Content uploaded by João Veríssimo
Author content
All content in this area was uploaded by João Veríssimo on Nov 21, 2018
Content may be subject to copyright.
https://doi.org/10.1177/1367006916688333
International Journal of Bilingualism
2018, Vol. 22(6) 619 –637
© The Author(s) 2017
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/1367006916688333
journals.sagepub.com/home/ijb
Aiming at the same target: A
masked priming study directly
comparing derivation and
inflection in the second language
Gunnar Jacob
University of Potsdam, Potsdam, Germany
Vera Heyer
University of Potsdam, Potsdam, Germany
João Veríssimo
University of Potsdam, Potsdam, Germany
Abstract
Aims and objectives/purpose/research questions: We compared the processing of
morphologically complex derived vs. inflected forms in native speakers of German and highly
proficient native Russian second language (L2) learners of German.
Design/methodology/approach: We measured morphological priming effects for derived and
inflected German words. To ensure that priming effects were genuinely morphological, the design
also contained semantic and orthographic control conditions.
Data and analysis: 40 native speakers of German and 36 native Russian learners of L2 German
participated in a masked-priming lexical-decision experiment. For both participant groups, priming
effects for derived vs. inflected words were compared using linear mixed effects models.
Findings/conclusions: While first language (L1) speakers showed similar facilitation effects for
both derived and inflected primes, L2 speakers showed a difference between the two prime types,
with robust priming effects only for derived, but not for inflected forms.
Originality: Unlike in previous studies investigating derivation and inflection in L2 processing,
priming effects for derived and inflected prime–target pairs were determined on the basis of the
same target word, allowing for a direct comparison between the two morphological phenomena.
In this respect, this is the first study to directly compare the processing of derived vs. inflected
forms in L2 speakers.
Significance/implications: The results are inconsistent with accounts predicting general L1/L2
differences for all types of morphologically complex forms as well as accounts assuming that L1
and L2 processing are based on the same mechanisms. We discuss theoretical implications for
L2 processing mechanisms, and propose an explanation which can account for the data pattern.
Corresponding author:
Gunnar Jacob, University of Potsdam, Haus 2, Campus Golm, Karl-Liebknecht-Str. 24-25, 14476 Potsdam, Germany.
Email: gujacob@uni-potsdam.de
688333IJB0010.1177/1367006916688333International Journal of BilingualismJacob et al.
research-article2017
Article
620 International Journal of Bilingualism 22(6)
Keywords
Morphological processing, derivation, inflection, second language (L2) learners, masked priming
Introduction
Morphological processing in second language (L2) learners1 has been subject to considerable
debate in the past decade. At the heart of the controversy is the question to what extent early pro-
cessing of morphologically complex forms in L2 speakers relies on the same fundamental mecha-
nisms as first language (L1) processing. To investigate this issue, a number of morphological
processing studies have made use of a methodological approach referred to as masked priming (see
Marslen-Wilson, 2007, for a comprehensive review of L1 processing studies using this method). In
a typical masked priming study, participants are confronted with a morphologically complex form
such as walked, and have to perform a lexical decision on a morphologically related target word
such as walk immediately afterwards. If the parser decomposes a prime word such as walked into
the stem walk and the affix -ed at an early stage of visual word recognition, this should facilitate
processing of the target word walk relative to a matched unrelated control trial in which the target
walk is preceded by an unrelated word because the stem walk has already been activated during the
processing of the prime. Crucially, prime words are presented for only a very brief period (typically
between 30 and 70 milliseconds (ms)) and embedded within a visual mask, preventing conscious
perception of the word. As a result, masked priming is considered to tap into a specific, particularly
early stage of automatic processing. This makes masked priming studies difficult to compare with
otherwise methodologically similar morphological priming studies in which the prime is con-
sciously perceived, such as overt visual priming or cross-modal priming. In the following, we thus
concentrate on masked priming studies.
With regard to morphological processing in L2 speakers, masked priming studies comparing
L1 and L2 processing have come to radically different conclusions. A number of studies have
found substantial differences between L1 and L2 speakers in morphological priming effects.
For example, Neubauer and Clahsen (2009) found significant morphological priming effects
for regular and irregular German participles in L1 speakers, but priming only for irregular
forms in their group of highly proficient Polish L2 learners of German. The authors concluded
that L2 processing might focus less on the morphological structure of a complex word, and
might instead rely more on storage and retrieval of whole-word-form representations in the
mental lexicon. In Clahsen and Neubauer (2010), a similar difference emerged for derived
forms (namely -ung nominalizations), again with significant facilitation effects in L1 speakers
of German but no priming in L2 speakers. In a set of masked priming experiments on English,
Silva and Clahsen (2008) found differences between L1 and L2 speakers for both inflected and
derived forms: For past tense inflection, L1 speakers showed significant morphological prim-
ing effects, while L2 speakers of English showed no priming; for deadjectival nominalizations,
facilitation effects, while reliable in both participant groups, were significantly stronger in the
L1 group than in the L2 groups.
A number of other studies, however, have questioned the claim that L2 processing is fundamen-
tally different from L1 processing. For example, Diependaele, Duñabeitia, Morris, and Keuleers
(2011) found parallel masked priming effects for derivationally related prime–target pairs such as
viewer – view in L1 speakers of English and in two L2 speaker groups (L1s: Dutch and Spanish).
In Dal Maso and Giraudo’s (2014) study on L2 morphological processing in Italian, L1/L2 differ-
ences emerged only for forms with infrequent derivational affixes; for derived forms with frequent
Jacob et al. 621
and productive affixes, L2 speakers of Italian showed similar priming effects as L1 speakers. The
authors thus argued that L2 speakers are in principle sensitive to morphological information, with
L1/L2 differences being restricted to phenomena which have not been fully acquired yet. In
Feldman, Kostić, Basnight-Brown, Filipović Đurđević, and Pastizzo’s (2010) masked priming
study on the processing of English inflected verbs, Serbian L2 speakers of English also showed
similar priming effects as L1 speakers of English, with significant priming effects for regular as
well as irregular past tense forms in both participant groups. In a subsequent post-hoc analysis
comparing high-proficient with low-proficient L2 speakers, significant priming effects emerged
only for the high-proficient group. Thus, the authors argued that differences between L1 and L2
processing might be restricted to lower proficiency levels, with L2 speakers developing native-like
processing mechanisms as they become more proficient in the L2. In a masked priming study
investigating the L2 processing of inflected French -er verbs, Coughlin and Tremblay (2015) found
a full priming effect (i.e. a similar amount of priming for morphologically related prime–target
pairs as for trials in which prime and target were identical) for both L1 speakers of French and
English learners of French.
Until recently, the discussion about why these studies have come to different conclusions has
focused mainly on either methodological differences between the studies, or on particular proper-
ties of the L2 groups, such as L2 proficiency, amount of exposure to the L2, or L1 background.
However, the studies described above do not only differ with regard to the particular L2 groups, but
also with regard to the investigated morphological phenomena, with some studies focusing on
derived and others on inflected stimuli.
Derived and inflected forms share a number of linguistic properties. In languages such as
English or German, both derivation and inflection typically involve affixation (walk–er and
walk–ed respectively), and in some theories of morphology, for example Distributed
Morphology (Harley & Noyer, 1999), the affixation of walk with the derivational suffix –er is
not considered to be different from the affixation with the inflectional suffix –ed. Nevertheless,
the two morphological processes add different types of information to the base or stem.
Derivation adds semantic information (walker = ‘person who walks’) and/or changes the word
category (walkV → walkerN; walkV → (the) walkN), while inflection adds abstract grammatical
information (e.g. tense or agreement; walkpresent → walkedpast). As a result, the products of deri-
vation and inflection differ in that derivation yields new lexemes (i.e. words with new seman-
tic and/or grammatical characteristics) whereas inflection yields new word forms of the same
lexeme. Realization-based accounts of morphology (e.g. Anderson, 1992) capture this differ-
ence in the nature and location of rules ‘realizing’ the two morphological processes: while
derivational rules map two derivationally related stem entries onto one another in the lexicon,
inflectional rules “fall ‘outside the lexicon’” (Anderson, 1992, p. 184) simply ‘spelling out’
the set of features required by the syntactic structure and not adding any additional semantic
information.
