You are on page 1of 24

Journal of ExperimentalPsychology: Copyright 2000 by the American PsychologicalAssociation, Inc.

learning, Memory,and Cognition 0278-7393/00/$5.00 I)131:10.1037//0278-7393.26.2.489


2000, Vol. 26, No. 2, 489-511

The Balance of Storage and Computation in Morphological Processing:


The Role of Word Formation Type, Affixal Homonymy, and Productivity
Raymond Bertram Robert Schreuder and R. Harald Baayen
University of Turku Universityof Nijmegen

This article is concerned with the way in which the balance of storage--storing and processing
words through full-form representations--and computation~storing and processing words
through morpheme-based representations--in lexical processing in the visual modality is
affected by the following 3 factors: word formation type (roughly, inflection vs. derivation),
productivity, and affixal homonymy. Experimental results for 5 different Dutch suffixes,
combined with previous results obtained for 4 comparable Finnish suffixes (R. Bertram, M.
Lalne, & K. Karvinen, 1999) and 2 Dutch suffixes (R. H. Baayen, T. Dijkstra, & R. Schreuder,
1997), show that none of these factors in isolation is a reliable cross-linguistic predictor of the
balance of storage and computation. The authors offer a general framework that outlines how
morphological processing is influenced by the interaction of word formation type, productiv-
ity, and affixal homonymy.

Current research on morphological processing in lan- ties. Using visual lexical decision, they measured response
guage comprehension reveals a growing awareness that latencies to nonwords consisting of a nonexisting base word
various linguistic properties of complex words profoundly and a real affLx. Reaction times to such words increased with
influence the way that these words are processed. Research- the increasing length of the affix. Also, affixes that occurred
ers have pointed out that prefixed words are processed in many different word types gave rise to longer response
differently than suffixed words Co16, Beanvillain, & Segui, latencies. Conversely, nonwords with affixes with a high
1989; Cutler, Hawkins, & Gilligan, 1985; Marslen-Wilson, confusability (defined as the percentage of word tokens for
Tyler, Waksler, & Older, 1994; Taft, 1994). Others have which the final orthographic string is homographic with the
called attention to the relevance of the distinction between affix but does not in fact denote the affix) were responded to
inflectional and derivational morphology in lexical process- faster than nonwords with low-confusability affixes. These
ing (Niemi, Laine, & Tuominen, 1994; Schriefers, Frie- distributional properties determine--what these authors
derici, & Graetz, 1992; Taft, 1994). Marslen-Wilson et al. call the orthographic salience of the affix. The more
(1994) point to the relevance of semantic transparency. More salient an affix is, the more wordlike the nonwords become,
recently, the individual properties of the afftxes themselves leading to longer rejection latencies.
have become subject to investigation. Landanna and Burani Bertram, Laine, and Karvinen (1999) show that linguistic
(1995) and Burani, DOvetto, Thornton, and Laudanna (1997) properties of individual alftxes, such as productivity and
investigated a number of distributional affix-specific proper- homonymy, are important factors in the processing of real
words. Derived Finnish words with an unambiguous, produc-
tive suffix were responded to faster than were monomorphe-
mic control words. However, derived words with either an
Raymond Bertram, Department of Psychology, University of unproductive or homonymic suffix (a suffix that is ambigu-
Turku, Turlm, Finland; Robert Schreuder and R. Harald Baayen, ous in that it serves more than one semantic function) were
Interfaculty Research Unit for Language and Speech, University of
responded to equally as fast as monomorphemic controls.
Nijmegen, Nijmegen, the Netherlands.
We thank four anonymous reviewers, Matti Laine, and Jukka The authors explain this finding in terms of statistical
Hytn~ for their helpful comments on an earlier version of this facilitation (Frauenfelder & Schreuder, 1992; Raab, 1962)
article, and Leonore Biegstraten for her help in conducting the that occurs for words with productive and unambiguous
experiments. This study was financially supported by the Academy affixes but that is absent for words with unproductive or
of Finland Grant 27774 to Matti Laine, by the Centre of Interna- homonymic affixes. Bertram et al. (1999) claim that derived
tional Mobility (CIMO, Grant to Raymond Bertram), by the Finnish words with an unambiguous, productive sttfftx are
Finnish Graduate School of Psychology Grant of the Finnish processed simultaneously on the basis of full-form represen-
Ministry of Education to Raymond Bera'am, and by the Dutch tations and on the basis of their morphological constituents.
National Research Council (Nederlandse Organisatie voor Weten- Because of overlap in the distributions of the processing
schappelijk Onderzoek, Persoonsgerichte Impuls voor Onder-
zoeksgroepen Met Nieuwe Ide~n voor Excellente Research, Grant times of the two lexical routes, these complex words can be
to R. Harald Baayen). responded to faster than can morphologically simplex words.
Correspondence concerning this article should be addressed to For simplex words, which by their nature have only
Raymond Bertram, Department of Psychology, University of full-form representations, no statistical facilitation can take
Turku, FIN-20520, Turku, Finland. Electronic mail may be sent to place. The fact that derived words with either an unproduc-
rayber@utu.fi. tive or homonymic suffix are responded to equally as fast as

489
490 BERTRAM,SCHREUDER,AND BAAYEN

monomorphemic words indicates that both word types are activation level of that word as a whole at some level of
processed in the same way, through full-form representa- representation. Effects of base frequency reveal that the
tions only. The authors concluded that homonymy induces constituents of a complex word, notably its base word, play a
storage, whereas productivity enhances morphological functional role in lexical processing. When such effects
parsing. arise, the common assumption is that the lexical representa-
The Italian and the Finnish studies show that distribu- tion of the base word is activated during lexical processing.
fional and linguistic properties of individual affixes affect the These two frequency effects are not mutually incompatible.
way in which words with these affixes are processed. An In parallel dual route models of morphological processing
additional complicating factor that may be relevant for the (Baayen, Dijkstra, & Schreuder, 1997; Schreuder & Baayen,
balance of storage and computation concerns the role of 1995), for instance, lexical access may take place both by
cross-linguistic differences between morphological systems. decomposition into the constituent morphemes (leading to a
Finnish, for instance, has a rich inflectional morphology, base frequency effect) and by retrieval from memory of the
whereas the inflectional system of Dutch is extremely full form (leading to a surface frequency effect).
simple. In Finnish, a given word can occur in hundreds or The levels of representation on which surface and base
even thousands of different inflectional variants (Karlsson & frequency effects are located differ from model to model.
Koskenniemi, 1985), whereas in Dutch at most 10 inflec- For instance, the base frequency effect is located at the
tional variants are available. With the combinatorial explo- access level and the surface frequency effect at a more
sion of possible forms in Finnish, the likelihood of extensive central leveP in the prefix stripping framework of Taft and
storage of such forms in the mental lexicon diminishes Forster (1976) and Taft (1979) as well as in the model
considerably. Indeed, until now hardly any empirical evi- outlined in Co16 et al. (1989). For the augmented addressed
dence has been obtained for storage of inflected words in morphology (AAM) model, Burani and Caramazza (1987)
Finnish (Bertram et al., 1999; Hy6n~ Laine, & Niemi, 1995; stated the following:
Niemi et al., 1994). By contrast, Baayen, Dijkstra, and Briefly, this model assumes that while lexical representations
Schrender (1997) reported evidence for extensive storage of in the orthographic(input) lexicon are morphologicallydecom-
regular noun plurals in Dutch (see also Serene & Jongman, l~sed--i.e., roots or stems are represented independently of
1997, for similar findings for English). This suggests that the inflectional affixes and, possibly, derivational affixes--the
access procedure operates with both whole-word and mor-
balance of storage and computation is co-determined by the pheme access units. Previously experienced words activate
lexical-stafistical properties of a given language. To explore whole-word access units while the recognition of novel words
the possible role of cross-linguistic factors on the balance of occurs through the activation of morphemic access units.
storage and computation, the present study addresses the We concluded that the AAM model locates base fre-
potential effects of morphological productivity, affix bom- quency effects at a more central level and takes surface
onymy, and word formation type (roughly, inflection vs. frequency effects for known words to arise at the level of
derivation) for Dutch, paralleling the study of Bertram et al. access representations. In our own theoretical framework,
(1999) for Finnish. We have investigated the following five we interpret both frequency effects as arising at the level of
Dutch suffixes: the productive inflectional suffix -re, which access representations.
marks singular past tense on verbs; the unproductive deriva- The experiments reported in the present article do not
fional suffix -te, with which abstract nouns are derived from allow for a decision between these three different theoretical
adjectives (e.g., warmte, warmth); the productive deriva- accounts. They have been designed to address a different
tional suffix -he/d, which like derivational -re creates issue, namely, to gain insight into the balance of storage and
abstract nouns from adjectives (e.g., leegheid, emptiness); computation for regular morphologically complex words.
the productive inflectional suffix -er, which forms compara- Traditional approaches to this issue have ranged from no
fives (e.g., warmer, warmer); and the productive deriva- storage under any circumstances to ubiquitous storage. For
tional suffix -er, wl'dch forms nouns from verbs (e.g., denker, instance, Pinker (1991) and Clahsen, Eisenheiss, and Sonnen-
thinker). smhl-Henning (1997) claimed that all regularly inflected
For each of these suffixes we have investigated the role of words are always processed on the basis of their base words
storage by varying the surface frequency of the complex and that full-form access representations are absent for such
words while keeping the base frequency constant; and we words. In terms of our frequency diagnostics, this view
have investigated the role of computation by varying the implies the presence of an effect of base frequency and the
frequency of the b a s e word while keeping the surface absence of an effect of surface frequency for regularly
frequency constant. This method was introduced by Taft
(1979) and has been used by various other researchers as
well (e.g,, Baayen, Dijkstra, & Sc.heuder, 1997; Bradley, 1979; t We use the terms access level and more central level to denote
Burani & CaramaT~,a, 1987; Burani, Salmaso, & Carumu~u: the early perceptual level of processing and representation on the
one hand, and the later linguistic levels of processing and represen-
1984; Col~ et al., 1989; Schreuder & Baayen, 1997). tation on the other hand. Similar distinctions are made by, for
The logic underlying this kind of method is as follows. instance, Taft (1979), who calls the first level the orthographic file
Effects of surface frequency reveal familiarity of the process- and the second level the master file, and by Burani and C..arsma~Ta
ing system with the complex word as a whole. A common ( 1987; e.g., the augmented addressed ~ o g y model), who call
theoretical assumption is that the information concerning the the first level the access level and the second level the orthographic
frequency of use of a complex word is coded into the resting input lexicon.
STORAOE AND COMvtrrAT~ON 491

