinatory violations and a lexical
N400 for misapplications of irregular patterns, are com-
patible with this distinction and suggest that the two
Hahne, Mueller, and Clahsen
129
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
f
.
/
o
n
1
8
M
a
y
2
0
2
1
processing routes posited by dual-mechanism models of
inflection (lexical storage and morphological decompo-
sition) are also accessible and employed by L2 learners.
This finding does not support Ullman’s (2001) original
claim that L2 processing is ‘‘largely dependent upon
declarative/lexical memory’’ (p. 105) according to which
one would expect to find the same ERP responses in L2
learners for regular inflectional processes as for irregular
inflection. Instead, our ERP results indicate that different
processes are involved in regular and irregular inflec-
tion. More recently, however, Ullman (2004) suggested
that high proficiency levels in L2 learners can lead to
enhanced reliance on the procedural processing system.
The differences we found in the L2 learners’ ERP re-
sponses for participles and plurals are indeed compati-
ble with this suggestion as the degree of automaticity
appears to be lower for the system, which is more
difficult for the L2 learners. Recall that in native speakers
of German, a left anterior negativity was found in three
different experiments on participles (Penke et al., 1997)
and in both the visual and the auditory ERP studies on
noun plurals (Lu¨ck et al., 2001; Weyerts et al., 1997). In
the L2 learners, an anterior negativity was found for
participles, but not for noun plurals. One possibility
would be that the working-memory demands incurred
by the task assigned to the participants in the two ERP
experiments (to verify a probe sentence) were higher for
L2 learners than for native speakers, but this only
provides a partial account of the ERP results. The probe
verification task was given in both the participle and the
plural experiment yielding similar task demands, and yet
the results were different. The L2 learners performed
similarly to native speakers for participles, but not for
plurals. We therefore do not think that the absence of
an anterior negativity in the L2 learners’ ERPs to plural
violations can be explained in terms of working-memory
limitations.1
Instead, we suggest that the L2 learners employ early
automatic processes of word-internal morphological
decomposition for participles (as reflected by the ante-
rior negativity), but not for plurals, and this contrast in
the ERP results coincides with different levels of profi-
ciency for participles and plurals. The German noun
plural system is rather unusual in that it has a default
rule (the plural -s) with an extremely low frequency,
which means that more than 95% of the German nouns
form their plurals according to one of the various
irregular patterns. The participle formation system has
a more common frequency distribution (in that irregular
verbs do not outnumber regular ones) and a relatively
small number of irregular patterns. Moreover, the plural
system is linguistically more diverse in that it comprises
five different endings, whereas participle formation only
involves the choice between -t and -n. These factors
make it easier for L2 learners to acquire German parti-
ciple formation in nativelike ways than to learn the noun
plural system of German. There are indications for that
in the behavioral data, as the elicited productions on the
critical items used in the two ERP experiments yielded
considerably worse correctness scores for plurals than
for participles. Clearly, plural formation in our L2 par-
ticipants is less stable than participle formation. Corre-
spondingly, the ERP results suggest that the L2 learners’
processing of participles is more automatized and na-
tivelike than of noun plurals.
Our results are also compatible with previous ERP
studies on L2 processing. Studies investigating lexical–
semantic processing in L2 learners and bilinguals ob-
tained N400 effects for semantic anomalies (Hahne,
2001; Hahne & Friederici, 2001; Weber-Fox & Neville,
1996, Ardal, Donald, Meuter, Muldrew, & Luce, 1990)
and pronounceable nonwords (McLaughlin, 1999) in
nonnative speakers, which were similar to those found
in studies with L1 speakers. Previous ERP studies of
syntactic processing in nonnative speakers demonstrat-
ed P600 effects for phrase-structure violations (Hahne,
2001; Weber-Fox & Neville, 1996) and violations of
subject–verb agreement and gender concord (Sabourin,
2003). The learners’ proficiency in the L2 seems to affect
the ERP findings. Hahne (2001) found a P600 effect for a
group of highly proficient Russian learners of German,
but when Hahne and Friederici (2001) administered the
same experiment to a group of Japanese L2 learners with
a lower proficiency in German, there were no significant
differences between the correct sentences and the
phrase-structure violations in any relevant time window.
