7/28/2019 Wurm (2000)
1/17
Auditory Processing of Polymorphemic Pseudowords
Lee H. Wurm
Wayne State University
This study compared models of auditory word recognition as they relate to the processing of
polymorphemic pseudowords. Semantic transparency ratings were obtained in a preliminary rating
study. The effects of morphological structure, semantic transparency, prefix likelihood, and morphe-
mic frequency measures were examined in a lexical decision experiment. Reaction times and errors
were greater for pseudowords carrying a genuine prefix, and this effect was largest for pseudowords
that also carried a genuine root. While results were grossly similar for bound and free root types, there
were also some important differences. Regression analyses provided additional support for decom-positional models: semantic transparency, prefix likelihood, prefix frequency, and root frequency all
affected pseudoword rejection times. The results are most compatible with a modification of Tafts
(1994) interactive-activation model or a dual-route model. 2000 Academic Press
Key Words: lexical decision; morphology-language; semantic transparency; speech perception;
word recognition.
Morphological effects in spoken word recog-
nition have been receiving increasing attention.
English is considered to have only limited and
irregular morphological structure (e.g., Hender-son, 1985; Jarvella & Meijers, 1983), but recent
studies have shown that morphological infor-
mation is used in perception (e.g., Marslen-
Wilson, Tyler, Waksler, & Older, 1994; Wurm,
1997). There is of course some morphological
structure to the language, and different ap-
proaches could be used by the perceptual sys-
tem in dealing with that structure.
The traditional view in formal linguistics isthat nonarbitrary items do not need to be stored
in the lexicon (Bloomfield, 1933; Chomsky,
1965; Lyons, 1977). According to this view, it
is unnecessary to store built, builds, rebuild, and
other complex relatives of these, because the
lexicon would already contain the root mor-
pheme build. The complex forms can be gener-
ated as needed through the use of word forma-
tion rules. Reduction of redundancy is the mostattractive feature of this approach; some lan-
guages have verbs that can assume thousands of
distinct surface forms even though they differ
only by inflection and are essentially the same
vocabulary item (Anderson, 1988). Only the
base form of such verbs needs to be stored.
A class of word-recognition models that cor-
responds to this view can be referred to as
decompositional (or discontinuous). Although
there are several examples of discontinuous
models (e.g., Cutler, Hawkins, & Gilligan,
1985; Cutler & Norris, 1988; Grosjean & Gee,
1987; Jarvella & Meijers, 1983; MacKay, 1978;
Morton, 1969, 1979), the most visible one has
been the prefix-stripping model of Taft and his
colleagues (1981, 1985; Taft & Forster, 1975;
Taft, Hambly, & Kinoshita, 1986). This modelwas developed to explain visual lexical decision
times for various classes of morphologically
complex pseudowords.
According to this model, saying NO should
take longer for pseudowords with genuine pre-
fixes than for those without. The difference
should be even larger when the root of the
Portions of this research were supported by a National
Research Service Award from the National Institute of
Mental Health (Grant F32 MH11721). I thank Cynthia
Connine, Albrecht Inhoff, Arthur Samuel, Robert Schreu-
der, Marcus Taft, and an anonymous reviewer for making
helpful criticisms of a previous version of this paper. Mark
Aronoff and Mark Pitt also provided useful advice.
Correspondence and reprint requests concerning this ar-
ticle should be addressed to Lee H. Wurm, Department of
Psychology, Wayne State University, 71 West Warren Av-
enue, Detroit, MI 48202. E-mail: [email protected].
wayne.edu.
255 0749-596X/00 $35.00Copyright 2000 by Academic Press
All rights of reproduction in any form reserved.
Journal of Memory and Language 42, 255271 (2000)
doi:10.1006/jmla.1999.2678, available online at http://www.idealibrary.com on
7/28/2019 Wurm (2000)
2/17
pseudoword is a real English root. This is be-
cause there is a successful prefix strip for Pre-
fix, Root pseudowords that requires time; for
Prefix, Root stimuli there is a successful
prefix strip plus a successful root look-up,
which requires still more time (deciding that thetwo legitimate morphemes cannot be combined
with each other to make a word also slows the
process).
One interesting aspect of this predicted pat-
tern concerns pseudowords that begin with non-
prefix strings (i.e., Prefix stimuli). Roots
should not even be recognizable as roots when
there are no prefixes to strip off, so root status
should not have an effect here (see Taft et al.,1986): reaction times (RTs) for the two condi-
tions should be equal. Taft (1994) later con-
cluded that it is theoretically possible to observe
a RT disadvantage for the Prefix, Root
items if the root is very common and easily
recognized, as in the visually presented
pseudoword IBPEOPLE (his example). I will
have more to say about this following the main
RT experiment.Some theorists feel that lexical redundancy
can be an advantage to be exploited, rather than
a burden (Henderson, 1985). Such authors pre-
fer the full-listing view, which states that all
words are stored in the lexicon (Bybee, 1985,
1995a, b; Jackendoff, 1975). Bybee (1988) feels
that theorists should not be concerned with stor-
age efficiency given the capacity of the human
brain and the widespread idiosyncrasies presentin all languages (see also Sandra, 1994).
Continuous processing models correspond to
this view. Words are processed on a strict left-
to-right basis, with no regard for internal struc-
ture. Morphological structure and morphologi-
cal variables cannot affect RTs or error rates.
The Cohort model (Marslen-Wilson, 1984,
1987; Marslen-Wilson & Welsh, 1978) is one
such model, and there have been several other
arguments in favor of continuous processing
(e.g., Henderson, Wallis, & Knight, 1984; Ru-
bin, Becker, & Freeman, 1979; Tyler, Marslen-
Wilson, Rentoul, & Hanney, 1988).
Pseudoword rejection should occur as soon as
the input becomes inconsistent with all words.
Some interactive-activation models explicitly
deny the existence of morpheme or word units
(e.g., McClelland & Elman, 1986; Rueckl,
Mikolinski, Raveh, Miner, & Mars, 1997; Sei-
denberg, 1987; 1989; for critical views, see
Dennet, 1987; Forster, 1994). On this view,
so-called morphological effects are in fact dueto semantic and form-based similarity, fre-
quency of occurrence of sublexical letter
strings, and so on. Most interactive-activation
models are characterized as continuous, but Taft
(1994) proposed an interactive-activation ver-
sion of the earlier prefix-stripping model. The
new model is behaviorally very similar to the
earlier one, but Taft found the interactive-acti-
vation framework more plausible and appealing(the major difference is that a prelexical prefix
store is not needed). The model has distinct
word and morpheme units and exhibits decom-
positional behavior. The equivalent of prefix
stripping takes place as a consequence of the
mapping process (acoustic-phonetic or visual-
orthographic).
Some researchers have argued that some
words are decomposed while others are not.Wurm (1997) proposed a dual-route model
based on the idea of parallel, competing pro-
cesses [cf. the Race model of Cutler and Norris
(1979)]. In his model, morphologically complex
words are processed simultaneously as full-
forms and as analyzed constituent morphemes.
The decompositional route of the model is sen-
sitive to variations in semantic transparency, the
likelihood that a given string is a prefix, andmorpheme frequencies (see below). Other vari-
ations on the dual-route theme have also been
proposed (e.g., Anshen & Aronoff, 1981; Berg-
man, Hudson, & Eling, 1988; Caramazza, Lau-
danna, & Romani, 1988; Frauenfelder &
Schreuder, 1992; Laudanna, Burani, & Cer-
mele, 1994; Laudanna & Burani, 1995; Lau-
danna, Cermele, & Caramazza, 1997; Schreuder
& Baayen, 1995; Stanners, Neiser, & Painton,
1979).