This difference between the outputs from inflectional and derivational processes is also reflected
in Aronoff’s (1994) realization pair formats, as adapted for derivation by Clahsen, Sonnenstuhl, and
Blevins (2003) as well as Clahsen and Felser (2006); see (1) and (2) below. The left-most part (in
square brackets) contains the relevant morphosyntactic properties or features and the adjacent part
specifies the realization rule that ‘spells out’ the features. For instance, if the verb walk is inserted for
the variable X in (1a), the spell-out is the inflected form walked; however, for the derivational rule in
(2a), the variable X is, as specified in the input, a lexeme (lex) which is then turned into a noun and
realized as walker, for example. Note that there are also rules that have a constant output rather than
containing a variable (see formats in (1b) and (2b)), reflecting unproductive forms.
622 International Journal of Bilingualism 22(6)
(1) Inflection
a. <[V, past], X + ed>
b. <[V, past, go], went>
(2) Derivation
a. <[V, lex], X> → <[N, lex], X + er>
b. <[V, refuse]> → <[N, refuse], refusal>
A number of overt priming studies have examined the processing of derived versus inflected
forms in native speakers. Some of these studies (e.g. Feldman, 1994; Stanners, Neiser, Hernon, &
Hall, 1979) have found stronger priming effects for inflected than for derived forms. Other studies,
in contrast (e.g. Clahsen et al., 2003; see Marslen-Wilson, 2007, for a comprehensive review),
report priming effects of a similar magnitude for derivation and inflection. Note, however, that
most of these studies are based on overt priming, which is considered to tap into a later, central-
lexical stage of processing (Marslen-Wilson, 2007). To our knowledge, the only masked priming
study that compared priming from inflected and derived words in native speakers was conducted
by Raveh (2002) and reported equivalent priming effects. Feldman, Barac-Cikoja, and Kostic
(2002), in a study on Serbian in which primes occurred without a mask, but were nevertheless
presented for a very short stimulus-onset asynchrony, instead found stronger priming effects for
inflection than for derivation, and explained this on the basis of stronger semantic relatedness
between primes and targets for inflected forms.
With regard to L2 processing, three masked-priming studies have also investigated both deriva-
tion and inflection. However, these studies have come to radically different conclusions. On the
one hand, the above-mentioned study by Silva and Clahsen (2008) revealed a difference between
derivation and inflection, with L2 speakers showing (in contrast to L1 speakers) no priming for
inflected forms, but significant priming for derived forms. Similarly, Kırkıcı and Clahsen (2013),
in a masked-priming study investigating the L2 processing of derived and inflected Turkish words,
found differences between L1 and L2 speakers only for regular Aorist inflection (with significant
morphological priming effects in native speakers of Turkish, but no priming for L2 speakers), but
not for derived nominalizations (with significant priming effects in both participant groups). These
findings are consistent with the linguistic distinction between inflection and derivation illustrated
in (1) versus (2). From a psycholinguistic perspective, this would mean that derivational and inflec-
tional priming are based on distinct sources, similarly to what Crepaldi, Rastle, Coltheart, and
Nickels (2010) proposed for regular versus. irregular inflectional priming. Masked priming effects
for regular inflection are indicative of decomposition, that is, walk is recognized faster following
walked because walked is decomposed into its component parts (walk + -ed) and, consequently, the
stem is accessed and preactivated during the processing of the prime word. Masked priming effects
for irregular inflection (e.g. fell priming fall) are due to lexical connections, the irregular form
being a subentry of the main lexeme fall. Following this reasoning, Kırkıcı and Clahsen (2013)
argued that L2 speakers do not decompose morphologically complex words. Instead, priming
effects for derived (but not inflected) forms are due to the lexical relations between the entries for
derived forms and their bases. On the other hand, however, a recent study by Voga, Anastassiadis-
Symeonidis, and Giraudo (2014) has challenged this account. In a study based on the same experi-
mental items as in Silva and Clahsen (2008), L2 speakers of English (with Greek as L1) showed
significant priming effects of a similar magnitude for both derivation and inflection. Voga et al.
(2014) thus argue against differences between derivation and inflection in L2 processing.
The present study aims to shed more light on this controversy by directly comparing the priming
effects of a derived and an inflected form on the same target, and by expanding the research to yet
another language, namely German.
Jacob et al. 623
The present study
Most L2 processing studies described above have only tested either derivation or inflection. Even
the three studies investigating both phenomena (i.e. Kırkıcı & Clahsen, 2013; Silva & Clahsen,
2008; Voga et al., 2014), while making far-reaching claims about the L2 processing of derived
versus inflected forms, have not directly compared the two. Instead, all three studies have essen-
tially conducted two separate experiments (one on derivation, the other on inflection) based on
different sets of prime–target pairs, with the priming effects from these two experiments later being
compared with each other. To illustrate this, consider the following derived and inflected sample
prime–target pairs from Silva and Clahsen (2008):
(3) boldness – BOLD (derivation)
(4) warned – WARN (inflection)
In this example, derived and inflected prime–target pairs are based on different targets (BOLD
vs. WARN). All three studies mentioned above determined priming effects for derivation and inflec-
tion by comparing (3) and (4) with their respective unrelated control conditions, and then com-
pared the resulting priming effects for derivation and inflection with each other. However, as
lexical decision times (and thus, also priming effects determined on the basis of these) also depend
on a number of basic properties of the target word such as frequency or orthography, such a com-
parison between two prime–target pairs that contain different targets is problematic. Also, because
the derivational and inflectional suffixes -ness and -ed differ in length, both the length of the
primes and the amount of surface form overlap between primes and targets are not controlled for
across derived and inflected items. Finally, because all three studies are based on separate exper-
iments for derivation and inflection, none of the studies felt the need to report the three-way
interactions between morphology type, prime type, and participant group. These, however, are
actually the critical test of the hypothesis.
A much better methodological approach is a design in which the same target word is preceded
by either an inflectionally related or otherwise similar derivationally related prime word. For a
similar question in L1 processing, Raveh and Rueckl (2000) made use of an experimental design
in which this is the case. For example, consider a set of morphologically related prime–target pairs
such as (5) and (6):
(5) boiler–BOIL (derivation)
(6) boiled–BOIL (inflection)
Both prime–target pairs are based on the same target word (BOIL), allowing for a direct com-
parison between them. In (5), the target is preceded by its derived nominalization and, in (6), by its
past tense form; in this respect, prime and target in (5) and (6) are both morphologically related, the
only difference being that the relationship between prime and target is derivational in nature in (5)
and inflectional in nature in (6). In L1 processing research, such designs have been used success-
fully to compare different morphological phenomena (e.g. Feldman et al., 2002; Fowler, Napps, &
Feldman, 1985; Raveh & Rueckl, 2000). We consider this methodological approach a promising
way to investigate whether L1/L2 differences in this area are specific to particular types of mor-
phologically complex forms.
The present study is, to our knowledge, the first masked priming study to directly compare the
L2 processing of derived and inflected forms. As in the above example, the experimental design is
624 International Journal of Bilingualism 22(6)
based on derived and inflected prime–target pairs in which the same target word (the infinitive
form of a German verb) is preceded by either a derivationally related prime (the verb’s -ung nomi-
nalization) or an inflectionally related prime (the verb’s past participle, -t form). We chose these
two particular morphological phenomena because German -ung nominalizations and past partici-
ples of a verb are typically of equal length (unlike other overtly inflected forms); also, both past
participles, in contrast to other inflected forms, and -ung nominalizations may also appear in isola-
tion in German usage (like in our experiment). Furthermore, both affixes add only one feature to
the base or stem: derivational -ung creates action nouns (e.g. Gründung ‘act of founding’) and the
participle encodes the feature [+ part] (Wunderlich & Fabri, 1995).