Likewise, Friederici et al. (2002) report findings from an
artificial grammar experiment in which adult subjects were
trained on an artificial language system (BROCANTO) to
a level at which they were highly proficient and pro-
duced hardly any errors. A subsequent ERP experiment
examining syntactic violations in BROCANTO revealed
the familiar biphasic ERP pattern known from com-
parable studies of natural languages in native speakers,
that is, an early negativity followed by a P600. Finally,
Sabourin (2003) found a P600 for gender concord viola-
tions, but only in the subgroup of German L2 learners of
Dutch, who Sabourin independently demonstrated to be
more proficient in L2 gender marking than the other
subgroups of L2 learners she tested. Taken together,
these results suggest that at least in domains in which
they are highly proficient, L2 learners can employ the
same processing mechanisms as L1 speakers.
Conclusion
The results of the present study show regular/irregular
contrasts in adult L2 learners’ processing of inflected
words. In processing combinatory violations, the L2
learners evidenced ERP components (an anterior nega-
tivity and/or a P600) that have been linked to morpho-
syntactic processing, and for misapplications of irregular
inflection, they showed an ERP effect (the N400) that
has been claimed to be characteristic of lexical process-
130
Journal of Cognitive Neuroscience
Volume 18, Number 1
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
.
/
f
o
n
1
8
M
a
y
2
0
2
1
ing. These results are consistent with the two processing
routes posited by dual-mechanism models of inflection
(lexical storage and morphological decomposition) and
suggest that they are also accessible in L2 processing of
inflected words. The similarities of the electrophysiolog-
ical responses of the L2 speakers to those previously
observed in native speakers provide evidence for largely
shared processing mechanisms in L1 and L2 inflectional
morphology.
METHODS
Eighteen learners of German with Russian as L1 were
tested (mean age 25 years, range 20–33; 5 men). They
had lived in Germany for an average of 4.5 years (range
0.5–12 years) and had their first exposure to German at
age 17 years (range 8–29). One of the participants
reported that he had first been exposed to German at
about the age of 8 years, but like the others, he did not
consider himself bilingual. All participants were right-
handed. They had no known hearing deficits and were
paid for their participation. When the experiments took
place, all learners were living and working in the Leipzig
area. All the participants reported using German on a
daily basis for interaction with native and nonnative
speakers, that is, all participants use German in their
work environment, and all of them have German friends
or partners, and they communicate with them in Ger-
man. To obtain a general measure of their proficiency,
they were asked to rate their German language profi-
ciency on a 6-point scale2 (6 = very good, 1 = hardly
any knowledge); this yielded a relatively high average
score of 5.0.
All the L2 learners participated in an elicited participle
production task (Experiment 1A), a plural acceptability
judgment task (Experiment 2A), and in the two ERP
experiments, one on participles (Experiment 1B) and
one on plurals (Experiment 2B). Experiment 2B was
conducted in the same session as Experiment 2A. The
two ERP experiments were separated by at least one
week, and the order in which the experiments were
administered was counterbalanced across participants.
Experiment 1A employs the elicitation task from
Clahsen (1997), in which participants were presented
with a nonce verb in the infinitive and first-person
simple past form (Step 1). They were then asked to
repeat the simple past form of the nonce verb (Step 2),
and finally they had to produce the participle form, that
is, an inflectional form of the nonce word that they had
not seen before (Step 3). All nonce verbs were pre-
sented as part of sentences in written form and in a
pseudorandomized order. Participants were asked to
carefully read the sentences and to fill
in blanks for
the various verb forms. There were 10 sentences each in
four experimental conditions: (a) regular (‘‘weak’’) nonce
verbs, (b) nonce verbs with a mixed inflectional pattern,
(c) irregular (‘‘strong’’) nonce verbs that rhymed with
existing ones, (d) nonrhyming irregular nonce verbs.
The simple past forms provided an unambiguous cue as
to whether a nonce verb was regular (in which case, the
simple past form had a -te suffix and no stem changes),
‘‘mixed’’ (simple past forms with -te and vowel changes)
or irregular (simple past forms with vowel changes and
no -te). One participant had to be excluded from the
analysis because of a high error rate of more than 90% in
Step 2. All other participants carried out Step 2 virtually
perfectly. There were three trials (0.4%) in which an
error occurred in Step 2, and these were excluded from
any further analysis.
