Assignment details: When Does Length Cause the Word Length Effect?
Annie Jalbert and Ian Neath Memorial University of Newfoundland
Tamra J. Bireta The College of New Jersey
Aimée M. Surprenant Memorial University of Newfoundland
The word length effect, the finding that lists of short words are better recalled than lists of long words, has been termed one of the benchmark findings that any theory of immediate memory must account for. Indeed, the effect led directly to the development of working memory and the phonological loop, and it is viewed as the best remaining evidence for time-based decay. However, previous studies investigating this effect have confounded length with orthographic neighborhood size. In the present study, Experi- ments 1A and 1B revealed typical effects of length when short and long words were equated on all relevant dimensions previously identified in the literature except for neighborhood size. In Experiment 2, consonant–vowel– consonant (CVC) words with a large orthographic neighborhood were better recalled than were CVC words with a small orthographic neighborhood. In Experiments 3 and 4, using two different sets of stimuli, we showed that when short (1-syllable) and long (3-syllable) items were equated for neighborhood size, the word length effect disappeared. Experiment 5 replicated this with spoken recall. We suggest that the word length effect may be better explained by the differences in linguistic and lexical properties of short and long words rather than by length per se. These results add to the growing literature showing problems for theories of memory that include decay offset by rehearsal as a central feature.
Keywords: word length, orthographic neighborhood, working memory
The word length effect, the finding that lists of short words (e.g., cat, boat, pear, etc.) are recalled better than lists of long words (e.g., gorilla, hovercraft, banana, etc.) has played such a signifi- cant a role in the development of theories of memory that it is now regarded as a “benchmark finding” that current theories of short- term or working memory must address (cf. Lewandowsky & Farrell, 2008). Indeed, the basic finding is one of the core phe- nomena that led directly to the development of the phonological loop component of working memory (Baddeley, 1992), has been termed the “best remaining solid evidence” for the existence of such temporary memory systems (Cowan, 1995, p. 42), and is the focus of many computational models (e.g., Brown & Hulme, 1995;
Burgess & Hitch, 1999; Hulme, Surprenant, Bireta, Stuart, & Neath, 2004; Neath & Nairne, 1995; Page & Norris, 1998). In this article, we consider evidence that questions the idea that length per se is the critical factor underlying the word length effect.
The first systematic exploration of the word length effect was reported by Baddeley, Thomson, and Buchanan (1975), although the basic finding was known earlier (e.g., Watkins, 1972). In a series of experiments, Baddeley et al. (1975) identified two ways in which word length can have an effect on memory performance. The time-based word length effect is shown with words that are equated on all dimensions, including the number of syllables and phonemes, but vary systematically only in the time required to pronounce the words. In contrast, the syllable-based word length effect is demonstrated when the short and long words vary not only in pronunciation time but also in the number of syllables and phonemes.
According to the working memory framework, both effects are explained in the same way: Items in the phonological loop decay within about 2 s if not refreshed by an articulatory control process. Given the assumption that there is a positive correlation between the rate of subvocal rehearsal and overt pronunciation time, it will take longer to refresh a list of long words than a list of short words, and therefore, the general prediction is that lists of items that take longer to pronounce will be worse recalled than otherwise com- parable lists of items that take less time to pronounce.
The Time-Based Word Length Effect
The time-based word length effect was established in two initial studies. In their Experiment 3, Baddeley et al. (1975) showed that
This article was published Online First December 20, 2010. Annie Jalbert, Ian Neath, and Aimée M. Surprenant, Department of
Psychology, Memorial University of Newfoundland, St. John’s, New- foundland and Labrador, Canada; Tamra J. Bireta, Psychology Depart- ment, The College of New Jersey.
Some of this work was presented at the 49th Annual Meeting of the Psychonomic Society, Chicago, IL, November 2008 and at the 19th Annual Meeting of the Canadian Society for Brain, Behaviour, and Cognitive Science, York, England, July 2009. This research was supported by grants from the National Sciences and Engineering Research Council to Annie Jalbert, Ian Neath, and Aimée M. Surprenant. We thank Caroline Barnes for assistance testing participants.
Correspondence concerning this article should be addressed to Annie Jalbert, Psychology Department, Memorial University of Newfoundland, St. John’s, Newfoundland and Labrador A1B 3X9, Canada. E-mail: annie.jalbert@mun.ca
Journal of Experimental Psychology: © 2010 American Psychological Association Learning, Memory, and Cognition 2011, Vol. 37, No. 2, 338 –353
0278-7393/10/$12.00 DOI: 10.1037/a0021804
338
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
lists of disyllabic words that could be said quickly (bishop, pectin ember, wicket, wiggle, pewter, tipple, hackle, decor, phallic) were recalled better than lists of disyllabic words that took longer to pronounce (Friday, coerce, humane, harpoon, nitrate, cyclone, morphine, tycoon, voodoo, zygote). In Experiment 4, a subset of these words was used such that the short and long words were equated for the number of syllables, the number of phonemes (given Scottish pronunciation), and frequency. Once again, a word length effect obtained: The words that took less time to say were recalled better than the words that took more time to say. These results were taken as support for the phonological loop component of working memory (Baddeley, 1986).