An advantage of such a within-items design is that potential confounds due to item characteris-
tics or response time (RT) differences between L1 and L2 groups can be avoided. First, RTs to the
same target are measured and compared to the same unrelated baseline, which avoids potential
difficulties due to baseline differences in the ‘traditional’ two-item-sets design. In other words, if
RTs following the unrelated prime in one but not the other set are exceptionally slow (due to char-
acteristics of the unrelated prime), the resulting priming effect is much larger in this set, creating
artificial differences between item sets. Second, differences in the magnitude of priming for the
two morphological primes can be compared within the L1 and L2 groups. A comparison of L1 and
L2 effects can be problematic as L2 learners’ lexical decision times are often considerably slower,
which, in turn, might result in numerically larger differences between conditions (i.e. priming
effects) in the L2 group. By comparing different experimental conditions within groups (besides
any between-group comparisons), the possible effect of general speed differences can be curtailed
to some degree.
If L2 processing is based on the same mechanisms as L1 processing, both L1 and L2 speakers
should show priming effects of a similar size for derivation and inflection. Assuming differences
between L1 and L2 processing, there might be two patterns of results: first, if L2 speakers are not
sensitive to morphological structure, they should not show any priming effects, neither for deriva-
tion nor for inflection; and second, if L1/L2 differences are specific to inflection, L1 speakers
should show priming for both morphological phenomena, while L2 speakers should show stronger
priming for derivation than for inflection.
Method
Participants
40 native speakers of German (mean age 23.93, age range 18–40, standard deviation (SD) 4.54, 28
female) and 36 Russian learners of German (mean age 26.72, age range 20–43, SD 4.79, 30 female),
who were living in Germany at the time of testing participated in the experiment for payment. Prior
to the experimental session, all participants from the L2 group completed the Goethe Institute
Placement Test, a 30-item multiple-choice cloze test, as a measure of their proficiency in L2
German. Participants obtained a mean score of 26.61 out of 30 (range 22–30, SD 2.07) in the
Goethe test, which corresponds to a mean group proficiency of ‘C1’, labelled as ‘effective opera-
tional proficiency’, in the Common European Framework for Languages.
Materials
The design contained three item sets: a morphological; an orthographic; and a semantic one.
Examples and item characteristics are provided in Table 1. The morphological set consisted
of 28 infinitival targets (e.g. ändern ‘to change’) preceded by either the corresponding -ung
Jacob et al. 625
nominalization as derived prime (e.g. Änderung ‘(the) change’), the past participle as inflected
prime (e.g. geändert ‘changed’), an identity prime, or a matched unrelated control prime. The deri-
vational and inflectional affixes both consisted of three letters added to the verbal stem (e.g. änder-),
thus affecting the length of derived and inflected primes in the same way.2 In the unrelated control
condition, the primes were nouns and adjectives (50% each, e.g. klein ‘small’), which were seman-
tically (as determined by a pretest in which native speakers were asked to spontaneously produce
associates when confronted with the prime words) and orthographically (i.e. prime and target did
not share any letters in the same position) unrelated to the target. Unrelated and identity primes
were matched with respect to length in letters (t(27) = −1.41, p = 0.170). Furthermore, all four
prime types were matched for word form frequency (all ps > 0.258), as taken from the webCELEX
database (http://celex.mpi.nl/).
Note that the prime–target pairs in the morphological item set are not only morphologically
related, but also related with regard to semantics and orthographic surface form. It is thus necessary
to check whether morphological priming effects are indeed morphological in nature, or instead
caused by the semantic or orthographic relatedness between prime and target. As the properties of
the language did not allow us to determine semantic and orthographic priming effects on the basis
of the same target words as in the morphological item set, we added two control item sets, a seman-
tic and an orthographic one (see Table 1). In both control item sets, targets were preceded by three
prime types: related; unrelated; and identity. The orthographic set consisted of 24 morphologically
simplex prime–target pairs which shared a similar amount of letters as the morphological items but
did not have a (pseudo-)morphological and/or semantic relationship. Half of the items mimicked
the word-initial shared letters (e.g. Kasten ‘box’–Kasse ‘cash register’) present in the derived and
half the overlap in later positions (e.g. Engel ‘angel’–Geld ‘money’) of inflected prime–target
pairs. The related prime–target pairs had the same amount of overlap as the morphological (derived
Table 1. Length and frequency information (from webCELEX) for the three item sets (WFF = word form
frequency per million, LF = lemma frequency per million, Length = length in letters).
Primes Targets
WFF LF Length WFF LF Length
Morphological items
Derived primes e.g. Änderung ‘change’
27.6 37.2 7.8
Inflected primes e.g. geändert ‘changed’ e.g. ändern ‘to change’
22.1 82.6 7.9 23.8 82.6 6.4
Unrelated primes e.g. klein ‘small’
23.1 62.8 6.3
Orthographic items
Related primes e.g. Kasten ‘box’
31.6 83.0 6.8 e.g. Kasse ‘cash register’
Unrelated primes e.g. Schwan ‘swan’ 36.3 71.5 5.5
32.0 91.2 6.5
Semantic items
Related primes e.g. Doktor ‘doctor’
107.6 142.8 5.1 e.g. Arzt ‘physician’
Unrelated primes e.g. Presse ‘press’ 86.3 120.1 4.8
83.0 150.0 5.1
626 International Journal of Bilingualism 22(6)
+ inflected) primes and the respective targets – both based on an absolute overlap (t <1) and spatial
coding measure (t <1), as determined with Davis’s (2000) Match Calculator. As unrelated primes,
morphologically simplex nouns or verbs were selected (here: Schwan ‘swan’; Stirn ‘forehead’),
matching the related prime in word category, length in letters (t(23) = −1.66, p = 0.110) as well as
word form and lemma frequency (both ts <1).
In the semantic set, there were 24 semantically related noun–noun pairs, half synonyms (e.g.
Doktor ‘doctor’–Arzt ‘physician’) and half associates (e.g. Wolke ‘cloud’–Himmel ‘sky’). The
semantic relatedness of these pairs was determined in a separate rating experiment with 20 L1
speakers of German, who rated the critical items with an average of 5.19 (SD=0.85) on a 7-point
Likert scale (1 = low semantic relatedness, 7 = high semantic relatedness). Morphologically sim-
plex nouns (here: Presse ‘press’; Kunde ‘customer’) matching the related primes in length as well
as word form and lemma frequency were selected as unrelated primes (all ts <1).
The targets in both the orthographic and semantic control items sets were comparable to the
morphological targets in terms of word form frequency (orthographic: t(33) = −1.27, p = 0.213;
semantic: t(24) = −1.79, p = 0.087) and lemma frequency (both ts <1). A full list of stimuli is pro-
vided in the Appendix.
To avoid participants making repeated lexical decisions on the same target verb, experimental
items were distributed across four presentation lists according to a Latin square design, with each
list containing exactly one of the four prime–target pairs from each item set.3 The 76 experimental
targets were mixed with 324 fillers in a pseudo-randomized order, with half of the targets being
nonwords, requiring a no-response in 50% of trials. Nonwords were created by changing 1 to 3
letters of existing German words. Targets and primes were a mixture of simplex and morphologi-
cally derived adjectives, verbs and nouns, in equal shares. Ten of the word–word fillers were -ung
nominalizations derived from verbs prefixed with be- and the corresponding simplex stem (e.g.
Bedienung ‘service, waiter’–(bedienen ‘to wait on someone’)–dienen ‘to serve’). Overall, the ratio
of items in which prime and target were related to each other in any way (i.e. morphologically,
semantically or orthographically) was 18.69%. In order to counterbalance training or fatigue
effects, each of the four lists was reversed for half of the participants.