inflected words. Conversely, Butterworth (1983) and Seiden- effects of parsing and no effects of storage for the inflec-
berg (1987) argued that complex words are extensively tional -re.
stored and that their base words have little or no role to play. By contrast, we expected the mirror image pattern for the
Thus, this view predicts no effect of base frequency and derivadonal -re. This suffix, although fully regular, is not
ubiquitous effects of surface frequency. productive, and it is much less frequent than its inflectional
The aim of the present article is to show that both of these homonym. Moreover, instead of adding only syntactically
views are too restricted and that the balance of storage and relevant information without changing the meaning of the
computation is co-determined by factors such as word base word, the derivational -re also involves meaning-
formation type, productivity, and affixal homonymy. We changing morphology. (For instance, words such as warmte
present a framework for how these factors interact in the [warmth] have a base word denoting a property, but they
General Discussion section using both the present experimen- themselves embody the notion of measurement.) All of these
tai results as well as data obtained for a number of factors conspire to favor storage rather than parsing.
comparable affixes of Finnish reported in Bertram et ai.
(1999). Experiment 1
Experiments 1 and 2 address the role of storage and
computation for words with the homonymic suffix -re, Method
which is used as an inflectional suffix to express the simple Participants. Eighteen undergraduate students from the Univer-
past tense and as an unproductive, but otherwise fully sity of Nijmegen were paid to participate in Experiment la.
regular, derivational suffix to express abstract nouns. Experi- Twenty-seven different undergraduate students from the same
ment 3 focuses on the processing of words with the university were paid to perform Experiment lb. All were native
productive derivational suffix -heid, which, like -re, forms speakers of Dutch and had normal or corrected-to-normal vision.
abstract nouns. In Experiments 4 and 5, we investigate the Target materials for Experiment l a. Forty inflected verb forms
way in which words with the homonymic suffix - e r are with the singular past tense suffix -re were selected from the
processed. As in English, the --er suffix forms either CELEX lexical database, of which 20 had a relatively high surface
comparatives or agent nouns. The task used in each experi- frequency average of 5.3, whereas the other 20 had a low surface
ment is the visual lexical-decision task. frequency average of 1.4 (all token frequency counts reported are
scaled to one million). The frequency difference between the two
conditions is significant, t(38) = 5.8, p < .00001. The two sets
E x p e r i m e n t s With the Suffix - t e were matched for base frequency 0aigh surface, 15.8; low surface,
15.4), family size of the bas¢~ (high surface, 15.8; low surface,
In Experiment la, we investigate the role of surface 15.4), geometric mean bigram frequency (high surface, 13.7; low
frequency for the inflectional singular past tense suffix -re surface, 13.9), and word length in letters (high surface, 6.5; low
using a factoriai design with base frequency kept constant in surface, 6.9). Here, and in all the other experiments reported, items
the mean between the high and low surface frequency were matched on an item-by-item basis. In other words, pairs of
conditions. We define the base frequency as the summed items were selected that were comparable on all the variables
mentioned previously except for the manipulated frequency value,
frequencies of the base word itself and its inflectional in this case the surface frequency.
variants. In Experiment Ib, we use a similar design to Target materials for Experiment lb. Forty inflected verb forms
investigate the role of base frequency for inflectional -re, with the singular past tense suffix -re were selected from the
this time keeping mean surface frequency constant across CELEX database, of which 20 had a high base frequency (23.0)
conditions. In Experiment 2, we investigate the role of and the other 20 a low base frequency (5.1). The two sets were
surface and base frequency for derivational -re using a matched for surface frequency (high base, 2.7; low base, 2.3),
correlational design because there are only 73 formations family size of the base (high base, 16.5; low base, 16.8), geometric
with -re in the CELEX lexical database of 42 million word mean bigram frequency (high base, 13.9; low base, 13.7), and word
tokens (Baayen, Piepenbrock, & Gnlikers, 1995), of which length in letters (high base, 6.4; low base, 6.6). The materials for
17 are so obscure that they cannot be used for experimenta- these (and all of the other experiments reported in this article) are
listed in Table A1 in the Appendix. Figures A1 and A2 in the
tion. The remaining items are too few to construct orthogo- Appendix show by means of scatterplots how our experimental
nal contrasts between surface and base frequency. data points in the experiments manipulating surface frequency
For the inflectional -re, we did not expect an effect of (Figure A1) and the experiments manipulating base frequency
surface frequency, but only an effect of base frequency, for (Figure A2) have been selected from the population of available
the following reasons. First, it is a fully regular and very words meeting the constraint of having an unambiguous Germanic
productive suffix. Second, it has a much higher frequency of monomorphemic base word. In both figures, the upper right panel
occurrence than its unproductive derivationai homonym. shows the scatterplots for the inflectional-re.
Third, the inflectional -re specifies deictic tense and person- Filler materials for Experiment 1. The same filler material was
number marking, adding only syntactically relevant informa- added for Experiments la and lb. The additional 100 filler words
tion without changing the meaning of the base word. We will
refer to this kind of inflection as meaning-invariant morphol- 2 The family size of the base is the number of derived and
ogy. For a similar inflectional suffix, -en, in its use as a plural compound words with a given base word as a constituent.
marker on past tense verbs, Baayen, Dijkstra, and Schreuder Schrender and Baayen (1997) show that this is a relevant factor in
(1997) reported the total absence of a surface frequency lexical processing and that the relevance is type-based rather than
effect. These considerations lead us to expect maximal token-based.
492 BF.,RTRAM, SCHREUDER, AND BAAYEN

consisted of 10 monomorphemic nouns; 40 inflected nouns; 30 and base frequencies of the individual inflected verb forms,
verbs, of which 10 were in the nominative case, whereas the others using linear models of the form,
were some kind of inflection (7 participles, 7 third-person singu-
lars, and 6 singular plurals); and 20 derived words, of which 10 Reaction Tune (RT) = a* log (base frequency + 1)
were with the feminine suffix -in, and 10 were with the productive
de-adjectival suffix -he/d. For each of the 140 words, a nonword
was obtained by changing one to three letters so that the phonotac- + b* log (surface frequency + 1) + c. (1)
tics of Dutch were not violated. The nonwords had a similar
structure as the real words in that 40 of them were formed with the We use log(frequency + 1) rather than simply log(fre-
target suffix -re, 20 were without morphological structure, and the quency) because our counts include frequencies equal to
remaining 80 nonwords involved exactly the same suffixes that zero, for which the logarithmic transformation is undefined.
were used for the filler words in exactly the same proportions. In what follows, we will refer to the effect of base frequency
Procedure. For the lexical-decision task, 2 participants were and surface frequency in the linear model with the tacit
tested at a time in noise-proof experimental booths. They were to assumption that the frequency counts (per 42 million) are
decide as quickly and as accurately as possible whether a letter logarithmically transformed as in Equation 1. A stepwise
string appearing on the computer screen was a real Dutch word or
regression analysis revealed a significant coefficient for base
not. Each stimulus was preceded by a fixation mark in the middle of
the screen for 500 ms. After 50 ms, the stimulus appeared at the frequency only (for the RTs: base frequency, a = - 4 4 . 6 7 ,
same position. Stimuli were presented on NEC MultiSync ® color p < .01; surface frequency, b = - 1 . 5 4 , p > .8; throughout
monitors in white lowercase 18-point Helvetica letters on a dark this paper we present Bonferroni-adjusted p values), Note
background, and they remained on the screen for 1,500 ms. The that by itseK our frequency contrast (5.3 vs. 1.4) was
maximum response time was 2,000 ms from stimulus onset. Fifteen powerful enough to yield substantial frequency effects in
practice trials, 8 words and 7 nonwords, preceded the actual visual lexical decision; Baayen, Dijkstra, and Schreuder
experiment. The experiment itself was divided into two blocks of (1997) reported a reliable 63-ms effect for regular plural
140 items (each block contained 70 words and 70 nonwords). There nouns with the suffix - e n for a similar frequency contrast of
was a short pause between the two blocks. The experiment lasted 4 vs. 1. Because our post hoc analysis also did not reveal any
approximately 20 rain.
dependency of the response latencies on surface frequency,
we concluded that surface frequency does not play a role for
Results and Discussion our inflected words with the regular past tense suffix -re. The
Experiment la. The data for all of the participants were next experiment tested whether the effect of base frequency
included in the analyses, as all performed with an overall observed in this post hoc analysis could be replicated with a
error rate below 15%. All items elicited error rates below factorial design.
30%, which is our criterion for inclusion in the analyses. The Experiment lb. The data for all of the participants were
observations were used to calculate the mean response included in the analyses because they all performed with an
latencies and error scores for the different test conditions overall error rate below 15%. Two items elicited error rates
(see Table 1). As expected, neither a paired t test for higher than 30% and were therefore excluded from the
participants nor a standard two-sample t test for items analyses. Here, and in all of the other experiments in this
showed a significant difference between the response laten- article, exclusion of error-prone items did not alter the
cies of two conditions, t~(17) = 1.5,p > .1 and t2(38) = 0.7, matching of the two target groups in a sitmiricant way. The
p > .1, two-tailed tests. Also the error scores did not reveal a remaining observations were used to calculate the mean
significant difference, t2(38) = 1.6, p > .1. In a post hoc response latencies and error scores for the two conditions, as
analysis, we further analyzed the dependence of the re- can be seen in Table 1. As expected, the verbs with a high
sponse latencies on the logarithmically transformed surface base frequency were recognized significantly faster than
were the verbs with a low base frequency, fi(26) = 3.8, p <
.001; t2(36) = 2.5, p < .01, one-tailed tests. In addition, the
low-frequency condition elicited significantly more errors
Table 1 than did the high-frequency condition, tz(36) = 2.88, p <
Mean Response Latencies With Standard Deviations and .01. 3 A post hoc analysis supported the dependence of the
Error Percentages for Inflected Verbs With the Singular response latencies on only the base frequency of the
Past Tense Su_~x-re, With High Versus Low Surface
Frequency (Experiment l a), and With High Versus Low
Base Frequency (Experiment 1b)
3 One might argue that no effect was found in Experiment la and
Frequency Reaction Error that a significant effect in Experiment lb was induced by a
(per million) Manipulated time SD (%) difference in power for the two experiments because we used 18
Experiment la participants in Experiment la and 27 participants in Experiment lb.
To exclude this possibility, we reanalyzed Experiment lb with the
High (5.3) Surface 651 54 1.7 first 18 participants only. The pattern of results was exactly the
Low (1.4) Surface 665 64 4.2 same as with 27 participants. Participants responded to high base
Experiment lb frequency words much faster than they did to low frequency words
(634 ms vs. 678 ms), ti(17) = 3.4,p < .005; t2(36) -- 1.9,p < .05)
High (23.0) Base 630 45 3.9 and elicited less errors as well (3.3% vs. 15.4%), tt(17) = 5.3,
Low (5.1) Base 685 63 14.1 p < .001; t2(36) = 2.9,p < .005).
SrORAnE AND coMptrrAriON 493

individual inflected verb forms. Stepwise regression analy- storage and computation for words with the derivational -te
ses revealed a significant coefficient for base frequency and has shifted toward storage. This dependency on storage may
a zero coefficient for surface frequency (for the RTs: base be induced by the presence of a very productive and frequent
frequency, a = - 2 4 . 4 5 , p < .0001; surface frequency, homonym on the one hand, and by a lack of productivity of
b = 0). We interpret these results as indicating that for the the suffix itself on the other hand. Bertram et al. (1999)
inflectional -re the balance of storage and computation is showed that words with the unproductive but unambignous
strongly in favor of computation. A similar result for another and regular loeational derivational Finnish suffix -1A (el.
verbal inflectional suffix was obtained in Baayen, Dijkstra, English --cry in bakery) are fully dependent on storage. This
and Schreuder (1997) for Dutch. For Finnish, Bertram et al. cross-linguistic comparison suggests that a lack of productiv-
(1999) reported evidence for full parsing with locative ity by itself is already sufficient to induce storage. The next
inflectional suffixes. For evidence of parsing for inflected experiment addresses whether the balance of storage and
words in English, see Taft (1994). Contrary evidence for computation is similar for the unambiguous and productive
storage of inflected words in English can be found in Sereno Dutch suffix -heid or whether storage is as ubiquitous as it is
and Jongman (1997). We return to this issue in the General for its near synonym, the derivational -re.
Discussion section.
In the next experiment, we shift our attention from the Experiments With the Suffix - h e i d
domain of inflection to the domain of derivation using a
suffix with the same orthographic form as in the preceding In Experiment 3a, we investigate the role of surface
experiments. Because the derivational -re is unproductive frequency for the derivational suffix -heid using a factorial
and has a very productive inflectional homonym, we ex- design with the base frequency kept constant in the mean
pected to find exactly the opposite to what we found in between the high- and low-surface-frequency conditions. In
Experiment 1, (i.e., no effect of base frequency and a solid Experiment 3b, we use a similar design to investigate the
effect of surface frequency). role of base frequency but now keeping mean surface
frequency constant across conditions.
Experiment 2
Experiment 3
Method
Method J