Many studies have since replicated this time-based word length effect with the original stimuli (e.g., Cowan, Day, Saults, Kellar, Johnson, & Flores, 1992; Longoni, Richardson, & Aiello, 1993; Lovatt, Avons, & Masterson, 2000; Nairne, Neath, & Serra, 1997). However, there are no other sets of stimuli that produce this result. For example, Neath, Bireta, and Surprenant (2003) tested four different sets of short and long words that were equated for the number of syllables and phonemes but differed in pronunciation time; only the original Baddeley et al. (1975) stimuli produced a word length effect. An additional set of English words (Lovatt et al., 2000) and a set of Finnish nonwords (Service, 1998) also failed to yield a time-based word length effect. Thus, whereas one set of words consistently produces the effect, five other sets of stimuli do not. Neath et al. concluded that the time-based word length effect was due to some unknown property of the original stimuli. They noted that unless a large number of other stimulus sets are shown to generate a difference in recall based solely on pronunciation time, it is reasonable to conclude that the time-based word length effect does not exist. This poses a problem for theories that incorporate something like the phonological loop.
The Syllable-Based Word Length Effect
In contrast to the time-based word length effect, the syllable- based word length effect is robust and has been demonstrated with numerous sets of stimuli. However, there are still disagreements about the cause of this effect. One class of theories, based on the phonological loop, invokes an explanation based on the trade-off between decay and pronunciation time (e.g., Baddeley, 1986, 2003; Burgess & Hitch, 1999; Page & Norris, 1998), the lack of a pure time-based effect notwithstanding. To generate evidence in support of this view, researchers began examining recall of short and long items in pure and mixed lists. Using a computational model that incorporates the assumptions of the phonological loop, Burgess and Hitch (1999; Figure 16) generated the prediction that recall of lists made up of a mixture of short and long lists would fall between that of pure short and pure long lists. The list that can be rehearsed most quickly, the pure short list, will be recalled best, and the list that takes the longest amount of time to rehearse, the pure long list, will be recalled worst. The mixed lists will take less time to rehearse than will the pure long lists but more time to rehearse than the pure short lists, and so recall level will be intermediate.
In contrast, theories based on item (rather than list) properties make quite different predictions. For example, the feature model (Neath & Nairne, 1995) assumes that long items have more seg- ments than short items and that at some point during the retrieval
process, the segments need to be assembled. If one assumes a fixed probability of making an assembly error, then a word length effect will obtain. According to this account, list composition does not matter: short items in mixed lists should be recalled just like short items in pure lists. Similarly, Brown and Hulme (1995) proposed a model in which rehearsal plays no role at all, but rather, differ- ential decay of individual items is what leads to the word length effect. Because items decay at their given rate regardless of list composition, this account also predicts that recall of short items will be identical whether presented in a pure list or mixed with long items.
Although the predictions of both classes of models are clear-cut, the empirical results are not. Cowan, Baddeley, Elliott, and Norris (2003) included lists of pure short words (1 syllable), pure long words (5 syllable), and mixed lists that contained three short and three long words. Although performance in the mixed lists was in between that of the pure lists, as predicted by the phonological loop accounts, recall of short words from mixed lists was still better than recall of long words from mixed lists, a result predicted by the item-based accounts. Hulme et al. (2004) reported a differ- ent pattern of results. They found, in two experiments, that recall of short items in mixed lists was equivalent to recall of long items in mixed lists, a result predicted by the list-based view, but recall of these items was equivalent to recall of short items in pure lists. The item-based view predicts that only short items from mixed lists would be recalled as well as short items from pure lists.
Bireta, Neath, and Surprenant (2006) argued that the difference in the pattern of results was attributable to particular properties of the stimulus sets used. Bireta et al. (2006) replicated the results reported by Cowan et al. (2003) when using Cowan et al.’s (2003) stimuli, and also replicated the results reported by Hulme et al. (2004) when using Hulme et al.’s (2004) stimuli. Bireta et al. (2006) noted that neither the item-based accounts nor the list-based accounts (i.e., the phonological loop) could predict either pattern in its entirety. Just as with the time-based word length effect, then, aspects of the syllable-based word length effect appear to vary depending on the particular stimuli used.
The Phonological Loop Model Revisited
As more and more results were being published that contra- dicted the central claims of the phonological loop hypothesis, Mueller, Seymour, Kieras, and Meyer (2003, p. 1353) published an article in which they argued that these earlier results may have been due to “less than ideal measurements of articulatory duration and phonological similarity.” To address the former, they intro- duced a different way of measuring the pronunciation time of the to-be-remembered items. To replace the various methods that have been used in the literature, Mueller et al. (2003) developed a procedure in which participants memorize a sequence of words and then produce the sequence from memory at least twice both “rapidly and accurately” (p. 1362). This procedure is then repeated with different orderings of the words, and the subsequent times analyzed.