Dual-route models have often provided the
context for the initial exploration of previously
ignored variables, such as semantic transpar-
ency (Bergman et al., 1988; Henderson, 1985;
Smith, 1988; Smith & Sterling, 1982). Words
such as unhappy are highly transparent, while
256 LEE H. WURM
7/28/2019 Wurm (2000)
3/17
those such as relate are opaque. There is also a
sizable middle ground (Wurm, 1997). Recent
data (Libben, 1998; Schreuder & Baayen, 1995;
Wurm, 1997) have shown that this variable
plays a role in word recognition and have more
generally called into question the defensibilityof an all-or-none theoretical position on com-
plex word recognition. For example, Marslen-
Wilson et al. (1994) found that suffixed words
that do not have a semantic relationship that is
obvious to current language users are treated as
monomorphemic.
In an investigation of visual processing of
Italian pseudowords, Laudanna et al. (1994)
introduced another important concept: the pro-portion of tokens beginning with a given letter
string that are prefixed (e.g., retold is prefixed,
realize is not). Schreuder and Baayen (1994)
found that the average value for this variable in
English was very low and rejected the notion of
prefix-stripping. Wurm (1997) reported a simi-
lar average value for this variable (which he
called prefix likelihood), but found that it inter-acted with several other variables in the recog-
nition of auditorily presented prefixed English
words. The nature of the interactions suggested
decompositional processing for some items.
The current study extends previous work in
many ways. First, most previous studies have
presented stimuli visually. Auditory presenta-
tion can inform theory in a unique way, because
the pieces of a polymorphemic stimulus arriveat the listener at different, specifiable times (cf.
Butterworth, 1983; Grosjean & Gee, 1987;
Henderson, 1985; Kempley & Morton, 1982;
Marslen-Wilson, 1984; Morton, 1979; Radeau,
Morais, Mousty, Saerens, & Bertelson, 1992).
Second, the critical stimuli in most experi-
ments have carried bound roots (e.g., -ceive in
receive and conceive). Overreliance on bound
roots is a potential problem given recent find-
ings about the importance of semantics; bound
roots are semantically empty, at least to nonlin-
guists, and thus they are not subject to phenom-
ena like semantic drift (Aronoff, 1976) to the
same extent that free roots are (free roots are
those that can stand alone as words, such as the
build in rebuild). Bound roots are also less
productive than free rootsthey cannot com-
bine with prefixes to make novel words.
Finally, studies of auditory pseudoword pro-
cessing have not included prefix likelihood or
semantic transparency, nor have they looked at
interactions between these variables and mor-phemic frequency measures. This is important,
because pseudowords are simply potential
words that happen not to be used; speakers and
writers coin new combinations as needed, but
this almost never causes problems for listeners
and readers (provided the new combination is
phonotactically legalsee Baayen, 1994;
Coolen, van Jaarsveld, & Schreuder, 1991;
Schreuder & Flores dArcais, 1989).Because many studies have used pseudowords
as critical stimuli (e.g., Caramazza et al., 1988;
Laudanna et al., 1994; Taft, 1994; Taft & Forster,
1975; Taft et al., 1986), the use of pseudowords
allows contact with a large body of literature. The
current study examines whether the same vari-
ables that influence word recognition also affect
the processing of pseudowords.
PRELIMINARY RATING STUDY
This study provides values on semantic trans-
parency for polymorphemic pseudowords.
Method
Participants
Twenty students from the Department of Psy-
chology subject pool participated. All were na-tive speakers of English. Participants received
extra credit in a psychology course.
Materials
Critical pseudowords in this study fell into
one of four groups, defined by crossing the
presence or the absence of a genuine prefix and
a genuine root (see Appendix A). Only the
Prefix, Root pseudowords with free roots
were included in this rating study. Pilot ratings
collected for Prefix, Root pseudowords with
bound roots were uniformly low. These items
were dropped from the rating study, but will be
included in the main part of the lexical decision
experiment.
Stimulus construction is described more fully
257POLYMORPHEMIC PSEUDOWORDS
7/28/2019 Wurm (2000)
4/17
under the next Method section. The stimuli to
be rated were printed in a rating packet in two
different random orders.
Procedure
Participants made their ratings by writing anumber from 1 to 7 in a blank next to each
pseudoword. Anchor points on the scale were
labeled Impossible to put this in a sentence
(1) and Very easy to put this in a sentence (7).
Participants were given an example of a
pseudoword that can easily be put into a mean-
ingful sentence: The band was interrupted mid-
song by a power failure (a sentence heard by
the author on a radio station in Binghamton,
NY) and one that cannot easily be put into a
meaningful sentence (transplay). This indirect
method is one way of getting at the construct of
semantic transparency, which in the case of
pseudowords concerns how easily interpretable
each stimulus is (see Caramazza et al., 1988;
Coolen et al., 1991).
Results
Median semantic transparency ratings are
shown in Appendix B. There was significant
variation on this dimension, even though the
stimuli were created by the random concatena-
tion of a prefix and a root. Median ratings
ranged from 2 (e.g., transfrost) to 7 (e.g., re-
bolt).
CALCULATION OF OTHER REGRESSOR
VARIABLES
Prefix likelihood is a ratio: the numerator is
the summed frequency (Francis & Kucera,
1982) of the truly prefixed words beginning
with a given phonetic string, and the denomi-
nator is the summed frequency of all words
beginning with that string in which removal of
the string leaves a pronounceable syllable or
syllables. For example, although real begins
with re-, this word was not considered a prefix-
stripping failure because the remainder of the
word (simply the phoneme /l/ in this case) is not
a syllable. Prefix likelihoods for each prefix
were taken from Wurm (1997). The value for
ad-, which was not used in that study, was
calculated by the methods described in that pa-
per.
Prefix likelihoods can range from 0 to 1. A
value of 0 would indicate that no words begin-ning with the string are truly prefixed. A value
of 1 would indicate that all words beginning
with the string are truly prefixed. Values for the
prefixes used in this study are listed in Appen-
dix C. They ranged from .005 (per-) to .283
(un-), averaging .07. Wurm (1997) found that
this variable played a role in word recognition
despite the fact that most of these values are
small.A measure of root morpheme frequency
was needed for Root pseudowords. The Bir-
mingham/Cobuild corpus (18 million tokens)
of the CELEX database (Baayen, Piepen-
brock, & van Rijn, 1993; Burnage, 1990) was
searched for each root. Frequencies were
summed across all cases where that root was
found (e.g., the frequencies of repay, prepay,
and so on are all included in the root fre-quency for pay). Root frequencies are shown
in Appendix B.
Prefix frequencies were calculated in essen-
tially the same way. Counts for words in the
Birmingham/Cobuild corpus beginning with
each (orthographic) prefix string were obtained.
From these, the frequencies for cases that were
instances of prefixation were summed (e.g., re-
play counts but reach does not). Appendix C
includes the prefix frequency for each prefix.
Summary statistics for both frequency measures
are shown in Table 1.
LEXICAL DECISION EXPERIMENT
There have been few explicit discussions of
possible processing differences for bound vs.
TABLE 1
Summary Statistics for Frequency Measures
M (SD) Range
Prefix frequency 481 (610) 91881
Root frequency (free) 192 (216) 4870
Root frequency (bound) 119 (134) 0619
Note. Per million tokens, from the CELEX database
(Baayen et al., 1993; Burnage, 1990).
258 LEE H. WURM
7/28/2019 Wurm (2000)
5/17
free roots. Most researchers who have ad-
dressed the issue have concluded that bound
elements are represented in the same way as
free ones (Bergman et al., 1988; Emmorey,
1989; Stanners et al., 1979; Taft, 1994). How-
ever, Marslen-Wilson et al. (1994) concludedthat bound roots do nothave the same represen-
tational status as free roots because they lack
reliable meanings. This experiment includes
stimuli with both root types.
Method
Participants
Participants were 88 students from the De-
partment of Psychology subject pool. All werenative speakers of English with no known hear-
ing problems. Participants received extra credit
in a psychology course for their participation.
Materials
Yoked quartets of critical pseudowords were
constructed by crossing /Prefix with
/Root. These quartets are listed in Appendix
A. Four lists of stimuli were prepared, eachconsisting of 480 items (240 words and 240
pseudowords). Each list contained 120 critical
pseudowords: 30 Prefix, Root pseudo-
words; 30 Prefix, Root pseudowords; 30
Prefix, Root pseudowords; and 30 Prefix,
Root pseudowords. One member of each
stimulus quartet was assigned to each list, so
that no participant heard more than one member
of the quartet. In each of the four conditions,half of the pseudowords came from a quartet
with bound roots and half came from a quartet
with free roots.