Participants. Sixteen undergraduate students from the Univer-


sity of Nijmegen were paid to participate in the lexical-decision Participants. Sixteen undergraduate students from the Univer-
experiment. All were native speakers of Dutch and had normal or sity of Nijmegen were paid to participate in Experiment 3a, and 16
corrected-to-normal vision. None had participated in any of the different undergraduate students were paid to perform in Experi-
previous experiments. ment 3b. All were native speakers of Dutch and had normal or
Target materials. As mentioned previously, there are only 73 corrected-to-normal vision. None had participated in the previous
words with the derivational suffix -te in the CELEX lexical experiments.
database. Fifty-six of these words could be used for a correlational Target materials for Experiment 3a. Forty nouns with the
experimental design. These 56 words include a few formations de-adjectival suffix -held were selected from the CELEX database,
with nontransparent readings (e.g., groente, green-tit i.e., veg- of which 20 had a high surface frequency (26.4), whereas the other
etables). The materials are listed in Table A1 in the Appendix. The 20 had a low surface frequency (0.9). The two sets were matched
upper left panel of Figures A1 and A2 in the Appendix summarize for base frequency (high surface, 291; low surface, 289), family
the frequential properties of our critical words. (Both panels plot size of the base (high surface, 53.0; low surface, 48.7), geometric
the same data of this correlational design with the variables on the mean bigram frequency (high surface, 13.4; low surface, 13.5), and
axes interchanged.) word length in letters (both 8.7). The materials are listed in the
Filler materials. In this experiment we used 84 filler words, all Appendix.
of which were also used in Experiment 1. Sixteen filler words (10 Target materials for Experiment 3b. Fifty nouns with the
nominal inflections and 6 derived words in the suffix -/n) that were de-adjectival suffix -heid were selected from the CELEX database,
used in Experiment 1 were not included in Experiment 2. The of which 25 had a high base frequency (207.2), whereas the other
nonwords in the present experiment were constructed along the 25 had a low base frequency (24.9). The two sets were matched for
same lines as those of Experiment 1. Of these nonwords, 56 ended surface frequency (both 1.0), family size of the base word (high
in the suffix -re to match the number of real words in -re. base, 14.2; low base, 14.0), geometric mean bigram frequency
Procedure. The procedure was identical to that of Experiment 1. (high base, 13.4; low base, 13.3), and word length in letters (high
base, 8.6; low base, 8.1). The materials are listed in Table A 1 in the
Results a n d Discussion Appendix. The center panels of Figures A1 & A2 in the Appendix
show how we have sampled the words of Experiments 3a and 3b.
The data for all of the participants were included in the Filler materials for Experiment 3. The same set of 100 filler
analyses, as they all performed with an overall error rate words were selected both for Experiment 3a and Experiment 3b.
below 15%. Eighteen items with an error rate exceeding The set of filler words was different from that used for the first two
experiments, but it was similar in design in that it consisted of
30% were excluded from the analyses. A stepwise regression
words belonging to a variety of word types and word categories.
analysis with Equation I as the underlying linear model There were 25 monomorphemic nouns (10 in Experiment 1); 40
revealed a significant coefficient for surface frequency only inflected nouns (40 in Experiment 1); 20 verbs (30 in Experiment
(for the RTs: surface frequency, a = - 21.61, p < .001; base 1), 5 of which were in the nominative case (10 were in the
frequency, b = 0). These results show that the balance of nominative case in Experiment 1); 5 participles (7 in Experiment
494 BERTRAM, SCHREUDER,AND BAAYEN

1); 5 third person singulars (7 in Experiment 1); 5 plural inflections the basis of full-form representations, with a subsidiary role
(6 in Experiment 1); 5 nominative adjectives (new for this for parsing only. To tease apart these possibilities, we used a
experiment); and 10 derived words (20 in Experiment 1) with the factorial design in Experiment 3b in which we controlled for
de-verbal suffixes --sel(5) and-ing (5). For each of the 140 words, surface frequency while maximizing the contrast in base
we constructed nonwords along the same lines as outlined in frequency.
Experiments 1 and 2.
Procedure. The procedurewas identicalto that of Experiment1. Experiment 3b. For the lexical-decision task, the data
for all of the participants were included in the analyses, as
they all performed with an overall error rate below 15%.
Results and Discussion Three target words elicited error rates higher than 30% and
Experiment 3a. The data for all of the participants were were excluded from the analyses. The remaining observa-
included in the analyses, as they all performed with an tions were used to calculate the mean response latencies and
overall error rate below 15%. One target word elicited an error scores for the two experimental conditions. Response
error rate higher than 30% and was excluded from the latencies and errors are listed in Table 2.
analyses. The remaining observations were used to calculate Words in -held with a high-frequency base word were
the mean response latencies and error scores for the different recognized faster than were words with a low-frequency
test conditions, as can be seen in Table 2. Both a paired t test base word. This effect is significant in the by-participant
for participants and a standard two-sample t test for items analysis but not in the by-item analysis, t~(15) = 2.2, p <
revealed a significant difference between the two conditions; .05; t2(45) = 0.9, p > .1. Also the error analysis showed no
the derived words with a high surface frequency were difference between the two conditions, t2(45) = 1.5, p > .1.
recot,niTed significantly faster than were those with a low However, a stepwise regression analysis for the RT data
surface frequency, tl(15) = 9.8,p < .001; t2(37) = 4.8,p < using Equation 1 as the basic model revealed significant
.001. Moreover, significantly more errors were made in the coefficients for both independent variables (base frequency,
low-frequency condition than were made in the high- a = - 1 3 . 1 3 , p < .05; surface frequency, b = - 1 7 . 3 3 ,
frequency condition, tz(37) = 3.4, p < .005. A stepwise p < .001). Note that the coefficient for surface frequency is
regression analysis with Equation 1 as the underlying linear larger than that for base frequency, as in Experiment 3a. This
model suggests that not only surface frequency but also base may explain why the effect of base frequency did not emerge
frequency determines the response latencies (base fre- reliably in the by-item factorial analysis; the variance
quency, a = -12.38, p < .01; surface frequency, introduced by surface frequency masked the variance be-
b = -18.27, p < .001). It is clear that surface frequency is cause of base frequency. As an effect of base frequency also
an important determinant of the response latencies for words emerged in the post hoc regression analysis of Experiment
in -he/d. At the same time, we also observed an effect of 3a, we concluded that base frequency indeed plays a role in
base frequency in the response latencies that, given its the lexical processing of nouns in -heid, albeit to a lesser
smaller coefficient in the linear model, appears to be extent than does surface frequency.
somewhat weaker than that of surface frequency. This may Some researchers have argued that morphological decom-
be a direct consequence of our experimental design, in position takes place only for neologisms (Caramazza, Lan-
which we have attempted to keep base frequency constant in danna, & Romani, 1988). Because our experimental materi-
the mean across the two surface frequency conditions. als contained 8 words that have a surface frequency of zero
Hence, it is possible that for words in -heid, in general, base in the CELEX lexical database, and which can be considered
frequency is a more important factor than surface frequency. as good approximations of neologisms for our participants, 4
Alternatively, the balance of storage and computation might we had the opportunity to investigate this possibility. If
in fact be in favor of lexical processing of nouns in -heid on Caramazza et al. (1988) are correct, base frequency should
have disappeared as an independent effect in our linear
model when words with a surface frequency of zero were
excluded from the analysis. Interestingly, the linear model,
Table 2 Equation 1, applied to the remaining words with frequency
Mean Response Latencies With Standard Deviations and greater than zero, still showed a reliable effect for base
Error Percentages for Derived Nouns With the frequency (base frequency, a = -15.32, p < .05; surface
De-Adjectival Sufftx-heid, With High Versus Low Surface frequency, b = -30.35, p < .01). This suggests that the role
Frequency (Experiment 3a), and With High Versus Low of parsing extends beyond the mere handling of neologisms.
Base Frequency (Experiment 3b) Curiously enough, excluding neologisms from the linear
Frequency Reaction Error model leads to a substantial increase in the coefficient of the
(per million) Manipulated time SD (%) surface frequency effect from -17.33 to -30.35. One
Experiment 3a would expect that excluding the lowest frequency items,
High (24.4) Surface 554 40 0.6
Low (0.9) Surface 644 72 6.5 4 A CELEX frequency of zero means that these words are
Experiment 3b registered in comprehensive dictionaries of Dutch but do not
appear in the corpus underlying the CELEX frequency counts with
High (207.2) Base 646 76 8.3 a probability greater than 1 in 42 million. Hence, these words were
Low (24.9) Base 664 72 4.9 prime candidates to be neologisms for our participants.
STORAGEAND COMPUTATION 495

which should elicit the longest response latencies, would full-form access representations for words with such idiosyn-
lead to a smaller coefficient and not a higher one. We cratic meanings is that they provide an efficient means for
therefore inspected the individual item means using a accessing these meanings. However, -he/d is a productive
scatterplot with log(surface frequency + 1) on the horizon- and unambiguous suffix that leads one also to expect the
tal axis, and with RT on the vertical axis, as shown in Figure effects of parsing, especially for the many fully composi-
1. A non-parametric, robust locally weighted scatterplot tional formations (see Baayen & Neijt, 1997, for a detailed
smoother (Cleveland, 1979) for the full frequency range, corpus-based study of this suffLx and its semantic properties)
plotted as a dotted line, shows an initial positive correlation for which parsing can provide all required semantics. From
followed by a negative correlation. In fact, this initial this point of view, it is not surprising that base frequency
positive correlation is supported by a t test comparing the appears as a significant second factor in our experiments.
item means of the neologisms with the item means of the Similarly, Bertram et ai. (1999) observed effects of both
words with log(surface frequency + 1) in the range of 1 to 2, surface and base frequency for the Finnish suffix -stO,
t2(22) = 2.3, p < .05. Possibly, we are dealing with two which is--as is the Dutch suffix -he/dmalso productive and
distinct subsets, as shown by the solid lines in Figure 1. This unambiguous.
initial positive correlation explains why the linear coefficient The next experiments investigated if affixal homonymy
of surface frequency was larger when the neologisms were shifts the balance of storage and computation further in the
excluded. direction of storage when two productive affixes are in-
At the same time, this finding raises the question of why it volved, as observed by Bertram et ai. (1999) for the
seems to be more difficult to process words in -heid with homonymic Finnish suffix -jA.
very low surface frequencies than to process neologisms.
Our tentative explanation is that we are observing the effects
Experiments With the Suffix - e r
of early lexical competition between the access representa-
tions of the higher frequency base words on the one hand As in English, the Dutch suffix - e r has two functions.
and the weak representations of the very low-frequency full Attached to adjectives it denotes the comparative (e.g.,
forms on the other hand. When such pairs of access warmer, warmer), and attached to verbs it denotes subject
representations are in competition with each other, lexical nouns (as with main semantic classes or agents, e.g., meier,
processing is slowed down. Conversely, when there is no rower; and instruments, e.g., opener, opener; for details see
fuU-form representation at all, as is the case for our Booij, 1986). Both homonyms are productive, but the
neologisms, no such competition takes place, leading to inflectional - e r accounts for a majority of the 64% of all
slightly faster response latencies. Clearly, further experimen- string tokens in the CELEX lexicai database with the suffix
tal research is required here. -er. Experiment 4a investigated the role of surface fre-
Summing up, Experiment 3 shows that there is a solid quency, and Experiment 4b investigated the role of base
effect of surface frequency for the suflZx -heicL This strong frequency for the derivationai -er. Experiments 5a and 5b do
effect of surface frequency, observed for one of the most the same for the inflectional -er.
productive derivationai suffixes of Dutch, is probably best There are two reasons why we expected to find more
understood in terms of the word formation type involved. evidence for parsing for the inflection --er than for the
Derivation involves meaning-changing morphology, which derivationai -er. First, inflected words are prime candidates
often goes hand in hand with the accretion of idiosyncratic for parsing, given their full compositionality and their
aspects of meaning (Sandra, 1994). The functionality of simple semantics, whereas subject-noun formation leads to
several related but distinguishable meanings such as agent
and instrument. Second, comparatives are encountered more
9O0 often than are derived words in --er. This may also be a
• QO 0 relevant factor given the results of Baayen, Dijkstra, and
Schreuder (1997) who showed for the Dutch plural suffix
-en that the numerically stronger homonym is parsed by
"~ 700 .......
default, whereas its rival is extensively stored.
6OO
Experiment 4
5OO
0 1 2 3 4 5 6 Method
Log(SurfaceFrequency+l) Participants. Sixteen undergraduate students from the Univer-
sity of Nijmegen were paid to participate in Experiment 4a, and 16
Figure 1. Reaction time (RT) as a function of log(surface different undergraduate students from the same university were
frequency + 1) for the derivational suffix -heid. The dotted line is paid to perform Experiment 4b. All were native speakers of Dutch
a non-parametric locally weighted scatterplot smoother for the full and had normal or corrected-to-normai vision. None had partici-
frequency range. The solid line represents the same smoother for pated in any of the previous experiments.
the nonzero surface frequencies. The horizontal line segment for Target materialsfor Experiment4a. Forty derived nouns with
the words with zero frequency represents the mean RT for these the de-verbal suffix--er were selected from the CELEX database, of
words. which 20 had a relatively high surface frequency (2.8), whereas the
496 BERTRAM, SCHREUDER,AND BAAYEN

