Lexical Representation of Complex Verbs
Running Head: THE LEXICAL REPRESENTATION OF COMPLEX VERBS
The Lexical Representation of Complex Verbs in German: Morphological Structure Overrides
Semantic Compositionality
Eva Smolka, Katrin Preller, and Carsten Eulitz
University of Konstanz
Correspondence to:
Eva Smolka
Department of Linguistics
University of Konstanz
78457 Konstanz
GERMANY
Email: [email protected]
Phone: +49-7531-88 4834
Fax:
+49-7531-88 4898
1
Lexical Representation of Complex Verbs
Abstract
The lexical representation of words in Indo-European languages like English is
generally assumed to be driven by meaning compositionality. This study examined the lexical
representation of complex verbs in German, which is a morphologically rich representative of
Indo-European languages. Three overt priming experiments manipulated prime-target
relations between morphological, semantic, and form relatedness. Base verbs (e.g., binden,
‘bind’) were preceded by derivations that were semantically related (zubinden, ‘tie’) or
semantically unrelated (entbinden, ‘deliver’), by purely semantically related (zuschnüren,
‘tie’), form-related (abbilden, ‘depict’), or unrelated (abholzen, ‘deforest’) verbs. To examine
whether lexical representation is affected by modality (specific vs. independent) or by
information delivery (simultaneous vs. sequential), prime presentation was varied between
visual-simultaneous (Exp. 1), auditory-sequential (Exp. 2), and visual-sequential (Exp. 3).
Results showed that information delivery affected semantic processing, while modality
affected form and morphological processing. In particular, semantic facilitation emerged only
under visual-sequential prime presentation (Exp. 3), while form-related inhibition surfaced
only under cross-modal priming (Exp. 2). In contrast, morphological facilitation emerged
under all prime conditions (Exp. 1-3): priming by semantically opaque derivations was less
than that by transparent derivations under cross-modal presentation (Exp. 2), but equivalent to
that by transparent derivations under visual presentations (Exp. 1, 3). Altogether, these
findings indicate that complex verbs refer to their base as lexical representation. Lexical
representations in German thus differ from those in other Indo-European languages. In light
of this new evidence for the existence of a morphological dimension to lexical representation
in German, we discuss such a model.
Keywords: morphological priming, semantic priming; lexical representation; semantic
transparency; simultaneous vs. sequential processing; modality-abstract representation
2
Lexical Representation of Complex Verbs
The Lexical Representation of Complex Verbs in German: Morphological Structure Overrides
Semantic Compositionality
Complex words like departmental comprise several morphemic constituents, de, part,
ment, al, that recur in the language in many other words. Since early research in the seventies
(Murrell & Morton, 1974; Taft & Forster, 1975), it is well established that morphemic units
affect the recognition of complex words. The much debated question—whether complex
words are decomposed into constituent units or retrieved as full forms from memory—is
tightly interconnected with the questions regarding the nature of morphological structure and
its representation and the locus at which morphological structure operates.
The first question is whether morphological structure differs from that of shared form
and meaning. On a distributed connectionist account, for instance, morphological structure
reflects the learned systematic relationships among word forms and their meanings. Hence,
morphological effects emerge from the combined activation of orthographic, phonological,
and semantic codes (Plaut & Gonnerman, 2000; Raveh, 2002; Rueckl, Mikolinski, Raveh,
Miner, & Mars, 1997). Indeed, morphological effects have been shown to vary according to
the gradual overlap of form and meaning between word pairs (Gonnerman, Seidenberg, &
Andersen, 2007). Since morphological structure is assumed to emerge as an epiphenomenon
of form and meaning overlap, it is not explicitly represented but rather (indirectly) captured
by the hidden units in between form codes and semantic codes.
In contrast, localist approaches assume that morphological constituents are explicitly
(locally) stored in lexical memory, given that morphological effects have been observed to
emerge independently of semantic or form effects (Feldman, 2000; Marslen-Wilson, Tyler,
Waksler, & Older, 1994; Schreuder & Baayen, 1995; Taft & Forster, 1975). For example,
morphological decomposition in English was not affected by orthographic differences
between prime and target, like the missing “e” in adorable vs. adore (McCormick, Rastle, &
3
Lexical Representation of Complex Verbs
Davis, 2009). Priming was also shown to be independent of phonological overlap (Fowler,
Napps, & Feldman, 1985), given that morphologically and phonologically related word pairs
(healer – heal) resulted in the same amount of priming as morphologically related but
phonologically unrelated word pairs (health – heal).
Similarly findings were observed in German (Smolka, Zwitserlood, & Rösler, 2007),
where morphologically related words whose stems differ with respect to their surface
appearance (gezogen – ziehen, ‘pulled – pull’) primed just as well as morphologically related
words with the same orthographic stem (gekauft – kaufen, ‘bought – buy’). However, many
localist approaches today admit that the morphological effects may be implemented in
interactive or connectionist accounts (Smolka et al., 2007; Taft, 1994).
Another intriguing question concerns the locus of morphological processes, that is,
whether morphological structure operates at the prelexical level that guides lexical access or
at the lexical level itself where meaning is represented. In the work presented here, we
specifically consider this latter issue. Hence, we will discuss models and findings regarding
the level of representation in more detail, starting with models that assume morphological
processes to occur at the prelexical level.
Prelexical Processing and Representation
Early work on prelexical processing has often used nonwords to explore whether
lexical access is gained via morphemic or whole-word units. In a seminal study, Taft and
Forster (1975) compared lexical decision latencies to nonwords like juvenate and pertoire, the
former holding a real stem, the latter a pseudostem, depending on whether or not the prefixed
letter cluster re added to the overall meaning of the word and thus represented a real prefix, as
in rejuvenate, or a pseudoprefix, as in repertoire. Real stems (juvenate) were harder to
classify as nonwords than pseudostems (pertoire). Similarly, illegal combinations of a prefix
4
Lexical Representation of Complex Verbs
with a real stem (dejuvenate) were harder to classify as nonwords than those with a
pseudostem (depertoire).
Taft and colleagues (Taft, 1979, 1994, 2004; Taft & Forster, 1975; Taft, Hambly, &
Kinoshita, 1986) concluded in both their affix-stripping model (1975) and the later interactive
activation model (1994) that words are automatically decomposed into prefix and stem (hence
‘obligatory decomposition’ account), and that stems (in later models also prefixes) are used as
lexical access code to search for lexical entries. In the case of pertoire, a nonword response
can be given as soon as the lexical search for the stem is unsuccessful. In contrast, the real
stem juvenate provides access to the lexical entry rejuvenate. However, additional
information about the stem needs to be consulted in order to decide whether the stem
represents a word by its own or whether it may combine with a particular prefix. This latter
information is gathered at the lexical entry level. The obligatory decomposition account thus
holds that the prelexical level is determined by morphological structure, whereas the lexical
level comprises whole words.
Most data gathered in recent years in support of prelexical morphological
decomposition come from masked priming experiments. In this technique (Forster, Mohan, &
Hector, 2003), a briefly presented visual prime (usually below 50 ms) is sandwiched between
the presentation of a forward mask and the presentation of the visual target that also functions
as backward mask. The extremely short exposure duration together with the masking
procedure make the prime unavailable for conscious report, so that its effects on the target are
assumed to reflect the automatic preactivation of shared representations.
Using the masked priming technique in combination with English or French lexical
decision tasks, morphological priming was obtained due to any kind of morphological relation
between prime and target (Longtin, Segui, & Hallé, 2003; Rastle, Davis, Marslen-Wilson, &
Tyler, 2000; Rastle, Davis, & New, 2004). Priming of a stem like depart was obtained by true
morphological derivations, be they transparently or opaquely related with it, as by departure
5
Lexical Representation of Complex Verbs
and department, respectively. However, also pseudoderivations, that is, words of apparent
morphological complexity like corner (er occurs as suffix in other English words but is a
pseudosuffix in the word corner) facilitated the recognition of corn. In contrast, words like
brothel that did not comprise a suffix-like ending did not facilitate the recognition of broth
(Longtin et al., 2003; Rastle et al., 2000; Rastle et al., 2004). The difference between the latter
two conditions—priming due to orthographic similarity with morpheme-like but not with
nonmorpheme-like endings—was henceforth called the ‘corner – corn effect’ and taken to
indicate that any morpheme-like ending induces the segmentation process. Most recently,
McCormick et al. (2009) further showed that this early segmentation process may as well be
generalized to nonwords that comprise a stem and an affix. However, other studies using
similar primes and priming conditions, namely masked priming at SOAs 32 ms to 66 ms,
observed facilitation due to pure orthographic similarity without morpheme-like endings, that
is, primes like tinsel facilitated the recognition of a target like tin (Diependaele, Sandra, &
Grainger, 2005; Feldman, 2000; Giraudo & Grainger, 2000; Rastle et al., 2000; Segui &
Grainger, 1990).
Together, these findings were interpreted to indicate that morphological segmentation
operates in early visual word recognition on a purely orthographic basis—not only
independent of semantic relatedness but also independent of true morphological relatedness.
Most importantly, yet, it is generally assumed that this segmentation process ceases to affect
the word recognition process as soon as the morphemic (or morpheme-like) constituents are
semantically integrated, as is the case when primes are consciously perceived under long
exposure durations (Rastle et al., 2000). For example, Meunier and Longtin (2007) argued for
a recognition process in two stages: in the early prelexical stage, complex words are
decomposed on an orthographic basis, followed by the second lexical stage, in which the
decomposed constituents are reappraised for semantic and syntactic information. If the words
6
Lexical Representation of Complex Verbs
were decomposed in an incorrect way, as is the case with semantically opaque words, this is
corrected in the second stage, the lexical level.
Other recent studies using the masked priming technique have found that
morphological effects are influenced by semantics and thus suggested a morpho-semantic
system in addition to the morpho-orthographic system. This is either assumed to be
incorporated at the lemma level (Taft & Kougious, 2004; Taft & Nguyen-Hoan, 2010), or
both systems are assumed to be independent processing systems that work in early visual
word recognition (Diependaele et al., 2005).
Be it one or two systems, the so-called ‘corner – corn effect’ has dominated research
over the past years by shifting the focus from lexical processing to prelexical processing
under early visual word recognition. One reason may be that the prevailing models—
predominantly based on English and French findings—assume similar lexical representations,
as will be discussed below. We now turn to models that explicitly explored the lexical
representation of complex words.
Lexical Processing and Representation
In contrast to masked priming, overt priming is regarded as tapping into lexical
processing and representation. The prime is presented either auditorily or visually at long
exposure durations (230 ms or 250 ms) and is thus consciously perceived, so that the meaning
of the word is retrieved.
Hence, most researchers consider overt priming to trigger morphological
decomposition as a high-level process, either following whole word access, as assumed by
supralexical accounts (Giraudo & Grainger, 2000), or constrained by semantic knowledge. In
any case, semantic relatedness between prime and target has been regarded as the
precondition for the occurrence of morphological priming. For instance, when prime and
target were both morphologically and semantically related, English and French prefixed
7
Lexical Representation of Complex Verbs
derivations like distrust primed their semantically related stems like trust, as well as other
prefixed or suffixed derivations like entrust or trustful. Also suffixed derivations like
production and productivité or confession and confessor, respectively, primed each other
(Feldman & Larabee, 2001; Feldman, Soltano, Pastizzo, & Francis, 2004; Marslen-Wilson et
al., 1994; Meunier & Segui, 2002). However, even though stems like confess were primed by
semantically transparent derivations like confessor, the recognition of stems like success was
not facilitated by morphologically related but semantically opaque derivations like successor,
this latter finding being replicated under both cross-modal (Longtin et al., 2003; MarslenWilson et al., 1994) and visual priming with long exposure durations at 230 ms or 250 ms
(Feldman & Soltano, 1999; Feldman et al., 2004; Rastle et al., 2000).