The Prefix, Root critical pseudowords
were constructed by randomly concatenating 1
of 10 English prefixes with 1 of 60 bound and
60 free roots. Each root was used once, and each
prefix was used 12 times (combined 6 times
with bound roots and 6 times with free roots).
To create the other 3 conditions, prefixes were
made into nonprefixes by the substitution of one
of the phonemes to a different phoneme from
the same broad class. Readers may notice that
Prefix strings were repeated more often than
Prefix strings throughout the experiment. I
will address this point below.
Roots were changed into nonroots by the
same procedure. Phoneme substitutions were
balanced among early, medial, and late posi-
tions within the individual morphemes. Item
durations were well matched across the eight
conditions (see Table 2).
Each list also contained 120 fillerpseudowords with no apparent internal structure
(e.g., *chormal), 120 prefixed filler words (e.g.,
enslave), and 120 unprefixed filler words (e.g.,
glutton). The 360 filler items were identical in
each list. Across the 480 stimuli heard by a
participant, 49% of words and 53% of
pseudowords had weak first syllables (the stress
of all critical items was weakstrong).
Two- or three-syllable filler words were cho-sen at random from a dictionary, subject to the
constraint that they be of sufficiently high fre-
quency to be familiar to the participant popula-
tion. Filler pseudowords were chosen the same
way. A randomly-selected two- or three-sylla-
ble word was changed into a pseudoword by the
substitution of one or two phonemes with a
phoneme or phonemes from the same class.
Three quarters of the filler pseudowords had one
phoneme change, and a quarter had two
changes. The position of these substitutions
were randomly determined. This mixture ap-
proximated the proportions established by the
critical pseudowords; matching the proportions
exactly was not possible, because a quarter of
the critical pseudowords (i.e., those in the
TABLE 2
Mean Item Durations (SD) in Milliseconds
Total Prefix Root
Items with free roots
Prefix, Root 879 (69) 203 (55) 676 (61)
Prefix, Root 877 (68) 200 (60) 677 (59)
Prefix, Root 880 (72) 196 (61) 684 (66)
Prefix, Root 879 (73) 212 (61) 667 (67)
Items with bound roots
Prefix, Root 877 (62) 198 (56) 679 (57)
Prefix, Root 875 (68) 186 (56) 689 (61)
Prefix, Root 876 (63) 197 (57) 679 (54)
Prefix, Root 878 (63) 200 (57) 678 (62)
Note. n 60 items per condition.
259POLYMORPHEMIC PSEUDOWORDS
7/28/2019 Wurm (2000)
6/17
Prefix, Root condition) had no phoneme
substitutions.
Stimuli were digitized at a sampling rate of
10 kHz, low-pass-filtered at 4.8 kHz, and stored
in disk files. A practice list of similar composi-
tion, consisting of 100 items, was used prior tothe main experiment. Visual feedback about
accuracy was given to the participants after each
trial, but only during the practice list.
Procedure
Participants (alone or in pairs) listened to stim-
uli over headphones in a sound-attenuating room.
Order of stimulus presentation was randomized
for each group of participants. An equal numberof participants heard each of the four stimulus
lists. On each trial, a participant heard a stimulus
and made a lexical decision by pressing a button
on a response board with his or her dominant
hand. Participants pressed one button for words
and another button for pseudowords.
RTs were measured from the acoustic offset
of each item. This approximates the measure-
ment method used by Taft et al. (1986) and wasnecessary given the goals of this paper. One
goal was to see if the RT pattern predicted by
the prefix-stripping model would emerge for
stimuli carrying free roots rather than bound.
Another goal, contingent on the first one, was to
see if models other than the prefix-stripping
model can explain that pattern of data. Taft et al.
(1986) reasoned that it did not make any differ-
ence where the RT measurement began, pro-vided that two conditions were met: First, the
RT measurement had to start somewhere in the
root portion of each stimulus, and second, the
starting position had to be the same point for
both stimuli that contained a given root. Thus,
Taft et al. (1986) chose an arbitrary point in
each root from which to measure RTs. The
current study uses the analogous method of
measuring from item offset: the offset of each
item equals the offset of each root, which is as
good a point as any according to this view (if
item durations are well matchedsee Table 2).
Results and Discussion
A participant was excluded from the experi-
ment if he or she had an error rate greater than
15% or a mean RT greater than 1000 ms. Eight
participants were excluded by these criteria.
Analyses reported here were conducted on the
remaining 80 participants. RTs for trials onwhich the participant incorrectly classified a
critical stimulus as a word were not included.
RTs were discarded if they were more than 2 SD
above the mean for a given participant in a
given condition (subject analyses) or for a par-
ticular item (item analyses).
Analyses of Variance (ANOVAs)
Mean RT as a function of root status, prefix
status, and root type (free vs. bound) is shown in
Fig. 1. The mean error rate for each condition is
shown above the bar in the figure.
A 2 (root type) 2 (prefix status) 2 (root
status) ANOVA was conducted. Pseudowords
with bound roots had slightly faster mean rejec-
tion times than those with free roots (292 ms vs.
FIG. 1. Mean reaction time (RT) as a function of root
status, prefix status, and root type, in milliseconds (ms).
Error bars show 1 SEM. Mean error rates are shown above
the bar for each condition. (A) Pseudowords with free roots;
(B) pseudowords with bound roots.
260 LEE H. WURM
7/28/2019 Wurm (2000)
7/17
271 ms), but this difference was not significant
by items: F1(1, 79) 11.67, p .001; F2(1,
472) 1.98, p .10.
Items with genuine prefixes (M 354 ms)
took longer to reject than those without [M
209 ms; F1(1, 79) 369.24, p .001;F2(1, 472) 165.35, p .001]. Items with
genuine roots took longer than those without
[321 ms vs. 242 ms; F1(1, 79) 120.54, p
.001; F2(1, 472) 55.86, p .001], and
the interaction between prefix status and root
status was also significant [F1(1, 79) 22.80,
p .001; F2(1, 472) 9.63, p .01]. As
can be seen in the figure, the disruptive effect of
a genuine root was even more pronounced in thecontext of a genuine prefix. These last two
effects are incompatible with continuous pro-
cessing models.1,2
Both portions of Fig. 1 fit the overall pattern
predicted by a prefix-stripping model, except
for one important difference: RTs for Prefix,
Root items (M 229 ms) were slower than
RTs for Prefix, Root items [M 189 ms;
F1(1, 79) 14.82, p .001; F2(1, 236)
10.40, p .001]. I will return to this point
under General Discussion.
The error rates shown in Fig. 1 follow the
same pattern as the RTs. The effect of root type
(bound vs. free, 3.6% vs. 4.4%, respectively)
was not significant: F1(1, 79) 3.58, p
.07; F2(1, 472) 1.67, p .10. There was
a 4.6% difference between Root items and
Root items [F1(1, 79) 84.59, p .001;
F2(1, 472) 49.68, p .001]. Similarly,
there was a 3.9% error rate increase for Prefix
items, compared to Prefix items [F1(1, 79) 71.10, p .001; F2(1, 472) 36.95, p
.001]. The interaction between prefix status and
root status was also significant [F1(1, 79)
36.41, p .001; F2(1, 472) 16.60, p
.001].
One of the more informative aspects of the
error data can be found in the Prefix condi-
tions, which we already focused on in the RT
analyses. Prefix, Root items had higher er-ror rates than Prefix, Root items [3% vs.
1%F1(1, 79) 23.38, p .001; F2(1,
236) 5.83, p .05]. This significant dif-
ference underscores the RT result: there appears
to be some activation of the root portion of a
Prefix, Root item, regardless of whether the
root is bound or free.