other 20 had a very low surface frequency (0.1). The two sets were experiment maximizes the contrast in base frequency for the
matched for base frequency (high surface, 75.7; low surface, 76.9), derivational --er, with the usual factorial design.
family size of the base word (high surface, 29.4; low surface, 27.1), Experiment 4b. The data for all of the participants were
geometric mean bigram frequency (high surface, 14.1; low surface, included in the analyses, as they all performed with an
14.0), and word length in letters (high surface, 6.3; low surface,
overall error rate below 15%. Six items elicited an error rate
6.4). The materials are listed in the Appendix.
Target materialsfor Experiment 4b. Forty derived nouns with higher than 30% and were excluded from the analyses. The
the de-verbal suffix -er were selected from the CELEX database, of remaining observations were used to calculate the mean
which 20 had a high base frequency (456.8), whereas the other 20 response latencies and error scores for the two test condi-
had a low base frequency (39.0). The two sets were matched for tions. Response latencies and errors are listed in Table 3.
surface frequency (high base, 0.7; low base, 0.9), family size of the Both a paired t test for participants and a standard two-
base (high base, 33.8; low base, 30.2), geometric mean bigram sample t test for items failed to reveal any significant
frequency (high base, 14.2; low base, 14.0), and word length in difference in response latencies between words with a high
letters (high base, 6.4; low base, 6.6). The materials are listed in base frequency and words with a low base frequency,
Table A 1 in the Appendix. The bottom left panels of Figures A 1 and h(15) = 1.2,p > .1; t2(32) = 0.3,p > .1. Post hoe stepwise
A2 in the Appendix show how we have sampled our target words
multiple regression analyses using Equation 1 as the under-
for Experiments 4a and 4b.
Filler materialsfor Erperiment 4. The same 100 filler words lying model revealed a n effect of surface frequency only
that were used in Experiment 3 were used for Experiments 4a and (RT: a = 0, b = - 15.23, p < .03). Similarly, the error rates
4b. The 140 nonwords were constructed in a similar way as in the of the high- and low-frequency conditions did not differ,
previous experiments. h(32) = 0.4, p > .1. We concluded that the presence of a
Procedure. The ~ was idcnticai to that of Experinaent 1. numerically stronger homonymic inflectional rival in combi-
nation with the semantic diversity of the derivational - e r had
Results and Discussion shifted the balance of storage and computation completely in
favor of storage. This result parallels the findings of Bertram
Experiment 4a. The data for all of the participants were et al. (1999) who found evidence for storage for the Finnish
included in the analyses, as they all performed with an equivalent of the derivational --er:. the suffix -jA. Tiffs suffix
overall error rate below 15%. Four items elicited an error also has a productive inflectional homonym, which is the
rate higher than 30% and were excluded from the analyses. partitive plural case marker. Our hypothesis was that the
The remaining observations were used to calculate the mean presence of a productive inflectional rival induced storage to
response latencies and error scores for the two test condi- avoid on-line resolution of the ambiguity of a given suffix.
tions. Response latencies and errors are listed in Table 3. As The next experiment studies the role of surface and base
expected for a derivational suffix with a wide range of frequency for the comparative -er, in which we expected to
meanings, both a paired t test for participants and a standard find a solid effect of base frequency and no effect of surface
two-sample t test for items showed that words with a high frequency, just as with the inflectional -re.
surface frequency were responded to significantly faster than
were words with a low surface frequency, tl(15) = 3.9, p <
.01; t2(34) = 2.6, p < .05.5 Post hoe linear regression models Experiment 5
using Equation 1 revealed an effect of surface frequency Method
only (RT: a = 0; b = - 14.93, p < .02). Note that this result
contrasts with that obtained for -heid in Experiment 3, in Participants. Seventeen undergraduate students from the Uni-
which we observed an effect of base frequency in the versity of Nijmegen were paid to participate in Experiment 5a,
regression analysis. The error rates of both conditions did whereas 16 different undergraduate students from the same univer-
sity were paid to perform Experiment 5b. All were native speakers
not differ significantly, t2(34) = 0.2, p > .1. The next
of Dutch and had normal or corrected-to-normal vision. None had
participated in any of the previous experiments.
Target materialsfor Experiment 5a. Forty comparatives were
Table 3
selected from the CELEX database, of which 20 had a high surface
Mean Response Latencies With Standard Deviations and frequency (18.9), whereas the other 20 had a low surface frequency
Error Percentages for Derived Nouns With the De-Verbal (1.1). The two sets were matched for base freqneney (high surface,
Sufftx-er, With High Versus Low Surface Frequency 147.6; low surface, 147.5), family size of the base (high surface,
(Experiment 4a), and With High Versus Low Base 17.8; low surface, 18.4), geometric mean bigram frequency (for
Frequency (Experiment 4b) both, 14.0), and word length in letters (for both, 6.7). The materials
are listed in the Appendix.
Frequency Reaction Error Target materialsfor Experiment 5b. Forty comparatives were
(per million) Manipulated time SD (%) selected from the CELEX database, of which 20 had a high base
Experiment 4a frequency (146.0), whereas the other 20 had a low base frequency
High (2.8) Surface 603 62 2.3
Low (0.1) Surface 659 66 2.6
Experiment 4b 5 Again, the significant effect was not dependent on the words
with zero-frequencies in the low-frequency condition. Excluding
High (505.5) Base 685 59 9.4 these items did not change the pattern of results: response latencies,
Low (48.3) Base 676 87 8.2 615 ms vs. 655 ms; tl(15) = 2.7,p < .02; t2(28) = 2.2,p < .04.
S ~ R A G B A N D COMPUTATION 497

(21.0). The two sets were matched for surface frequency (both 1.3), analysis. The remaining observations were used to calculate
family size of the base (high base, 14.9; low base, 15.0), geometric the mean response latencies and error scores for the two
mean bigram frequency (both 14.0), and word length in letters conditions, as can be seen in Table 4. Both a paired t test for
(high base, 7.1; low base, 6.3), The materials are listed in Table A1 participants and a standard two-sample t test for items did
in the Appendix. The bottom right panels of Figures A1 and A2 in not reveal any significant difference between words with a
the Appendix show how we have sampled the target words of
Experiment 5. high base frequency and words with a low base frequency,
Filler materials for Experiment 5. The 100 filler words were neither in the response latencies, tl(15) = 0.2, p > 0.1;
the same as the ones used in the previous two experiments, except t2(33) = 0.1,p > 0.1, nor in the error rates, t2(33) = 1.2,p >
for the 40 inflected nouns, of which 30 were replaced with 0.1. Post hoe linear regression analyses using Equation 1 as
adjectives (20 nominatives and 10 superlatives) to ensure that the the underlying model confirmed that the RTs depend on
target words were not the only adjectives in the experiment. The surface frequency only (a = 0, b = -26.19, p < .001). The
nonwords were constructed along the same lines as in the previous presence of solid effects of storage and the absence of any
experiments. effects of parsing in the reaction time data for the inflectional
Procedure. The procedurewas identicalto that of Experiment1. suffix - e r was probably due to affixal homonymy. Unlike in
the case of the inflectional -re, for which the derivational
Results and Discussion rival is completely unproductive, the rival derivational suffix
of the inflectional --er is fully productive. For productive
Experiment 5a. The data for all of the participants were
rival suffixes, Baayen, Dijkstra, and Schreuder (1997)
included in the analyses, as they all performed with an
argued that a mere subcategorization conflict can already
overall error rate below 15%. One item elicited an error rate
induce storage. Within the framework of parallel dual-route
higher than 30% and was excluded from the analyses. The
modelling, our hypothesis was that the resolution of affixal
remaining observations were used to calculate the mean
homonymy between two productive affixes is so time costly
response latencies and error scores for the different test
that the direct route is almost always first to complete, not
conditions, as can be seen in Table 4.
only for the derivational --er, but also for the comparative
Surprisingly, both a paired t test for participants and a
-er. With response latencies depending on the processing
standard two-sample t test for items showed that compara-
times of the first route to win the race---for -er, always the
fives with a high surface frequency were recognized signifi-
direct route---no effect of parsing was visible. Our general
cantly faster than were comparatives with a low surface
conclusion with respect to the results for the comparative
frequency, h(16) = 3.0, p < .01; t2(37) = 2.2, p < .05.
- e r is that the disambiguation problem caused by affixal
Moreover, the error analysis indicated that the low-
homonymy was so severe that massive storage was induced
frequency condition elicited more errors than did the high-
even for this completely regular and compositional suffix.
frequency condition, t2(37) = 2.0, p = .06. In addition, post
What we do not yet know is if affixal homonymy might
hoc linear regression analyses on the response latencies
likewise induce storage for the Finnish inflectional suffix
using Equation 1 as the model resulted in a significant effect
for surface frequency only (RT: a = 0, b = - 1 6 . 3 4 ,
-jA, the partitive plural, and the homonymic with the
derivational subject noun marker -jA. We are currently
p < .001). These results suggest extensive storage and no
discernable effect of parsing for comparatives, a pattern that investigating whether there is no storage for this inflectional
is completely opposite to that for the inflectional -re. suffi× just as we observed no storage for the past tense suffix
Experiment 5b. The data for all of the participants were -re, or whether homonymy induces massive storage just as
for the comparative suffi× -er. Our hypothesis was that the
included in further analyses, as they all performed with an
lexical statistics of Finnish would not allow such massive
overall error rate below 15%. Five items elicited error rates
storage.
higher than 30% and were therefore excluded from further

Methodological Issues
Table 4
Mean Response Latencies With Standard Deviations and Before proceeding to the General Discussion section,
Error Percentages for Inflected Words With the three methodological issues require discussion. First, the
Comparative Sufftx -er, With High Versus Low Surface absence of a surface frequency effect for inflected words in
Frequency (Experiment 5a), and With High Versus Low -re (Experiment 1) does not allow us to conclude that
Base Frequency (Experiment 5b ) storage does not occur at all for such words. Although our
experimental words cover a substantial range of surface
Frequency Reaction Error frequencies, as shown in the upper right panel of Figure A1,
(per million) Manipulated time SD (%)
a small number of very high-frequency-inflected words were
Experiment 5a not included in the experiment. Although the conclusion that
High (18.9) Surface 577 38 2.4 storage does not seem to play a role for inflected words in-re
Low (1.1) Surface 612 60 6.2 holds for the bulk of such words, the high-frequency outliers
Experiment 5b might still be stored with their own access representations,
which would then be an exceptional property given the
High (146.1) Base 624 61 3.1 morphological category of all inflected words in -re as a
Low (21.3) Base 626 58 5.5
whole.
498 BV_J~'~, scI-meuD~P., AND BAAY~