Experiment 1B adopts the design and materials of
the ‘‘sentence’’ experiment from Penke et al. (1997) ex-
cluding, however, the additional nonce verb condition.
Our version has a 2 (cid:2) 2 design: participle forms were
suffixed with -t (= regular) or with -n (= irregular) and
were correct or incorrect. Two types of violations were
tested: regularizations, a regular -t suffixed to a verb
that requires an irregular form (*gelauft vs. gelaufen
‘‘run’’ past participle), and irregularizations, a regular
(‘‘weak’’) verb incorrectly suffixed with -n (*getanzen
vs. getanzt ‘‘danced’’ participle). All the strong verbs
tested were of the so-called A–B–A class, which have
stem changes in the simple past, but not in their
participle forms. Hence, the participle forms we tested
only differed in terms of their endings, -t versus -n.
There were 50 regular and 50 irregular participles, which
were matched for participle and lemma frequencies.
These critical participles were embedded in simple
declarative sentences, which all consisted of six words.
The participle was the final word of the sentence (see
Penke et al., 1997, for further details). Sentences were
distributed in two blocks such that a correct and an
incorrect version of a given participle occurred in differ-
ent lists. The two blocks were pseudorandomized with
the general constraints that there were less than four
consecutive trials containing either regular or irregular
participles and less than four consecutive trials contain-
ing either correct or incorrect participle forms. Three
different combinations of presentation order of the two
blocks were created. Three more presentation lists were
generated by reversing the order of the original lists.
Each of these six lists was presented with a probe
verification task (see below), which required a button
press either with the left or with the right hand, thus
yielding a total number of 12 different lists. None of the
lists was presented more than twice.
Participants were tested in a soundproof booth. Sen-
tences were presented on a computer screen word by
word in yellow letters on a blue background. The words
were presented for 700 msec with no blank interval in
between.3 The stimulus onset asynchrony was as in
Penke et al. (1997) with a 2300-msec interval between
two stimulus sentences. The words subtended 0.758 of
visual angle in height and a maximum of 4.858 in width.
Hahne, Mueller, and Clahsen
131
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
f
.
/
o
n
1
8
M
a
y
2
0
2
1
After 10 sentences, a ‘‘probe sentence’’ was presented in
red letters. Half of these probe sentences were exact
repetitions of one of the 10 sentences shown before and
the other half differed with regard to the content words
of any of the previous sentences.4 True and incorrect
probes were equally distributed across the four experi-
mental conditions. Participants received a written in-
struction in which they were asked to carefully read
each sentence and to indicate via a push-button response
whether a probe sentence was a repetition of one of the
previous 10 sentences. Furthermore, they were asked to
minimize eye and body movements during the presen-
tation of the sentences. The experiment was subdivided
in four blocks of 55 sentences each. Subsequent to the
EEG session, a list of all critical verbs was read out to the
participants in their infinitive form and they were asked
to generate the participle forms.
The EEG was recorded from 25 scalp sites (Sharbrough,
1991) by means of Ag/AgCl electrodes attached to an
elastic cap: FP1/2, F7/8, F3/4, FT7/8, FC3/4, T7/8, C3/4,
CP5/6, P7/8, P3/4, O1/2, Fz, Cz, and Pz. Recordings
were referenced to the left mastoid. To control for eye
movement artifacts, the horizontal electrooculogram
(EOG) was monitored from electrodes at the outer
canthus of each eye and the vertical EOG from two
electrodes located above and below the participant’s
right eye. Electrode impedances were kept below 5 k(cid:1).
EEG and EOG channels were recorded continuously
with a band pass from DC to 40 Hz with a digitization
rate of 250 Hz. ERPs were filtered off-line with 10 Hz low
pass for the plots, but all statistical analyses were com-
puted on nonfiltered data. ERPs on the critical participle
forms were recorded for 1200 msec separately for each
participant, electrode, and condition. All ERP averages
were aligned to a 200-msec baseline before the onset of
the critical word. Trials with ocular or amplifier satura-
tion artifacts were excluded from the averages. The
average percentage of rejected trials was 16%. All statis-
tical analyses were performed on the mean ERP ampli-
tudes. Based on visual inspection and previous studies,
we chose three different time windows: 250–600 msec
for the anterior negativity, 600–1000 msec for the late
positivity, and 450–600 msec for possible N400 effects.