Procedure
Participants were tested in a quiet laboratory room. Each participant was randomly assigned to one
of the presentation lists, and tested on this list only. Items were presented in the center of a com-
puter screen, with lexical decision times being measured using the DMDX experiment software
(Forster & Forster, 2003). Participants were instructed that they would see a number of existing
German words and nonwords on the computer screen, with the task being to decide as quickly as
possible whether a word was an existing German word or not, by pressing one of two different
buttons on a gamepad as quickly as possible.
An individual trial started with a 500-ms blank screen. Next, a forward mask consisting of a
number of hashes equal to the length of the following prime word appeared in the center of the
screen for 500 ms. Directly following the mask, the prime word was shown for 50 ms. The prime
was immediately followed by its corresponding target word. The target remained on screen until
the participant had made a lexical decision or until 500 ms had passed, with lexical decisions still
being possible for up to an additional 4500 ms after the target had disappeared.
Following the experiment, participants were given a short biographic questionnaire about their
language background. Additionally, all participants from the L2 group also completed the Goethe
Institute Placement Test.
Jacob et al. 627
Data analysis
Data cleaning was first carried out based on accuracy rates of participants and items. Two items
(semantic: Schlips ‘tie’; orthographic: Scheck ‘check’) were removed from all subsequent analyses
due to very low accuracy (70% or lower) in both the L1 and the L2 groups. No participants were
excluded (accuracy ranged from 88% to 100%). Further data cleaning involved removing incorrect
responses and timeouts (L1: 3.48% of the L1 remaining data; L2: 3.72%). Finally, in order to nor-
malize RT distributions and to reduce the influence of outliers (see Baayen & Milin, 2010; Ratcliff,
1993), analyses were conducted on the logarithm of RTs and cutoffs were applied to exclude
extreme latencies (under 300 ms or over 1,700 ms; L1: 8 data points, 0.04% of all L1 correct
responses; L2: 24 data points, 0.94%).4
Response times were analyzed with mixed-effects linear regression models with crossed
random effects for participants and items (see, e.g. Baayen, Davidson, & Bates, 2008). Parameters
were estimated with restricted maximum likelihood or, in the case of model comparisons, likeli-
hood ratio tests were conducted to compare models fitted with maximum likelihood (Baayen,
2008; Pinheiro & Bates, 2000).
Model selection
We first fitted a mixed-effects model to log RTs for targets in the morphological set, with random
intercepts for participant and for item, as well as two fixed effects and their interaction: Group (L1
and L2) and Prime Type (Unrelated, Inflected, Derived, Identity). The Group factor was coded
with “main effect” deviation contrasts (i.e. −0.5 and 0.5). Prime Type was coded with successive
differences contrasts, in which each level is compared to the one before (Inflected–Unrelated,
Derived–Inflected, and Identity–Derived). These contrasts were chosen because they allowed us to
assess the presence of inflectional priming, as well as to specifically test whether the derivational
priming is greater than inflectional priming. In addition, Trial (the rank of the item in the experi-
mental list; centered) was also included as a predictor (and significantly improved model fit; χ2(1)
=9.09, p = 0.003), both to remove residual auto-correlation and to control for trial-level task effects
(Baayen & Milin, 2010).
In order to determine a random structure justified by the data, random slopes for the key experi-
mental effects (Prime Type, by participant and item; Group, by item) were tested for inclusion via
likelihood ratio tests (Baayen, 2013). If one or more random slopes produced significantly better
models, we included the random slope that produced the greatest improvement in fit (as measured
by Akaike information criterion) and repeated the selection procedure for the remaining candi-
dates. Only the by-item random slope for Group produced a model with a better fit (χ2(2) =7.82,
p = 0.020; all other ps > 0.688).
The same model selection procedure was applied to the logged RT data from the control sets
(semantic and orthographic). For the data in the semantic set, only the random slope for Group
produced a better model (χ2(2) =11.58, p = 0.003; all other ps > 0.507). For the data in the ortho-
graphic set, random slopes for Group (χ2(2) = 11.43, p = 0.003) and Prime Type (by-; χ2(2) =16.57,
p = 0.005) led to better models. In order to maintain consistency across conditions, we only
included the random slope for Group. Nevertheless, the analyses of the orthographic set were
repeated with the additional random slope for Prime Type and produced the same pattern of signifi-
cant and non-significant results.
Accuracy data were analyzed with generalized linear models with a binomial link function, and
were subjected to the same model selection procedure.
628 International Journal of Bilingualism 22(6)
Results
Table 2 shows mean accuracy rates, as well as by-participant means and SDs of RTs to targets for
each condition (after removal of extreme values, as described above). The accuracy analysis
revealed no interactions between Prime Type and Group, and showed only more accurate responses
after identical primes.
In order to assess whether the L1 and L2 groups showed different facilitation effects, the best-fit
mixed-effects model (described above) was compared to a minimally different model, from which
the critical interaction between Prime Type and Group was removed. This comparison revealed
that including the interaction parameter significantly improved model fit (χ2(3) = 9.91, p = 0.019).
In contrast, for the control sets, the interaction between Prime Type and Group did not improve
model fit, neither for items in the semantic set (χ2(2) = 0.09, p = 0.957), nor for items in the ortho-
graphic set (χ2(2) = 2.49, p = 0.288). That is, whilst priming effects for the morphological set dif-
fered between the L1 and L2 participant groups, the two groups displayed similar priming effects
in the semantic and orthographic control sets.
In order to further investigate differences between the L1 and L2 groups in their responses to
items in the morphological set, we fitted separate mixed-effects models to the L1 and L2 latency
data (see Table 3). Fitting separate models to the data of each group allows more precise estimates
of the random effects, given the existence of large group differences in RTs, as evidenced by
substantially slower (and more variable) responses in the L2 group.
For the L1 group, inflectionally related primes elicited significantly faster responses than unre-
lated primes. Response latencies following inflected versus derived primes did not differ from each
other and responses after derived primes were slower than after the presentation of identical primes.
Table 2. Mean lexical decision times (and standard deviations) and mean accuracy by condition for first
language (L1) and second language (L2) speakers.
Morphological.
Unrelated Inflected Derived Identity
L1 Response time (RT) 635 (126) 608 (109) 617 (113) 568 (119)
Accuracy 93.9% 96.4% 97.5% 96.4%
L2 RT 767 (154) 751 (172) 725 (150) 711 (151)
Accuracy 93.7% 95.2% 96.4% 96.4%
Orthographic.
Unrelated Related Identity
L1 RT 592 (85) 601 (96) 543 (114)
Accuracy 98.2% 97.5% 96.8%
L2 RT 701 (113) 683 (114) 636 (122)
Accuracy 97.8% 98.6 94.2
Semantic.
Unrelated Related Identity
L1 RT 591 (89) 589 (91) 547 (104)
Accuracy 96.5% 98.5% 93.3%
L2 RT 685 (116) 685 (122) 635 (118)
Accuracy 97.1% 98.2% 94.9%
Jacob et al. 629
In addition, relevelling the Prime Type factor also showed that derivational primes elicited faster
responses than unrelated primes (b = −0.0309, standard error (SE) = 0.0144, t = −2.14). That is, for
the L1 group, the previous presentation of morphologically related forms, inflected or derived,
facilitated the recognition of target verbs and produced priming effects of similar magnitude.
The coefficients for the L2 group revealed a different pattern. First, responses after inflected
primes were not significantly faster than after unrelated primes. Second, RTs following the presen-
tation of derived primes were significantly shorter than for the inflected primes, and similar to
those after identical primes. Consistently, RTs following derived primes were also found to be
shorter than those following unrelated primes (b = 0.0615, SE = 0.0160, t = 3.83). In other words,
the L2 group showed a significant priming effect only for derived, but not for inflected forms.