Table 5 target words without any fillers or nonwords. The results of


Mean Log Frequencies (With Standard Deviation and this separate series of factorial subjective frequency experi-
Range) and Pearson Correlation Coe~icients of Base ments, summarized in Table 6, exactly mirrors the pattern of
F~:equencyand Reaction Tone (rBase) and of Surface the ~factorial RT data. The congruence between this off-line
Frequency and Reaction ~ne (rSurface) for All of the task and the lexical-decision task shows that list composition
Suffzxes Used Within the Surface Frequency Range [2:6] in is an unlikely source of our pattern of results. Furthermore,
the Surface Frequency F_~eriments and Within the Base the congruence of the two tasks shows that this pattern is
Frequency Range [4:8] in Base Frequency Experiments quite robust.
Mean A final objection against our methodology raised by one
Suffix frequency SD Range rBase rSurface of the reviewers is that two samples of words that differ
markedly in surface frequency might in fact differ in base
surm. r u ~ m ~ ma~ [2:6] ~at ~ ~ rm~mey eximinmus
frequency when remeasured and vice versa. In other words,
Vte (V) 4.46 0.98 2.48-5.96 -0.51 [-0.29] the phenomenon of regression toward the mean might affect
Ate (N) 4.01 1.31 2.20--5.97 [-0.11] -0.67 the matching for base frequency of sets of words with
A-he/d(N) 4.51 10.3 2.48--5.94 [-0.18] -0.51
V--er(N) 3.98 1.23 2.08--5.91 [-0.01] -0.47 different surface frequencies and vice versa. In response to
A--er(A) 4.16 1.29 2.08-5.97 [-0.16] -0.42 this potential problem, we have to distinguish between
measurement error on the one hand and the phenomenon of
B~ ~ n u ~ [4:8]in me t~sc fiequcncyc,~oaim~ts
regression toward the mean on the other hand.
V--ce(V) 6.06 1.06 4.25-7.83 -0.38 [-0.22] First consider measurement error. The question that we
Ate (N) 7.05 0.83 5.08-8.00 [-0.48] -052 should ask ourselves here is how our surface frequencies and
A-heM(N) 6.69 0.79 5.20-7.98 [-0.34] -0.51
V--er(N) 6.62 1.09 4.03-7.89 [-0.54] -0.61' our base frequencies might change ff we counted these
A--er(A) 6.31 1.11 4.51-7.90 -0.52 -0.60 frequencies for exactly the same experimental words in
Note. V -- verb; N ffi noun; A = adjective. another corpus of 42 million words. Fortunately, lexical
'One outlier was t~-movedfrom the analysis. statistics estimates for the expected frequency of a word in a
new corpus with the same size are available (Baayen, 1996;
Church & Gale, 1991; Good, 1953). For a word that occurs
Second, we have studied 5 affixes with different frequen- m times in a corpus of N tokens, the expected frequency in
tial properties. Our experimental materials reflect these another corpus of N tokens, m*, is estimated by
differences, leading to differences in the range of surface and
base frequencies used in the individual experiments. This m* = (m + 1)*E[V(m + 1, N)]/E[V(m, N)],
raises the question of to what extent it remains possible to
compare results across experiments. To make sure that our in which E[.] is the expectation operator and V(tm N)
results are not a consequence of these differences in denotes the number of types in a corpus of N tokens that
frequency ranges, we re-analyzed all of the experiments occur with frequency n~ Various techniques are ava)iAhle for
using multiple regression for fixed frequency ranges, chosen estimating the expectations E[V(m + 1, N)] and E[V(m, N)]
such that a maximum number of words from all experiments (see Chitashvili & Baayen, 1993; Church & Gale, 1991;
could be included. For the experiments manipulating surface Gale & Sampson, 1995). For word frequency data, m* is
frequency, we selected the log frequency range 2:6 (see smaller than m. This is because corpora are samples that do
Figure AI); for the experiments manipulating base fre- not exhaustively sample all possible word types. These
quency, we selected the log frequency range 4:8 (see Figure unseen word types have a joint probability P of being
A2). Table 5 summarizes mean, standard deviation, and sampled equal to P -- E[V( 1,N)]/N, the total number of types
range for each suffix in both kinds of experiments together occurring once only divided by the size in tokens of the
with the Pearson correlation coefficients of base frequency
and RT (rBase) and of surface frequency and RT (rSurface).
6 In the past, subjective frequency (or familiarity) ratings have
Correlations listed between square brackets were removed in been used to check whether corpus-based frequency counts were
a stepwise analysis. The basic pattern of results that emerged reliable. However, recent experiments have revealed that subjec-
from this limited data set is identical to that observed on the tiv~efrequency ratings are not only sensitive to surface frequency,
basis of the factorial studies. The inflectional -re revealed an but also to the morphological family size (see Footnote 1) both in
effect of base frequency and no effect of surface frequency. Dutch (Schreuder & Baayen, 1997) and in English(Baayan, Lieher,
Nominal --re and the - e r suffixes revealed only surface & Schreuder, 1997). Although it is an off-line task, the subjective
frequency effects, and --he/d revealed the effects of both. We frequency rating task apparently taps into various aspects of lexical
concluded that the differences in the frequency ranges representation and processing. The experiments summarized in
between our experiments were unlikely to have caused the Table 6 show that, just as lexical decisions, subjective frequency
ratings are also occasionally sensitive to differences in base
specific pattern of results that we have observed. frequency when surface frequency and morphological family size
Third, differences in list composition (selection of filler are controlled for. We assume that if access happens through the
materials and nonwords) might in some way have influenced base, the frequency of the base is felt in the rating. If access takes
our results. However, each factorial RT experiment was place through the whole-word form, the frequency of the base does
paralleled by a subjective frequency ratine (on a 7-point not matter, and the rating should be a function of the surface
scale) with other participants using only the experimental frequency only.
STORAOEAND coMvtrrAr~oN 499

Table 6
Mean Frequency Ratings (With Standard Deviation and Range) for All of the Sufftxes
Used in the Surface Frequency and Base Frequency Experiments
Mean rating for Mean rating for
high-frequency low-frequency
Suffix condition SD condition SD Test statistic
Surface frequency experiments
V-re (V) 4.8 1.2 4.2 1.1 t2(38)= 1.6,p > .1
A t e (N) stepwise: a = .60, p < .001
A-he/d (N) 5.9 0.9 3.1 1.0 t2(37)*= 8.9, p < .001
V--er(N) 5.1 1.3 3.5 0.8 t2(37)*= 4.7, p < .001
A-er (A) 5.4 0.9 3.7 1.2 t2(38)= 5.3, p < .001
Base frequency experiments
V-re (V) 4.2 1.0 3.1 1.4 t2(38)= 2.9, p < .005
A t e (N) stepwise: b = 0.00,p > .1
A-heM(N) 3.3 1.3 3.2 1.2 t2(47)*= 0.3,p > .1
V-er (N) 3.6 1.4 3.8 1.4 t2(37)*= 0.4, p > . 1
A--er (A) 4.7 1.0 4.0 1.7 t2(38)= 1.6, p > .05
Note. *Because of a programming error, one item that was used in the corresponding reaction time
(RT) experiment was dropped in the subjective frequency rating experiment. V = verb; N = noun;
A = adjective.

corpus (Good, 1953). When we estimate word probabilities to the question of ecological validity, significant by-item
on the basis of their sample relative frequencies, we leave no analyses allowed us to assume that similar results would be
probability space for these unseen types. Therefore, all observed for words with similar frequential properties. We
relative sample frequencies slightly overestimate the popula- make no claims with respect to the lexicon as a whole, as our
tion probabilities. Given that the probability of a word W aim was to examine very specific parts of the lexicon,
with frequency m is less than m/N, its frequency in another exactly those parts that allowed us to test our hypotheses
corpus of size N will be less than m as well. For instance, for concerning lexical processing.)
our corpus of 42 million tokens, a word appearing with a
frequency of 4 is estimated to have an expected frequency of General Discussion
3.851 in other corpora of 42 million words. For a higher
frequency such as 600, the Good-Turing estimate equals The question that we have addressed in this paper is how
598.978. For the large corpus that underlies our counts, the the three factors, affixal homonymy, productivity, and word
measurement error is very small indeed; it minimally affects formation type, affect the balance of storage and computa-
our frequency counts, and it does not at all affect between- tion in visual lexical processing. Experiments 1 and 2
set contrasts under pairwise matching. It will be clear that showed that the productive inflectional past tense suffax -re
measurement error does not invalidate our methodology. is processed exclusively by means of the parsing mute,
Next consider the potential problem of regression toward whereas its unproductive derivational homonym is pro-
the mean. Imagine two words, A and B, with the same base cessed exclusively by the direct mute. Experiment 3 showed
frequency, but with A having a higher surface frequency than that for the productive derivational suffix -he/d, both routes
B. Might it not be the case that the probability that the base operated in parallel. Experiments 4 and 5 revealed extensive
frequency of A has been underestimated will be higher than storage and no effect of parsing for the productive deriva-
the probability that it has been overestimated? If we look at tional suffix - e r and its productive inflectional homonym,
the very same words A and B and their expected frequencies the comparative - e r
of occurrence in other corpora, the answer to this question is Table 7 summarizes the present results that we obtained
clearly no. As we have seen previously, the base frequencies for Dutch, as well as the results of Baayen, Dijkstr~ and
of A and B (which are matched in this example) are both Schreuder (1997) for the Dutch plural suffax-en and those of
slightly overestimated, as are, in fact, their surface frequen- Bertram et al. (1999) for Finnish. Our present results are
cies. It is only when we randomly select other words, say C remarkably similar to those for Finnish, and as of yet do not
and D, with the same surface frequency contrasts as A and B, support the hypothesis that the balance of storage and
that it is likely that one will find a higher base frequency for computation is a priori biased toward computation for
Word C compared with Word D. But this is a completely Finnish, the language with the more complex morphological
different question that concerns the ecological validity of our system. Table 5 tabulates word formation type, productivity,
data sets and not the reliability of our methodology and the and affixal homonymy, and whether the experiments show
way it uses contrasts in one frequency measure and match- evidence for storage, parsing, or both. In what follows we
ing for another frequency count to gain insight in the role of will trace the effect of each of these factors on the balance of
storage and computation in lexical processing. (With respect storage and computation.
500 BERTRAM, $CHREUDER, AND BAAYEN

Table 7
Summary of Affixal Properties and the Role of Storage and Computation for the Dutch
and Finnish Languages
Word formation
Suffix type Productive Homonymic Storage Computation
Dutch
V-re Invariant yes yes no yes
A-re Changing no yes yes no
A--~/d Changing yes no yes yes
V-er Changing yes yes yes no
A-er Adding yes yes yes no
N-en Adding yes yes yes yes
V-en Invariant yes yes no yes
Finnish
N-ssA Invariant yes no no yes
V-jA Changing yes yes yes no
N-stO Changing yes no yes yes
N-/A Changing no no yes no
Note. This table summarizes data from the present article and from Baayen, Dijkstra, and
Schreuder (1997), which both focus on Dutch, and it summarizes data from Bertram et al. (1999),
which focuses on Finnish. V = verb; N = noun; A = adjective.