To address the second issue, Mueller et al. (2003) developed a new measure of phonological dissimilarity called PSIMETRICA (Phonological Similarity Metric Analysis). Phonological dissimi- larity between words is multidimensional and based on relevant dimensions like stress patterns and syllable onset. In order to
339LENGTH AND THE WORD LENGTH EFFECT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
compare words for dissimilarity using PSIMETRICA, each word is first decomposed into phonemes. Each syllable of a word is assumed to be composed of three different phoneme clusters; the onset (first consonants), the nucleus (vowels), and the coda (last consonants). The next step is to align the phoneme clusters in pairs of words. After the clusters have been aligned, the phonological dissimilarity is measured to obtain a dissimilarity profile. Two identical clusters have a dissimilarity value of 0 and two very different clusters have a dissimilarity value closer to 1. The dis- similarity values for different phonemes can be calculated using a table of phonological features based on Chomsky and Halle’s (1968) system. For a given set of words, the dissimilarity measure is comprised of the average of the dissimilarity value of all possible word pairs from the set. For a more detailed description of PSIMETRICA, see pages 1357–1361 of Mueller et al. (2003).
Mueller et al. (2003) reported two experiments, one of which they stated demonstrated a time-based word length effect and the other of which demonstrated a syllable-based word length effect. They argued that these “confirm and extend the predictions of the phonological-loop model” (Mueller et al., 2003, p. 1353).
However, the results are not as unambiguous as they initially appear, for three reasons. First, their method of measuring pronun- ciation time has been criticized. For example, Lewandowsky and Oberauer (2008) noted that by using the time to reproduce the lists from memory as their measure of duration, Mueller et al. (2003) are “predicting accuracy in immediate serial recall from speed in immediate serial recall” (p. 879). This makes it difficult to claim it as a true prediction, as both measures—accuracy and latency—are typically highly correlated.
A second issue is that by one measure, Mueller et al. (2003) did not, in fact, demonstrate a time-based word length effect. The experiment involved three sets of words: simple short (Set 7), simple long (Set 8), and complex long (Set 9). For a pure effect, there needs to be a difference between simple short and simple long words because the complex long differs from the simple short in at least two ways (i.e., length and complexity). Although mem- ory span for Set 7 was 5.21, compared with 5.05 for Set 8, this difference was not reported as statistically significant (see Mueller et al., 2003, p. 1371).
The third issue involves the evidence for a syllable-based word length effect. Like other researchers, Mueller et al. (2003) used a set of short and long words that confounded length with ortho- graphic neighborhood size, and thus it is not clear which difference is driving the effect. Of importance, the confound is the same one prevalent in the literature, and we now turn to this issue.
Stimulus Set Specificity and Neighborhood Effects
Despite the empirical and theoretical disagreements in the word length effect literature, one aspect has become increasingly appar- ent: the particular stimulus set used can critically determine whether effects of length will be seen (e.g., Bireta et al., 2006; Lovatt et al., 2000; Neath et al., 2003; see also Lewandowsky & Oberauer, 2008). Researchers do attempt to equate the short and long words on all relevant dimensions, but it is difficult, if not impossible, to anticipate every dimension of importance.
One factor rarely considered in such studies concerns the lexical neighbors of the to-be-remembered items. Words that are similar to a target word are referred to as its neighbors, and the set of these
words is referred to as the target word’s neighborhood (cf. M. Coltheart, Davelaar, Jonasson, & Besener, 1977). Similarity can be defined on the basis of a word’s orthography (M. Coltheart et al., 1977) or by its phonology (Luce & Pisoni, 1998). An orthographic neighbor is a word of the same length as the target that differs by only one letter. For example, given the word cat, the words bat, fat, cot, cut, cab, can, and so on, are all considered orthographic neighbors. A phonological neighbor is one that differs from the target word by the substitution of a single phoneme at any position (Roodenrys, Hulme, Lethbridge, Hinton, & Nimmo, 2002).1
Two published articles have demonstrated better recall of words with a large neighborhood than otherwise comparable words with a small neighborhood. In their Experiment 1, Roodenrys et al. (2002) used consonant–vowel– consonant (CVC) words, manipu- lating both neighborhood size (small vs. large) and frequency of the target words. The task was memory span, using spoken recall, and the words were presented auditorily. Memory span was higher for words with larger neighborhoods than those with smaller neighborhoods. In Experiment 3, they used a second set of CVC words, this time manipulating frequency of items that comprise the neighborhood as well as neighborhood size. Again, span was higher for words with larger neighborhoods. Finally, Experiment 4 used a third set of CVC words, manipulating word frequency, neighborhood size, and neighborhood frequency. The beneficial effect of neighborhood size was replicated.
Allen and Hulme (2006, Experiment 2) used the stimuli from Experiment 1 of Roodenrys et al. (2002), but with a slightly different task. Their participants heard a list of seven words and were then given a spoken immediate serial recall test. Despite the change in test, memory was again better for words with a larger neighborhood than those with a smaller neighborhood.
This beneficial effect of neighborhood size is not limited to just words; it is also observed with pronounceable nonwords (for a review, see Roodenrys, 2009). The neighborhood of a nonword can be defined as all of the valid words that can be produced by the substitution of a letter (for orthographic neighborhood) or pho- neme (for phonological neighborhood). Roodenrys and Hinton (2002, Experiment 2) asked subjects to listen to lists of four nonwords and then immediately repeat them back in order. Per- formance was better for nonwords with large neighborhoods than those with small neighborhoods. Thus, three sets of English words and one set of nonwords produce a recall advantage for items with a large neighborhood over those with a small neighborhood.2
Of relevance to the word length effect, short English words tend to have more neighbors— both orthographic and phonological— than do long words, and so neighborhood size is likely to be
1 There is a subtle difference between the Luce and Pisoni (1998) definition of a phonological neighbor and the M. Coltheart et al. (1977) definition of an orthographic neighbor. The former also includes all words that differ from the target word by the addition or deletion of a single phoneme in any position. Thus, the Luce and Pisioni definition includes scat and at as (phonological) neighbors of cat, whereas the M. Coltheart et al. definition does not include either as (orthographic) neighbors of cat.