In summary, under experimental conditions, in which the prime is consciously
perceived, morphological priming was obtained only if prime and target were also
semantically related.
Interestingly, all accounts that are based on English, French, or Dutch—regardless of
whether they hold that morphological decomposition precedes or follows whole word access
at the lexical level—agree that lexical representation depends on semantic compositionality:
semantically transparent words, whose meaning can be derived from the meaning of their
parts, possess a lexical entry that corresponds to their base. A word like departure will thus be
represented as the base {depart} and the suffix {ure}. In contrast, semantically opaque words
like department, whose meaning cannot be derived from the meaning of the parts, must be
represented in their full form {department}, which in turn may be affixed with other suffixes,
such as {al} (Diependaele et al., 2005; Marlsen-Wilson, Bozic, & Randall, 2008; Taft &
Nguyen-Hoan, 2010).
However, this concept of lexical representation is in stark contrast to our previous
findings in German that morphologically related prefixed verbs primed their base to the same
extent regardless of whether they were both semantically and morphologically related
8
Lexical Representation of Complex Verbs
(mitkommen – kommen, ‘come along – come’) or only morphologically related (umkommen –
kommen, ‘perish – come’). These findings in German suggest – unlike those in English and
French – that a complex verb like depart is not only decomposed into {de} and {part} during
early visual word recognition but is also lexically represented in such a way (Smolka,
Komlósi, & Rösler, 2009). Given that there are hardly any studies on German, we seek to
investigate this issue more fully. German is an interesting test language to explore lexical
representations, since it belongs to the same language family as English and French do and
thus possesses concatenative morphology of all Indo-European languages, but it differs from
many Indo-European languages regarding its morphological richness and productivity, as is
demonstrated by the productivity of German complex verbs.
The test case for shared lexical representations are word pairs that are morphologically
related but semantically unrelated. Such pairs are plentiful among German complex verbs. We
thus turn to discuss their main characteristics.
German Complex Verbs
German complex verbs are very productive and frequently used in standard German
and are thus a particularly useful means by which the effects of meaning relatedness to the
same base verb can be studied. For instance, derivations of the base tragen (‘carry’), like
hintragen (‘carry to’), forttragen (‘carry away’), zurücktragen (‘carry back’), abtragen (‘carry
off’), auftragen (‘apply’), vertragen (‘get along’), ertragen (‘suffer’), alter the meaning
relatedness from fully transparent to fully opaque with respect to the base.
In German, word formation of verbs involves few suffixes and is rather dominated by
prefixation (Eisenberg, 2004). The process of word formation via prefixation is very
productive and analogous to that of suffixation. For example, the prefix ent- derives not only
verbs from verbs (laufen, ‘run’ – entlaufen, ‘run away’), but also verbs from nouns (Staub,
‘dust’ – entstauben, ‘dust off’).
9
Lexical Representation of Complex Verbs
The linguistic literature (Eisenberg, 2004; Fleischer & Barz, 1992; Olsen, 1996)
distinguishes two word formations: prefix verbs and particle verbs. Both consist of a verbal
root and either a verbal prefix or a particle. According to Eisenberg (2004), the core verbal
prefixes are bound morphemes that do not necessarily possess meaning, be-, ent-, er-, ver-,
zer-, and five prepositional prefixes that are homonyms with particles and thus have lexical
meaning, durch (‘through’), hinter (‘behind’), über (‘above’), um (‘around’), unter (‘under’).
Verbal prefixes are inseparable from their base in finite forms (Truppen umstellen das
Gebäude, ‘Troops surround the building’) and substitute the typical ge- prefix in participle
formation (umstellt, ‘surrounded’).
In contrast to the limited number of verbal prefixes, the number of verbal particles,
such as an (‘at’), auf (‘on’), ab (‘off’), ein (‘into’), nach (‘after’), vor (‘before’), mit (‘with’),
um (‘around’), zurück (‘back’), and the like, is in the seventies, since they are free morphemes
in the function of prepositions or adverbs. They are separated from the verb stem in finite
forms (Sie stellen die Stühle um, ‘They relocate the chairs’) and enclose the prefix ge- in
participle formation (umgestellt, ‘relocated’). Prefix and particle verbs also differ from a
prosodic perspective: the former carry the main word stress on the base (umSTELLEN), the
latter on the particle (UMstellen).
In spite of these prosodic and morphosyntactic differences, prefix and particle verbs
share similar semantic properties. Both can be transparently and opaquely related to the
meaning of their base. For example, the particle an (‘at’) only slightly alters the meaning of
the base führen (‘guide’) in the derivation anführen (‘lead’), but radically does so with respect
to the base schicken (‘send’) in the opaque derivation anschicken (‘get ready’). Similarly, the
prefix ver- produces the transparent derivation verschicken (‘mail’) as well as the opaque
derivation verführen (‘seduce’).
Besides some of these distinctive prosodic and morphosyntactic characteristics,
though, the distinction between prefix and particle verbs is not unequivocal among linguists
10
Lexical Representation of Complex Verbs
and rather suggests that they share some formal and functional similarities. Moreover, effects
of prefix and particle verbs were found to be alike in both German (Drews, Zwitserlood, &
Neuwinger, 2000) and Dutch (Schriefers, Zwitserlood, & Roelofs, 1991). We therefore do not
differentiate between these types in the present study and, henceforth, refer to them as
‘complex verbs’ or ‘derived verbs’.
In the following experiments, we sought to explore the lexical representation of such
complex verbs: In all experiments, we measured priming by purely semantically related,
semantically and morphologically related, purely morphologically related complex verbs
relative to either unrelated or form-related complex verbs. Experiment 1 was closely modeled
after our previous experiments in Smolka et al. (2009) and thus applied visual prime
presentation at 300 ms SOA and thus tested intra-modal lexical representation, Experiment 2
used cross-modal priming with auditory primes and visual targets and thus tested modality
abstract lexical representation, and Experiment 3 aimed at differentiating between modality
(specific vs. abstract) and information delivery (simultaneous vs. sequential) and combined
the visual with sequential prime presentation by presenting visual primes in sequential word
presentation.
Experiment 1
The purpose of the first experiment was, first, to test whether the findings reported in
Smolka et al. (2009) could be replicated with improved materials and design, and second, to
bridge between those and the forthcoming (cross-modal and visual sequential) experiments.
Using long visual prime-exposure durations (of 300 ms SOA), Smolka et al. (2009)
tested the effect of semantic transparency on the size of morphological priming relative to an
unrelated as well as an identity condition. Morphologically related primes were complex verb
derivations that were either transparently or opaquely related to their base verb that served as
target. Contrary to the view that semantic meaning presides over conscious word processing,
11
Lexical Representation of Complex Verbs
they found that the magnitude of morphological priming was not modulated by semantic
transparency. Instead, morphologically related but semantically opaque verbs (umkommen,
‘perish’) primed their base (kommen, ‘come’) to the same extent as did semantically
transparent (mitkommen, ‘come along’) or identity (kommen, ‘come’) primes, while
semantically associated verbs (nahen, ‘approach’ – kommen, ‘come’) did not induce priming.
Using the same prime-exposure durations, the second experiment tested whether the
morphological effects could be reduced to the form overlap between complex verbs and their
base. However, orthographically similar primes (kämmen, ‘comb’) hindered target recognition
(kommen, ‘come’), whereas, again, semantically opaque derivations facilitated the recognition
of the base to the same extent as did transparent derivations. In the same experiment,
facilitation due to semantic associations was significant but nevertheless weaker in magnitude
that due to morphological relations.
To tease out semantic effects, the third experiment expanded prime-exposure durations
to 1000 ms. At last, priming due to semantic associates was as strong as that due to
morphological relatedness, and a small semantic transparency effect showed in form of
accuracy (but not latency) data in favor of semantically transparent over opaque derivations.
On the whole, the three experiments displayed strong morphological facilitation that is (a) not
modulated by semantic transparency (with the exception of accuracy at extreme primeexposure durations), (b) stronger than semantic facilitation (at SOA 300), and (c) different
from (partly nonsignificant) form inhibition.
Even though these results are equivocal with regard to prelexical decomposition, they
are hard to explain within the prelexical framework assuming that morphological structure
affects only prelexical processing. In particular, morphological effects under long primeexposure durations contradict the notion of a prelexical morphological decomposition
mechanism that operates in early visual word recognition and disappears as soon as meaning
integration comes into play. Smolka and colleagues interpreted these data as indicating that, in
12
Lexical Representation of Complex Verbs
German, pure morphological effects last longer than usually observed in other Indo-European
languages, where morphological effects are restricted to early visual word recognition and
vanish once participants become aware of the prime, as substantiated under visual priming at
long SOAs in English (Feldman et al., 2004; Rastle et al., 2000; Raveh, 2002) and Serbian
(Feldman, Barac-Cikoja, & Kostić, 2002).
Smolka et al. (2009) assumed that the morphological richness of German determined
the occurrence of morphological effects (independent of semantic transparency). They
interpreted their results in the framework of an interactive-activation model similar to those of
distributed connectionist accounts (Plaut & Gonnerman, 2000; Raveh, 2002; Rueckl & Raveh,
1999). In this architecture, orthographic codes are mapped onto meaning codes, and
morphological regularities are assumed to emerge as the result of statistical regularities
between form and meaning overlap. The amount of such morphological regularities will
determine the performance of the system: the more form and meaning overlap (that is, the
higher morphological regularity) the more efficient the system. In a morphologically rich
system where mappings between form and meaning are straightforward, morphological
regularities will dominate language processing so that language processing (and thus also
priming effects) will become independent of semantic transparency (Plaut & Gonnerman,
2000). However, in contrast to connectionist models that assume morphemes to be indirectly
represented by means of hidden layers in between orthographic and semantic codes, we
argued for explicit morphemic representation.
The present study examined how morphological regularities are represented in lexical
memory, in particular, whether they are influenced by semantic and form relatedness.
Experiment 1 of the present study was closely modeled after Experiment 2a of the
previous study (Smolka et al., 2009) and was thus conducted with overt visual priming at long
prime exposure durations of 300 ms SOA. Second, to avoid word category effects, only verbs
were used as materials. Third, large amount of fillers were inserted to prevent expectancy and
13
Lexical Representation of Complex Verbs
strategic effects. Most importantly, we again applied a within-target manipulation which
allowed us to directly compare the effect of each type of prime on the same target. We further
used simple verbs as targets, like binden (‘bind’), and complex verbs as morphologically
related primes, hence, the conditions of semantically transparent (zubinden, ‘tie together’) and
opaque (entbinden, ‘deliver’) derivations were the same as those used in the original study.
Again, we compared the priming by semantically transparent derivations to that by purely
semantically related verbs, and measured priming relative to unrelated and form-related
conditions. Notwithstanding, the design of the present study improved on several
methodological aspects of the original research.