While the RT patterns were similar for both
free and bound roots, the prefix status roottype interaction was significant by subjects and
approached significance by items: F1(1, 79)
17.23, p .001; F2(1, 472) 3.28, p
.08. However, this test includes all items, and
the free vs. bound manipulation has no real
meaning for the Root items. Therefore, I reran
the interaction including only the Root items
(i.e., those in the right half of both panels of
Fig. 1).Looking first at the Prefix, Root items,
one sees a small RT advantage for items that
carried free roots (221 ms 238 ms 17
ms); for Prefix, Root items, the effect is
large and inhibitory (442 ms 381 ms 61
ms in the opposite direction). For this subset of
the data, the interaction was significant, but only
by subjects [F1(1, 79) 10.52, p .01;
F2(1, 236) 1.75, p .10]. The corre-
sponding interaction on error rates was also
significant in the subjects analysis only: F1(1,
79 ) 5.48, p .05; F2(1, 236) 1.30,
p .10.
The major difference between Prefix items
and Prefix items is that in the former case, the
perceptual system is not expected to attempt
1
To ensure that the results shown are not due to differ-ences in the number of auditory neighbors each kind of
pseudoword has, I calculated the number of words that
differ from each pseudoword by a single phoneme substi-
tution. Zero was the median and modal value for all com-
binations of/Prefix and /Root (78.3% of the stimuli
had 0 neighbors). Furthermore, number of neighbors did not
differ significantly across the eight types of pseudowords,
whether the analysis included only the number of
pseudowords having 0 neighbors (Kolmogorov-Smirnov
Z 1.18, p .10) or data for all of the pseudowords
[2 9.38 (df 7), p .10].2 As mentioned previously, the stimulus-initial phoneme
strings in the Prefix conditions were repeated more often
(12 times each) than those in the Prefix conditions (M
2.4 times each). To ensure that the results obtained were not
due to this difference in repetition, I recalculated perfor-
mance in the Prefix conditions using only each partici-
pants first two encounters with each prefix. While perfor-
mance on the early trials was slower and more variable, the
data patterns are consistent with the results shown in Fig. 1.
261POLYMORPHEMIC PSEUDOWORDS
7/28/2019 Wurm (2000)
8/17
decomposition. Therefore, root type should not
have an influence here. Decomposition is ex-
pected in the Prefix cases, and that is where a
large RT difference was observed. The fact that
it was items with free roots that suffered such a
large inhibitory effect in the context of a genu-ine prefix may illustrate the importance of se-
mantics, discussed earlier in this paper (Libben,
1998; Marslen-Wilson et al., 1994; Schreuder &
Baayen, 1995; Wurm, 1997). The regression
analyses to be reported below lend additional
support to this idea.3
Regression Analyses
RTs were also analyzed using hierarchicalmultiple regression. Only Prefix, Root
pseudowords were analyzed, because these are
the only pseudowords for which it was possible
to get values on all of the regressors. Regression
models assume independence of observations,
which does not hold for the current experiment
because each participant provided more than
one observation. In repeated-measures regres-
sion analyses, this is controlled by the inclusionofN-1 dummy variables (79 in the present case)
that represent the participants. The interested
reader can refer to Cohen and Cohen (1983) for
more details.
After entering the 79 dummy variables, prefix
frequency and root frequency were found to
have inhibitory effects on RTs [F(1, 1967)
8.80, p .01; and F(1, 1967) 8.44, p
.01, respectivelythe large df value in the de-nominator equals the number of participants
times the number of relevant stimuli minus the
number of incorrect critical trials and the num-
ber of previous factors in the model].
The prefix frequency effect can be viewed
one of two ways, both of which rest on the idea
that processing is more difficult in portions of
lexical space that are densely populated (see
Goldinger, Luce, & Pisoni, 1989; Luce, Pisoni,
& Goldinger, 1990). The inhibitory effect of
prefix frequency may be a byproduct of contin-uous processing. Prefix frequency is necessarily
correlated with neighborhood density, so words
with high-frequency prefixes have more neigh-
bors than words with low-frequency prefixes.
Alternatively the prefix frequency effect may
have decompositional underpinnings. A high-
frequency prefix usually attaches to more roots
than a low-frequency prefix does [Wurm (1996)
found a correlation of .75 for these quantities].In general, then, a pseudoword response should
require more time if its prefix has high fre-
quency, because the pool of root candidates
would be relatively large.
The root frequency effect agrees with
Wurms (1997) finding for real prefixed words
and fits with his conclusion that high-frequency
roots compete with the full-forms that carrythem. This conclusion, if correct, would suggest
that the prefix frequency effect is due to the size
of the pool of root candidates and is not simply
a byproduct of continuous processing.
The next effect assessed was that of root type.
Included in the model ahead of root type were
the N-1 dummy variables, prefix frequency, and
root frequency. Pseudowords with free roots
had slower RTs than those with bound roots[439 ms vs. 374 ms; F(1, 1965) 20.48, p
.001; this analysis only considers items from
the Prefix, Root conditionthe overall ad-
vantage for items with bound roots was 21 ms,
significant only by subjects].
The next effect assessed was that of prefix
likelihood. Items higher on prefix likelihood
had slower RTs [F(1, 1965) 4.12, p
.05]. This agrees with the finding of Laudanna
et al. (1994) for Italian pseudowords, presented
visually. Higher semantic transparency was also
associated with slower RTs [F(1, 926)
9.49, p .01this analysis was done for
items with free roots only]. These effects argue
against strict continuous and strict decomposi-
tional models. The behavior of the perceptual
3 The possibility that the prefix status root type inter-
action was due in part to some unidentified aspect of the
materials cannot be completely ruled out, because the in-teraction was also present in the subjects analysis of Root
items [i.e., those in the left half of Fig. 1: F1(1, 79) 7.79,
p .01; F2(1, 236) 1.57, p .10]. This was
unexpected, because as noted above, the root type manipu-
lation is meaningless for Root items. However, the effect
was weaker for these items (3 ms and 42 ms 45 ms)
than it was for the Root items (17 ms and 61 ms 78
ms), and both F ratios were less than 1.0 for the correspond-
ing effect on error rates.
262 LEE H. WURM
7/28/2019 Wurm (2000)
9/17
system seems to be more flexible than those
models suggest.
The next three analyses assessed the interac-
tions between root type and the main effects of
prefix frequency, root frequency, and prefixlikelihood. Root type (bound vs. free) interacted
with prefix likelihood [Fig. 2: F(1, 1963)
6.13, p .05] and root frequency [Fig. 3: F(1,
1963) 13.51, p .001]. The figures show
high and low values based on median splits, but
these dichotomies were not used in the analyses.
This is merely a convenient way to show the
nature of each interaction. A significant interac-
tion indicates that the slope of the relationshipbetween one independent variable and RT
changes as a function of the other independent
variable (Aiken & West, 1991; Cohen & Cohen,
1983, Tabachnick & Fidell, 1989).
Figure 2 shows that the cost in processing
time for items that are good candidates for de-
composition (by virtue of their high prefix like-
lihoods) is more pronounced if the accompany-
ing root is free rather than bound. As was
suggested in connection with Fig. 1, free roots
tend to pay a price for their meaningfulness; the
exact price depends on whether the carrier item
is a good candidate for decomposition, as de-
termined by high or low prefix likelihood.
Figure 3 shows the interaction between root
type and root frequency. The interaction sug-
gests that any free root will slow down rejection
times, but a more complicated situation holds
for bound roots. First, bound roots never slow
down processing to the same extent that free
ones do. Second, the amount of interference
caused by a bound root is related to that rootsfrequency: the higher the frequency, the more
interference.
One three-way interaction was also signifi-
cant. Figure 4 [F(1, 1960) 8.07, p .01]
shows that the two-way interaction shown in
Fig. 3 depends additionally on prefix likelihood.
One interpretation of this interaction, based on
the results of Wurm (1997) for prefixed real
words with free roots, takes as its starting pointthe assumption that the perceptual system learns
over time to associate high prefix likelihood (in
conjunction with other variables) with success-
ful decomposition. Low prefix likelihood would
therefore signal an item that the perceptual sys-
tem should not be inclined to decompose.
We can understand this interaction by look-
ing at the fastest and slowest RTs. The fastest
RTs were for items that are low on prefix like-lihood and carry bound, low-frequency roots.