2 Goh and Pisoni (2003) found better recall of low neighborhood words than high neighborhood words. However, there are a number of differences in stimuli and experimental design which make it difficult to reconcile the results with those of Roodenrys et al. (2002); Allen and Hulme (2006), and those reported in the current article.
340 JALBERT, NEATH, BIRETA, AND SURPRENANT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
confounded in word length effect experiments. To assess this, we examined published studies on the syllable-based word length effect that used English words. For those studies that report the stimuli used, we obtained measures of orthographic neighborhood size using the Medler and Binder (2005) database, which is based on the CELEX database.3 Table 1 lists the results. In all studies examined, short words had a larger orthographic neighborhood than did long words. One study in particular is highly suggestive: V. Coltheart, Mondy, Dux, and Stephenson (2004, Experiment 1) had three sets of stimuli: short one-syllable words (four letters), long one-syllable words (six or seven letters), and three-syllable words (six or seven letters). The task was immediate serial recall of five-item lists presented at a rate of one item per second. The orthographic neighborhood size for the three types of items was 7.80, 1.03, and 0.48, respectively. Recall level followed both word length (defined by the number of letters) and orthographic neigh- borhood size: .76 for the shortest words, .62 for the intermediate length words, and .56 for the longest words.
Given the confound between word length and orthographic neighborhood (see Table 1) and given that words with a large neighborhood are better recalled than words with a small neigh- borhood (Roodenrys et al., 2002; Allen & Hulme, 2006), the present series of studies was designed to assess the extent to which neighborhood size affects the word length effect. The first two experiments were designed to show that a syllable-based word length effect (Experiments 1A and 1B) and an orthographic neigh- borhood effect (Experiment 2) are observable in our paradigm. Experiments 3 and 4 each used a different set of short and long words, but this time the short and long words were equated for orthographic neighborhood size. In Experiment 5, we replicated the method from Experiment 3 using spoken recall instead of a strict reconstruction of order test to see if the results could be replicated with a different recall method. To anticipate the results, effects of length were observed when length was confounded with orthographic neighborhood size but were not observed when or-
thographic neighborhood size was equated for short and long items.
Experiment 1A
The purpose of Experiment 1A was to demonstrate that typical word length effects are observable with our methodology. We created a new set of short (one syllable) and long (three syllable) items that were equated for frequency, concreteness, imageability, and familiarity, as well as for phonological dissimilarity as mea- sured by PSIMETRICA. We did not equate for orthographic neighborhood size or frequency. Second, we included mixed lists in addition to pure lists to provide additional data on the effects of list composition on the word length effect. Third, we used written serial recall.
Method
Participants. Sixteen undergraduate students from Memorial University of Newfoundland participated in exchange for a small honorarium. All participants were native English speakers.
Stimuli. A set of 15 short words and 15 long words was created (see Appendix A). The words were equated for familiarity, frequency (both Kucera-Francis and Thorndike-Lorge), concrete- ness, and imageability using the Medical Research Council Psy- cholinguistics database (http://www.psy.uwa.edu.au/mrcdatabase/ uwa_mrc.htm). In addition, the set of short and long words were equated for phonological dissimilarity using Mueller et al.’s (2003) PSIMETRICA. The short words had a dissimilarity measure of .31, compared with .30 for the long words. However, the short and long words differed significantly in orthographic neighborhood size, t(28) � 5.456, p � .001, with values typical of those in previous studies (9.00 vs. 0.22).
Design and procedure. There were four types of lists: Pure lists that contained only short words, pure lists that contained only long words, mixed lists that alternated short and long words (i.e., short long short long short long), and mixed lists that alternated long and short words (i.e., long short long short long short). There were 15 trials for each type of list, randomly ordered for each participant.
On each trial, six words were randomly selected from the pool and were presented at a rate of one item per second on a computer screen. At the end of list presentation, the participants were asked to write down, in order, the words they had just seen. Strict serial recall instructions were given, such that participants were in- structed to write the items in their exact order of presentation, beginning with the first one. They were told to leave a blank line if they could not recall an item at a given serial position and were instructed not to backtrack to fill a blank. There was no time limit for recall. Once the participant had finished recalling the words, he or she clicked on a button on the computer to begin the next list.
3 We focus on orthographic rather than phonological neighborhood effects for simplicity, as one need not worry about differences in pronun- ciation (and thus phonemes) as a function of geographic region. The available data suggest both phonological and orthographic neighborhoods are highly correlated, and indeed, the measures are often confounded (cf. Yates, Locker, & Simpson, 2004).