Word Formation Type AffixalHomonymy


It has been argued that inflected words are always parsed Bertram et al. (1999) claimed that affixal h o m o n y m y
and that derived words are always stored (Niemi et al., 1994; triggers storage. Table 7 shows that this statement is
Taft, 1994). Our data show that inflection is not a necessary somewhat too strong in the sense that an unproductive
nor a sufficient condition for parsing. As can be seen in Table homonymic rival suffix (the derivational -te) does not
5, parsing can take place for derivational suffixes (-he/d, induce storage for words with its productive homonymic
-stO), whereas at the same time massive storage can be counterpart (the inflectional-te).
present for inflectional suffixes (inflectional -er). Instead of It is clear that the balance of storage and computation
strictly distinguishing between inflection and derivation, it cannot be captured in terms of broad single-factor generaliza-
seems more useful to consider word formation type as a tions. From this perspective, full-parsing models (e.g., Taft
scalar dimension with meaning-invariant morphology at the and Forster, 1975) and full-storage models (Butterworth,
oneextreme (e.g,, person and number marking on verbs) and 1983, Caramazza et al., 1988) describe logical possibilities
meaning-changing morphology (e.g., subject-noun forma- that are indeed realized in Dutch and Finnish, but in certain
tion) at the other extreme, with meaning-adding morphology circumstances only. As theories of morphological processing
(e.g., diminutive and comparative formation, noun pluraliza- in general, they are too restrictive.
tion) at an intermediate position (see Baayen, Lieber, & Figure 2 presents the decision tree that depicts how the
Schreuder, 1997, for extensive discussion of noun pluraliza- three factors interact to determine the balance between
tion as meaning-adding morphology). With this more fine- storage and computation. First, note that productivity is a
grained analysis of the dimension, wordformation type, we necessary but not a sufficient condition for parsing. Second,
can formulate the hypothesis that parsing is more likely to the existence of a productive rival homonym induces storage
take place for words with productive meaning-invariant irrespective of word formation type. Only for productive
affixes without productive semantic homonyms. affixes without a productive rival homonym does the distinc-
tion between meaning-invariant morphology on the one
hand and meaning-adding or meaning-changing morphol-
Productivity ogy on the other hand seem to become relevant. Under these
It has similarly been argued that words with productive circumstances, complex words with meaning-invariant mor-
affixes are always parsed and never stored, whereas words phology revealed effects of parsing only, whereas complex
with unproductive suffixes are always stored and never words with meaning-adding or meaning-changing morphol-
parsed (Anshen & Aronoff, 1988, 1997). Table 7 indeed ogy showed effects of both storage and parsing.
shows no evidence for parsing for unproductive suffixes (the We have been able to come to these conclusions by
derivafional -te, the derivationa1-1A), but storage can take combining the data from research involving Dutch and
place for productive suffixes. It is rampant for the hom- Finnish, which typologically are very different and unrelated
onymic suffixes --er and -jA, and it also plays an extensive languages, and it is encouraging to find that these cross-
role alongside parsing for the suffixes --hem and -stO. linguistic data converge in a consistent manner. It is clear,
STORAGEAND COMPUTATION 501

growth curve of the vocabulary. Computational Linguistics, 22,


productive~? 455-480.
Baayen, R. H., Dijkstra, T., & Schreuder, R. (1997). Singulars and
plurals in Dutch: Evidence for a parallel dual mute model.
Journal of Memory and Language, 36, 94-117.
Baayen, R. H., Lieber, R., & Schreuder, R. (1997). The morphologi-
[storage I (homonym?~ cal complexity of simplex nouns, Linguistics, 35, 861--877.
Baayen, R. H., & Neijt, A. (1997). lh'oductivity in context: A case
study of a Dutch suffix. Linguistics, 35, 565-587.
Baayen, R. H., Piepenbrock, R., & Gulikers, L. (1995). The
CELEX lexical database [CD-ROM]. University of Pennsylva-
storage I nia, Philadelphia, PA: Linguistic Data Consortium.
Bertram, R., Laine, M., & Karvinen, K. (1999). The interplay of
word formation type, affixal homonymy, and productivity in
V-er, A-er lexical processing: Evidence from a morphologically rich lan-
I parsing V-jA guage. Journal of Psycholinguistic Research, 28, 213-226.
Booij, G. E. (1986). Form and meaning in morphology: The case of
Dutch 'agent nouns.' Linguistics, 24, 503-517.
-heid, -stO V-re, -ssA Bradley, D. C. (1979). Lexical representation of derivational
N-en V-en relations. In M. Aronoff & M. L. Kean (Eds.), Jlmcture (pp.
37-55). Cambridge, MA: Mrr Press.
Figure 2. Decision tree for parsing and storage. Productive: Is the Burani, C., & Caramazza, A. (1987). Representation and process-
affix productive (+) or unproductive (-); Homonym: Does the ing of derived words. Language and Cognitive Processes, 2,
have a productive rival affix with a different semantic function 217-227.
(+) or not (-); Invariant: Is the Word Formation Type meaning- Burani, C., Dovetto, M., Thornton, A. M., & Laudanna, A. (1997).
invariant (+) or not (-). Accessing and naming suffixed pseudo-words. In G. E. Booij &
J. van Made (Eds.), Yearbook of Morpholosy 1996 (pp. 55-73).
Norwell, MA: Kluwer Academic.
Burani, C., Salmaso, D., & Caramazza, A. (1984). Morphological
however, that our conclusions are only based on 11 suffixes, structure and lexical access. ~sible Language XVIII, 4, 342-
and that the theory embedded in the decision tree of Figure 2 352.
requires further empirical justification. More affixes from a Butterworth, B. (1983). Lexical representation. In B. Butterworth
wider range of languages need to be investigated experimen- (Ed.), Language production: Vol. II. Development, writing and
tally. Moreover, there are other factors that need to be taken other language processes (pp. 257-294). London: Academic
Press.
into account, such as the prefix-suffix distinction, the Caramazza, A., Laudanna, A., & Romani, C. (1988). Lexical access
specific semantic properties of individual affixes, the distri- and inflectional morphology. Cognition, 28, 297-332.
butional properties of affixes, and the computational complex- Chitashvili, R. J., & Baayen, R. H. (1993). Word frequency
ity of different morphological systems. For instance, Figure distributions. In G. Altmann & L. Hl'ebf~ek(Eds.), Quantitative
2 predicts extensive storage for the Finnish inflectional text analysis (pp. 54-135). Trier, Germany: Wissenschaftlicher
suffix -jA, which realizes the partitive plural, whereas the Verlag Trier.
extreme productivity of the Finnish morphological system Church, K., & Gale, W. (1991). A comparison of the enhanced
suggests that massive storage of so many different inflected Good-Turing and deleted estimation methods for estimating
words would be highly surprising. We are currently investi- probabilities of English bigrams. Computer Speech and Lan-
gating if storage is indeed absent for Finnish partitive guage, 5, 19-54.
Clahsen, H., Eisenbeiss, S., & Sounenstuhl-Henning, L (1997).
plurals, which would falsify our theory. The theory that we Morphological structure and the processing of inflected words.
have offered here is only a first approximation that, in its Theoretical Linguistics, 23, 201-249.
simplicity, is probably wrong. Subsequent research will Cleveland, W. S. (1979). Robust locally weighted regression and
undoubtedly reveal other factors that co-determine the smoothing scatterplots. Journal of the American Statistical
balance of storage and computation as well. At the same Association, 74, 829--836.
time, the present results suggest that in spite of the complex- Colt, P., Beauvillain, C., & Segui, J. (1989). On the representation
ity of the various factors and their interrelations, the and processing of prefixed and suffixed derived words: A
beginnings of a coherent cross-linguistictheory
logical processing are emerging.
of morpho- differential frequency effect. Journal of Memory and Language,
28, 1-13.
Cutler, A., Hawkins, J. A., & Gilligan, G. (1985). The suffixing
preference: A processing explanation. Linguistics, 23, 723-758.
References Frauenfelder, U. H., & Schreoder, R. (1992). Constraining psycho-
linguistic models of morphological processing and representa-
Anshen, E, & Aronoff, M. (1988). Producing morphologically tion: The role of productivity. In G. E. Booij & J. van Made
complex words. Linguistics, 26, 641-655. (Eds.), Yearbookof Morphology 1991 (pp. 165-183). Dordrecht,
Anshen, E, & Aronoff, M. (1997). Morphology in real time. In the Netherlands: Kluwer Academic.
G.E. Booij & J. van Marie (Eds.), Yearbookof Morphology 1996 Gale, W. A., & Sampson, G. (1995). Good-Taring frequency
(pp. 9-13). Norwell, MA: Kluwer Academic. estimation without tears. Journal of Quantitative Linguistics, 2,
Baayen, R. H. (1996). The effect of lexical specialization on the 217-237.
502 BERTRAM,SCHREUDmt,AND BAAYEN

Good, L J. (1953). The population frequencies of species and the Schreuder, R., & Baayen, R. H. (1995). Modelling morphological
estimation of population parameters. Biometrika, 40, 237-264. processing. In L. B. Feldman (Ed.), Morphological aspects of
Hy6nli, J., I.afine, M., & Niemi, J. (1995). Effects of a word's language processing (pp. 131-154). Hillsdale, NJ: Erlbaum.
morphological complexity on readers' eye fixation pattern. In J. Schreuder, R., & Baayen, R. H. (1997). How complex simple
Findlay, R. Kentridge, & R. Walker (Eds.), Eye movement words can be. Journal of Memory and Language, 36, 118--139.
research: Mechanisms, processes and applications (pp. 445- Schriefers, H., Friederici, A., & Graetz, R (1992). Inflectional and
452)./umterdam: Elsevier. derivational morphology in the mental lexicon: Symmetries and
Kerlsson, E, & Koskeuniemi, K. (1985). A process model of asymmetries in repetition priming. QuarterlyJournal of Experi-
morphology and lexicon. Folia Linguistica, 29, 207-231. mental Psychology: Human Experimental Psychology, 44A,
Landanna, A., & Burani, C. (1995). DisUibutional properties of 373-390.
derivational affixes: Implications for processing. In L. B. Feld- Seidenberg, M. S. (1987). Sublexical structures in visual word
recognition: Access units or orthographic redundancy? In M.
man (Ed.), Morphological aspects of language processing (pp.
Coltbeart (Ed.), Attention and performance XII: Reading (pp.
345--364). Hillsdale, NJ: Erlbaum. 245-263). Hillsdale, NJ: Erlbaum.
Marslen,grtlson, W., Tyler, L. IC, Waksler, R., & Older, L. (1994). Sereno, J. A., & Jongman, A. (1997). Processing of English
M ~ l o g y and meaning in the English mental lexicon. Psycho- inflectional morphology. Memory and Cognition, 25, 425-437.
logical Review, 101, 3-33. Taft, M. (1979). Recognition of affixed words and the word
Niemi, J., Laine, M.,& Tuominen, J. (1994). Cognitive morphol- frequency effect. Memory and Cogni'tion, 7, 263-272.
" ogy in Finnish: Foundations of a new model. Language and Taft, M. (1994). Interactive-activation as a framework for under-
Cognitive Processes, 9, 423--446. standing morphological processing. Language and Cognitive
Pinker, S. (1991). Rules of languase. Science, 253, 530--535. Processes, 9, 271-294.
R a ~ ~ H. (1962). Statistical facilitation of simple reaction time. Taft, M., & Forster, ILI. (1975). Lexical storage and retrieval of
~tions of the New York Academy of Sciences, 24, 574- prefixed words. Journal of VerbalLearning and VerbalBehavior,
590. 14, 638--647.
Sandra, D, (1994). The morphology of the mental lexicon: Internal Taft, M., & Forster, K. I. (1976). Lexical storage and retrieval of
word ~ viewed from a psycholinsuistic perspective. polymorpbemic and polysyllabic words. Journal of Verbal
Language and Cognith,e Processes, 9, 227-269. Learning and VerbalBehavior, 15, 607-620.
STORAGE AND COMPUTATION 503

Appendix
-re (derivation) -re (inflection)
12

j :
0 2 4 6 8 10 0 2 4 I 8 10
Surface Frequency SuN~ Frequency

-held
14

,ot- .,~** . * * . . - ~ * . . - ~

0 2 4 6 IB
Surface Fr~luency

-er (derivation) -er (inflection)

!il, .....,.. == ,o" •


== . ..,,- .... ..