In order to assess the possible role of proficiency in modulating priming effects, we fitted
additional statistical models to the lexical decision times of the L2 group for the items in the
morphological set. We first employed a regression model which included proficiency score (as
measured by the Goethe Institute Placement Test) as an additional (centered) predictor, as well
as the interaction between proficiency and Prime Type. This model revealed exactly the same
pattern of effects as the original analysis, that is, a significant difference between derivational
and inflectional priming (b = −0.0342, SE = 0.0160, t = −2.14), and a non-significant effect of
inflectional priming (b = −0.0270, SE = 0.0161, t = −1.68). Neither the effect of proficiency, nor
the interactions between proficiency and the contrast between derivation and inflection, reached
significance (both |t|s < 1.09). Furthermore, model comparisons revealed that adding proficiency
and its interaction with Prime Type as additional predictors did not produce a model with greater
goodness of fit than the model shown on Table 3 (χ2(4) = 2.96, p = 0.565).
A similar analysis was conducted to examine whether years of experience with an L2 had an
influence on morphological priming effects. As in the previous analysis, the number of years since
the onset of German acquisition was included as a (centered) predictor, as well as its interaction with
Prime Type. This model also revealed the same pattern of effects as the original analysis, namely, a
significant difference between derivational and inflectional priming (b = −0.0339, SE = 0.0160,
t = −2.13), whereas the effect of inflected primes failed to reach significance (b = −0.027, SE = 0.0161,
t = −1.70). Neither the effect of years of experience, nor the interaction between years of experience
and the contrast between derivation and inflection, reached significance (both |t|s<1.67). Model
comparisons showed that adding years of experience (and interactions with Prime Type) did not
improve goodness of fit, relatively to the model presented in Table 3 (χ2(4) = 6.78, p = 0.148).
In sum, neither language proficiency nor years of experience with an L2 modulated the comparison
between derivation and inflection. In addition, the crucial difference between priming produced by
derived and inflected forms remained significant, even when these variables were controlled.
Priming effects for the semantic and orthographic sets were also examined by fitting mixed-
effects models to the L1 and L2 data. For the semantic set, RTs following related primes were not
significantly faster than responses following unrelated primes, neither in the L1 group (b = −0.0071,
SE = 0.0131, t = −0.54) nor in the L2 group (b = −0.0001, SE = 0.0157, t = −0.02), but the pres-
entation of identical primes elicited faster RTs than after related primes in both the L1 group
(b = −0.0810, SE = 0.0132, t = −6.14) and the L2 group (b = −0.0893, SE = 0.0159, t = −5.63).
The same pattern was obtained for the orthographic set, that is, no facilitation after the presentation
of related primes in either participant group (L1: b = 0.0112, SE = 0.0140, t = 0.80; L2: b = −0.0178,
SE = 0.0156, t = −1.14), but faster responses after identical primes (L1: b = −0.1151, SE = 0.0141,
t = −8.16; L2: b = −0.0900, SE = 0.0157, t = −5.73). In sum, the presentation of semantically or
orthographically related forms did not produce facilitation effects in either participant group.
While the orthographic priming effects did not reach significance in either participant group,
we nevertheless also determined whether the priming effects in the morphological items were
630 International Journal of Bilingualism 22(6)
significantly different from the orthographic priming effects.5 Mixed-effects models fit to both the
morphological and orthographic data sets showed significant Relation Type (morphological vs.
orthographic) X Prime Type (related vs. unrelated) interactions. In particular, for the L1 group both
inflectional (b = 0.0500, SE = 0.0201, t = 2.49) and derivational priming (b = 0.0428, SE = 0.0204,
t = 2.10) were found to be larger than orthographic priming. For the L2 group, derivational priming
was stronger than orthographic priming (b = 0.0447, SE = 0.0224, t = 2.00), but inflectional prim-
ing was not (b = 0.0082, SE = 0.0225, t = 0.37).
Finally, within the orthographic set, we also compared priming effects for items in which ortho-
graphic overlap between prime and target was word-initial (mimicking the orthographic overlap in
the derivational primes, e.g. Kasten ‘box’–Kasse ‘cash register’), with items in which orthographic
overlap was word-central (as was the case for inflectional primes, for example, Engel ‘angel’–Geld
‘money’). To this end, we analyzed the L1 and L2 data in the orthographic data set with regression
models that included an Overlap Position factor (initial vs. middle), interacting with the Prime
Type factor. The results for the L1 group revealed no significant orthographic priming effects (ini-
tial: b = 0.0222, SE = 0.0208, t = 1.06; central: b = 0.0005, SE = 0.02070, t = 0.03), as well as no
significant differences between the priming effects of word-initial and word-central overlap items
(i.e. no Prime Type X Overlap Position interaction; b = −0.0216, SE = 0.0306, t = −0.71). The same
results were obtained for the L2 group, that is, no orthographic priming for initial (b = −0.0119,
SE = 0.0224, t = −0.53) or central-overlap items (b = −0.0246, SE = 0.0245, t = −1.00), and no dif-
ference between the two types of items (b = −0.0126, SE = 0.0350, t = −0.36).
Discussion
The present study revealed an interesting L1/L2 difference with respect to morphological priming
effects. The L1 group showed equally strong facilitation following derived and inflected primes. In
the L2 group, by contrast, differences between the two types of morphologically related item sets
emerged, with a robust priming effect for derivation, but no significant priming for inflection. For
both the L1 and L2 groups, the semantic and orthographic control conditions did not show any
Table 3. Fixed effects from the linear mixed-effects model for morphological items.
First language.
Fixed effects Estimate Standard error t
Intercept 6.378 0.02800 227.78
Prime Type (inflected - unrelated) –0.03767 0.01448 −2.60*
Prime Type (derived - inflected) 0.006753 0.01434 0.47
Prime Type (identity - derived) –0.08639 0.01432 −6.03*
Trial (centered) –0.00006 0.00004 −1.41
Second language.
Fixed effects Estimate Standard error t
Intercept 6.569 0.03529 186.14
Prime Type (inflected - unrelated) –0.02721 0.01609 −1.69
Prime Type (derived - inflected) –0.03430 0.01597 −2.15*
Prime Type (identity - derived) –0.01953 0.01591 −1.23
Trial (centered) –0.00014 0.00005 −2.95*
Jacob et al. 631
significant priming effects; in this respect, the priming effects in the derived and inflected condi-
tions were unique to morphologically related prime–target pairs.
To what extent could the difference between derivational and inflectional priming be explained
through non-morphological factors? Several masked-priming studies (e.g. Diependaele et al.,
2011; Feldman et al., 2010; Heyer & Clahsen, 2015) have found differences between L1 and L2
speakers with regard to the role of orthographic overlap between prime and target, with L2 speak-
ers focusing relatively more on orthographic surface form properties. As morphologically related
prime–target pairs are also orthographically related, the possibility that L2 processing focuses
relatively more on orthographic surface form could potentially cause L1/L2 differences in prim-
ing effects. However, while the role of orthography in the L2 processing of complex words cer-
tainly deserves attention, we consider it difficult to explain the difference between derivation and
inflection in our L2 data in this way. First, our orthographic control condition neither showed any
orthographic priming effects in the L2 group nor any significant differences between the L1 and
L2 groups with regard to orthographic priming. While it is true that our orthographic and morpho-
logical conditions are based on different items, a form-based account would nevertheless predict
both a significant orthographic priming effect in the L2 group, and L1/L2 differences in the ortho-
graphic control condition. Second, the degree of overall orthographic overlap (i.e. the number of
letters shared between prime and target) was actually the same for derived and inflected items, the
only difference being that the overlap was relatively more word-initial for derived than for
inflected items. Thus, in order to explain the difference between derivation and inflection on the
basis of orthography, it would at least be necessary to make the additional assumption that L2
speakers react differently to word-initial and word-central overlap. Our orthographic control con-
dition, however, did not show any difference between items with word-initial and word-central
overlap. Finally, the difference between derivational and inflectional priming did not only occur
in our study, but also in the two above-mentioned L2 processing studies by Silva and Clahsen
(2008) and Kırkıcı and Clahsen (2013) on English and Turkish respectively. In these studies,
derived and inflected prime–target pairs did not differ with regard to word-initial versus word-
central overlap. As a result, a form-based account struggles to explain the difference between
derivation and inflection in these cases.