Repeated measure ANOVAs were performed on all lat-
eral electrodes with three within-subject variables: con-
dition (correct vs. incorrect), hemisphere (left vs. right),
and site (anterior vs. central vs. posterior). The variables
hemisphere and site were completely crossed, yielding
six regions-of-interest with three electrodes each: left
anterior (F7, FT7, and FC3), right anterior (F8, FT8, and
FC4), left central (T7, C3, and CP5), right central (T8,
C4, and CP6), left posterior (P7, P3, and O1), and right
posterior (P8, P4, and O2). The Greenhouse–Geisser
correction was applied whenever effects with more than
one degree of freedom in the numerator were evaluated.
Below, we report uncorrected degrees of freedom and
corrected probabilities.
Experiment 2A employs the paper-and-pencil
judg-
ment task from Marcus et al. (1995), in which partic-
ipants are asked to judge the acceptability of plural
forms of nonce nouns, 12 monosyllabic items that
rhymed with existing German nouns that take irregular
plural forms (e.g., Pund on analogy with Hund–Hunde
‘‘dog–dogs’’) and 12 nonrhymes. Each item was first
presented in a context sentence in its singular form in
masculine or feminine gender, followed by 8 test sen-
tences each containing one of the plural
forms of
German. Participants were asked to rate each test sen-
tence on a 5-point scale (1 = perfectly natural, 5 = very
unnatural) for acceptability. The gender of the items
was counterbalanced across participants as was the
sequential order in which the plural forms were pre-
sented. Thus, four different experimental versions were
presented. The experimental items were presented in a
pseudorandomized order such that there were no more
than three rhyme or nonrhyme items in direct succes-
sion. In the analysis, each rating was subtracted from 6,
so that the higher the score, the more natural a given
plural form. Following Marcus et al., we took the highest
rated irregular for comparison with the ratings of the
regular -s plural. A two-way ANOVA with Regularity
(-s plurals vs. irregular plurals) and Rhyme (rhyming vs.
nonrhyming nouns) as within-subjects factors was con-
ducted on the mean rating scores.
in which the critical
Experiment 2B adopts the design and materials of the
auditory ERP violation experiment from Lu¨ck et al.
(2001),
items were matched for
the same frequency and morphophonological criteria as
detailed by Weyerts et al. (1997). There were 48 critical
nouns forming their plural with -(e)n, 24 masculine/
neuter and 24 feminine nouns; the latter were disyllabic
and ended in a schwa. Another 48 critical nouns re-
quired -s plurals, none of which had a stem-final vowel,
most of which were loan words. In addition, 24 surnames
served as critical items in the ‘‘surname condition,’’ all of
which require -s plurals in German. Each noun was
presented twice, in its correct plural form and with an
incorrect plural form. Two types of violation were tested:
regularizations, a regular -s suffixed to a noun that re-
quires an -n plural form (*Vases vs. Vasen ‘‘vases’’), and
irregularizations, nouns that require -s plurals incorrectly
suffixed with -n (*Waggonen vs. Waggons ‘‘wagons’’).
The critical words were embedded in sentences and
always appeared in direct object position followed by an
adverbial or a prepositional phrase. Four different ex-
perimental sentences were constructed for each noun,
and each of these 480 sentences was recorded in two
versions, once with the correct and once with the
incorrect plural form of the critical noun. The different
experimental sentences were counterbalanced across
subgroups of participants, and each participant received
480 sentences. The order of presentation was pseudo-
randomized such that no more than three consecutive
items belonged to the same condition or contained
132
Journal of Cognitive Neuroscience
Volume 18, Number 1
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
/
.