Table 1 Orthographic Neighborhood Size for Short and Long Words in Syllable-Based Word Length Studies and the Current Study
Study
Word length
Short Long
Baddeley et al. (1975, Experiment 6) 2.88 0.00 Baddeley et al. (2002, Experiment 1) 7.20 0.30 V. Coltheart et al. (2004, Experiment 1) 7.80 0.48 Cowan et al. (1994) 10.00 0.17 Cowan et al. (1997, Experiment 2) 14.17 0.17 Cowan et al. (2003) 6.33 0.33 Hulme & Tordoff (1989) 9.83 0.00 LaPointe & Engle (1990, Experiment 5) 8.37 0.31 McNeil & Johnston (2004, Experiment 1) 8.63 0.25 Mueller et al. (2003, Experiment 1) 8.42 0.17 Romani et al. (2005, Experiment 1) 7.25 0.38 Russo & Grammatopoulou (2003, Experiment 6) 8.40 0.00 Tehan & Turcotte (2002, Experiment 1) 12.60 0.60
M 8.61 0.24 Experiments 1A and 1B 9.00 0.20 Experiments 3 and 5 1.00 1.00 Experiment 4 2.00 2.00
341LENGTH AND THE WORD LENGTH EFFECT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
Participants were tested individually, and an experimenter was present throughout to ensure compliance with the instructions.
Results
A word was considered correctly recalled only if it was written in the correct order. Following Hulme et al. (2004), we constructed derived lists for short and long words presented in mixed lists. Thus, short words in mixed lists combined the first, third, and fifth words from the short long short long short long list and the second, fourth, and sixth words from the long short long short long short list. In this and all subsequent analyses, the .05 level of signifi- cance was adopted.
As the left panel of Figure 1 shows, a classic word length effect was observed in the pure lists, with substantially better recall of short words than long words. However, recall of short and long words from mixed lists did not differ from each other, with performance intermediate between that of short words in pure lists and long words in pure lists.
A 2 � 2 repeated measures analysis of variance (ANOVA) with word length (short and long) and list type (pure and mixed) as within subject factors confirmed these observations. There was a main effect of word length, F(1, 15) � 45.05, MSE � 0.003, partial �2 � .750, with more short words correctly recalled in order than long words (.715 vs. .616, respectively). There was also a main effect of list type, F(1, 15) � 6.12, MSE � 0.004, partial �2 � .290, with slightly more words correctly recalled in order in mixed lists than pure lists (.685 vs. .646, respectively). These two factors also interacted, F(1, 15) � 46.14, MSE � 0.003, partial �2 � .755. This was due to a large difference between recall of short and long words in pure lists (.745 vs. .547) and no difference between short and long words in mixed lists (.685 for both types of items). A Tukey honestly significant difference (HSD) test confirmed that there was a reliable effect of word length in the pure lists but not in the mixed lists. Finally, a repeated measures analysis of covariance (ANCOVA) was conducted with neighbor- hood size as the covariate. The previously significant effect of length was now removed (F � 1).
Another way of assessing the results is to see how many par- ticipants show a word length effect and how many do not. In pure lists, all 16 participants recalled more short words than long words (significant by a sign test). For the mixed lists, seven participants recalled more short words than long words, with eight showing the reverse and one tie, which is not significant by a sign test.
Discussion
The syllable-based word length effect observed in Experiment 1A are exactly as predicted by phonological loop accounts (see Burgess & Hitch, 1999). Pure lists of short words were recalled more accurately than lists of long words, even though the words were equated for frequency, familiarity, concreteness, imageabil- ity, and phonological dissimilarity. In addition, recall of mixed lists was better than recall of pure long lists but worse than recall of pure short lists. According to accounts based on the phonolog- ical loop hypothesis, it takes longer to refresh a list of long words than short words, and therefore, more long words will have de- cayed too far to be recallable at the time of test than short words. Similarly, it takes more time to rehearse a list consisting of both long and short words than it takes to rehearse lists of short words, and consequently, pure lists of short word are better recalled than mixed lists. Conversely, mixed lists are rehearsed faster than pure lists of long words, making mixed lists easier to recall than pure lists of long words.
One possible problem with Experiment 1A is that written serial recall was used, which could cause a confound: Because it takes longer to write long words (telegraph, sympathy, . . .) than short words (sale, rose, . . .), output time is not equal in the two condi- tions. Experiment 1B removed this confound by using a strict serial reconstruction of order test rather than a strict written serial recall test.
Experiment 1B
Output time has been shown to be related to accuracy, with longer times associated with lower performance (e.g., Bireta et al.,
Figure 1. Proportion of short and long words correctly recalled in Experiment 1A (left panel) and Experiment 1B (middle panel), and proportion of words from large and small neighborhoods correctly recalled in Experiment 2 (right panel) as a function of list type. Error bars show the standard error of the mean.
342 JALBERT, NEATH, BIRETA, AND SURPRENANT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
2010; Dosher & Ma, 1998; Surprenant, Neath, & Brown, 2006). The purpose of Experiment 1B was to demonstrate that typical word length effects are observable even when the confound of differential output time is removed. We used the same items as in Experiment 1A, but we used a strict serial reconstruction of order test rather than written serial recall. This test yields results com- parable with those observed with written serial recall, including not only word length effects (e.g., Neath et al., 2003), irrelevant speech, and phonological similarity effects (e.g., Surprenant, Neath, & LeCompte, 1999), but also modality and suffix effects as well as effects of articulatory suppression (e.g., Surprenant, LeCompte, & Neath, 2000). More importantly, it permits output time to be equated: Unlike written or spoken recall, it takes the same amount of time to click on a button labeled with a long word as a button labeled with a short word.