First, we matched the morphological structure of all critical primes. In the previous
study, unrelated and form controls, as well as semantically related verbs were simple verbs. In
the present study, all were complex verbs and thus had the same morphological structure (and
the same length and number of syllables) as the verbs in the morphologically related
conditions: Unrelated controls like abholzen (‘deforest’) were neither morphologically, nor
semantically nor form-related with the target, and form-related controls like abbilden
(‘depict’) possessed a base with the same onset or first syllable as that of the target like binden
(‘bind’), but were otherwise neither morphologically nor semantically related with it. Purely
semantically related primes like zuschnüren (‘tie together’) were neither morphologically nor
form-related with the target.
Moreover, to provide for a direct comparison between semantically transparent
derivations and purely semantically related verbs, we matched these two conditions with
respect to their meaning relatedness with the target like binden (‘bind’) and included in both
conditions only synonyms of the base, like zubinden (‘tie together’) and zuschnüren (‘tie
together’) in the morphological and semantic condition, respectively. Thus we actually
compared verbs in two synonym conditions, once with and once without a morphological
relation to the base.
14
Lexical Representation of Complex Verbs
Semantic priming has often been tested using pairs such as cello and violin, which are
different in many respects from the base targets used in the morphologically relevant
conditions like departure – depart or apartment – apart (Marslen-Wilson et al., 1994; Rastle
et al., 2000). In our previous study (Smolka et al., 2009), we used semantically associated
verbs as primes like nahen – kommen (‘approach – come’). Notwithstanding, the semanticsynonym condition of the present study excels previously used semantic conditions.
In summary, the primes in all conditions were complex verbs with the same
morphological structure and were thus (a) of the same word category, and (b) closely matched
on distributional variables like lemma frequency, number of syllables and letters. They
differed only with respect to the morphological, semantic, or form-relatedness with the target.
Prime conditions are exemplified in Table 1; all critical items are listed in the Appendix.
We further improved the experimental design to avoid possible episodic memory
effects. In contrast to the original study, where a participant saw all prime-target
combinations, in the present study, each participant saw a target base only once.1
In summary, we carried out an overt visual priming experiment to test whether or not
complex verbs in German are processed by means of their morphological base.
If lexical representation in German is organized according to meaning computability
as it is in other Indo-European languages (Marlsen-Wilson et al., 2008; Marslen-Wilson et al.,
1994; Rastle et al., 2000; Rastle et al., 2004; Taft & Nguyen-Hoan, 2010), semantically
transparent derivations will share a lexical entry with their base, whereas semantically opaque
derivations will have their own representation in lexical memory. Accordingly, the former
will induce priming to their base, but not the latter. If, however, our assumption holds that
lexical representations in German comprise the base (Smolka et al., 2009), we should obtain
priming by both semantically transparent and opaque derivations. To anticipate the results,
Experiment 1 revealed strong morphological effects, and no effect of semantic transparency.
15
Lexical Representation of Complex Verbs
Methods Experiment 1
Participants
Sixty-four students of the University of Konstanz were paid for their participation. All
were monolingual native speakers of German, were not dyslexic, and reported normal or
corrected-to-normal vision.
Materials
Critical stimuli. Forty base verbs were selected from the CELEX German lexical
database (Baayen, Piepenbrock, & van Rijn, 1993). All were monomorphemic, had no prefix,
and took only one meaning.
Each base verb was combined with five primes, resulting in 200 prime-target pairs (all
critical items are listed in the Appendix). In contrast to our previous experiments (Smolka et
al., 2009), all conditions comprised complex verbs, that is, prefix or particle verbs, that
differed in their relation with the base verb. Derivations of the base were, by definition,
morphologically and thus also form related with the target, all other prime conditions were
morphologically unrelated with the target. Specifically, prime-target relations were defined by
three factors: morphological, semantic, and form relatedness with the base verb (e.g., binden,
‘bind’; see also Table 1): (a) M+S+F+, semantically transparent derivations of the base (e.g.,
zubinden, ‘tie’), importantly, they were synonyms of the base like the primes in condition (c)
below; (b) M+S–F+, semantically opaque derivations of the base (e.g., entbinden, ‘deliver’);
(c) M–S+F–, morphologically unrelated synonyms of the base verb (e.g., zuschnüren, ‘tie’)
were selected by means of the online synonym dictionaries http://www.canoo.net/ and
http://synonyme.woxikon.de/; (d) M–S–F+, form-related verbs (e.g., abbilden, ‘depict’) were
unrelated complex verbs with bases whose onset or first syllable matched that of the target
base and differed from the target by a single grapheme (1 or 2 letters); one complex verb was
16
Lexical Representation of Complex Verbs
put in the preterit to create form-relatedness; and (e) M–S–F–, unrelated verbs (e.g., abholzen,
‘deforest’) were neither morphologically nor semantically nor form related.
__________________
Table 1 about here
__________________
Note that our within-target design of the present study – by definition – controls for
target frequency, target length, target neighborhood size, and target family size across
conditions.
Table 1 provides all prime characteristics. All morphological derivations (except for a
semantically opaque one) were less frequent than their base. Primes across conditions were
matched on lemma frequency according to CELEX (Baayen et al., 1993). A one-way
ANOVA conducted on lemma frequencies indicated that there was no difference between
prime conditions, F < 1. Primes were further matched on number of letters and syllables. To
control for length effects, all but six prefixes consisted of a single syllable. The critical set of
200 prime-target pairs was selected from a pool of verb pairs that were subjected to the
semantic association test described below.
Semantic association test. A semantic association test was conducted to establish the
relatedness between primes and targets for all prime conditions. Sixty-six candidate verbs
were selected from the CELEX database (Baayen et al., 1993), for each of which two
synonyms, two semantically transparent derivations, two semantically opaque derivations,
and two form related verbs were distributed across eight lists, so that each list contained only
one prime for the same target verb. The verb intended as the prime preceded the target, and
both were presented in citation form (stem/-en). In total, 528 prime-target pairs were tested.
Hundred-thirteen participants who did not take part in the experiments proper rated the
relationship between the verbs of each pair on a 7-point scale from completely unrelated (1) to
highly related (7).
17
Lexical Representation of Complex Verbs
The following criteria determined whether or not a base verb and its primes were
included in the critical set: The mean ratings for a semantically-related pair (M+S+F+ and M–
S+F–) had to be higher than 4, and those for a semantically unrelated pair (M+S–F+ and M–
S–F+) lower than 3. Mean ratings of the final set were 5.4 for synonyms (M–S+F–), 5.6 for
semantically transparent derivations (M+S+F+), 2.2 for semantically opaque derivations
M+S–F+), and 1.3 for form-related pairs (M–S–F+).
A one-way ANOVA was performed on mean ratings with items as random variables.
The repeated measures factor prime type was highly significant, F(3, 117) = 633.75, p <
.0001. Scheffé post hoc comparisons indicated that the ratings for synonyms (M–S+F–) and
semantically transparent derivations (M+S+F+) did not differ from each other, but were
significantly higher than those for semantically opaque derivations (M+S–F+) or form-related
pairs (M–S–F+). The difference between the latter two was significant as well.
Fillers. In order to prevent strategic effects, a total of 264 prime-target pairs was added
as fillers. All had complex verbs as primes, 112 had verbs and 152 had pseudoverbs as targets.
Pseudoverbs were constructed by exchanging one or two letters in real verbs, while
preserving the phonotactic constraints of German. Forty of the pseudoverb targets were
closely matched to the critical verb targets by keeping the onset of the verbs’ first syllable
(e.g. binden – binken). Corresponding to the critical set, twenty-four of these pseudoverbs
were preceded by a form-related prime (e.g., umwerben – wersen). These form-related primepseudoverb pairs were inserted to deter participants from ‘automatic’ ‘word’ responses on
seeing a form-related prime-target pair.
Overall, the large amount of fillers reduced the proportion of (a) critical prime-target
pairs to 13%, (b) semantically related prime-target pairs to 5%, and (c) form-related (verb and
pseudoverb) pairs to 16% of the whole material set. All filler items differed from those of the
critical set.
18
Lexical Representation of Complex Verbs
Throughout the experiment, all primes and targets (except for one form-related prime,
see above) were presented in the infinitive (stem/-en), which is also the citation form in
German.
Apparatus
Stimuli were presented on a 18.1” monitor, connected to an IBM-compatible AMD
Atlon 1.4 GHz personal computer. Stimulus presentation and data collection were controlled
by the Presentation software developed by Neurobehavioral Systems (http://nbs.neurobs.com/). Response latencies were recorded from the left and right buttons of a push-button
box.
Procedure
Primes of the same target were rotated over five lists according to a Latin Square
design, with one of the five prime-target combinations in each list. Participants received only
one experimental list and therefore participated in all priming conditions (eight items per
condition) but saw each target word only once. Each list was divided into four blocks, each
block containing the same amount of stimuli per condition. The form-related primepseudoverb pairs and the remaining filler pairs were evenly allocated to the blocks. In total, an
experimental session comprised 304 prime-target pairs presented in four experimental blocks,
with 76 prime-target pairs per block. Within blocks, prime-target pairs were randomized
separately for each participant. Sixteen additional prime-target pairs were used as practice
trials.
Participants were tested individually, seated at a viewing distance of about 60 cm from
the screen. Stimuli were presented in white Sans-Serif letters on a black background. To make
primes and targets physically distinct stimuli, primes were presented in uppercase letters,
point 26, 20 points above the center of the screen, targets were presented centrally in
lowercase letters, point 30.
19
Lexical Representation of Complex Verbs
Each trial started with a fixation cross in the center of the screen for 200 ms. This was
followed by the presentation of the prime for 200 ms, followed by an offset (=blank screen)
for 100 ms, resulting in a stimulus onset asynchrony (SOA) of 300 ms. After the offset, the
target immediately followed and remained on the screen until a participant’s response. The
intertrial interval was 1500 ms. Participants were instructed to make lexical decisions to the
targets, as fast and as accurately as possible. ‘Word’ responses were given with the index
finger of the dominant hand, ‘pseudoword’ responses with the subordinate hand. Feedback
was given on both correct (‘richtig’) and incorrect (‘falsch’) responses during the practice
session, and on incorrect responses during the experimental session. The experiment lasted for
about 30 minutes. Participants self-administered the breaks between blocks, and took at least
one longer break.
Results Experiment 1
Analyses were performed on latency and accuracy data with participants (F1) and
items (F2) as random variables. Mean response latencies were calculated for correct
responses. Only responses times (RTs) within 3 standard deviations from a participant’s mean
were included; and three participants whose error rates (>16%) exceeded three standard
deviations of the overall error mean (5.6%) were removed, so that the data of 61 participants
were included in the analyses. There was no speed-accuracy trade-off (RT and error means
over participants were not correlated, r = .00977, p = .8793).
_________________________
Table 2 about here
________________________
A one-way ANOVA with repeated measures factor prime type (M–S+F–/ M+S+F+/
M+S–F+/ M–S–F+/ M–S–F–) was run on latency and accuracy data. RT and error means are
provided in Table 2. The effect of prime type was highly significant for RT data, F1(4, 240) =
20
Lexical Representation of Complex Verbs
6.51, p < .0001, F2(4, 156) = 5.22, p = .0006. Given that response accuracy did not reveal any
significant effects, in the following only effects with response latency will be reported.
A set of planned comparisons was performed to examine the differences among the
prime conditions. These priming effects are depicted in the mid panel of Figure 1. Relative to
the unrelated condition, neither synonyms (M–S+F–), F1(1, 240) = 1.18, p = .2785, F2(1, 156)
= 1.68, p = .1973, nor form-related primes (M–S–F+), F1(1, 240) = 1.13, p = .2894, F2 < 1,
differed significantly from the unrelated condition, even though they showed numerically
faster or slower responses, respectively.