Assuming that the difference between derivation and inflection is related to morphological
properties of the items, we can think of three accounts to explain the divergent effects for derived
and inflected forms in our as well as previous studies’ L2 groups. One possibility is differences in
complexity between inflected and derived forms. While derivational suffixes typically have one
meaning or function, inflectional affixes may encode several morphosyntactic features at the same
time. This is particularly the case for fusional inflection; for example, the 2nd sg. affix -st in
German (as in änderst ‘you change’) encodes person, number and tense. Affix complexity may
increase processing costs, particularly in a late-learned L2. This contrast between inflection and
derivation can indeed explain Kırkıcı and Clahsen’s (2013) finding of more priming for -lIk deri-
vational forms than for tense + aspect inflection (Aorist) in L2 Turkish. However, this reasoning
cannot account for the current results, because German participles, just as derived nominalizations,
encode only a single feature, [+ part] (Wunderlich & Fabri, 1995).
A second possibility, as suggested by Kırkıcı and Clahsen (2013), assumes that the difference
between derivation and inflection is not caused by properties of derivational versus inflectional
affixes, but instead by properties of derived versus inflected words. The account makes use of the
distinction between derivation and inflection as proposed in realization-based models of morphol-
ogy (e.g. Anderson, 1992). Recall that, in these models, derivation creates new lexical entries while
inflection spells out or realizes morphosyntactic features. Kırkıcı and Clahsen (2013) propose that
the different priming effects for derivation and inflection in the L2 can be explained in terms of this
632 International Journal of Bilingualism 22(6)
distinction. Consider, for example, the realization rules in (7) and (8) for the two phenomena we
investigated. While both involve the combination of a variable (‘X’) with another component (-t
and -ung respectively), the output of (8) is a new lexeme (‘LEX’) and the inflectional -t is just the
spell-out of a morphosyntactic feature (‘part’). Hence, if the recognition of -ung forms (e.g.
Änderung) involves a structured representation such as (8) (here: [[änder]V,LEX -ung]N, LEX), this
co-activates the corresponding base entry (here: [änder]V,LEX ), which then causes a lexically medi-
ated priming effect. Given representations such as (7) for regular inflection, priming via the lexical
route is not available in these cases. As mentioned above, L1 priming effects for regular inflection
are commonly explained in terms of stem reactivation due to affix stripping. For L2 processing,
however, several studies have shown that the L2 comprehension system relies less on this morpho-
logical decomposition route than the L1 system (e.g. Neubauer & Clahsen, 2009; Silva & Clahsen,
2008). Thus, regular inflection produces particularly clear L1/L2 differences.
(7) <[V, part], X + t>
(8) <[V,
lex], X> → <[N, lex], X + ung>
The account posits two routes through which masked priming effects may emerge: affix strip-
ping; and lexical mediation. As pointed out by Kırkıcı and Clahsen (2013), these two routes have
previously been proposed for L1 processing to explain facilitation effects for both affixal and non-
decompositional (irregular) inflection in masked priming (Crepaldi et al., 2010). Thus, priming
effects for derived words, in both the L1 and L2, are assumed to be due to lexical relatedness rather
than morphological decomposition. For instance, Änderung may prime ändern not because
Änderung is decomposed, but because the lexical entries for the two words partially overlap in the
mental lexicon. However, note that while Crepaldi’s model indeed explains priming effects for
irregular inflection through co-activation of lexical entries, it explicitly rules out this possibility for
derivation. In other words, while Kırkıcı and Clahsen (2013) refer to Crepaldi’s model to explain
their results, their explanation is actually inconsistent with the assumptions of Crepaldi’s model.
Also, Kırkıcı and Clahsen’s account implicitly assumes that derived words, in both L1 and L2
processing, are not decomposed, with priming effects for derived words instead arising through
lexical mediation. This is a highly controversial claim, especially given that previous L1 research
has even shown decomposition effects for pseudo-derived words such as brother (e.g. Rastle,
Davis, & New, 2004). Finally, if we assume that the priming effects for derived words in L2
speakers are due to lexical mediation, this raises the question why the L2 speakers do not show
lexically-mediated priming for inflected forms as well.
An alternative, third possibility is that L2 speakers can, in principle, strip off affixes from
morphologically complex words, but struggle to do so for inflected forms due to the particular
properties of inflectional versus derivational affixes. One such property is that derivational affixes,
unlike inflectional ones, contain semantic information. For example, the English deadjectival suf-
fix -ness refers to ‘the state of being adjective’ or the German deverbal suffix -ung describes ‘the
act of verbing’. The fact that derivational affixes contain meaning might allow the L2 processor to
decompose a derived word in a similar way as a compound. Inflectional affixes in contrast, can be
considered less salient, which potentially constitutes a problem for the L2 decomposition mecha-
nism. This account is consistent with previous L2 research on compound processing, which has
found similar decomposition effects for compound words in L1 and L2 speakers (Li, Jiang, & Gor,
2015).
A challenge for our results, at least prima facie, comes from studies that have claimed native-
like processing of inflected word forms in L2 learners. For example, both Feldman et al. (2010) and
Jacob et al. 633
Coughlin and Tremblay (2015) found significant priming effects for inflected forms in both L1 and
L2 speakers, suggesting that L1 and L2 processing are based on the same processing mechanisms.
However, as shown by our findings, the mere presence of significant inflectional priming effects in
L2 speakers does not necessarily mean that there are no differences between L1 and L2 morpho-
logical processing. Our L2 results show a numerical trend for inflectional priming, and are in this
respect not necessarily inconsistent with findings from these previous studies; despite this, L1/L2
differences still emerged when comparing derivation and inflection. As discussed in the introduc-
tion, the lack of a difference between L1 and L2 effects in the previous studies might have been due
to group differences in overall lexical decision speed. For instance, priming effects for regular
inflection in Feldman et al.’s (2010) L1 and L2 groups were numerically comparable (L1: 30 ms;
L2: 32 ms); however, in relation to the overall RTs for each participant group (L1: 606–648 ms vs.
L2: 768–800 ms), the facilitation was proportionally weaker in the L2 group. In our study, it is
possible to avoid this problem because the crucial effect involves a within-group comparison
between the derived and inflected conditions instead of a comparison of priming effects across dif-
ferent participant groups.
Conclusion
The present study has revealed a distinction between the processing of derived and inflected forms,
thus supporting linguistic theories that distinguish between these two morphological processes.
Crucially, this distinction became apparent in the L2 data only. Therefore, it is important to inves-
tigate morphological processing in different participant groups and not only in L1 speakers.
Furthermore, future research should employ designs that directly compare different linguistic phe-
nomena within groups rather than across groups in order to avoid confounds resulting from low-
level differences (e.g. response speed) between groups.
Author’s Note
Vera Heyer is now affiliated to the University of Braunschweig, Germany.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or
publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or pub-
lication of this article: The research reported in this manuscript was supported by a Humboldt professorship
grant to Harald Clahsen and the Potsdam Research Insitute for Multilingualism.
Notes
1. The term second language (L2) is used in its general sense here, referring to a late-learned (i.e., after the
age of 3) foreign language, even though the investigated language might in some cases have been the
third, fourth,… language a participant learned.