Method
Participants. Sixteen different undergraduate students from Memorial University of Newfoundland participated in exchange for a small honorarium. All participants were native English speak- ers.
Stimuli, design, and procedure. The stimuli, design, and procedure were the same as in Experiment 1A except for the recall procedure. Following the presentation of the list, the six words from the current trial appeared as labels, in alphabetical order, on buttons on the computer screen, and participants were asked to reconstruct the order in which the words were presented by click- ing on the appropriately labeled buttons with the mouse. Partici- pants were asked to click on the first word first, the second word second, and so on.
Results
Despite the change in test, the results of Experiment 1B were almost identical to those of Experiment 1A. As the middle panel of Figure 1 shows, short words were better recalled than were long words in the pure lists, but recall of short and long words in mixed lists was equivalent, and between that of the short and long words from mixed lists. The results are exactly what models based on the phonological loop (e.g., Burgess & Hitch, 1999; Page & Norris, 1998) predict.
The data were analyzed with a 2 � 2 repeated measures ANOVA with word length (short vs. long) and list type (pure vs. mixed) as within subject factors, which confirmed the observations noted above. There was a significant main effect of word length, with more short words correctly recalled than long words (.687 vs. .610, respectively), F(1, 15) � 44.871, MSE � 0.002, partial �2 � .749. There was no difference in recall of pure or mixed lists (.650 vs. .647, respectively, F � 1).
Of importance, the interaction between word length and list type was significant, F(1, 15) � 19.110, MSE � 0.003, partial �2 � .562. This was due to finding a word length effect (i.e., better recall of short than long items) only in pure lists (.719 vs. .582) and not in the mixed lists (0.656 vs. 0.639). A Tukey HSD test confirmed that there was a reliable effect of word length in the pure lists but not in the mixed lists. As in Experiment 1A, a repeated measures ANCOVA was conducted with neighborhood size as the covariate. The previously significant effect of length was now removed (F � 1).
In pure lists, 15 participants recalled more short words than long words, one showed the reverse pattern, and there were no ties (significant by a sign test). For the mixed lists, 10 participants recalled more short words than long words, with four showing the reverse and two ties, which is not significant by a sign test ( p � .15).
Each time a participant clicked on a response button, the time of the click was recorded. These output time data were analyzed with a 2 � 2 repeated measures ANOVA with word length (short vs. long) and list type (pure vs. mixed) as within subject factors. There was no difference in output time as a function of word length (F � 1), with equivalent mean output time for short and long words (1.488 s vs. 1.479 s). There was also no difference in output time as a function of list type (F � 1), with equivalent mean output times in pure and mixed lists (1.471 s vs. 1.496 s). The interaction was not reliable, F(1, 15) � 1.141, MSE � 0.022, partial �2 � .170.
Discussion
Experiment 1B demonstrated that a robust word length effect is observable in our paradigm with a strict serial reconstruction of order test. Short words were better recalled than were long words in pure lists, but not in mixed lists; here, recall was between that of pure short and pure long lists, and recall did not differ between mixed short and mixed long lists. This pattern is exactly what the Burgess and Hitch (1999) model, based on the phonological loop, predicts. This pattern also differs subtly from previous patterns seen with pure versus mixed lists. Unlike the results using the stimuli of Cowan et al. (2003), no word length effect was seen in mixed lists. Unlike the results using the stimuli of Hulme et al. (2004), recall of short and long items from mixed lists was worse than that of pure short lists.
There are several possible reasons for the difference. First, we were successful in equating output time, as the analysis above showed. Bireta et al. (2010) also measured output time and ob- served a word length effect with pure lists when output times did not differ (see also Surprenant et al., in press). It is not known whether output times differed in the other studies, but this could easily be a factor. Second, it is possible that differences in the stimulus sets is the cause, particularly as the current set of stimuli were equated on more dimensions than either the Cowan et al. (2003) or the Hulme et al. (2004) stimuli. Given that serial recon- struction of order tasks remove a potential confound, relative to written or spoken recall, and given only one minor and theoreti- cally unimportant difference—whether the main effect of list type was significant— between Experiments 1A and 1B, Experiments 2– 4 used reconstruction of order.
Experiment 2
A recall advantage for words with large orthographic neighbor- hoods has been demonstrated with three different sets of CVC words (Roodenrys et al., 2002; see also Allen & Hulme, 2006), as well as with nonwords (Roodenrys & Hinton, 2002). Each dem- onstration had auditory presentation and spoken recall. The pur- pose of Experiment 2 was to determine whether the beneficial effect of a larger orthographic neighborhood is observable with visual presentation and strict serial reconstruction of order.
343LENGTH AND THE WORD LENGTH EFFECT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
Method
Participants. Sixteen different undergraduate students from Memorial University of Newfoundland volunteered to participate in exchange for a small honorarium. All participants were native speakers of English.