f
o
n
1
8
M
a
y
2
0
2
1
either only correct or incorrect plural forms, and repe-
titions of the same noun were separated by at least 27
intervening sentences. Sentences were spoken by a
trained female speaker of German and digitized at a sam-
pling rate of 44 kHz. Participants were seated in a
comfortable chair in a soundproof booth 130 cm in
front of a computer monitor. Sentences were presented
via loudspeaker while a fixation signal appeared on the
screen. Trials were separated by an intertrial interval of
2500 msec. After eight sentences, participants listened to
a warning tone followed by a ‘‘probe sentence’’ while a
question mark appeared on the screen, and participants
were instructed to give a push-button response indicat-
ing whether this sentence had been presented as one of
the previous eight trials. Half of these probe sentences
were exact repetitions of one of the eight sentences
shown before, and the other half differed with regard to
the content words of any of the previous sentences.
Participants were given a visual feedback on their re-
sponses. After every 48 sentences, participants were
given a short break.
The procedures for ERP recording and data analysis
were as described for Experiment 1B. In the present
experiment, ERPs on the critical nouns were computed
for 1500 msec relative to word onset separately for
each participant, electrode, and condition. Trials with
ocular or amplifier saturation artifacts were excluded
from the averages, with an average of 18% rejected trials.
Subsequent to the EEG session, all critical nouns were
read out to the participants in their singular form, and
they were asked to generate the corresponding plural
forms.
Acknowledgments
Supported by the Leibniz Prize awarded to Angela D. Friederici
by the German Science Foundation and two Visiting Research
Scholarships awarded by the Max-Planck-Institute for Human
Cognitive and Brain Sciences to H. C. We are grateful to Angela
Friederici for her support of this research, to Peter Indefrey,
Edith Kaan, Laura Sabourin and the anonymous reviewers for
helpful comments on previous versions of this paper, and to
Andrew Spencer for providing information on Russian mor-
phology. We also thank Sonja Rossi for assistance in design-
ing Experiment 1A and Monika Lu¨ck for helping us prepare
Experiment 2B.
Reprint requests should be sent to Anja Hahne, Max Planck In-
stitute for Human Cognitive and Brain Sciences, Stephanstrasse
1A, D-04103 Leipzig, Germany, or via e-mail: hahne@cbs.mpg.de.
Notes
1. Moreover, anterior negativities have been argued to
increase with working-memory demands (Felser, Mu¨nte, &
Clahsen, 2003). Thus, if the task we assigned to participants
affected the L2 learners’ working memory more than the native
speakers, we should have seen more pronounced anterior
negativities in the L2 learners than in the native speakers, the
opposite of what we found. For these reasons, the pattern of
results is unlikely to be due to the task assigned to participants.
2.
Self-ratings have been used in many previous L2 process-
ing studies as an index of general language proficiency (see
e.g., Dussias, 2003; Sabourin, 2003; Frenck-Mestre, 2002), and
self-rating scores have been shown to correlate with other
proficiency measures (Ferna´ndez, 2002).
3. This procedure deviates slightly from Penke et al. (1997)
where words were shown for 300 msec, followed by a blank
screen for 400 msec. For the present study, we chose longer
presentation times, as word recognition processes are likely to
be slower in L2 learners than in native speakers.
4. Note that we created new incorrect probe sentences, as
the ones used by Penke et al. (1997) included only very slight
modifications of the correct sentences, which we thought were
too hard for the L2 learners to detect.
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
.
f
/
o
n
1
8
M
a
y
2
0
2
1
REFERENCES
Ardal, S., Donald, M. W., Meuter, R., Muldrew, S., & Luce, M.
(1990). Brain responses to semantic incongruity in
bilinguals. Brain and Language, 39, 187–205.
Beck, M.-L. (1997). Regular verbs, past tense and frequency:
Tracking down a potential source of NS/NNS competence
differences. Second Language Research, 13, 93–115.
Brovetto, C., & Ullman, M. (2001). First vs. second language:
A differential reliance on grammatical computations
and lexical memory. Poster presented at the 14th Annual
CUNY Conference on Human Sentence Processing,
University of Pennsylvania.
Clahsen, H. (1997). The representations of participles in the
German mental lexicon: Evidence for the dual-mechanism
model. Yearbook of Morphology, 1996, 73–95.