In contrast, morphologically related primes significantly facilitated responses relative
to the unrelated condition. This was true for both semantically transparent (M+S+F+), F1(1,
240) = 10.15, p = .0016, F2(1, 156) = 8.35, p = .0044, and opaque derivations (M+S–F+),
F1(1, 240) = 7.73, p = .0059, F2(1, 156) = 6.90, p = .0095. Most importantly, the contrast
between the two conditions, namely that between 28 ms and 24 ms, was not significant (F1
and F2 < 1), indicating that morphological priming was unaffected by the semantic relation
between prime and target.
Given that form relatedness slightly (though not significantly) decelerated responses
relative to the unrelated condition, calculating priming against the form-related condition
enhanced the effects even more: priming of semantically transparent derivations, F1(1, 240) =
18.05, p < .0001, F2(1, 156) = 13.79, p = .0003, and semantically opaque derivations, F1(1,
240) = 14.76, p = .0002, F2(1, 156) = 11.91, p = .0007. Relative to the form control also
synonym priming reached significance, F1(1, 240) = 4.61, p = .0327, F2(1, 156) = 4.49, p =
.0358. Note, however, that this comparison is not straightforward, since the two conditions
differ in both semantic and form aspects, M–S–F+ versus M–S+F–.
The contrast between the two semantically-related prime-target pairs—synonyms and
semantically transparent derivations—was significant in the analysis by participants, F1(1,
240) = 4.41, p = .0368, though not by items, F2(1, 156) = 2.54, p = .1127, and indicated that
21
Lexical Representation of Complex Verbs
morphologically related primes more strongly facilitated responses than did semantically
related primes.
__________________
Figure 1 about here
__________________
Discussion of Results
In Experiment 1, we examined whether the factors morphological, semantic, and form
relatedness affect target recognition under visual prime presentations with long exposure
durations. The results were straightforward: lexical decision latencies were facilitated by
morphological relatedness, but not by the pure meaning or form relatedness between prime
and target. Most importantly, morphological facilitation was not affected by semantic
transparency. In this respect, we replicated our previous findings (Exp. 1 and 2 in Smolka, et
al., 2009) under the same prime conditions, but in different labs, with different materials,
improved design and controls. We thus believe that we have a strong case in point with these
data. In contrast to morphological priming, semantic priming was not significant. A
potentially critical difference between our primes and those used by both our previous and
other studies relates to the fact that we used synonyms of the base rather than semantic
associates. However, also semantically transparent derivations were synonyms of the base
verbs and did induce priming to the base. Moreover, even studies that used semantically
associated verbs in German did not find significant facilitation (Smolka, Khader, Wiese,
Zwitserlood, & Rösler, 2010; Smolka et al., 2009; Zwitserlood, Drews, Bolwiender, &
Neuwinger, 1996). As to why German verbs do not show stable semantic facilitation effects is
not clear. Anyways, the fact that priming was greater in the morphological than in the
semantic-synonym condition provides strong evidence that morphological priming is robust
and independent of meaning relatedness.
22
Lexical Representation of Complex Verbs
Form inhibition did not reach significance. Nevertheless, response latencies were
numerically decelerated, which concurs with the previous finding that the orthographic
similarity of a clearly visible prime (at long SOAs of 230 ms or more) hinders target
recognition in English (Feldman, 2000; Pastizzo & Feldman, 2002; Rastle et al., 2000), Dutch
(Drews & Zwitserlood, 1995; Segui & Grainger, 1990), German (Drews & Zwitserlood,
1995; Smolka et al., 2009), French (Grainger, 1990; Segui & Grainger, 1990), and Serbian
(Feldman & Andjelkovic, 1992). In summary, we may conclude that complex verbs in
German show strong morphological effects that are independent of semantic and form
relatedness.
According to models on visual word recognition, this provides evidence for shared
lexical representations between morphological derivations and their base, even for
semantically opaque ones. That is, we may assume that both the semantically transparent
derivation mitkommen and the semantically opaque derivation umkommen are lexically
represented as {kommen} and the prefixes {mit} and {um}.
Experiment 2
In Experiment 1, we presented visual primes at long exposure durations of 300 ms and
obtained morphological effects without any difference between semantically transparent and
opaque derivations. However, given that primes and targets were presented within the same
modality, it is possible that these effects are restricted to modality-specific representations and
to not extend to abstract representations. Experiment 2 thus examined whether the
morphological effects obtained in Experiment 1 can be substantiated under cross-modal
(auditory-visual) priming. Specifically, we tested whether complex verbs in German possess
not only modality-specific but also abstract lexical representation.
Abstract lexical representations are assumed to be traced under overt cross-modal
priming conditions (Marslen-Wilson et al., 1994), where an auditory (and thus consciously
23
Lexical Representation of Complex Verbs
perceived) prime precedes a visually presented target, or where an overt visual prime precedes
an auditory target. Shared lexical representations thus need to be modality-independent.
Nevertheless, some authors (Feldman & Larabee, 2001) doubt that cross-modal priming is
lexically neutral because of the asymmetrical patterns of morphological facilitation and
inhibition obtained under auditory-visual and visual-auditory priming conditions. In the
following, mostly the former type (auditory-visual) will be referred to with the term crossmodal priming.
So far, data gathered under cross-modal priming in English (Marslen-Wilson et al.,
1994), French (Longtin et al., 2003), and Polish (Reid & Marslen-Wilson, 2000) support the
notion that semantic relatedness and compositionality is a prerequisite for shared lexical
representations. Suffixed derivations like production and productivité primed each other in
French (Meunier & Segui, 2002), though not in English (Feldman & Larabee, 2001; Feldman
et al., 2004; Marslen-Wilson et al., 1994). Both English and French prefixed derivations like
distrust, that are most similar to those of the present study, primed semantically related stems
like trust, as well as other prefixed or suffixed derivations like entrust or trustful (Feldman &
Larabee, 2001; Marslen-Wilson et al., 1994; Meunier & Segui, 2002). Most importantly,
whenever derivations were semantically transparent they produced priming to their base, but
not when they were semantically opaque, that is, confessor primed confess but successor did
not prime success (Marslen-Wilson et al., 1994).
Using cross-modal priming, only studies in Arabic obtained equivalent morphological
effects between semantically related and unrelated derivations (Boudelaa & Marslen-Wilson,
2004a, 2004b). Interestingly, even in Hebrew, another Semitic language, morphological
priming was strongly influenced by semantic relatedness. Though semantically unrelated
primes facilitated the recognition of morphologically related targets, this facilitation was
significantly less than that induced by morphologically and semantically related primes
(Frost, Deutsch, Gilboa, Tannenbaum, & Marslen-Wilson, 2000).
24
Lexical Representation of Complex Verbs
In summary, research on morphological processing, so far, has recognized that lexical
organization is guided by morphological structure in Semitic languages and by semantic
compositionality in Indo-European languages. However, Experiment 1 of the present study as
well as previous experiments (Drews et al., 2000; Smolka et al., 2009) provide evidence that
lexical representation not only in Arabic but also in German is guided by morphological
structure. To explore whether morphological structure of German complex verbs is
represented via modality-specific or modality-independent (e.g. lexical, semantic,
phonological) representations, we carried out a cross-modal priming experiment (with
auditory primes preceding visually presented targets).
Several studies have shown processing asymmetries that support the assumption of
both modality-specific and modality-independent subprocesses for lexical representation (e.g.
(Feldman & Larabee, 2001; Grainger & Ferrand, 1994). For example, in the framework of a
bi-modal interactive activation model, Grainger and Ferrand (1994) suggested parallel routes
for visual and auditory word recognition that are heavily interconnected both at the prelexical
and lexical level.
Although made on different grounds, sublexical, lexical, and supralexical accounts
predict that—under overt cross-modal priming conditions—morphological priming should
surface only if primes and targets share a semantic relationship: In the sublexical account, the
morphological structure of a word is relevant for the blind parsing routine at the prelexical
level, but is not relevant when meaning integration comes into play under overt priming
conditions. Hence, even though morphologically related but semantically unrelated words
share the same prelexical decomposition units, semantically opaque derivations should not
prime their base, since they do not share the same lexical representation.
In the supralexical account, morphological structure is retrieved through a lexical
representation. Hence, (cross-modal) morphological priming should only be obtained with
25
Lexical Representation of Complex Verbs
primes and targets that have a semantically transparent relationship, since only these will
share representations at the morpheme level (Diependaele et al., 2005).
If, however, complex verbs in German share some lexical representation, we should
obtain priming—at least to some degree—by semantically opaque derivations. To anticipate
the results, Experiment 2 indeed showed morphological priming, though it was modulated by
semantic transparency.
Methods Experiment 2
Participants
Ninety-three students at the University of Konstanz, who did not take part in
Experiment 1, were paid for their participation. All were native speakers of German and
reported normal or corrected-to-normal vision, and no hearing impairment.
Materials
All verb stimuli and prime conditions of Experiment 1 were used except for some
form-related primes (M–S–F+). Seven orthographically but not phonologically related primes
were exchanged by complex verbs whose base had the same onset (and the same vowel
quality) as the target, but a different phoneme in the rhyme (for exchanged items see
Appendix). Allocation of experimental items (with the exchanged form-related primes) to five
lists was the same as in Experiment 1; each participant saw one list, with all conditions and
eight items per condition.
Apparatus
Recording of auditory stimuli. A female (native speaker of German) speech-therapist
was recorded for the primes. Recording took place in a sound-attenuated cabin by means of a
digital audio recorder (Tascam HD-P2; sampling rate 44.1kHz, 16Bit, mono). To avoid
26
Lexical Representation of Complex Verbs
intonation effects, each word was recorded twice in randomized order. Words were
segmented and normalized at 95% of the maximal amplitude.
Visual stimuli were presented on a 19” monitor IPS-panel, connected to an IBMcompatible Dual Core 2 personal computer. Auditory primes were presented via headphones
Sennheiser (HD25II 70Ω). Stimulus presentation and data collection were controlled by the
Presentation software developed by Neurobehavioral Systems (http://nbs.neuro-bs.com/).
Response latencies were recorded from the left and right buttons of a push-button box.
Procedure
Besides the stimulus presentation, the procedure was the same as in Experiment 1.
Each trial started with a fixation cross in the center of the screen. After 1000 ms, the auditory
prime was presented via headphones. The fixation cross remained on the screen until 100 ms
before the end of the auditory prime, followed by an offset (blank screen) for 100 ms. The
target appeared immediately at the offset of the auditory prime in the center of the screen (in
lowercase, white Sans Serif letters, point 30, on a black background), and remained on the
screen until a participant made a lexical decision (‘word’ responses were made with the
dominant, ‘pseudoword’ responses with the subordinate hand). The intertrial interval was
1500 ms.
Results Experiment 2
Analyses were performed on latency and accuracy data with participants (F1) and
items (F2) as random variables. Mean response latencies were calculated for correct
responses. Only RTs within 3 standard deviations from a participant’s mean were included;
and three participants whose error rates (>16%) exceeded three standard deviations of the
overall error mean (5.3%) were removed, so that the data of 90 participants were included in
the analyses. There was no speed-accuracy trade-off (RT and error means over participants
were not correlated, r = .02809, p = .5953).