Stimuli. The stimuli were the 32 low neighborhood frequency three-phoneme CVC words from Experiment 3 of Roodenrys et al. (2002; see their Appendix D for details). Although initially se- lected for a manipulation of phonological neighborhood size— half the words had large neighborhoods and half had small neighbor- hoods—the set of words used by Roodenrys et al. also differ in terms of orthographic neighborhood size. Orthographic neighbor- hood size and frequency were calculated with the MCWord Da- tabase (Medler & Binder, 2005) and the values are shown in Table 2. We also verified that the small and large neighborhood words did not differ in terms of the PSIMETRICA measure of phono- logical dissimilarity: This value was .30 for the small neighbor- hood set, compared with .31 for the large neighborhood set.
Design and procedure. Except for the substitution of ortho- graphic neighborhood size for word length, the design and proce- dure were identical to that in Experiment 1B. That is, neighbor- hood size (small vs. large) and list type (pure vs. mixed) were both within-subjects variables, and all lists contained six words. Pure small lists contained only words with small neighborhoods, and pure large lists contained only words with large neighborhoods. Mixed lists alternated words with different neighborhood sizes. Half the mixed lists began with a small neighborhood word (i.e., small, large, small, large, small, large), and the other half began with a large neighborhood word (i.e., large, small, large, small, large, small). To construct each list, six words were drawn ran- domly from the appropriate pool. There were 15 trials for each of the four types of list, and these were randomly ordered for each participant.
Results
As can be seen in the right panel of Figure 1, words with a large neighborhood were recalled better than were words with a small neighborhood in pure lists, replicating the basic effect observed by Roodenrys et al. (2002) and Allen and Hulme (2006). Recall of large and small neighborhood words did not differ in the mixed lists. This pattern is reminiscent of that observed in Experiments 1A and 1B, in which word length was manipulated.
The data were analyzed with a 2 � 2 repeated measures ANOVA with neighborhood size (large vs. small) and list type
(pure vs. mixed) as within-subject factors. The main effect of neighborhood size was significant, F(1, 15) � 17.566, MSE � 0.004, partial �2 � .539, with better recall of words with large neighborhoods than those with smaller neighborhoods (.719 vs. .656, respectively). The main effect of list type was not significant (F � 1), with approximately equivalent recall of pure and mixed lists (.682 vs. .693, respectively).
The interaction was significant, F(1, 15) � 13.801, MSE � 0.004, partial �2 �.479, due to an effect of neighborhood size in pure lists (.742 vs. .626) but no such effect in mixed lists (.697 vs. .689). A Tukey HSD test confirmed that there was a reliable effect of neighborhood size in pure lists but not in mixed lists.
Again, we determined how many participants showed the or- thographic neighborhood effect and how many did not. For pure lists, 13 participants recalled more large than small neighborhood words, two showed the reverse pattern, and one showed no differ- ence. This is significant by a sign test. For the mixed lists, six participants recalled more large than small neighborhood words, with 10 showing the reverse and no ties; this is not significant by a sign test ( p � .40).
The output time data were analyzed with a 2 � 2 repeated measures ANOVA with neighborhood size (large vs. small) and list type (pure vs. mixed) as within-subject factors. There was a difference in output time as a function of neighborhood size, F(1, 15) � 5.396, MSE � 0.039, partial �2 � .265, with faster mean output times for words with more neighbors than those with fewer neighbors (1.356 s vs. 1.470 s). There was no difference in output time as a function of list type (F � 1), with equivalent mean output times in pure and mixed lists (1.407 s vs. 1.418 s). The interaction was not reliable (F � 1).
Discussion
With pure lists, Experiment 2 replicated the neighborhood size effect reported by Roodenrys et al. (2002) and did so despite the many changes in design and procedure. Words with a large pho- nological or orthographic neighborhood are better recalled on immediate tests of serial recall than words with smaller neighbor- hoods. It does not matter if presentation is auditory or visual or if the test is memory span with spoken recall, immediate spoken serial recall, or strict reconstruction of order.
In mixed lists, however, there was no effect of neighborhood size. Performance in these lists was between that of the pure large and pure small conditions. This result may be seen as surprising, in that neighborhood size would appear to be an item property, and as such, list type should not matter. However, other lexical properties, such word frequency, have been shown to have different effects depending on whether the two kinds of items (e.g., low frequency vs. high frequency) appear in the same or in separate lists (e.g., Watkins, LeCompte, & Kim, 2000). If it is the case that neigh- borhood size is driving the word length effect then the finding of no neighborhood size effect in mixed lists is to be expected: Experiments 1A and 1B showed effects of word length in pure lists but not in mixed lists, and this disappearance is exactly what models based on the phonological loop predict (e.g., Burgess & Hitch, 1999).
The results of Experiment 2 provide more evidence that the confound between length and neighborhood size shown in Table 1 could be important. Although the length of the items may be
Table 2 Orthographic and Phonological Neighborhood Size and Orthographic and Phonological Frequency for the Words Used in Experiment 2
Measure
Small Large
M SD M SD
Orthographic size 3.8 3.2 12.6 3.7 Orthographic frequency 50.0 114.6 31.9 62.2 Phonological size 7.4 4.0 24.3 2.0 Phonological frequency 55.8 3.3 56.7 2.7
344 JALBERT, NEATH, BIRETA, AND SURPRENANT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
causing the word length effect, it is also possible that the difference in neighborhood size plays a role. We test this in Experiment 3.