2. Note that while the derived form has the three-letter suffix -ung, the participle’s affix appears to con-
sist of a prefix and a suffix at first sight. However, ge- is a piece of prosody in order to insure that the
resulting participle has an unstressed first syllable (cf. Clahsen, 1999, p. 1011; Wunderlich, 1996, p. 98;
Wunderlich & Fabri, 1995, p. 265). For instance, ˈlieben ‘to love’ takes ge- (geˈliebt), whereas the first
syllable in verˈlieben (‘to fall in love’) and spaˈzieren (‘to go for a walk’) is already unstressed, resulting
in participles without ge- (verˈliebt and spaˈziert respectively).
634 International Journal of Bilingualism 22(6)
3. This was due to the morphological items having four prime types. As the orthographic and semantic
items had only three different primes, a third of prime–target pairs from each of the first three lists was
repeated in the fourth.
4. The cut-off point for long outliers was determined by its effect in normalizing RT distributions. We
applied successively smaller cut-offs at each 100 millisecond (ms) interval (starting at 3000 ms) and
chose the first cut-off point for which both the first language (L1) and second language (L2) distributions
of log RTs in the morphological set were sufficiently close to a Gaussian (indicated by values of skew-
ness and excess kurtosis below |1|).
5. Separate models tested whether inflectional and derivational priming was different from orthographic
priming, so that, as was the case in the orthographic data set, the Prime Type factor could have three
levels (Unrelated, Test, and Identity).
References
Anderson, S. R. (1992). A-Morphous morphology. Cambridge, UK: Cambridge University Press.
Aronoff, M. (1994). Morphology by itself. Cambridge, MA: MIT Press.
Baayen, R. H. (2008). Analyzing linguistic data: A practical introduction to statistics using R. Cambridge,
UK: Cambridge University Press.
Baayen, R. H. (2013). Multivariate statistics. In: R. Podesva, & D. Sharma (Eds.), Research methods in lin-
guistics (pp. 337–372). Cambridge, UK: Cambridge University Press.
Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects modelling with crossed random effects
for subjects and items. Journal of Memory and Language, 59, 390–412.
Baayen, R. H., & Milin, P (2010). Analyzing reaction times. International Journal of Psychological Research,
3(2), 12–28.
Clahsen, H. (1999). Lexical entries and rules of language: A multidisciplinary study of German inflection.
Behavioral and Brain Sciences, 22, 991–1060.
Clahsen, H., & Felser, C. (2006). Grammatical processing in language learners. Applied Psycholinguistics,
27(1), 3–42.
Clahsen, H., & Neubauer, K. (2010). Morphology, frequency, and the processing of derived words in native
and nonnative speakers. Lingua, 120, 2627–2637.
Coughlin, C. E., & Tremblay, A. (2015). Morphological decomposition in native and non-native French
speakers. Bilingualism: Language & Cognition, 18, 524–542. doi: 10.1017/S1366728914000200.
Crepaldi, D., Rastle, K., Coltheart, M., & Nickels, L. (2010). ‘Fell’ primes ‘fall’, but does ‘bell’ rime ‘ball’?
Masked priming with irregularly-inflected primes. Journal of Memory and Language, 63(1), 83–99.
Dal Maso, S., & Giraudo, H. (2014). Morphological processing in L2 Italian: Evidence from a masked prim-
ing study. Lingvisticae Investigationes, 37, 322–337.
Davis, C. J. (2000). Match calculator. Software. Retrieved from http://www.pc.rhul.ac.uk/staff/c.davis/
Utilities/MatchCalc/index.htm
Diependaele, K., Duñabeitia, J.A., Morris, J., & Keuleers, E. (2011). Fast morphological effects in first and
second language word recognition. Journal of Memory and Language, 64, 344–358.
Feldman, L. B. (1994). Beyond orthography and phonology: Differences between inflections and derivations.
Journal of Memory and Language, 33, 442–470.
Feldman, L. B., Barac-Cikoja, D., & Kostic, A (2002). Semantic aspects in morphological processing:
Transparency effects in Serbian. Memory & Cognition, 30, 629–636.
Feldman, L. B., Kostić, A., Basnight-Brown, D. M., Filipović Đurđević, D., & Pastizzo, M. J. (2010).
Morphological facilitation for regular and irregular verb formations in native and nonnative speakers:
Little evidence for two distinct mechanisms. Bilingualism: Language and Cognition, 13, 119–135.
Fowler, C. A., Napps, S. E., & Feldman, L. B. (1985). Relations among regular and irregular morphologically
related words in the lexicon as revealed by repetition priming. Memory & Cognition, 13, 241–255.
Forster, K. I., & Forster, J. C. (2003). DMDX: A Windows display program with millisecond accuracy.
Behavior Research Methods Instruments and Computers, 35, 116–124.
Harley, H., & Noyer, R. (1999). Distributed morphology. Glot International, 4(4), 3–9.
Jacob et al. 635
Heyer, V., & Clahsen, H. (2015). Late bilinguals see a scan in scanner AND in scandal: Dissecting formal
overlap from morphological priming in the processing of derived nouns. Bilingualism: Language and
Cognition, 18, 543–550.
Kırkıcı, B., & Clahsen, H. (2013). Inflection and derivation in native and non-native language processing:
Masked priming experiments on Turkish. Bilingualism: Language and Cognition, 16, 776–794.
Li, M., Jiang, N., & Gor, K. (2015). L1 and L2 processing of compound words: Evidence from masked
priming experiments in English. Bilingualism: Language and Cognition, FirstView, 1–19.
Marslen-Wilson, W. D. (2007). Morphological processes in language comprehension. In: G. Gaskel (Ed.),
The Oxford handbook of psycholinguistics (pp. 175–193). Oxford, UK: Oxford University Press.
Neubauer, K., & Clahsen, H. (2009). Decomposition of inflected words in a second language: An experimen-
tal study of German participles. Studies in Second Language Acquisition, 31, 403–435.
Pinheiro, J. C., & Bates, D. M. (2000). Mixed-effects models in S and S-PLUS. New York, NY: Springer.
Rastle, K., Davis, M. H., & New, B. (2004). The broth in my brother’s brothel: Morpho-orthographic segmen-
tation in visual word recognition. Psychonomic Bulletin & Review, 11, 1090–1098.
Ratcliff, R. (1993). Methods for dealing with reaction time outliers. Psychological Bulletin, 114, 510–532.
Raveh, M. (2002). The contribution of frequency and semantic similarity to morphological processing. Brain
and Language, 81(1–3), 312–325.
Raveh, M., & Rueckl, J. G. (2000). Equivalent effects of inflected and derived primes: Long-term morpho-
logical priming in fragment completion and lexical decision. Journal of Memory and Language, 42,
103–119.
Silva, R., & Clahsen, H. (2008). Morphologically complex words in L1 and L2 processing: Evidence from
masked priming experiments in English. Bilingualism: Language and Cognition, 11, 245–260.
Stanners, R., Neiser, J., Hernon, W., & Hall, R. (1979). Memory representation for morphologically related
words. Journal of Verbal Learning and Verbal Behavior, 18, 399–412.
Voga, M., Anastassiadis-Symeonidis, A., & Giraudo, H. (2014). Does morphology play a role in L2 process-
ing? Two masked priming experiments with Greek speakers of ESL. Lingvisticae Investigationes, 37,
338–352.
Wunderlich, D. (1996). Minimalist morphology: The role of paradigms. In: G. Booij, & J. van Marle (Eds.),
Yearbook of Morphology 1995 (pp. 93–114). Dordrecht, The Netherlands: Kluwer Academic Publishers.
Wunderlich, D., & Fabri, R. (1995). Minimalist morphology: An approach to inflection. Zeitschrift für
Sprachwissenschaft, 14, 236–264.
Author biographies
Gunnar Jacob is a Postdoctoral Researcher at the Potsdam Research Institute for Multilingualism, University
of Potsdam. His research focusses on grammatical processing in adult second-language learners.