~ r! .i - -

I11 "
0 2 4 6 8 10 0 2 4 (I 8 10
Surface Frequency Surface Frequency

Figure A1. Scatterplots of log(surface fiequency + 1) by log(base frequency + 1) for the


derivafional -re (upper left), the inflectional -re (upper fight), the derivational suffix -he/d (center),
and the derivational --er (bottom left) and the inflectional --er (bottom right). The small dots represent
all of the available bi-morphemic words in the CELEX lexical dAtAbase.The superimposed large dots
represent the words selected for the experiments manipulating surface frequency. The solid lines are
non-paramelric robust locally weighted scatterplot smoothers (Cleveland, 1979) highlighting the
dependencies between the two frequency counts.
504 BERTRAM, SCHREUDER, AND BAAYEN

-te (derivation) -te (inflection)

14. • t V~l . / ~ e
| • 4

== -'IP .am- • ~ 0
m o 2 i 6 8 10 12 14 ff~ 0 2 4 6 8 10 12 14
~, Base Frequency Base Frequency

-held
c•"10
|. 6

..:. .-"
I=: 2
~o
o = 4 e e lo n 1:~
Base Frequency

-er (derivation) -er (inflection)

8 .." ~ 8
e .o e
U.

o
'
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Base Frequency Base Frequency

Figure A2. Scatterplots of log(surface frequency + 1) by log(base frequency + 1) for the


derivational --re (upper left), the inflectional -te (upper fight), the derivational suff~ -he/=/(center),
and the derivational - e r (bottom left) and the inflectional - e r (bottom right). The small dots represent
all of the available hi-morphemic words in the CELEX lexical database. The superimposed large dots
represent the words selected for the experiments manipulating base fiequency. The solid lines are
non-parametric robust locally weighted scatterplot smoothers (Cleveland, 1979) highlighting the
dependencies between the two frequency counts.
SrORAO~ Am) COMPtrrA_~OS 505

Table A 1

Materials Used in the Experiments


Mean reaction Mean error score Base Surface
Experiment Word (in English) time (%) frequency frequency
Verbs with a high surface frequency
la blafte (barked) 622 0 696 218
glipte (slipped) 567 0 415 184
groette (greeted) 602 5.6 1128 386
hakte (chopped) 654 0 875 119
klikte (told tales) 635 5.6 279 114
knoopte (knotted) 674 0 916 307
kuchte (coughed) 612 0 306 192
plofte (plopped) 674 0 252 97
propte (crammed) 716 0 230 85
raapte (picked) 636 0 594 272
rookte (smoked) 556 0 2198 456
schopte (kicked) 662 0 983 332
schraapte (scraped) 671 0 556 305
smakte (smacked) 738 5.6 296 120
spitste (pricked up) 685 5.6 318 119
stampte (stamped) 626 5.6 572 154
startte (started) 603 0 888 302
stokte (broke down) 773 0 332 219
stroopte (poached) 691 0 219 72
tastte (touched) 631 5.6 1187 413
Verbs with a low surface frequency
la dampte (steamed) 609 16.7 407 36
dempte (dimmed) 690 16.7 477 49
kweekte (cultivated) 573 0 1016 47
kwetste (wounded) 586 0 606 33
lustte (liked) 724 0 287 55
poetste (polished) 630 0 677 92
prikte (stung) 576 0 683 201
raspte (grated) 770 0 203 13
schetste (sketched) 712 5.6 1066 81
schikte (arranged) 615 5.6 2024 180
schorste (suspended) 663 5.6 182 11
sloopte (demolished) 706 0 329 17
splitste (spli0 694 5.6 427 37
spotte (mocked) 648 0 1249 122
staakte (went on strike) 640 0 958 58
strookte (was in accordance) 751 0 267 38
testte (tested) 747 0 544 22
toetste (tried out) 580 0 982 13
twistte (quarreled) 644 5.6 148 3
wreekte (avenged) 739 22.2 431 30
Verbs with a high base frequency
lb blufte (boasted) 625 3.7 100 17
dankte (thanked) 620 0 2222 154
jatte (nabbed) 602 0 121 14
knarste (crunched) 715 3.7 335 66
kookte (cooked) 589 0 2436 177
kraakte (cracked) 606 3.7 1296 262
kweekte (cultivated) 598 0 1016 47
kwetste (hutted) 622 0 606 33
lapte (shimmied) 609 0 183 16
lootte (drew lots for) 731 18.5 47 2
mikte (aimed) 608 0 323 104
piepte (peeped) 576 0 635 125
rustte (rested) 607 0 2518 372
schetste (sketched) 693 11.1 1066 81
schikte (arranged) 611 3.7 2024 180
sloopte (demolished) 611 3.7 329 17
stookte (burned) 613 7.4 302 23
twistte (quarreled) 622 3.7 148 3

(table continues)
BF.RTRAM, SCHREUDER,AND BAAYEN

Table A 1(confinue~
Mean reaction Mean error score Base Surface
Experiment Word (in English) time (%) frequency frequency
Verbs with a high base frequency (continued)
wekte (woke up) 633 3.7 3171 538
wreekte (avenged) 713 14.8 431 30
Verbs with a low base frequency
Ib pookte (poked)* 811 40.7 38 26
ventte (hawked)* 754 33.3 29 7
balkte (brayed) 718 14.8 38 7
blikte (glanced) 639 18.5 279 119
dweepte (idolized) 704 25.9 93 32
flapte (flapped) 619 0 149 86
floepte (Impped) 716 29.6 51 22
ka~e (teased) 718 14.8 15 13
klakte (clacked) 753 25.9 95 52
klapte (clapped) 620 0 923 402
mestte (manured) 675 0 32 5
pare (blazed) 766 11.1 28 11
propte (crammed) 637 7.4 230 85
schepte (created) 559 0 353 184
schraapte (scraped) 617 14.8 556 305
spurtte (spurted) 759 18.5 55 26
startte (started) 630 0 888 302
stokte Caroke down) 622 3.7 332 219
strecpte (striped) 618 0 70 30
trompette (trumpeted) 756 22.2 3 3
Nouns in --re
zoelte (mildness)* 621 93.8 53 4
dwarste (crossness)* 757 75.0 1899 0
dunte (thinness)* 778 75.0 1990 0
8anwte (quickness)* 1075 62.5 5132 16
scheefte (crookedness)* 858 56.2 867 2
klaarte (clarity)* 854 50.0 5270 63
puurte (pureness)* 700 50.0 1521 0
grofte (rudeness)* 784 43.8 1214 0
ruigte (roughness)* 909 43.8 368 8
scheelte (cross-eyedness)* 624 43.8 142 0
schuinte (obliqueness)* 909 43.8 1367 5
sfljfte (stiffness)* 696 43.8 1658 6
fijnte (fineness)* 681 37.5 3890 0
friste (freshness)* 813 37.5 1584 4
graagte (eagerness)* 807 37.5 11221 61
kleinte (sinai Iness)* 574 37.5 21535 0
kromte (crookedness)* 749 31.2 394 0
loomte (languor)* 782 31.2 350 2
donkerte (darkness) 746 25.0 8982 109
kaalte (baldness) 617 25.0 1594 0
lauwte (tepidness) 838 25.0 582 5
ruwte (roughness) 751 25.0 1623 0
vuilte (dirtyness) 613 25.0 1996 0
engte (strait) 692 18.8 856 16
schraalte (leanness) 801 18.8 325 0
smalte (narrowness) 628 18.8 3044 0
krapte (tighmess) 668 12.5 205 8
nanwte (defile) 671 12.5 2628 0
slapte (slackness) 758 12.5 1156 24
sombette (somberness) 824 12.5 1696 9
steilte (steepness) 717 12.5 982 8
droogte (dryness) 586 6.2 2716 167
koelte (coolness) 626 6.2 2344 227
luwte (shelter) 688 6.2 10 67
magerte (meagerness) 748 6.2 2032 8
schaarste (scarcity) 684 6.2 831 194
breedte (breadth) 578 0 5558 445
dikte (thickness) 591 0 6689 180
flauwte (faint) 618 0 1001 30
gekte (foolishness) 650 0 5820 28
STORAGE A N D COMPUTATION 507

Table A1 (continued)
Mean reaction Mean error score Base Surface
Experiment Word (in English) time (%) frequency frequency
Nouns in -re (continued)
2 gewoonte (custom) 550 0 13509 2459
(cont'~ groente (vegetables) 532 0 2978 977
grootte (size) 564 0 51667 1314
hoogte (height) 553 0 17053 5164
kalmte (calmness) 577 0 2047 391
kilte (chilliness) 616 0 907 142
leegte (emptiness) 556 0 5290 679
mimte (space) 549 0 4678 6355
sterkte (strength) 608 0 16548 482
stilte (silence) 613 0 6995 4283
verte (distance) 637 0 23125 2362
warmte (warmth) 529 0 6680 1995
wijdte (width) 721 0 2205 13
ziekte (sickness) 499 0 3783 5174
zwaarte (weight) 633 0 9812 193
zwoelte (sultriness) 794 0 160 3
Nouns in --hekt with a high surface frequency
3a bitterheid (bitterness) 546 0 1807 228
dichtheid (density) 529 0 9137 205
droefheid (sadness) 558 0 269 263
dwaasheid (absurdity) 507 0 1129 308
grootheid (greatness) 528 0 51667 426
hardheid (hardness) 526 6.2 9923 231
hoogheid (highness) 544 0 17067 378
ijdelheid (vanity) 567 0 310 310
kuisheid (chastity) 643 6.2 230 90
lafheid (cowardice) 542 0 430 181
oudheid (antiquity) 593 0 20545 537
schoonheid (beauty) 547 0 2884 1856
snelheid (speed) 532 0 16716 1694
tederheid (tenderness) 627 0 773 513
veiligheid (safety) 552 0 3340 1395
vrijheid (freedom) 521 0 11678 5298
vroomheid (piety) 632 0 406 148
waarheid (troth) 515 0 69252 4752
wreedheid (cruelty) 552 0 842 324
zekerheid (certainty) 521 0 24640 3049
Nouns in -he/d with a low surface frequency
3a halfheid 0udfness)* 833 50.0 10588 11
bitsheid (tartness) 728 19.0 337 3
broosheid (fragility) 641 19.0 395 18
bruutheid (brutality) 770 6.2 273 6
eigenheid (ownness) 605 0 31731 82
forsheid (robustness) 682 25.0 8O5 2
goedheid (kindness) 553 0 83272 473
goorheid (dinginess) 700 13.0 254 2
groenheid (greenness) 717 6.2 2996 3
koelheid (coolness) 575 0 2701 37
kortheid (shortness) 543 0 14846 11
lichtheid (lightness) 562 6.2 18700 20
losheid (looseness) 667 13 8935 13
netheid (decency) 543 0 19595 60
nieuwheid (newness) 598 6.2 32846 15
puurheid (purity) 692 6.2 1547 5
rechtheid (straightness) 584 0 11726 0
scheefheid (crookedness) 733 0 867 3
serieusheid (seriousness) 631 0 1853 2
troebelheid (turbidity) 703 6.2 316 5
Nouns in -he/d with a high base frequency
3b linksheid (leftness)* 610 31.2 3873 6
dikheid (fatness) 668 18.8 6724 3
felheid (fierceness) 599 0 2585 111
(table continues)
508 BeS~, SCHI~UDEa, AND BAAY~