These are items that the perceptual system
should be disinclined to decompose because of
the low value of prefix likelihood. In addition,
the roots of these items are bound and low in
frequency. Therefore, these stimuli can be re-
FIG. 3. Mean reaction time (RT) as a function of root
type and root frequency, in milliseconds. Error bars show
1 SEM.
FIG. 2. Mean reaction time (RT) as a function of root
type and prefix likelihood, in milliseconds. Error bars show
1 SEM.
263POLYMORPHEMIC PSEUDOWORDS
7/28/2019 Wurm (2000)
10/17
jected quickly. The slowest RTs were for items
high on prefix likelihood that carry high-fre-
quency free roots. These are items that the per-
ceptual system should be inclined to decom-pose, and the resulting root is easily
recognizable. Pseudowords like this are partic-
ularly difficult to reject.
GENERAL DISCUSSION
One finding of the current study that should
be explored more fully is the significant perfor-
mance disadvantage for Prefix, Root items,
compared to Prefix, Root items. It is hard to
determine whether the roots used in the current
study meet Tafts (1994) underspecified crite-
rion for root recognizability: the mean fre-
quency for free roots was 192 (range 4 to
870), while the mean frequency for bound roots
was 119 (range 0 to 619). For comparison,
people [Tafts (1994) example root] has a fre-
quency of 847. It is also worth noting in this
context that the observed performance disad-
vantage was just as large for roots that are
bound (47 ms, vs. 33 ms for free roots).4 This
may be inconsistent with Tafts (1994) hypoth-
esis, insofar as bound roots in general do not
appear to be as recognizable as free ones. In any
event, one specific part of the pattern predicted
by the prefix-stripping model appears to be in-
correct: genuine roots elevate both RTs anderror rates even for pseudowords that do not
begin with genuine prefixes.
The later version of Shortlist (Norris, Mc-
Queen, & Cutler, 1995) might be able to predict
this effect. The Metrical Segmentation Strategy
of Cutler and Norris (1988) was implemented in
Shortlist to accommodate experimental findings
(e.g., Vroomen & de Gelder, 1997; see also
McQueen, Norris, & Cutler, 1994; McQueen,Cutler, Briscoe, & Norris, 1995). Strong sylla-
bles help determine alignment, which is rele-
vant because roots in the current study were
stressed. If strong syllables are used in deter-
mining alignment between a word candidate
and the stimulus input, and if such alignment is
not absolutely crucial in determining activation,
then Shortlist might indeed predict partial acti-
vation for the root of a Prefix, Root item. Aswith Tafts (1994) account, though, this expla-
nation becomes less attractive when we con-
sider that the root effect held for bound roots,
too. Individual morphemes are not represented
in Shortlist unless they also happen to be words,
so it would be hard to explain the origin of this
effect for bound roots.
Another question addressed by this study was
whether pseudowords with free roots are pro-cessed in the same way as those with bound
roots. At a fairly gross level of analysis, one
finds that the performance data in the current
study were quite similar across root types. How-
ever, it would be premature to conclude any-
thing on the basis of those results alone. The
interactions shown in Figs. 24 and the 65-ms
main effect of root type in the regression anal-
ysis indicate that there is something different
about the processing of the two root types. In
addition, the potentially very interesting inter-
action between prefix status and root type4 These effects may in fact be nearly identical in magni-
tude, because of a small difference in the deviation points
for stimuli in these two conditions (the deviation point is the
point at which a pseudoword diverges from all real words in
the language). If RTs are adjusted to reflect this difference,
the sizes of the root effects for Prefix stimuli become 34
ms for items with bound roots and 35 ms for items with free
roots.
FIG. 4. Mean reaction time (RT) as a function of root
frequency, prefix likelihood, and root type (free or bound),
in milliseconds. PL stands for prefix likelihood. Error bars
show 1 SEM.
264 LEE H. WURM
7/28/2019 Wurm (2000)
11/17
for Root items is worthy of further investiga-
tion.
The current study suggests that while bound
roots probably are recognizable entities, their
representations may not be as meaningful or
richly interconnected as those for free roots.The semantic fields normally associated with
lexical entries are essentially empty for bound
roots because they have no clear definitions.
This would predict less computation time for
rejecting an item carrying a bound root, which
is what was found for Prefix, Root
pseudowords.
Tafts (1994) interactive-activation proposal
offers an attractive starting position from whichto explain these data. That model has a level of
representation for bound morphemes (i.e., pre-
fixes and bound roots) and one where all free-
standing words (including polymorphemic
words) are represented. Elements that combine
to make larger words, whether free or bound,
are interconnected. This would predict the
muted effects observed for bound roots; they are
recognizable elements with their own represen-tations, but do not have the same degree of
interconnectedness or combinatorial flexibility
as free elements. A number of different dual-
route models can also accommodate the current
results (e.g., Caramazza et al. 1988; Frauen-
felder & Schreuder, 1992; Laudanna et al.,
1994, 1997; Schreuder & Baayen, 1995; Wurm,
1997).
Future research efforts might use a varietyof strategies to extend what has been learned.
Manipulating the stress pattern of the critical
stimuli in different ways would help deter-
mine whether a model such as Shortlist (par-
ticularly in its second version, which incor-
porates the Metrical Segmentation Strategy
Norris et al., 1995) can be reconciled with
perception data.
Another strategy that may prove useful
would be to use common word beginnings that
are not prefixes. This would help tease apartvarious classes of models. TRACE (McClelland
& Elman, 1986), Cohort (Marslen-Wilson,
1984, 1987; Marslen-Wilson & Welsh, 1978),
and Shortlist (Norris, 1994; Norris et al., 1995)
all predict that common, nonprefix word begin-
nings will have the same processing conse-
quences as prefixes do, because prefixation ef-
fects in those models are essentially cohort
effects. On the other hand, prefix-stripping and
dual-route models predict that there is some-
thing special about prefixes; common word be-
ginnings that are not prefixes will not have the
same perceptual consequences as actual pre-
fixes.
Finally, it should be noted that roots cannot
always be classified as unambiguously as those
used in the current study (e.g., Scalise, 1984;Selkirk, 1982; Siegel, 1979). For example, al-
though English has a free-standing word vent,
many theorists consider it to be a different mor-
pheme than the one found in words like invent
and convent because there is no relationship in
meaning between those words. Taft and Forster
(1975) performed one experiment looking at
this type of root and concluded that the bound
morpheme -vent and the free morpheme ventwere separate entities, stored separately in
memory. It might prove interesting for fu-
ture studies to use not only clearly bound roots,
such as -ceive, but also some of these less clear
cases.