Vera Heyer conducted her PhD research at the Potsdam Research Institute for Multilingualism, and is now a
Postdoctoral Researcher and a Lecturer at the University of Braunschweig, Germany. Her research focusses
on morphological processing in native and non-native adult speakers, spelling, and language acquisition.
João Veríssimo is a Postdoctoral Researcher at the Potsdam Research Institute for Multilingualism of the
University of Potsdam, Germany. His research focuses on lexical representation and morphological pro-
cessing in native and non-native speakers.
636 International Journal of Bilingualism 22(6)
Appendix. List of critical items.
(1) Morphological items.
Derived prime Inflected prime Unrelated prime Target
Änderung ‘change’ geändert ‘changed’ klein ‘small’ ändern ‘to change’
Äußerung ‘utterance’ geäußert ‘uttered’ Teufel ‘devil’ äußern ‘to utter’
Bohrung ‘drilling’ gebohrt ‘drilled’ Pfanne ‘pan’ bohren ‘to drill’
Drehung ‘rotation’ gedreht ‘turned’ Teppich ‘carpet’ drehen ‘to turn’
Drohung ‘threat’ gedroht ‘threatened’ schräg ‘crooked’ drohen ‘to threaten’
Ehrung ‘distinction’ geehrt ‘honored’ Dampf ‘steam’ ehren ‘to honor’
Forderung ‘request’ gefordert ‘demanded’ Artikel ‘article’ fordern ‘to demand’
Führung ‘leadership’ geführt ‘led’ Brief ‘letter’ führen ‘to lead’
Fütterung ‘feeding’ gefüttert ‘fed’ Scheune ‘barn’ füttern ‘to feed’
Kündigung ‘written notice’ gekündigt ‘discharged’ Batterie ‘battery’ kündigen ‘to give notice’
Kürzung ‘reduction’ gekürzt ‘shortened’ Storch ‘stork’ kürzen ‘to shorten’
Lenkung ‘steering’ gelenkt ‘steered’ privat ‘personal’ lenken ‘to steer’
Lieferung ‘delivery’ geliefert ‘delivered’ scharf ‘sharp’ liefern ‘to deliver’
Lösung ‘solution’ gelöst ‘solved’ knapp ‘scarce’ lösen ‘to solve’
Öffnung ‘opening’ geöffnet ‘opened’ streng ‘strict’ öffnen ‘to open’
Prüfung ‘check’ geprüft ‘checked’ Spiegel ‘mirror’ prüfen ‘to check’
Rettung ‘rescue’ gerettet ‘rescued’ schwarz ‘black’ retten ‘to rescue’
Sammlung ‘collection’ gesammelt ‘collected’ hübsch ‘pretty’ sammeln ‘to collect’
Schenkung ‘gift’ geschenkt ‘given (gift)’ Knochen ‘bone’ schenken ‘to make a gift’
Sicherung ‘safeguarding’ gesichert ‘secured’ Fleisch ‘meat’ sichern ‘to secure’
Steigerung ‘increase’ gesteigert ‘increased’ Flasche ‘bottle’ steigern ‘to increase’
Störung ‘disruption’ gestört ‘disturbed’ frisch ‘fresh’ stören ‘to disturb’
Täuschung ‘deception’ getäuscht ‘deceived’ schlank ‘slim’ täuschen ‘to deceive’
Trennung ‘separation’ getrennt ‘separated’ schlimm ‘bad’ trennen ‘to separate’
Übung ‘practice’ geübt ‘practised’ Wolf ‘wolf’ üben ‘to practise’
Wanderung ‘hike’ gewandert ‘hiked’ komplett ‘complete’ wandern ‘to hike’
Warnung ‘warning’ gewarnt ‘warned’ flach ‘flat’ warnen ‘to warn’
Weigerung ‘refusal’ geweigert ‘refused’ kaputt ‘broken’ weigern ‘to refuse’
(2) Orthographic items.
Related prime Unrelated prime Target
Apfel ‘apple’ Knopf ‘button’ Fell ‘fur’
brauchen ‘to require’ schreiben ‘to write’ tauchen ‘to dive’
denken ‘to think’ hören ‘to hear’ danken ‘to thank’
Engel ‘angel’ Stirn ‘forehead’ Geld ‘money’
Eule ‘owl’ Pilz ‘mushroom’ heulen ‘to wail’
grinsen ‘to grin’ stinken ‘to stink’ Insel ‘island’
Kartoffel ‘potato’ Petersilie ‘parsley’ Karte ‘card, map’
Kasten ‘box, chest’ Schwan ‘swan’ Kasse ‘cash register’
löschen ‘to extinguish’ trösten ‘to comfort’ Scheck ‘check’
Regel ‘rule’ Prinz ‘prince’ Regen ‘rain’
reiten ‘to ride’ malen ‘to paint’ Reise ‘journey’
Schachtel ‘case, box’ Trompete ‘trumpet’ Schal ‘scarf’
Schaf ‘sheep’ Biene ‘bee’ schaffen ‘to accomplish’
Jacob et al. 637
Related prime Unrelated prime Target
schauen ‘to look’ lächeln ‘to smile’ Haus ‘house’
schlafen ‘to sleep’ zwingen ‘to force’ Schlange ‘snake’
schneiden ‘to cut’ flüchten ‘to flee’ Schnee ‘snow’
Schule ‘school’ Technik ‘technology’ Schuh ‘shoe’
schweigen ‘to be silent’ springen ‘to jump’ Wein ‘wine’
Schwein ‘pig’ Schmutz ‘dirt’ weinen ‘to cry’
schwindeln ‘to swindle’ schnarchen ‘to snore’ Wind ‘wind’
sprechen ‘to speak’ finden ‘to find’ rechnen ‘to calculate’
Tante ‘aunt’ Sturm ‘storm’ Tanne ‘fir tree’
Tasche ‘bag’ Stoff ‘fabric’ Tasse ‘cup’
Wache ‘guard’ Alarm ‘alarm’ wachsen ‘to grow’
Orthographic items (continued)
(3) Semantic items.
Related prime Unrelated prime Target
Banane ‘banana’ Forelle ‘trout’ Affe ‘monkey’
Bild ‘picture’ Land ‘country’ Foto ‘photo’
Boot ‘boat’ Burg ‘castle’ Schiff ‘ship’
Doktor ‘doctor’ Presse ‘press’ Arzt ‘physician’
Frau ‘woman’ Teil ‘part’ Mann ‘man’
Gardine ‘curtain’ Gazelle ‘gazelle’ Vorhang ‘curtain’
Gürtel belt’ Sperre ‘barricade’ Hose ‘pants’
Herd ‘stove’ Kuss ‘kiss’ Ofen ‘oven’
Hut ‘hat’ Lob ‘praise’ Mütze ‘cap’
Kirche ‘church’ Woche ‘week’ Gebet ‘prayer’
Krawatte ‘tie’ Passage ‘passage’ Schlips ‘tie’
Matrose ‘sailor’ Tablette ‘tablet, pill’ Seemann sailor’
Motor ‘engine’ Name ‘name’ Auto ‘car’
Soldat ‘soldier’ Etappe ‘stage’ Krieg ‘war’
Stadt ‘city’ Grund ‘reason’ Dorf ‘village’
Tisch ‘table’ Licht ‘light’ Stuhl ‘chair’
Tonne ‘barrel’ Wiese ‘meadow’ Fass ‘barrel’
Truhe ‘chest, box’ Ampel ‘traffic light’ Kiste ‘box’
Vase ‘vase’ Nonne ‘nun’ Blume ‘flower’
Wolke ‘cloud’ Kunde ‘customer’ Himmel ‘sky’
Wüste ‘desert’ Zitat ‘quote’ Sand ‘sand’
Zeit ‘time’ Herr ‘mister’ Uhr ‘clock’
Zug ‘train’ Hof ‘yard, court’ Bahn ‘rail’, ‘lane’
Zweig ‘twig, branch’ Wucht ‘impact’ Baum ‘tree’