27
Lexical Representation of Complex Verbs
_________________________
Table 3 about here
________________________
A one-way ANOVA with repeated measures factor prime type (M–S+F–/ M+S+F+/
M+S–F+/ M–S–F+/ M–S–F–) was run on latency and accuracy data. RT and error means are
provided in Table 3. Again, the effect of prime type was highly significant for RT data, F1(4,
356) = 13.35, p < .0001, F2(4, 156) = 9.67, p < .0001, but not for accuracy data. Hence, only
effects with response latency will be reported.
Planned comparisons were performed to establish effects relative to the unrelated
condition. Priming effects are depicted in the left panel of Figure 1. As in Experiment 1,
synonyms (M–S+F–), even though they slightly speeded responses, did not induce significant
facilitation, F1(1, 356) = 1.80, p = .1808, F2(1, 156) = 1.18, p = .2796. In contrast, formrelated primes significantly inhibited responses, F1(1, 356) = 5.56, p = .0189, F2(1, 156) =
4.05, p = .0458.
As in Experiment 1, morphological priming was significant, by both semantically
transparent (M+S+F+), F1(1, 356) = 21.00, p < .0001, F2(1, 156) = 14.94, p = .0002, and
opaque derivations (M+S–F+), F1(1, 356) = 5.15, p = .0239, F2(1, 156) = 4.11, p = .0442.
However, semantically transparent derivations induced faster responses than semantically
opaque derivations, F1(1, 356) = 5.35, p = .0212, F2(1, 156) = 3.37, p = .0682, indicating that
morphological priming was affected by the semantic relatedness between prime and target.
Given that form relatedness inhibited responses, semantic and morphological
relatedness were, of course, highly significant, when contrasted with form relatedness:
semantically transparent derivations, F1(1, 356) = 48.17, p < .0001, F2(1, 156) = 43.55, p <
.0001, and semantically opaque derivations, F1(1, 356) = 21.41, p < .0001, F2(1, 156) = 16.33,
p < .0001; synonyms, F1(1, 356) = 13.68, p = .0003, F2(1, 156) = 9.60, p = .0023.
28
Lexical Representation of Complex Verbs
Finally, also the contrast between synonym primes and semantically transparent
derivations was significant, F1(1, 356) = 10.51, p = .0013, F2(1, 156) = 7.73, p = .0061, and
indicated that morphological relatedness outreached semantic relatedness.
Discussion of Results
Experiment 2 examined whether German complex verbs share a modality-abstract
lexical representation with their base, or whether abstract representations are determined by
semantic compositionality, as is assumed in other Indo-European languages (Marslen-Wilson
et al., 1994).
Most importantly, and in contrast to the predictions of previous cross-modal priming
studies in Indo-European languages (Longtin et al., 2003; Marslen-Wilson et al., 1994), we
obtained morphological priming effects for both types of derivations, semantically transparent
as well as opaque ones. That is, not only zubinden (‘tie together’) but also entbinden
(‘deliver’) produced priming to their base binden (‘bind’). Most importantly, the cross-modal
testing conditions provide evidence that modality-independent representations comprise
morphological structure. This is at odds with models assuming that semantic relatedness and
compositionality is a prerequisite for shared lexical representations (Diependaele et al., 2005;
Marlsen-Wilson et al., 2008; Rastle et al., 2000).
Nevertheless, semantic transparency affected the size of priming. The 14 ms
facilitation due to entbinden (‘deliver’) was significantly smaller than the 27 ms facilitation
due to zubinden (‘tie together’), which suggests that semantic relatedness indeed influences
modality-independent processing. This pattern of results is in fact very similar to that obtained
in Hebrew under cross-modal priming (Frost et al., 2000), where morphologically related but
semantically unrelated primes induced less facilitation to the targets (20 ms; relative to control
primes) than did morphologically and semantically related primes (36 ms). We will discuss
modality-specific and modality-independent representations in more detail below.
29
Lexical Representation of Complex Verbs
Significant form inhibition confirmed that the morphological effects cannot be
attributed to mere form overlap between morphologically related primes and targets. Form
inhibition under cross-modal priming conditions further fits with previous findings that
phonological similarity hinders target recognition (Pastizzo & Feldman, 2002), though
phonological relatedness did not necessarily induce significant inhibition in several crossmodal experiments (Frost et al., 2000; Marslen-Wilson et al., 1994; Pastizzo & Feldman,
2002).
In spite of the semantic transparency effect that surfaced in the morphological prime
conditions, semantic priming due to synonyms was not significant and thus replicates the
nonsignificant semantic effect in Experiment 1. Interestingly, the priming patterns of the two
semantically related conditions (transparent derivations and synonyms) in Experiment 2
closely resembled those of the first experiment in spite of the modality difference across
experiments. Given that not only the semantically related primes but also the semantically
transparent derivations were synonyms of the base, we do not know what caused the lack of
semantic facilitation, which contrasts previous cross-modal priming studies that did observe
priming by synonyms (Marslen-Wilson et al., 1994). In spite or because of these findings—
form inhibition together with a lack of semantic facilitation on the one hand and
morphological facilitation on the other hand—we may altogether conclude that the
morphological priming effects did not emerge as the combination of form and meaning
overlap, as is suggested by the connectionist approaches (Plaut & Gonnerman, 2000).
Experiment 3
In summary, Experiments 1 and 2 yielded different morphological effects:
independent of semantic transparency in the former and affected by semantic transparency in
the latter.
30
Lexical Representation of Complex Verbs
This difference may have two origins, (a) modality (specific vs. independent) and (b)
information delivery (simultaneous vs. sequential presentation). Priming is usually assumed to
be affect by modality, with modality-specific effects being larger than those that are modality
independent. However, modality also intrinsically refers to different information delivery.
While visually presented words are usually presented at once, so that the whole stimulus may
be perceived and processed at once, auditory word recognition is sequential in nature,
phoneme-by-phoneme.
So far, studies have referred to modality-specific versus modality-independent features
of lexical representation. To our knowledge it has not been tested yet in how far lexical
representation is influenced by the simultaneous versus sequential perception of a stimulus,
that is, whether the word stimulus is perceived at once—as is usually the case with visual
word recognition (be it masked or unmasked) where all letters are presented and thus
perceived at once—or whether the word is perceived sequentially—as is the case with
auditory word recognition, where one phoneme is spoken and thus also heard one after the
other. Therefore, information modality and delivery are two confounded sources.
The purpose of Experiment 3 was to tear apart effects of modality-specificity versus
modality-independence and effects of simultaneous versus sequential word presentation. To
our knowledge, this is the first study that tries to disentangle these two possibilities. Prime
presentation in Experiment 3 thus combined the modality-specific features of Experiment 1
and the sequential prime presentation of Experiment 3. That is, visual primes were presented
sequentially, letter-by-letter.
If morphological effects are induced by a modality-specific component, semantically
transparent and opaque priming should be similar in magnitude, as was the case in
Experiment 1. If, however, morphological effects are a consequence of consecutive prime
perception, priming of semantically transparent derivations should be greater in magnitude
than that of opaque derivations, as was the case in Experiment 2.
31
Lexical Representation of Complex Verbs
Methods Experiment 3
Participants
Fifty-five previously untested students of the University of Konstanz were paid for
their participation. All were monolingual native speakers of German, and reported normal or
corrected-to-normal vision.
Materials
The material and design of this experiment were identical to those in Experiment 1.
Apparatus
The apparatus was identical to that used in Experiment 1.
Procedure
The procedure was identical to that of Experiment 1 except for the trial sequence,
since Experiment 3 simulated a sequential prime presentation in the visual mode. Each trial
started with a fixation cross in the center of the screen for 1000 ms. To adapt the duration of
the sequential visual presentation to that of the auditory presentation (mean duration 959 ms),
the duration of each auditory prime (in ms) was divided by the number of its letters. Averaged
across primes, this proportion was 102 ms/letter. We thus used an exposure-duration of 100
ms for the presentation of each letter in the visual mode. Specifically, each letter appeared for
100 ms (and remained on screen) before the next letter appeared. Prime length in letters was
counted, and the presentation of the prime’s first letter was shifted to the left in such a way
that the complete prime was seen centrally. After the offset of the last letter, the target
immediately followed and remained on the screen until a participant made a lexical decision
(‘word’ responses with the dominant, ‘pseudoword’ responses with the subordinate hand).
The intertrial interval was 1500 ms. The experiment lasted for about 30 minutes. Participants
self-administered the breaks in-between blocks.
32
Lexical Representation of Complex Verbs
Results Experiment 3
Analyses were performed on latency and accuracy data with participants (F1) and
items (F2) as random variables. Mean response latencies were calculated for correct
responses, mean error rate was 5.1%. Outliers were treated as in the previous experiments:
Only RTs within 3 standard deviations from a participant’s mean were included. There was no
speed-accuracy trade-off (RT and error means over participants were not correlated, r = .10467, p = .1213).
_________________________
Table 4 about here
________________________
As in the previous experiments, the effect of prime type (M–S+F–/ M+S+F+/ M+S–
F+/ M–S–F+/ M–S–F–) was highly significant for RT data, F1(4, 216) = 18.34, p < .0001,
F2(4, 156) = 9.61, p < .0001, but not for error data, so that only latency data will be reported.
RT and error means are provided in Table 4, the priming effects (measured relative to the
unrelated condition) are depicted in the right panel of Figure 1.
As in Experiments 1 and 2, a set of planned comparisons was performed to examine
the differences among the prime conditions. Relative to the unrelated condition, significant
facilitation effects were observed in both morphological conditions: semantically transparent
derivations (M+S+F+), F1(1, 216) = 30.39, p < .0001, F2(1, 156) = 16.21, p < .0001, and
semantically opaque derivations (M+S–F+), F1(1, 216) = 25.42, p < .0001, F2(1, 156) =
13.92, p = .0003. As in Experiment 1, there was no difference between the two (F1 and F2 <
1), indicating that semantic transparency did not affect responses. In addition (and different
from the previous experiments), also synonyms (M–S+F–) significantly facilitated responses
relative to the unrelated condition, F1(1, 216) = 19.50, p < .0001, F2(1, 156) = 10.02, p =
.0019, even to the same amount as did semantically transparent derivations, F1(1, 216) = 1.20,
p = .2738, F2 <1.
33
Lexical Representation of Complex Verbs
However, as in Experiment 1, form-related primes (M–S–F+) did not significantly
inhibit responses (F1 and F2 < 1). All contrasts with the form-related condition were highly
significant: semantically transparent derivations, F1(1, 216) = 41.40, p < .0001, F2(1, 156) =
21.44, p < .0001, semantically opaque derivations, F1(1, 216) = 35.57, p < .0001, F2(1, 156) =
18.79, p < .0001, and synonyms, F1(1, 216) = 28.48, p < .0001, F2(1, 156) = 14.21, p = .0002.
Discussion of Results
Experiment 3 was conducted to differentiate between effects of modality (specific vs.
independent) and effects of information delivery (simultaneous vs. sequential) and applied
sequential presentation of modality-specific stimuli.
We hypothesized that semantically transparent and opaque priming should be similar
in magnitude (as in visual Exp. 1), if morphological effects include a modality-specific
component. This was indeed the case.
Nevertheless, sequential prime presentation had its effect as well: overall, facilitation
effects were greater in magnitude than in the previous experiments, both for morphologically
related and for synonym primes, the latter producing significant priming for the first time in
this study. Most interestingly, these semantic effects were similar in magnitude to the
morphological ones.