APPENDIX A: CRITICAL PSEUDOWORDS
Prefix Prefix
Root Root Root Root
Stimuli with free roots
adlay [&dleI] adloo [lu] aflay [&f ] afloo
adlead [&dlid] adlod [lAd] udlead [Vd] udlod
adlease [&dlis] adlose [lOUs] aklease [&k ] aklose
265POLYMORPHEMIC PSEUDOWORDS
7/28/2019 Wurm (2000)
12/17
APPENDIX AContinued
Prefix Prefix
Root Root Root Root
adlive [&dlIv] adlave [leIv] odlive [OUd] odlaveadseal [&dsi@l] adseaf [si@f] idseal [Id] idseaf
adstate [&dsteIt] adstote [stOUt] agstate [&g] agstote
cobend [kOUbend] cogend [gend] pobend[pOU] pogend
cobind [kOUbaInd] cobund [bVnd] dobind [dOU] dobund
cocast [kOUk&st] cocaft [k&ft] jocast [dZOU] jocaft
codate [kOUdeIt] codape [deIp] todate [tOU] todape
comix [kOUmIks] cobix [bIks] chomix [tSOU] chobix
coscreen [kOUskrin] coscrone [skrOUn] cooscreen [cu] cooscrone
decap [d@k&p] depap [p&p] pecap [p@] pepap
defit [d@fIt] defot [fAt] sefit [s@] sefot
dejoin [d@dZOIn] depoin [pOIn] doojoin [du] doopoin
depay [d@peI] depoe [pOU] kepay [k@] kepoe
detaste [d@teIst] dedaste [deIst] getaste [g@] gedaste
detreat [d@tSrit] detroot [tSrut] daytreat [deI] daytroot
disact [dIs&kt] diseect [ikt] dosact [dAs] doseect
disbrace [dIsbreIs] disblace [bleIs] kisbrace [kIs] kisblace
disclog [dIsklOg] disclig [klIg] doosclog [dus] doosclig
discook [dIskUk] discoop [kUp] tiscook [tIs] tiscoop
displug [dIsplVg] disklug [klVg] pisplug [pIs] pisklug
distest [dIstest] distesht [teSt] gistest [gIs] gistesht
enclaim [enkleIm] englaim [gleIm] onclaim [An] onglaimenfund [enfVnd] enfunt [fVnt] esfund [es] esfunt
enphrase [enfreIz] enphrooze [fruz] ekphrase [ek ] ekphrooze
enread [enrid] enreat [rit] elread [el] elreat
ensell [ensel] enchell [tSel] oonsell [un] oonchell
ensort [ensOUrt] ensart [sArt] ersort [er] ersart
percount [p@rkaUnt] perpount [paUnt] dercount [d@r] derpount
perjudge [p@rdZVdZ] perjadge [dZ&dZ] terjudge [t@r] terjadge
perlight [p@rlaIt] perlighp [laIp] gerlight [g@r] gerlighp
perplace [p@rpleIs] perprace [preIs] kerplace [k@r] kerprace
persearch [p@rsertS] perfearch [fertS] pensearch [p@n] penfearch
perset [p@rset] persep [sep] pelset [p@l pelsep
preblock [priblAk] preglock [glAk] dreblock [dri] dreglock
prebuild [pribIld] prevuild [vIld] trebuild [tri] trevuild
prechain [pritSeIn] precheen [tSin] plechain [pli] plecheen
prename [prineIm] prenane [neIn] brename [bri] brenane
preprove [pripruv] preprooz [pruz] greprove [gri] greprooz
pretouch [pritVtS] pretaich [teItS] kretouch [kri] kretaich
rebar [r@bAr] redar [dAr] lebar [l@] ledar
rebolt [r@bOUlt] rebalt [bAlt] sebolt [s@] sebalt
recool [r@kul] recoor [kU@r] tecool [t@] tecoor
respeak [r@spik] resteek [
stik] kespeak [k@
] kesteek restress [ristres] restreff [stref] roostress [ru] roostreff
retrust [ritrVst] reprust [prVst] getrust [gi] geprust
transtring [tr&nstrIN] transkring [krIN] tronstring [trOUn] tronskring
transcut [tr&nskVt] transvut [vVt] pranscut [pr&ns] pransvut
transfrost [tr&nsfrOst] transfrest [frest] kransfrost [kr&ns] kransfrest
transprint [tr&nsprInt] transprant [pr&nt] gransprint [gr&ns] gransprant
transtrace [tr&nstreIs] transkrace [kreIs] branstrace [br&ns] branskrace
transword [tr&nswerd] transwurt [wurt] troonsword [truns] troonswurt
unform [VnfOUrm] unforn [fOUrn] ainform [eIn] ainforn
unheat [Vnhit] unkeat [kit] ulheat [Vl] ulkeat
266 LEE H. WURM
7/28/2019 Wurm (2000)
13/17
APPENDIX AContinued
Prefix Prefix
Root Root Root Root
unplay [VnpleI] unploe [plOU] usplay [Vs] usploeunsoak [VnsOUk] unsoat [sOUt] udsoak [Vd] udsoat
unview [Vnviu] unvai [veI] onview [OUn] onvai
unweigh [VnweI] unzeigh [zeI] eenweigh [in] eenzeigh
Stimuli with bound roots
adlect [&dlekt] admect [mekt] aklect [ak ] akmect
adlude [&dlud] adluche [lutS] aflude [&f ] afluche
adnounce [&dnaUns] adnounch [naUntS] odnounce [OUd] odnounch
adstruct [&dstrVkt] adstroct [strAkt] alstruct [&l] alstroct
advince [&dvIns] adzince [
zIns] idvince [Id
] idzinceadvive [&dvaIv] adveve [viv] udvive [Vd] udveve
cofide [kOUfaId] cokide [kaId] pofide [pOU] pokide
cofuse [kOUfius] copuse [pius] tofuse [tOU] topuse
copone [kOUpOUn] copene [pin] dopone [dOU] dopene
coprive [kOUpraIv] coproav [prOUv] cooprive [cu] cooproav
coturb [kOUterb] coturp [terp] choturb [tSOU] choturp
cozert [kOUzert] cozerch [zertS] gozert [go] gozerch
defess [d@fes] dejess [dZes] kefess [k@] kejess
degress [d@gres] degless [gles] tegress [t@] tegless
demit [d@mIt] demip [mIp] gemit [g@] gemip
depel [d@pel] dekel [kel] sepel [s@] sekeldevade [d@veId] dezade [zeId] doovade [du] doozade
devulse [d@vVls] develse [vels] dayvulse [deI] dayvelse
disdict [dIsdIkt] disdect [dekt] gisdict [gIs] gisdect
disfect [dIsfekt] dischect [tSekt] kisfect [kIs] kischect
displode [dIsplOUd] displud [plVd] tisplode [tIs] tisplud
dissume [dIssum] dissule [sul] doossume [dus] doossule
distect [dIstekt] dispect [pekt] pistect [pIs] pispect
distrive [dIstraiV] distroov [truv] dostrive [dAs] dostroov
encise [ensaIs] enfise [faIs] elcise [el] elfise
endain [endeIn] enzain [zeIn] ondain [An] onzain
enpand [enp&nd] engand [g&nd] espand [es] esgand
entain [enteIn] enyain [jeIn] ertain [er] eryain
entract [entSr&kt] entroct [tSrAkt] oontract [un] oontroct
envenge [envendZ] envenche [ventS] ekvenge [ek ] ekvenche
perflect [p@rflekt] perslect [slekt] derflect [d@r] derslect
perpute [p@rpiut] perpite [paIt] terpute [t@r] terpite
persult [p@rzVlt] pervult [vVlt] kersult [k@r] kervult
pervect [p@rvekt] pervoct [vAct] gervect [g@r] gervoct
pervise [p@rvaIz] pervose [vOUz] pelvise [p@l] pelvose
pervolve [p@rvOlv] pervolze [vOlz] penvolve [p@n] penvolze
preject [pridZekt] pregect [gekt] pleject [pli] plegectpreplore [priplOUr] preplere [plI@r] breplore [bri] breplere
preproach [priprOUtS] preproce [prOUs] greproach [gri] greproce
prespect [prispekt] preskect [skekt] krespect [kri] kreskect
prespond [prispAnd] prespond [spOUnd] trespond [tri] trespond
prevert [privert] preverp [verp] drevert [dri] dreverp
recept [r@sept] rechept [tSept] lecept [l@] lechept
reclude [r@klud] reclode [clOUd] rooclude [ru] rooclode
rejunct [r@dZuNkt] repunct [puNkt] gejunct [g@] gepunct
rerupt [r@rVpt] reroopt [rupt] kerupt [k@] keroopt
resorb [r@zOUrb] resork [zOUrk] tesorb [t@] tesork
267POLYMORPHEMIC PSEUDOWORDS
7/28/2019 Wurm (2000)
14/17
APPENDIX B: SEMANTIC TRANSPARENCY AND ROOT FREQUENCY
FOR PREFIX, ROOT PSEUDOWORDS
a
Rated on a 17 scale, for items with free roots only.b Per million tokens, from the CELEX database (Baayen et al., 1993; Burnage, 1990).