Table A 1 (continued)
Mean reaction Mean error score Base Surface
Experiment Word (in English) time (%) frequency frequency
Nouns in --he/d with a high base frequency (continued)
3 • fraaiheid (beauty) 659 0 1660 6
(c~mt'd) heelheid (wholeness) 748 18.8 55138 19
heetheid (hotness) 787 18.8 9873 4
jongheid (youngness) 610 6.2 15000 13
juistheid (correctness) 605 6.2 16928 385
koudheid (coldness) 577 0 5789 0
leegheid (emptiness) 575 0 5451 36
lenkheid (niceness) 709 18.8 4026 3
magerheid (thinness) 576 0 2032 7
moeheid (fatigue) 550 6.2 2546 212
mooiheid (beauty) 653 25.0 15065 7
nieuwheid (newness) 568 6.2 32846 15
puurheid (purity) 655 6.2 1547 5
scheetheid (crookedness) 729 6.2 867 3
schuinheid (obliqueness) 747 25.0 1371 0
serieusheid (seriousness) 793 18,8 1853 2
steilheid (steepness) 716 12.5 982 0
strengheid (strictness) 560 0 2508 78
traagheid (slowness) 547 0 1993 144
trotsheid (pride) 649 0 3103 0
witheid (whiteness) 611 6.2 14016 38
zwaarheid (heaviness) 616 0 9812 0
Nouns in --hem with a low base frequency
3b zatheid (satiety)* 667 31.2 37 8
ziltheid (saltishness)* 837 31.2 94 3
brosheid (fragility) 829 18.8 95 5
bruutheid (brutishness) 699 6.2 273 6
dartelheid (friskiness) 742 12.5 131 2
droefheid (sadness) 623 0 269 263
droogheid (dryness) 580 0 4185 11
dufheid (mustiness) 817 6.2 90 3
edelheid (nobleness) 650 0 534 0
effertheid (smoothness) 708 0 287 4
gaarheid (cookedness) 680 6.2 252 0
geilheid (lubricionsness) 685 0 341 64
ijdelheid (vanity) 604 0 310 310
kleflleid (clamminess) 685 0 91 0
koelheid (coolness) 602 0 2701 37
lafheid (cowardliness) 562 0 430 181
losheid (looseness) 583 25.0 8935 13
makheid (tameness) 650 0 101 0
matheid (weariness) 580 0 632 18
plompheid (rudeness) 710 18.8 186 0
rotheid (rottenness) 610 0 900 3
schuwheid (shyness) 654 0 569 39
soberheid (soberness) 684 6.2 329 96
vaalheid (fadedness) 727 6.2 292 6
vlugheid (quickness) 607 6.2 4059 8
Nouns in -er with a high surface frequency
4a dweper (idolater)* 851 50.0 93 16
turner (gymnast) 641 18.8 47 68
drijver (cattle driver) 641 6.2 3360 13
schenker (pourer) 624 6.2 3745 38
ruiker (bouquet) 637 6.2 1702 41
breker (breaker) 586 0 4762 55
gieter (wateringcan) 580 0 980 39
heerser (ruler) 582 0 2511 221
jager (hunter) 548 0 • 2168 363
kenuer (connoisseur) 630 0 21514 148
lijder (sufferer) 739 0 7188 45
redder (saver) 586 0 3076 99
meier (rower) 530 0 537 8
stoker (stoker) 719 0 302 143
robber (worrier) 679 6.2 213 15
wekker (alarm clock) 536 0 3171 249
STORAGE AND COMPUTATION 509

Table A1 (continued)
Mean reaction Mean error score Base Surface
Experiment Word (in English) time (%) frequency frequency
Nouns in - e r with a high surface frequency (continued)
4a zanger (singer) 542 0 4234 138
(cont'd) zender (sender) 543 0 1752 369
zwemmer (swimmer) 544 0 1567 52
zwerver (wanderer) 571 0 692 138
Nouns in - e r with a low surface frequency
4a smeder (forger)* 833 43.8 295 0
looier (tanner)* 752 37.5 55 0
schelder (slanger)* 807 31.2 659 0
buiger (bower) 678 12.5 3234 0
bieder (bidder) 696 6.2 7689 18
blusser (extinguisher) 707 6.2 148 3
kaatser (fives-player) 759 6.2 128 0
vreter (glutton) 691 6.2 555 0
wasser (washer) 671 6.2 2337 7
bidder (prayer) 645 0 1782 6
binder (binder) 724 0 1979 10
glijder (slider) 713 0 3193 0
grijper (grasper) 641 0 4155 8
hijger (gasper) 606 0 1489 0
hurler (crier) 660 0 4580 3
kmiper (creeper) 741 0 2836 6
sluiper (sneaker) 599 0 925 5
vanger (catcher) 546 0 3435 0
volger (follower) 598 0 21720 4
winner (winner) 527 0 3383 3
Nouns in - e r with a high basefrequency
4b :ronker (snorer)* 982 81.2 231 0
duider (suggestor)* 798 50.0 1310 77
brenger (bearer) 738 25.0 36536 15
janker (yelper) 732 18.8 495 0
smelter (melter) 766 18.8 816 0
dwinger (forcer) 731 12.5 3451 0
glijder (slider) 653 12.5 3193 0
hurler (crier) 613 12.5 4580 3
smijter (dasher) 650 12.5 947 0
stijger (riser) 602 12.5 3335 0
vinder (finder) 730 12.5 51631 24
bieder (bidder) 713 6.2 7689 18
blijver (stayer) 611 6.2 52837 0
ligger (liar) 688 6.2 40957 0
wasser (washer) 650 6.2 2337 7
zieuer (seer) 706 6.2 98375 82
heerser (ruler) 642 0 2511 221
kelmer (connoisseur) 586 0 21514 148
voeler (feeler) 771 0 29234 4
volger (follower) 743 0 21720 4
Nouns in - e r with a low base frequency
4b zifter (sifter)* 1383 81.2 14 0
schranzer (gormandizer)* 852 75 2O 0
kleumer (slaiverer)* 930 62.5 29 0
schender (violator)* 1087 37.5 226 2
binder (binder) 666 6.2 1979 10
gieter (watering can) 633 0 980 39
grijper (grasper) 697 0 4155 8
looier (tanner) 911 2.5 55 0
ordener (arranger) 787 25.0 637 0
rijder (rider) 669 0 8591 11
sluiper (sneaker) 630 0 925 5
snijder (cutter) 615 0 2682 10
strooier (strewer) 625 0 470 0
temmer (tamer) 731 25.0 192 3
robber (worrier) 705 18.8 213 15
(tablecon6nues)
510 BERTRAM, SCHREUDER,AND BAAYEN

Table A1 (continued)
Mean reaction Mean error score Base Surface
Experiment Word (in English) time (%) frequency frequency
Nouns in -er with a low base frequency (continued)
4b turner (gymnast) 695 0 47 68
(cont'd) vuller (filler) 711 18.8 3245 2
wekker (alarm clock) 556 0 3171 249
winner (winner) 618 12.5 3383 3
zender (sender) 571 0 1752 369
Comparatives with a high surfacefrequency
5a dieper (deeper) 553 0 11667 1957
dikker (fatter) 593 0 6689 357
dunner (thinner) 603 0 1801 189
enger (creepier) 589 0 856 153
feller (fiercer) 585 5.9 2585 338
fijuer (finer) 552 0 3890 182
geringer (smaller) 631 5.9 3539 489
kalmer (calmer) 542 5.9 2047 167
kouder (colder) 576 0 5789 170
langer (taller) 527 0 42525 7438
milder (milder) 621 5.9 909 196
rijper (riper) 569 5.9 1089 115
scherper (sharper) 590 5.9 4875 503
slechter (worse) 551 0 7977 542
slimmer (smarter) 539 0 1095 121
sueller (faster) 528 0 16312 1921
soepler (suppler) 679 11.8 849 92
strenger (stricter) 595 0 2423 168
vlugger (quicker) 573 0 4059 390
zwakker (weaker) 546 0 3006 383
Comparatives with a low surface frequency
5a zeldener (rarer)* 749 52.9 3719 4
banger (more afraid) 496 0 7921 98
blanker (whiter) 614 5.9 967 7
blauwer (bluer) 599 0 5281 30
blijer (happier) 599 5.9 4391 20
doller (wilder) 669 5.9 1059 3
ttinker (firmer) 674 11.8 2908 36
ganwer (sooner) 727 29.4 5132 25
juister (more right) 590 0 16298 173
linkser (more left) 716 17.6 3873 9
naakter (hurler) 637 0 2181 17
hatter (wetter) 574 5.9 2572 24
nieuwer (newer) 551 0 32729 137
rauwer (rawer) 624 11.8 882 8
schuiner (more oblique) 672 0 1367 7
lrotser (prouder) 576 0 2079 15
valser (falser) 533 11.8 1566 8
vreemder (suanger) 589 0 6646 99
witter (whiter) 614 11.8 12989 76
zwarter (blacker) 569 0 6846 56
Comparatives with a high base frequency
5b zelduer (rarer)* 861 43.8 3719 4
blonder (blonder)* 609 37.5 1291 7
banger (more afraid) 577 0 7921 98
flinker (firmer) 689 6.2 2908 36
gauwer (sooner) 698 18.8 5132 25
jaloerser (more jealous) 597 6.2 1097 4
juister (more right) 679 6.2 16928 173
linkser (more left) 670 0 3873 9
naakter (nuder) 614 0 2181 17
nieuwer (newer) 542 0 32729 137
schuiner (more oblique) 648 6.2 1367 7
smaller (narrower) 531 0 3044 155
stommer (more stupid) 619 0 1362 26
strikter (stricter) 755 6.2 1101 16
triester (drearier) 652 0 674 19
STORAGEAND COMPUTATION 511

Table A1 (continued)
Mean reaction Mean error score Base Surface
Experiment Word (in English) time (%) frequency frequency
Comparatives with a high base frequency (continued)
5b trotser (prouder) 605 6.2 2079 15
(cont'd) valser (falser) 609 0 1566 8
vreemder (stranger) 546 0 6646 99
witter (whiter) 640 0 12989 76
zwarter (blacker) 562 0 6846 56
Comparatives with a low base frequency
5b driester (more reckless)* 741 62.5 80 22
ranker (crankier)* 739 31.3 189 2
penibeler (more awkward)* 920 31.3 49 3
bonter (more variegated) 662 12.5 673 22
dommer (dumber) 596 0 2096 72
edeler (nobler) 584 6.2 534 28
enger (creepier) 546 6.2 856 153
koeler (cooler) 555 0 2344 164
krommer (more crooked) 672 6.2 394 13
lakser (slacker) 637 0 43 3
leuker (nicer) 553 0 4026 172
luxer (more luxurious) 644 6.2 180 2
matter (paler) 712 25.0 327 51
nobeler (nobler) 638 6.2 268 16
reiner (purer) 687 6.2 356 15
rijper (riper) 596 0 1089 115
soepeler (suppler) 570 0 849 92
spitser (sharper) 694 0 294 5
taaler (more tenacious) 582 6.2 506 36
wankeler (shakier) 716 12.5 352 2
Note. Words marked with an * were not included in the analyses because of their high error rates.

Received December 12, 1997


Revision received May 4, 1999
Accepted August 12, 1999 •

Low Publication Prices for APA Members and Affiliates


K e e p i n g y o u u ~ t o - c l a t e . All APA Fellows, Members, Associates, and Student Affiliates
rex~ive---as part of their annual dues---subscriptions to the American Psychologist and
APA Monitor. High School Teacher and International Affiliates receive subscriptions to
the APA Monitor, and they may subscribe to the American Psychologist at a significantly
reduced rate. In addition, all Members and Student Affiliates are eligible for savings of up
to 60% (plus a journal credit) on all other APA journals, as well as significant discounts on
subscriptions from cooperating societies and publishers (e.g., the American Association for
Counseling and Development, Academic Press, and Human Sciences Press).

Essential r e s o u r c e s . APA members and affiliates receive special rates for purchases of
APA books, including the Publication Manual of the American Psychological Association,
and on dozens of new topical books each year.

Other benefits of membership. Membership in APA also provides eligibility for


competitive insurance plans, continuing education programs, reduced APA convention fees,
and specialty divisions.

More i n f o r m a t i o n . Write to American Psychological Association, Membership Services,


750 First Street, NE, Washington, DC 20002-4242.

You might also like