Clahsen, H. (1999). Lexical entries and rules of language: A
multidisciplinary study of German inflection. Behavioral
and Brain Sciences, 22, 991–1060.
Dussias, P. E. (2003). Syntactic ambiguity resolution in L2
learners. Studies in Second Language Acquisition, 25,
529–557.
Felser, C., Mu¨nte, T., & Clahsen, H. (2003). Storage and
integration in the processing of filler-gap dependencies: An
ERP study of topicalisation and wh-movement in German.
Brain & Language, 87, 345–354.
Ferna´ndez, E. M. (2002). Bilingual sentence processing.
Amsterdam: Benjamins.
Frenck-Mestre, C. (2002). An on-line look at sentence
processing in a second language. In R. Heredia &
J. Altarriba (Eds.), Bilingual sentence processing
(pp. 217–236). Amsterdam: North-Holland.
Friederici, A. D. (2002). Towards a neural basis of auditory
sentence processing. Trends in Cognitive Sciences, 6,
78–84.
Friederici, A. D., Steinhauer, K., & Pfeifer, E. (2002). Brain
signatures of artificial language processing: Evidence
challenging the critical period hypothesis. Proceedings of the
National Academy of Sciences, U.S.A., 99, 529–534.
Gross, M., Say, T., Kleingers, M., Mu¨nte, T. F., & Clahsen, H.
(1998). Human brain potentials to violations in
morphologically complex Italian words. Neuroscience
Letters, 241, 83–86.
Hahne, A. (2001). What’s different in second language
processing? Evidence from event-related brain potentials.
Journal of Psycholinguistic Research, 30, 251–266.
Hahne, A., & Friederici, A. D. (1999). Electrophysiological
evidence for two steps in syntactic analysis: Early automatic
and late controlled processes. Journal of Cognitive
Neuroscience, 11, 194–205.
Hahne, A., & Friederici, A. D. (2001). Processing a second
language: Late learners’ comprehension mechanisms as
Hahne, Mueller, and Clahsen
133
revealed by event-related brain potentials. Bilingualism:
Language and Cognition, 4, 123–141.
Haznedar, B., & Schwartz, B. (1997). Are there optional
infinitives in child L2 acquisition? In E. Hughes, M. Hughes,
& A. Greenhill (Eds.), Proceedings of the 21st Annual
Boston University Conference on Language Development
(pp. 257–268). Somerville, MA: Cascadilla Press.
Klein, W. (1986). Second language acquisition. Cambridge:
Cambridge University Press.
Lalleman, J. A., van Santen, A. J., & van Heuven, V. J. (1997). L2
processing of Dutch regular and irregular verbs. Review of
Applied Linguistics, 115/116, 1–26.
Lu¨ck, M., Hahne, A., Friederici, A., & Clahsen, H. (2001,
November). Developing brain potentials in children:
An ERP study of German noun plurals. Paper presented
at 26th Boston University Conference on Language
Development.
Marcus, G. F., Brinkmann, U., Clahsen, H., Wiese, R., & Pinker,
S. (1995). German inflection: The exception that proves
the rule. Cognitive Psychology, 29, 189–256.
McLaughlin, J. (1999). Event related potentials reflect the early
stages of second language lexical acquisition. Unpublished
PhD thesis, University of Washington.
Meisel, J. M. (1991). Principles of Universal Grammar and
strategies of language use. In L. Eubank (Ed.), Point–
counterpoint: Universal grammar in the second language
(pp. 231–276). Amsterdam: Benjamins.
Mu¨nte, T. F., Say, T., Schiltz, K., Clahsen, H., & Kutas, M.
(1999). Decomposition of morphologically complex words
in English: Evidence from event-related brain potentials.
Cognitive Brain Research, 7, 241–253.
Newman, A., Izvorski, R., Davis, L., Neville, H., & Ullman, M. T.
(1999). Distinct electrophysiological patterns in the
processing of regular and irregular verbs. Journal of
Cognitive Neuroscience, Supplement, 47.
Osterhout, L. (1997). On the brain response to syntactic
anomalies: Manipulations of word position and word
class reveal individual differences. Brain and Language,
59, 494–522.