Pseudoword
Median semantic
transparencya Root frequencyb
adlay 2 144
adlead 2 555
adlease 2 8
adlive 2 526
adseal 2 18
adstate 2.5 152
cobend 3.5 77
cobind 5.5 26
cocast 5.5 148
codate 5 97
comix 6 133coscreen 4 48
decap 6 44
defit 4.5 151
dejoin 6 158
depay 4 441
detaste 4 22
detreat 5 168
disact 5 717
disbrace 5 49
disclog 6.5 5
discook 4 118
displug 5 16
distest 4 55
enclaim 5 202
enfund 4 70
enphrase 4 50
enread 3 571
ensell 3 153
ensort 4 40
Pseudoword
Median semantic
transparencya Root frequencyb
percount 3 153
perjudge 3 160
perlight 2.5 536
perplace 3 870
persearch 2 236
perset 3.5 347
preblock 5.5 83
prebuild 5.5 453
prechain 4 54
prename 5.5 407
preprove 6 290pretouch 5 120
rebar 4 114
rebolt 7 20
recool 6 79
respeak 6 402
restress 6.5 26
retrust 6 76
transcut 4 251
transfrost 2 16
transprint 4 103
transtrace 2.5 26
transtring 2 5
transword 2 4
unform 5.5 771
unheat 6 168
unplay 5 650
unsoak 5.5 23
unview 4 66
unweigh 5 35
APPENDIX AContinued
Prefix Prefix
Root Root Root Root
revide [r@vaId] revike [vaIk] dovide [dOU] doviketranceive [tr&nssiv] transfeive [fiv] tronceive [trOUns] tronsfeive
tranzide [tr&nzaId] tranzipe [zaIp] granzide [gr&n] granzipe
transcline [tr&nsklaIn] transcrine [kraIn] troonscline [truns] troonscrine
transflict [tr&nsflIkt] transfleect [flikt] bransflict [br&ns] bransfleect
transhort [tr&nshOUrt] transhorp [hOUrp] kranshort [kr&ns] kranshorp
transtinct [tr&nsstiNkt] transpinct [piNkt] pranstinct [pr&ns] pranspinct
uncess [Vnses] unpess [pes] aincess [eIn] ainpess
unfract [Vnfr&kt] unflact [fl&kt] eenfract [in] eenflact
ungest [Vngest ungesk [dZesk] usgest [Vs] usgesk
untrude [Vntrud] ungrude [grud] ultrude [Vl] ulgrude
unvoke [VnvOUk] unveke [vik] onvoke [An] onveke
unzerve [Vnzerv] unzorve [zOUrv] udzerve [Vd] udzorve
268 LEE H. WURM
7/28/2019 Wurm (2000)
15/17
APPENDIX C: PREFIX LIKELIHOOD
AND FREQUENCY
Prefix Prefix likelihood Prefix frequencya
ad- .023 9
co- .010 18de- .008 104
dis- .092 766
en- .092 371
per- .005 484
pre- .013 56
re- .067 1881
trans- .092 47
un- .283 1072
a
Per million tokens, from the CELEX database (Baayenet al., 1993; Burnage, 1990).
REFERENCES
Aiken, L. S., & West, S. G. (1991). Multiple regression:
Testing and interpreting interactions. Newbury Park,
CA: Sage.
Anderson, S. R. (1988). Morphology as a parsing problem.
Linguistics, 26, 521544.
Anshen, F., & Aronoff, M. (1981). Morphological produc-
tivity and phonological transparency. Canadian Jour-
nal of Linguistics, 26, 6372.Aronoff, M. (1976). Word formation in generative gram-
mar. Cambridge, MA: MIT Press.
Baayen, R. H. (1994). Productivity in language production.
Language and Cognitive Processes, 9, 447469.
Baayen, R. H., Piepenbrock, R., & van Rijn, H. (1993). The
CELEX lexical database. [CD-ROM]
Bergman, M. W., Hudson, P. T. W., & Eling, P. A. T. M.
(1988). How simple complex words can be: Morpho-
logical processing and word representations. Quarterly
Journal of Experimental Psychology, 40A, 4172.Bloomfield, L. (1933). Language. New York: Holt, Rine-
hart & Winston.
Burnage, G. (1990). CELEX: A guide for users. Nijmegen,
Netherlands: CELEX.
Butterworth, B. (1983). Lexical representation. In B. But-
terworth (Ed.), Language production: Development,
writing, and other language processes (Vol. 2, pp.
257294). New York: Academic Press.
Bybee, J. L. (1985). Morphology: A study of the relation
between meaning and form. Philadelphia, PA: Ben-
jamins.Bybee, J. L. (1988). Morphology as lexical organization. In
M. Hammond & M. Noonan (Eds.), Theoretical mor-
phology: Approaches in modern linguistics (pp. 119
141). New York: Academic Press.
Bybee, J. L. (1995a). Diachronic and typological properties
of morphology and their implications for representa-
tion. In L. B. Feldman (Ed.), Morphological aspects of
language processing (pp. 225246). Hillsdale, NJ: Erl-
baum.
Bybee, J. L. (1995b). Regular morphology and the lexicon.
Language and Cognitive Processes, 10, 425455.
Caramazza, A., Laudanna, A., & Romani, C. (1988). Lex-
ical access and inflectional morphology. Cognition, 28,
297332.
Chomsky, N. (1965). Aspects of the theory of syntax. Cam-
bridge, MA: MIT Press.Cohen, J., & Cohen, P. (1983). Applied multiple regression/
correlation analysis for the behavioral sciences (2nd
ed.). Hillsdale, NJ: Erlbaum.
Coolen, R., van Jaarsveld, H., & Schreuder, R. (1991). The
interpretation of isolated novel nominal compounds.
Memory & Cognition, 19, 341352.
Cutler, A., Hawkins, J. A., & Gilligan, G. (1985). The
suffixing preference: A processing explanation. Lin-
guistics, 23, 723758.
Cutler, A., & Norris, D. G. (1979). Monitoring sentence
comprehension. In W. E. Cooper & E. C. T. Walker
(Eds.), Sentence processing: Psycholinguistic studies
presented to Merrill Garrett (pp. 113134). Hillsdale,
NJ: Erlbaum.
Cutler, A., & Norris, D. (1988). The role of strong syllables
in segmentation for lexical access. Journal of Experi-
mental Psychology: Human Perception and Perfor-
mance, 14, 113121.
Dennet, D. C. (1987). The intentional stance. Cambridge,
MA: MIT Press.
Emmorey, K. D. (1989). Auditory morphological priming inthe lexicon. Language and Cognitive Processes, 4,
7392.
Forster, K. I. (1994). Computational modeling and elemen-
tary process analysis in visual word recognition. Jour-
nal of Experimental Psychology: Human Perception
and Performance, 20, 12921310.
Francis, W. N., & Kucera, H. (1982). Frequency analysis of
English usage: Lexicon and grammar. Boston: Hough-
ton Mifflin.
Frauenfelder, U. H., & Schreuder, R. (1992). Constraining
psycholinguistic models of morphological processing
and representation: The role of productivity. In G. E.
Booij & J. Van Marle (Eds.), Yearbook of morphology
1991 (pp. 165183). Dordrecht, Netherlands: Kluwer.
Goldinger, S. D., Luce, P. A., & Pisoni, D. B. (1989).
Priming lexical neighbors of spoken words: Effects of
competition and inhibition. Journal of Memory and
Language, 28, 501518.
Grosjean, F., & Gee, J. (1987). Prosodic structure and
spoken word recognition. Cognition, 25, 135155.
Henderson, L. (1985). Toward a psychology of morphemes.In A.W. Ellis (Ed.), Progress in the psychology of
language (Vol. 1, pp. 1572). Hillsdale, NJ: Erlbaum.
Henderson, L., Wallis, J., & Knight, K. (1984). Morphemic
structure and lexical access. In H. Bouma & D. G.
Bouwhuis (Eds.), Attention and performance: Control
of language processes (Vol. X, pp. 211226). Hills-
dale, NJ: Erlbaum.
Jackendoff, R. (1975). Morphological and semantic regu-
larities in the lexicon. Language, 51, 639671.