Given that experimental conditions in which primes are consciously perceived are
supposed to increase semantic effects (Feldman & Prostko, 2002; Feldman et al., 2004; Raveh
& Rueckl, 2000), the question arises why semantic relatedness was not significant in the
previous experiments, but only under prime conditions that converge intra-modal and
sequential presentation. It seems as if complex verbs in German require especially long
processing to activate pure meaning relatedness. In this respect, the priming patterns in
Experiment 3 correspond to those in Experiment 2b in Smolka et al. (2009), where visual
primes were presented for 1000 ms, thus allowing for extremely long processing of the prime.
34
Lexical Representation of Complex Verbs
Form relatedness, again, hindered target recognition, but to an insignificant extent.
Since this correlates with the nonsignificant effect in (intra-modal) Experiment 1 and differs
from the significant inhibition effect in (cross-modal) Experiment 2, one could assume that
form inhibition is modality-independent. However, we are not sure whether such a strong
conclusion is warranted.
General discussion
The focus of this series of experiments was on the lexical representation of
morphologically complex verbs in German. Specifically, we explored the effects of
morphological, semantic, and form relatedness between prime and target and what they reveal
about the lexical representation of complex verbs. Effects of semantic transparency and
semantic association are difficult to detect in either the masked or the long-term priming task
but dominate findings in the immediate repetition priming paradigm (Feldman & Prostko,
2002; Feldman et al., 2004; Frost et al., 2000; Raveh & Rueckl, 2000). Since we aimed at
exploring lexical representation, we applied experimental conditions that are sensitive to
semantic processing (i.e., an immediate repetition priming paradigm) and examined lexical
decisions to visual targets following overt visual or auditory primes. To further examine
whether lexical representation is influenced by modality or information delivery, we
manipulated the presentation of the prime by displaying (a) visual primes with simultaneous
word exposure at SOA 300 ms in Experiment 1, (b) auditory primes (with sequential word
presentation) in Experiment 2, and (c) visual primes with sequential word presentation in
Experiment 3. Most importantly, all prime conditions represent priming methods that are used
to examine not processes of early visual word recognition but the representation of lexical
processes themselves.
We thus tested whether lexical representation is guided not only by semantic
compositionality but also by morphological structure. If morphological units affect not only
35
Lexical Representation of Complex Verbs
prelexical processing in early visual word recognition (as is assumed by most models) but
also lexical representation, morphological relatedness should affect target recognition
independent of meaning and form relatedness.
This is indeed what we found: in spite of a lack of semantic and form effects,
morphological relatedness facilitated target recognition under all prime presentation modes.
We thus provide evidence for the existence of a morphological dimension to lexical
organization that cannot be reduced to formal or semantic relations between primes and
targets. Most importantly, this indicates that morphological structure needs to be incorporated
in the modeling of lexical representation.
Before we discuss the morphological effects and their indication for lexical
representation in more detail, we will look at the effects of semantic and form relatedness on
target recognition, since most models assume that either semantic relatedness alone (MarslenWilson et al., 1994; Rastle et al., 2000; Taft & Nguyen-Hoan, 2010) or both semantic and
form relatedness strongly guide lexical representation (Gonnerman et al., 2007; Plaut &
Gonnerman, 2000).
Semantic Relatedness
Our results showed that semantic priming—in contrast to morphological priming—
was rather unstable across prime conditions: it showed almost nil facilitation under crossmodal and visual priming at SOA 300 (8 ms and 9 ms, respectively), but proved to be
significant (38 ms) under visual sequential word presentation. Besides, only under the latter
presentation mode was semantic facilitation (38 ms) statistically equivalent to morphological
facilitation (48 ms).
Given that both semantically associated primes and semantically transparent
derivations were synonyms of the base, the difference in effects between the two cannot be
attributed to any difference in meaning relatedness between primes and targets. Furthermore,
36
Lexical Representation of Complex Verbs
the semantic association test confirmed that purely semantically related primes (5.4) were
rated as being as strongly meaning connected with the base as semantically transparent
derivations (5.6).
So far, several recent studies on German verbs have indicated that semantic priming is
generally hard to detect, with the number of nonsignificant effects (see Exp. 1 and 2 of the
present study; Exp. 1 and 2 in Smolka, et al., 2009; Smolka, et al., 2010; Zwitserlood et al.,
1996) outnumbering the significant ones (see Exp. 3 in Smolka, et al., 2009). However, we
are not aware of a study in another language that particularly examined the priming effects
between semantically associated verbs. Most studies that contrasted semantic with
morphological priming effects used nouns like wealth and fortune as semantic associates.
Even if verbs occurred as associates, many of these, particularly in English, may as well be
interpreted as nouns, like circle and ring or gamble and risk (Marslen-Wilson et al., 1994;
Rastle et al., 2000; Taft & Nguyen-Hoan, 2010). Therefore, to assert whether the instable
activation between semantically associated verbs represents a German phenomenon or may be
generalized to other languages will need further exploration.
If we summarize our findings on semantic effects from the present and previous
studies (Smolka, Komlósi, et al., 2009; see also Table 5) with respect to prime modality, we
get the following picture: semantic relatedness induce significant priming, neither under (a)
cross-modal presentations (8 ms) nor under (b) visual prime presentations at 300 ms SOA (9
ms, 0 ms, and 18 ms); (c) under visual priming at SOA 1000 ms, semantic facilitation was
significant (30 ms) and of equivalent size to that by semantically transparent (37 ms) and
opaque (28 ms) derivations; finally, a similar pattern emerged (d) under visual priming with
sequential word presentation, where semantic facilitation was significant (38 ms) and of the
same size as the morphological facilitation by semantically transparent (48 ms) and opaque
(44 ms) derivations. Significant semantic effects can thus be attributed neither to modality
37
Lexical Representation of Complex Verbs
(specificity or abstractness) nor to the type of information delivery (simultaneous or
sequential).
Rather, it seems that pure semantic relatedness between verbs requires extremely long
prime exposure duration to be as effective as morphological relatedness. That is, we observed
significant semantic facilitation of equivalent strength to that of morphological facilitation
only in two out of six experiments: when visual primes were presented at 1000 ms SOA (in
Exp. 3 in Smolka, et al., 2009) and under sequential word presentation (Exp. 3 of the present
study), where each letter was presented for 100 ms, so that the whole prime unfolded over a
time span between 700 ms and 1200 ms (semantically related primes held on average 9.28
letters; range 7 - 12).
What does this tell us about the lexical representation of complex verbs in German?
The fact that the semantic activation between verbs relating to different bases requires longer
processing time than that between verbs relating to the same base—even if these are NOT
semantically related—indicates that the network is related not via pure semantic/meaning
association but rather via base forms. We will return to this issue below in the description of a
model on lexical representation.
Form Relatedness
In contrast to semantic relatedness, form relatedness generally hindered target
recognition, though its strength varied with prime condition: form inhibition was significant
under cross-modal (-14 ms), but not under intra-modal prime presentation (-9 ms and -8 ms
under simultaneous and sequential presentation, respectively), the latter replicating
nonsignificant effects of two previous experiments (Smolka, et al., 2009; see Table 5).
Summarizing across languages, it seems that form relatedness of the prime generally
hinders target recognition, but whether or not this inhibition reaches significance is not stable.
This holds for prime conditions similar to those in Experiment 1 of the present study, that is,
38
Lexical Representation of Complex Verbs
visual priming at SOAs of 230 ms or more (Drews & Zwitserlood, 1995; Feldman, 2000;
Feldman & Andjelkovic, 1992; Grainger, 1990; Pastizzo & Feldman, 2002; Rastle et al.,
2000; Segui & Grainger, 1990), as well as for prime conditions similar to those in Experiment
2, that is, cross-modal priming (Marslen-Wilson et al., 1994; Pastizzo & Feldman, 2002).
Our experiments indicate that form relatedness is less effective when representations
are modality-specific (under intra-modal priming) and reaches its utmost inhibitory effect
when representations are abstract (under cross-modal priming). The effects of cross-modal
inhibition may stem from two different sources—modality-abstractness or sequential
(phoneme-by-phoneme) information delivery of the auditory prime—so that each of these
could determine the form effects.
It is possible that sequential phoneme presentation in auditory word recognition causes
the selection of a specific cohort earlier on than simultaneous word presentation (assuming
that the selection of a cohort inhibits other cohorts). However, if early cohort selection
depended on sequential information delivery alone, also sequential visual word presentation
(in Exp. 3) should have singled out the relevant cohort early on, which in turn should inhibit
other cohorts. However, this was not the case. Sequential word presentation did not
significantly delay target recognition (-8 ms), so that we may conclude that it is not the
information delivery (i.e., letter-by-letter or phoneme-by-phoneme) but rather the modality
that strongly influences form inhibition. Specifically, this indicates that abstracted form
representations develop stronger inhibitory connections than modality-specific form
representations.2
Overall, with respect to the above discussed relatively small and instable (significant
and nonsignificant) semantic and form effects, we may conclude that the morphological
effects we obtain with German complex verbs cannot be reduced to simple semantic and form
relatedness between words.
39
Lexical Representation of Complex Verbs
Morphological Relatedness
In contrast to semantic and form relatedness, morphological relatedness strongly
facilitated target recognition under all prime presentation modes. Under visual priming, with
both simultaneous (Exp. 1) and sequential (Exp. 3) word presentation, morphological
facilitation was not affected by semantic transparency. That is, semantically opaque complex
verbs like entbinden (‘deliver’) produced the same amount of priming to their base binden
(‘bind’) as did semantically transparent verbs like zubinden (‘tie together’): 24 ms versus 28
ms in Experiment 1 and 44 ms versus 48 ms in Experiment 3. Only under cross-modal
priming (Exp. 2) did semantic transparency affect morphological facilitation in that the
priming by semantically opaque verbs was about half the size (14 ms) of that by semantically
transparent verbs (27 ms).
Taken together with our previous findings (Smolka, et al., 2009; see also Table 5), we
have in fact evidence from six data sets (five visual and one cross-modal). Unequivocally, all
show that morphological structure determines word recognition: (a) under cross-modal
priming, morphological effects were influenced by semantic compositionality, whereas (b)
under visual prime presentation—including 300 ms and 1000ms SOA as well as simultaneous
and sequential word presentation—semantically transparent and opaque derivations yielded
the same amount of priming in five out of five experiments. In summary, we believe that we
have a strong case in point with these data: Morphological structure strongly affects lexical
representation in German and needs to be incorporated in modeling lexical representation – a
German verb like depart will be represented as the base {part} and the prefix {de}.
Furthermore, modality-specific representations seem to be more strongly connected with their
base than modality-abstract ones.
Morphological structure is usually taken to emerge as the mapping from form codes
onto meaning codes (Plaut & Gonnerman, 2000). With regard to orthographic input, present
models on early visual word recognition (Longtin et al., 2003; Marlsen-Wilson et al., 2008;
40
Lexical Representation of Complex Verbs
Rastle et al., 2000; Rastle et al., 2004) assume that morphological segmentation occurs on a
purely orthographic basic, simply because words have a morphological structure. For
example, (Rastle et al., 2004) showed that both semantically transparent pairs like cleanerclean and semantically opaque pairs like corner-corn produced significant and equivalent
priming, even though the latter type had only the appearance of a morphological relationship.
Similarly, German orthographic strings like zubinden (‘tie together’) and entbinden (‘deliver’)
are segmented into their constituent morphemes regardless of meaning compositionality: zu-,
ent-, bind, -en. This initial process comprises reiterative orthographic segmentation, which
maps form to meaning. Given that morphemes are the smallest meaningful units, they emerge
as the product of early form to meaning mappings.