Osterhout, L., & Holcomb, P. (1992). Event-related brain
potentials elicited by syntactic anomaly. Journal of Memory
and Language, 31, 785–804.
Penke, M., Weyerts, H., Gross, M., Zander, E., Mu¨nte, T.,
& Clahsen, H. (1997). How the brain processes
complex words: An event-related potential study of
German verb inflections. Cognitive Brain Research, 6,
37–52.
Pinker, S. (1999). Words and rules. The ingredients of
language. New York: Basic Books.
Pre´vost, P., & White, L. (2000). Missing surface inflection or
impairment in second language acquisition? Evidence
from tense and agreement. Second Language Research, 16,
103–133.
Rodriguez-Fornells, A., Clahsen, H., Lleo, C., Zaake, W., &
Mu¨nte, T. F. (2001). Event-related brain responses to
morphological violations in Catalan. Cognitive Brain
Research, 11, 47–58.
Rodriguez-Fornells, A., Mu¨nte, T., & Clahsen, H. (2002).
Morphological priming in Spanish verb forms: An ERP
repetition priming study. Journal of Cognitive
Neuroscience, 14, 443–454.
Rugg, M. (1987). Dissociation of semantic priming, word and
non-word repetition by event-related potentials. Quarterly
Journal of Experimental Psychology, 39A, 123–148.
Sabourin, L. (2003). Grammatical gender and second
language processing: An ERP study. Doctoral dissertation,
University of Groningen.
Sharbrough, F. (1991). American Electroencephalographic
Society Guidelines for standard electrode position
nomenclature. Journal of Clinical Neurophysiology, 8,
200–202.
Sonnenstuhl, I., & Huth, A. (2002). Processing and
representation of German -n plurals: A dual mechanism
approach. Brain and Language, 81, 276–290.
Timberlake, A. (1993). Russian. In B. Comrie & G. Corbett
(Eds.), The Slavonic languages. London: Routledge,
pp. 827–886.
Ullman, M. T. (2001). The neural basis of lexicon and grammar
in first and second language: The declarative/procedural
model. Bilingualism: Language and Cognition, 4, 105–122.
Ullman, M. T. (2004). Contributions of memory circuits to
language: The declarative/procedural model. Cognition, 92,
231–270.
Weber-Fox, C. M., & Neville, H. J. (1996). Maturational
constraints on functional specializations for language
processing: ERP and behavioral evidence in bilingual
speakers. Journal of Cognitive Neuroscience, 8, 231–256.
Weyerts, H., Penke, M., Dohrn, U., Clahsen, H., & Mu¨nte, T.
(1997). Brain potentials indicate differences between regular
and irregular German plurals. NeuroReport, 8, 957–962.
Weyerts, H., Penke, M., Mu¨nte, T. F., Heinze, H. J., & Clahsen,
H. (2002). Word order in sentence processing: An
experimental study on verb placement in German. Journal
of Psycholinguistic Research, 31, 211–268.
D
o
w
n
l
o
a
d
e
d
l
l
/
/
/
/
/
j
f
/
t
t
i
t
.
:
/
/
f
r
o
m
D
o
h
w
t
n
t
p
o
:
a
/
d
e
/
d
m
i
f
r
t
o
p
m
r
c
h
.
p
s
i
l
d
v
i
r
e
e
c
r
t
c
.
m
h
a
i
e
r
d
.
u
c
o
o
c
m
n
/
j
a
o
r
t
c
i
c
n
e
/
-
a
p
r
d
t
1
i
8
c
l
1
e
1
-
2
p
1
d
1
f
9
/
3
1
5
8
4
/
4
1
4
/
0
1
8
2
9
1
8
/
9
1
2
7
9
5
0
5
6
8
7
9
7
7
5
2
/
5
0
0
8
0
9
6
8
7
9
p
2
d
9
0
b
6
y
7
g
7
u
5
e
2
s
5
t
0
o
0
n
6
0
7
7
.
p
S
d
e
f
p
e
b
m
y
b
e
g
r
u
2
e
0
s
2
t
3
t
.
f
/
o
n
1
8
M
a
y
2
0
2
1
134
Journal of Cognitive Neuroscience
Volume 18, Number 1