269POLYMORPHEMIC PSEUDOWORDS
7/28/2019 Wurm (2000)
16/17
Jarvella, R. J., & Meijers, G. (1983). Recognizing mor-
phemes in spoken words: some evidence for a stem-
organised mental lexicon. In G. B. Flores dArcais &
R. J. Jarvella (Eds.), The process of language under-
standing (pp. 81112). London: Wiley.
Kempley, S., & Morton, J. (1982). The effects of priming
with regularly and irregularly related words in auditoryword recognition. British Journal of Psychology, 73,
441454.
Laudanna, A., & Burani, C. (1995). Distributional proper-
ties of derivational affixes: Implications for processing.
In L. B. Feldman, (Ed.), Morphological aspects of
language processing (pp. 345364). Hillsdale, NJ: Erl-
baum.
Laudanna, A., Burani, C., & Cermele, A. (1994). Prefixes as
processing units. Language and Cognitive Processes,
9, 295316.
Laudanna, A., Cermele, A., & Caramazza, A. (1997). Mor-
pho-lexical representations in naming. Language and
Cognitive Processes, 12, 4966.
Libben, G. (1998). Semantic transparency in the processing
of compounds: Consequence for representation, pro-
cessing, and impairment. Brain and Language, 61,
3044.
Luce, P. A., Pisoni, D. B., & Goldinger, S. D. (1990).
Similarity neighborhoods of spoken words. In G. T. M.
Altmann (Ed.), Cognitive models of speech processing:
Psycholinguistic and computational perspectives (pp.122147). Cambridge, MA: MIT Press.
Lyons, J. (1977). Semantics. Cambridge, England: Cam-
bridge Univ. Press.
MacKay, D. G. (1978). Derivational rules and the internal
lexicon. Journal of Verbal Learning and Verbal Be-
havior, 17, 6171.
Marslen-Wilson, W. D. (1984). Function and process in
spoken word recognition. In H. Bouma & D. G. Bou-
whuis (Eds.), Attention and performance: Control of
language processes (Vol. X, pp. 125150). Hillsdale,
NJ: Erlbaum.
Marslen-Wilson, W. D. (1987). Functional parallelism in
spoken word-recognition. Cognition, 25, 71102.
Marslen-Wilson, W. D., Tyler, L. K., Waksler, R., & Older,
L. (1994). Morphology and meaning in the English
mental lexicon. Psychological Review, 101, 333.
Marslen-Wilson, W., & Welsh, A. (1978). Processing inter-
actions and lexical access during word recognition in
continuous speech. Cognitive Psychology, 10, 2963.
McClelland, J. L., & Elman, J. L. (1986). The TRACE
model of speech perception. Cognitive Psychology, 18,186.
McQueen, J. M., Cutler, A., Briscoe, T., & Norris, D.
(1995). Models of continuous speech recognition and
the contents of the vocabulary. Language and Cogni-
tive Processes, 10, 309331.
McQueen, J. M., Norris, D., & Cutler, A. (1994). Compe-
tition in spoken word recognition: Spotting words
within other words. Journal of Experimental Psychol-
ogy: Learning, Memory, and Cognition, 20, 621638.
Morton, J. (1969). The interaction of information in word
recognition. Psychological Review, 76, 165178.
Morton, J. (1979). Facilitation in word recognition: Exper-
iments causing change in the logogen model. In P. A.
Kolers, M. E. Wrolstad, and H. Bouma (Eds.), Pro-
cessing of visible language (Vol. 1, pp. 259268). New
York: Plenum.Norris, D. (1994). Shortlist: A connectionist model of con-
tinuous speech recognition. Cognition, 52, 189234.
Norris, D., McQueen, J. M., & Cutler, A. (1995). Compe-
tition and segmentation in spoken-word recognition.
Journal of Experimental Psychology: Learning, Mem-
ory, and Cognition, 21, 12091228.
Radeau, M., Morais, J., Mousty, P., Saerens, M., & Bertel-
son, P. (1992). A listeners investigation of printed
word processing. Journal of Experimental Psychology:
Human Perception and Performance, 18, 861871.
Rubin, G. S., Becker, C. A., & Freeman, R. H. (1979).
Morphological structure and its effects on visual word
recognition. Journal of Verbal Learning and Verbal
Behavior, 18, 399412.
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, 382405.
Sandra, D. (1994). The morphology of the mental lexicon:
Internal word structure viewed from a psycholinguistic
perspective. Language and Cognitive Processes, 9,227269.
Scalise, S. (1984). Generative morphology. Cinnaminson,
NJ: Foris.
Schreuder, R., & Baayen, R. H. (1994). Prefix stripping
re-revisited. Journal of Memory and Language, 33,
357375.
Schreuder, R., & Baayen, R. H. (1995). Modeling morpho-
logical processing. In L. B. Feldman, (Ed.), Morpho-
logical aspects of language processing (pp. 131154).
Hillsdale, NJ: Erlbaum.
Schreuder, R., & Flores dArcais, G. B. (1989). Psycholin-
guistic issues in the lexical representation of meaning.
In W. D. Marslen-Wilson, (Ed.), Lexical representa-
tion and process (pp. 409 436). Cambridge, MA: MIT
Press.
Seidenberg, M. S. (1987). Sublexical structures in visual
word recognition: Access units or orthographic redun-
dancy? In M. Coltheart (Ed.), Attention and perfor-
mance: The psychology of reading (Vol. XII, pp. 245
263). Hillsdale, NJ: Erlbaum.
Seidenberg, M. S. (1989). Reading complex words. In G.N.Carlson & M.K. Tanenhaus (Eds.), Linguistic structure
in language processing (pp. 53105). Amsterdam: Re-
idel.
Selkirk, E. O. (1982). The syntax of words. Cambridge, MA:
MIT Press.
Siegel, D. (1979). Topics in English morphology. New
York: Garland.
Smith, P. T. (1988). How to conduct experiments with mor-
phologically complex words. Linguistics, 26, 699714.
270 LEE H. WURM
7/28/2019 Wurm (2000)
17/17
Smith, P. T., & Sterling, C. M. (1982). Factors affecting the
perceived morphemic structure of written words. Journal
of Verbal Learning and Verbal Behavior, 21, 704721.
Stanners, R. F., Neiser, J. J., & Painton, S. (1979). Memory
representation for prefixed words. Journal of Verbal
Learning and Verbal Behavior, 18, 733743.
Tabachnick, B. G., & Fidell, L. S. (1989). Using multivar-iate statistics (2nd ed.). New York: Harper Collins.
Taft, M. (1981). Prefix stripping revisited. Journal of Verbal
Learning and Verbal Behavior, 20, 289297.
Taft, M. (1985). The decoding of words in lexical access: A
review of the morphographic approach. In D. Besner,
T.G. Waller, & G.E. Mackinnon (Eds.), Reading re-
search: Advances in theory and practice (Vol. 5, pp.
83123). New York: Academic Press.
Taft, M. (1994). Interactive-activation as a framework for
understanding morphological processing. Language
and Cognitive Processes, 9, 271294.Taft, M., & Forster, K. I. (1975). Lexical storage and
retrieval of prefixed words. Journal of Verbal Learning
and Verbal Behavior, 14, 638647.
Taft, M., Hambly, G., & Kinoshita, S. (1986). Visual and
auditory recognition of prefixed words. Quarterly
Journal of Experimental Psychology, 38A, 351366.
Tyler, L. K., Marslen-Wilson, W. D., Rentoul, J., & Han-
ney, P. (1988). Continuous and discontinuous access in
spoken word-recognition: The role of derivational pre-
fixes. Journal of Memory and Language, 27, 368381.Vroomen, J. & de Gelder, B. (1997). Activation of embed-
ded words in spoken word recognition. Journal of
Experimental Psychology: Human Perception and
Performance, 23, 710720.
Wurm, L. H. (1996). Auditory processing of prefixed En-
glish words. Unpublished doctoral dissertation, State
Univ. of New York at Stony Brook.
Wurm, L. H. (1997). Auditory processing of prefixed En-
glish words is both continuous and decompositional.
Journal of Memory and Language, 37, 438461.
(Received January 12, 1999)
(Revision received August 30, 1999)
271POLYMORPHEMIC PSEUDOWORDS
Top Related