However, in contrast to models on English and French which confine morphological
effects to prelexical stages of word recognition, a model of lexical representation in German
needs to explain the resilient morphological effects under overt priming conditions in
German. It must allow derivations of the same verb to be processed via their base and
relatively independent of their meaning.
Model on Lexical Representation in German
Indeed, such representations may be best accommodated in a model that allows for the
base as lexical entry – for both inflections and derivations, without any regard to regularity
and semantic transparency (Smolka et al., 2009; Smolka et al., 2007). That is, a semantically
opaque verb like entbinden is not only segmented into ent-, bind, and -en at the prelexical
level, but is also lexically represented via its base {bind} and constituents {ent} and {en}. The
lexical entries of the base and constituents activate their corresponding underlying concepts at
the conceptual level: separate, bind, and infinitive, respectively. Hence, both the base {bind}
and its related concept bind are activated to the same degree by the derived verbs zubinden
(transparent) and entbinden (opaque), independently of their semantic transparency. This
41
Lexical Representation of Complex Verbs
activation facilitates the recognition of the target binden (‘bind’) that activates the same units
both at the prelexical level, lexical level {bind}, and at the conceptual level, bind.
In contrast to a morphologically related prime, a semantically related but
morphologically unrelated word like zuschnüren (‘tie together’) does not, as a whole, directly
activate another semantically related verb. Following its decomposition at the prelexical level,
its base {schnür} is activated at the lexical level and consequently its concept (tie) at the
conceptual level. Only spreading of activation at the conceptual level from the concept tie to a
semantically related concept like bind can facilitate the recognition of the target verb bind.
That is, meaning activation between different bases occurs not at the lexical level but rather at
the conceptual level. Since the conceptual level is activated only after the lexical level,
activation between different German verb bases is rather unstable and takes longer processing
times than that at the lexical level. This accounts for the findings of the present study that the
activation of morphologically related words (comprising the same base) is stronger and
independent of semantic transparency than that of semantically related words comprising a
different base.
If semantically opaque words are represented via their base just as transparent ones are
– how is the opaque meaning derived? Similar to the general assumption that the meaning of a
semantically transparent derivation is assembled from the meaning of its constituents, also the
joint activation of the constituents of an opaque derivation may activate the specific
underlying (noncompositional) concept. Note that even the components of a transparent
derivation usually generate the meaning of a very specific concept, so that the meaning of
opaque derivations may be accessed and stored in similar ways.
It is generally acknowledged that languages of different language families hold
structural features that lead to different morphological effects. For a long time, morphological
concatenativity has been taken as the structural dichotomy to explain the different
morphological effects between concatenative Indo-European languages and nonconcatenative
42
Lexical Representation of Complex Verbs
Semitic languages, namely that semantic relatedness is a precondition for morphological
effects in the former but not in the latter. However, the fact that German belongs to the IndoEuropean language family and shows morphological effects that are independent of semantic
compositionality indicates that the lines may not be drawn that simply. Our findings rather
indicate that language characteristics may differ strongly within a language family. German is
morphologically rich and productive and will thus show strong morphological effects. This is
predicted by connectionist accounts assuming that morphological structure is nothing more
than the learned systematic relationships among surface forms and meaning. Morphological
regularities thus emerge in morphologically rich systems, where mappings between form and
meaning are straightforward and dominate the system, as is the case in German.
Indeed, evidence for the acquisition of morphological structure is provided by most
recent data from our lab. Ninety 11-12 year old children were tested under visual priming at
long SOAs (500 ms) and showed morphological effects similar to those of adults, namely up
to 40 ms priming by semantically opaque derivations. However, unlike with adults, this effect
was about half the size of that by semantically transparent derivations. This indicates that the
children’s system still needs more exposure to morphological regularities so as to generalize
morphological structure above and beyond meaning compositionality, as is the case in the
adult system.
To summarize, we have shown that morphological structure represents an important
aspect in language processing in German and must be incorporated in the lexical
representation of German words. We have taken the first step by suggesting that the lexical
representation refers to the base of a complex word. Admittedly, the exact nature underlying
the activation of a lexical representation, in particular, whether its activation is symmetrical or
not, is not clear yet. That is, will the base also activate semantically transparent and opaque
derivations to the same extent, and will semantically transparent derivations prime
43
Lexical Representation of Complex Verbs
semantically opaque ones to the same degree as opaque derivations will prime transparent
ones? This issue deserves further attention in future studies.
44
Lexical Representation of Complex Verbs
References
Baayen, H. R., Piepenbrock, R., & van Rijn, H. (1993). The CELEX lexical database (on CDROM). Philadelphia, PA: Linguistic Data Consortium, University of Pennsylvania.
Boudelaa, S., & Marslen-Wilson, W. (2004a). Abstract morphemes and lexical representation:
The CV-Skeleton in Arabic. Cognition, 92, 271-303.
Boudelaa, S., & Marslen-Wilson, W. (2004b). Allomorphic variation in Arabic: Implications
for lexical processing and representation. Brain and Language, 90, 106-116.
Diependaele, K., Sandra, D., & Grainger, J. (2005). Masked cross-modal morphological
priming: Unravelling morpho-orthographic and morpho-semantic influences in early
word recognition Language and Cognitive Processes, 20(1/2), 75-114.
Drews, E., & Zwitserlood, P. (1995). Morphological and orthographic similarity in visual
word recognition. Journal of Experimental Psychology: Human Perception and
Performance, 21, 1098-1116.
Drews, E., Zwitserlood, P., & Neuwinger, E. (2000). How semantic is morphological
priming? Evidence from derivationally prefixed verbs in German. Unpublished
manuscript.
Eisenberg, P. (2004). Grundriß der deutschen Grammatik: Das Wort (2nd ed. Vol. 1).
Stuttgart-Weimar: J. B. Metzler.
Feldman, L., & Larabee, J. (2001). Morphological facilitation following prefixed but not
suffixed primes: Lexical architecture or modality-specific processes? Journal of
Experimental Psychology: Human Perception and Performance, 27, 680-691.
Feldman, L. B. (2000). Are morphological effects distinguishable from the effects of shared
meaning and shared form? Journal of Experimental Psychology: Learning, Memory,
and Cognition, 26, 1431-1444.
45
Lexical Representation of Complex Verbs
Feldman, L. B., & Andjelkovic, D. (1992). Morphological analysis in word recognition. In R.
Frost & L. Katz (Eds.), Orthography, phonology, morphology, and meaning (pp. 343360). Amsterdam: North Holland.
Feldman, L. B., Barac-Cikoja, D., & Kostić, A. (2002). Semantic aspects of morphological
processing: Transparency effects in Serbian. Memory and Cognition, 30, 629-636.
Feldman, L. B., & Larabee, J. (2001). Morphological facilitation following prefixed but not
suffixed primes: Lexical architecture or modality-specific processes? Journal of
Experimental Psychology: Human Perception and Performance, 27(3), 680-691.
Feldman, L. B., & Prostko, B. (2002). Graded aspects of morphological processing: Task and
processing time. Brain and Language, 81, 12-27.
Feldman, L. B., & Soltano, E. G. (1999). Morphological priming: The role of prime duration,
semantic transparency, and affix position. Brain and Language, 68, 33-39.
Feldman, L. B., Soltano, E. G., Pastizzo, M. J., & Francis, S. E. (2004). What do graded
effects of semantic transparency reveal about morphological processing? Brain and
Language, 90, 17-30.
Fleischer, W., & Barz, I. (1992). Wortbildung der deutschen Gegenwartssprache. Tübingen:
Max Niemeyer Verlag.
Forster, K. I., Mohan, K., & Hector, J. (2003). The mechanics of masked priming. In S.
Kinoshita & S. J. Lupker (Eds.), Masked Priming. The state of the art (pp. 3-37). New
York: Psychology Press.
Fowler, C. A., Napps, S. E., & Feldman, L. (1985). Relations among regular and irregular
morphologically related words in the lexicon as revealed by repetition priming.
Memory & Cognition, 13, 241-255.
Frost, R., Deutsch, A., Gilboa, O., Tannenbaum, M., & Marslen-Wilson, W. (2000).
Morphological priming: Dissociation of phonological, semantic and morphological
factors. Memory and Cognition, 28, 1277-1288.
46
Lexical Representation of Complex Verbs
Giraudo, H., & Grainger, J. (2000). Effects of prime word frequency and cumulative root
frequency in masked morphological priming. Language and Cognitive Processes,
15(4/5), 321-592.
Gonnerman, L. M., Seidenberg, M. S., & Andersen, E. S. (2007). Graded semantic and
phonological similarity effects in priming: Evidence for a distributed connectionest
approach to morphology. Journal of Experimental Psychology: General, 136(2), 323345.
Grainger, J. (1990). Word frequency and neighborhood frequency effects in lexical decision
and naming. Journal of Memory and Language, 29, 228-244.
Grainger, J., & Ferrand, L. (1994). Phonology and Orthography in Visual Word Recognition:
Effects of Masked Homophone Primes. Journal of Memory and Language, 33, 218233.
Longtin, C., Segui, J., & Hallé, P. (2003). Morphological priming without morphological
relationship. Language and Cognitive Processes, 18, 313-334.
Marlsen-Wilson, W., Bozic, M., & Randall, B. (2008). Early decomposition in visual word
recognition: Dissociating morphology, form, and meaning. Language and Cognitive
Processes, 23(3), 394-421.
Marslen-Wilson, W., Tyler, L. K., Waksler, R., & Older, L. (1994). Morphology and meaning
in the English mental lexicon. Psychological Review, 101, 3-33.
McCormick, S. F., Rastle, K., & Davis, M., H. (2009). Adore-able not adorable? Orthographic
underspecification studied with masked repetition priming. European Journal of
Cognitive Psychology, 21(6), 813-836.
Meunier, F., & Longtin, C.-M. (2007). Morphological decomposition and semantic
integration in word processing. Journal of Memory and Language, 56(4), 457-471.
Meunier, F., & Segui, J. (2002). Cross-modal priming in French. Brain and Language, 81, 89102.
47
Lexical Representation of Complex Verbs
Murrell, G. A., & Morton, J. (1974). Word recognition and morphemic structure. Journal of
Experimental Psychology, 102, 963-968.
Olsen, S. (1996). Partikelverben im Deutsch-Englischen Vergleich. In E. Lang & G. Zifonum
(Eds.), Deutsch - typologisch (pp. 261-288). Berlin: De Gruyter.
Pastizzo, M., & Feldman, L. B. (2002). Does prime modality influence morphological
processing? Brain and Language, 81, 28-41.
Pastizzo, M. J., & Feldman, L. B. (2002). Discrepancies between orthographic and unrelated
baselines in masked priming undermine a decompositional account of morphological
facilitation. Journal of Experimental Psychology: Learning, Memory, and Cognition,
28, 244-249.
Plaut, D. C., & Gonnerman, L. M. (2000). Are non-semantic morphological effects
incompatible with a distributed connectionist approach to lexical processing?
Language and Cognitive Processes, 15, 445-485.
Rastle, K., Davis, M., H., Marslen-Wilson, W., & Tyler, L. K. (2000). Morphological and
semantic effects in visual word recognition: A time-course study. Language and
Cognitive Processes, 15, 507-537.
Rastle, K., Davis, M., H., & New, B. (2004). The broth in my brother's brothel: Morphoorthographic segmentation in visual word recognition. Psychonomic Bulletin and
Review, 11(6), 1090-1098.
Raveh, M. (2002). The contribution of frequency and semantic similarity to morphological
processing. Brain and Language, 81, 312-325.
Raveh, M., & Rueckl, J. (2000). Equivalent effects of inflected and derived primes: Longterm morphological priming in fragment completion and lexical decision. Journal of
Memory and Language, 42, 103-119.
48
Lexical Representation of Complex Verbs
Reid, A., & Marslen-Wilson, W. (2000). Organising principles in lexical representation:
Evidence from Polish. Paper presented at the Twenty-Second Annual Conference of
the Cognitive Science Society Mahwah, NJ.
Rueckl, J., & Raveh, M. (1999). The influence of morphological regularities on the dynamics
of a connectionist network. Brain and Language, 68, 110-117.
Rueckl, J. G., Mikolinski, M., Raveh, M., Miner, C. S., & Mars, F. (1997). Morphological
priming, fragment completion, and connectionist networks. Journal of Memory and
Language, 36, 382-405.
Schreuder, R., & Baayen, H. R. (1995). Modeling morphological processing. In L. Feldman,
Beth (Ed.), Morphological aspects of language processing (pp. 131-154). Hillsdale,
NJ: Erlbaum.
Schriefers, H., Zwitserlood, P., & Roelofs, A. (1991). The identification of morphologically
complex spoken words: Continuous processing or decomposition? Journal of Memory
and Language, 30, 26-47.
Segui, J., & Grainger, J. (1990). Priming word recognition with orthographic neighbors:
Effects of relative prime-target frequency. Journal of Experimental Psychology:
Human Perception and Performance, 16, 65-76.
Smolka, E., Khader, P., Wiese, R., Zwitserlood, P., & Rösler, F. (2010). Continuous
Processing of Linguistic Categories: Behavioral and Electrophysiological Responses
to Regular and Irregular Verb Inflection in German. Manuscript submitted for
publication.
Smolka, E., Komlósi, S., & Rösler, F. (2009). When semantics means less than morphology:
Processing German prefixed verbs. Language and Cognitive Processes, 24(3), 337375.
49
Lexical Representation of Complex Verbs
Smolka, E., Zwitserlood, P., & Rösler, F. (2007). Stem access in regular and irregular
inflection: Evidence from German participles. Journal of Memory and Language, 57,
325-347.
Taft, M. (1979). Recognition of affixed words and the word frequency effect. Memory &
Cognition, 7, 263-272.
Taft, M. (1994). Interactive-activation as a framework for understanding morphological
processing. Language and Cognitive Processes, 9, 271-294.
Taft, M. (2004). Morphological decomposition and the reverse base frequency effect. The
Quarterly Journal of Experimental Psychology, 57A, 745-765.
Taft, M., & Forster, K. I. (1975). Lexical storage and retrieval of prefixed words. Journal of
Verbal Learning and Verbal Behavior, 14, 638-647.
Taft, M., Hambly, G., & Kinoshita, S. (1986). Visual and auditory recognition of prefixed
words. The Quarterly Journal of Experimental Psychology, 38A, 351-366.
Taft, M., & Kougious, P. (2004). The processing of morpheme-like units in monomorphemic
words. Brain and Language, 90, 9-16.
Taft, M., & Nguyen-Hoan, M. (2010). A sticky stick? The locus of morphological
representation in the lexicon. Language and Cognitive Processes, 25(2), 277-296.
Zwitserlood, P., Drews, E., Bolwiender, A., & Neuwinger, E. (1996). Kann man Geschenke
umbringen? Assoziative Bahnungsexperimente zur Bedeutungsheterogenität von
Verben. In C. Habel & S. Kanngießer (Eds.), Perspektiven der kognitiven Linguistik:
Modelle und Methoden (pp. 211-232).
50
Lexical Representation of Complex Verbs
Author Notes
This study was supported by the German Research Foundation (DFG), grant EU 39/71 awarded to Carsten Eulitz.
Correspondence concerning this article should be addressed to Eva Smolka,
Department of Linguistics, University of Konstanz, 78457 Konstanz, Germany. Email:
[email protected]
51
Lexical Representation of Complex Verbs
Footnotes
Footnote 1: In Smolka et al. (2009), we conducted post-hoc analyses on first-block
presentations only and obtained the same strength of morphological effects as those calculated
with repeated prime presentation.
Footnote 2: It could still be possible that auditory perception builds up stronger form
representations than visual perception – regardless of sequential or simultaneous information
delivery. To unequivocally ascertain if this is the case, intra-modal experiments with auditoryauditory prime-target presentation would be required.
52
Lexical Representation of Complex Verbs
Figure Captions
Figure 1. Synonym (S), semantically transparent (SM), semantically opaque (M), and
form priming effects (relative to unrelated verbs) in Experiment 1 (visual-simultaneous at
SOA 300), Experiment 2 (cross-modal), and Experiment 3 (visual-sequential) in the mid, left,
and right panel, respectively. The y-bars provide the standard errors of the means.
53
Lexical Representation of Complex Verbs
Table 1
Stimulus Characteristics of Primes that were Synonyms (M−S+F−), Semantically
Transparent Derivations (M+S+F+) or Semantically Opaque Derivations (M+S−F+), Form
Relations (M−S−F+), or Controls (M−S−F−) of Simple Verb Targets in Experiments 1 to 3
Lemma
Frequency
Word
Length
Syllable
Length
Relatedness
Score
Target
binden (‘bind’)
144.6
6.88
2.00
-
M−S+F−
zuschnüren (‘tie’)
6.55
9.28
3.03
5.38
M+S+F+
zubinden (‘tie’)
6.75
9.48
3.00
5.60
M+S−F+
entbinden (‘deliver’)
6.65
9.75
3.08
2.16
5.98/6.70
9.60
3.03
1.28
6.15
9.08
3.03
-
M−S−F+
abbilden (‘depict’)
M−S−F−
abholzen (‘deforest’)
Note. In Experiment 2, some orthographically related primes were replaced with
phonologically related ones. Statistics are given for the total set of stimuli, sample stimuli are
italicized. Frequencies are from the CELEX database (Baayen et al., 1993), count is per
million.
54
Lexical Representation of Complex Verbs
Table 2
Mean Response Latencies and Accuracies in Experiment 1 (Visual SOA 300)
Prime Type
RT (SD)
Effect
Error (SD)
M−S+F−
zuschnüren (‘tie’)
581 (103)
(9)
1.4 (5.6)
M+S+F+
zubinden (‘tie’)
563 (88)
28***
1.0 (3.5)
M+S−F+
entbinden (‘deliver’)
567 (84)
24**
1.4 (4.0)
M−S−F+
abbilden (‘depict’)
600 (100)
(-9)
1.6 (4.8)
M−S−F−
abholzen (‘deforest’)
591 (97)
1.6 (4.8)
Note. Mean RTs in milliseconds (SD in parentheses) and mean percentage of errors (SD) for
verb targets like binden (‘bind’) preceded by a visually presented synonym (M−S+F−), a
semantically transparent derivation (M+S+F+), a semantically opaque derivation (M+S−F+),
a form-related verb (M−S−F+), or by an unrelated control (M−S−F−). Priming effects were
calculated by subtracting the mean of the related condition from that of the unrelated
condition. Nonsignificant effects are given in parentheses. ** p < .05, *** p < .01 indicate
significance levels for analyses of both participants and items.
55
Lexical Representation of Complex Verbs
Table 3
Mean Response Latencies and Accuracies in Experiment 2 (Cross-Modal)
Prime Type
RT (SD)
Effect
Error (SD)
M−S+F−
zuschnüren (‘tie’)
592 (71)
(8)
0.8 (3.7)
M+S+F+
zubinden (‘tie’)
573 (77)
27****
0.6 (3.2)
M+S−F+
entbinden (‘deliver’)
587 (79)
14*
0.8 (3.2)
M−S−F+
abbilden (‘depict’)
614 (79)
-14**
1.7 (4.7)
M−S−F−
abholzen (‘deforest’)
600 (77)
1.1 (3.7)
Note. Mean RTs in milliseconds (SD in parentheses) and mean percentage of errors (SD) for
verb targets like binden (‘bind’) preceded by an auditorily presented synonym (M−S+F−), a
semantically transparent derivation (M+S+F+), a semantically opaque derivation (M+S−F+),
a form-related verb (M−S−F+), or by an unrelated control (M−S−F−). Priming effects were
calculated relative to the unrelated condition. Nonsignificant effects are given in parentheses,
* p < .05 indicates the significance level for the analysis of participants, ** p < .05, **** p <
.0001 indicate significance levels for analyses of both participants and items.
56
Lexical Representation of Complex Verbs
Table 4
Mean Response Latencies and Accuracies in Experiment 3 (Visual Sequential)
Prime Type
RT (SD)
Effect
Error (SD)
M−S+F−
zuschnüren (‘tie’)
619 (82)
38****
0.7 (2.9)
M+S+F+
zubinden (‘tie’)
609 (99)
48****
0.2 (1.7)
M+S−F+
entbinden (‘deliver’)
613 (109)
44***
0.9 (3.3)
M−S−F+
abbilden (‘depict’)
665 (113)
(-8)
1.6 (4.8)
M−S−F−
abholzen (‘deforest’)
657 (115)
0.7 (2.9)
Note. Mean RTs in milliseconds (SD in parentheses) and mean percentage of errors (SD) for
verb targets like binden (‘bind’) preceded by a visually unfolding synonym (M−S+F−), a
semantically transparent derivation (M+S+F+), a semantically opaque derivation (M+S−F+),
a form-related verb (M−S−F+), or by an unrelated control (M−S−F−). Priming effects were
calculated relative to the unrelated condition. Nonsignificant effects are given in parentheses.
** p < .05, *** p < .01, **** p < .001 indicate significance levels for analyses of both
participants and items.
57
Lexical Representation of Complex Verbs
Table 5
Summary of Semantic (S), Semantically Transparent (SM), Semantically Opaque (M), and Form (F) Effects over the Present and Previous Study
(Smolka, Komlósi et al., 2009) with respect to Prime Modality.
Study
Exp.
Modality
present
2
auditory
sequential
(8)
<
27**
present
1
visual
SOA 300
(9)
<
previous
1
visual
SOA 300
(0)
previous
2
visual
SOA 300
previous
3
visual
present
3
visual
Mode
S
SM
M
Form
>
14*
-14**
28**
=
24**
(-9)
<
30*
=
43*
-
18†
<
36**
=
36**
-14†
SOA 1000
30**
=
37**
=
28**
(-11)
sequential
38**
=
48**
=
44**
(-8)
Note. Target presentation was always visual, modality of the prime either auditory or visual in either sequential presentation (of phonemes or letters)
or simultaneous presentation of all letters at different stimulus onset asynchrony (SOA). Effects were calculated relative to an unrelated baseline;
</=/> indicate statistical differences between effects of semantic associates (S) and semantically transparent derivations (SM), as well as between
effects of semantically transparent and opaque (M) derivations. Effects of the previous study are taken from the first-block analyses (without target
58
Lexical Representation of Complex Verbs
repetition), those effects that reached significance due to target repetition only are marked with †; nonsignificant effects are given in parentheses; * p
< .05 indicates the significance level for analyses of participants, ** p < .05 for both participants and items.
59
Lexical Representation of Complex Verbs
60
50
priming (unrelated-related) in ms
40
30
20
S
SM
10
M
Form
0
-10
-20
-30
cross-modal
visual SOA 300
visual sequential
60