Experiment 3
Experiments 1A and 1B demonstrated that a word length effect is observed in pure but not mixed lists with both written recall and reconstruction of order tests. However, length and neighborhood size were confounded in Experiments 1A and 1B, and it is not clear which factor is driving the effect. Given that Experiment 2 showed a beneficial effect of neighborhood size in pure lists but not in mixed lists, the purpose of Experiment 3 was to determine what happens when short and long items are equated for orthographic neighborhood size and frequency, in addition to word frequency, concreteness, imageability, familiarity, phonological dissimilarity, and output time. If orthographic neighborhood size plays no role in the word length effect and the effects observed in Experiment 1b are due to length per se, Experiment 3 should replicate Experiment 1B. If, on the other hand, the effects observed in Experiment 1B are due to neighborhood characteristics, there should be no differ- ence in recall of short and long words in either pure or mixed lists.
Method
Participants. Thirty-two undergraduate students from Me- morial University of Newfoundland and The College of New Jersey participated in exchange for a small honorarium or course credit. All participants were native speakers of English, and none had been in previous experiments.
Stimuli. A set of 13 short and 13 long words was created (see Appendix B) in which the short and long words were equated on the same dimensions as in Experiments 1A and 1B as well as being equated for orthographic neighborhood size and frequency. For these measures, the smallest p value associated with a t test was p � .48. The measure of phonological dissimilarity was .33 for the short words, compared with .28 for the long words.
Design and procedure. With the exception of the stimuli used, the design and procedure were the same as in Experiment 1B.
Results
As can be seen in the left panel of Figure 2, the word length effect observed in Experiment 1B was not present in Experiment 3. Recall of short words, whether in pure or mixed lists, did not differ from recall of long words, whether in pure or mixed lists. That is, there was no effect of word length with a set of short and long words that were equated for neighborhood size.
A 2 � 2 repeated measures ANOVA with word length (short vs. long) and list type (pure vs. mixed) as within subject factors confirmed these trends. There was no effect of word length, F(1, 31) � 1.532, MSE � 0.005, partial �2 � .047, with short and long words recalled equivalently (0.708 vs. 0.723, respectively). There was a significant effect of list type, F(1, 31) � 5.799, MSE � 0.005, partial �2 � .158; the proportion of items recalled from mixed lists was .730, compared with .700 for pure lists. The interaction between length and list type was not significant (F � 1).
Because no effect of length was observed, either overall or just in the pure lists, it is possible that participants adopted a different strategy than in previous experiments. In particular, it is possible that participants were focusing on just the first letter of each word rather than on the whole word. If the participants were adopting this strategy, a list with words sharing the same first letter (i.e., tree, table, soap, sack) would be harder to recall than a list of words with a different first letter (i.e., tree, chair, soap, bag). To assess this possibility, we reanalyzed the data including shared first letter among the items as a covariate. A word was considered as having a shared first letter if it shared its first letter with another short or long word in the list. A 2 � 2 ANCOVA was conducted with word length (short vs. long) and list type (pure vs. mixed) as factors and shared first letter as a covariate. It is important to note that the homogeneity of the regression assumption was respected. When shared first letter was controlled for, there was still no effect of word length (F � 1), with short and long words recalled equivalently (.737 vs. .718, respectively). There was also no effect of list type (F � 1); the proportion of items recalled from mixed lists was .736, compared with .719 for pure lists. The interaction
Figure 2. Proportion of short and long words correctly recalled in Experiment 3 (left panel), Experiment 4 (middle panel), and Experiment 5 (right panel) as a function of list type. Error bars show the standard error of the mean.
345LENGTH AND THE WORD LENGTH EFFECT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
between length and list type was also not significant, F(1, 62) � 1.291, MSE � 0.010, partial �2 � 0.020, p � .25.
Another possibility for the lack of a word length effect in the experiment may be that the short and long words differed on a dimension that we did not consider when creating the stimulus sets. Although the set of short and long words had similar PSIMETRICA values, a measure of dissimilarity, we examined three other mea- sures of similarity that have been shown to affect memory perfor- mance: bigram frequency, neighbor overlap, and Levenshtein sim- ilarity (Freeman, Heathcote, Chalmers, & Hockley, 2010). Bigram frequency measures were obtained from the Medler and Binder (2005) database, and they refer to the averaged frequency of specific two-letter combinations within the CELEX database. Neighbor overlap is the number of times that words in the ortho- graphic neighborhood of the short items also appear in the ortho- graphic neighborhood of the long items. Levenshtein similarity is the inverse of Levenshtein distance, which is a number that rep- resents the total number of additions, deletions, and substitutions required to change one word into another.
The short and long items did not differ significantly in bigram frequency, t(22) � 1.39, p � .15, and there was no neighbor overlap. The short and long words did, however, differ signifi- cantly in terms of Levenshtein similarity: The value for short items was .21, compared with .13 for the long items, t(152) � 15.5, p � .01. Therefore, an ANCOVA was conducted on recall performance with word length as a factor and the Levenshtein similarity as a covariate. The homogeneity of the regression assumption was respected. When Levenshtein similarity was controlled for, there was no difference in recall performance as a function of length, F(1, 23) � 1.48, MSE � 0.001, partial �2 � .060, p � .20.
In pure lists, 15 participants recalled more short than long words, 17 showed the reverse pattern, and there were no ties. In the mixed list, the same pattern was observed. Neither was significant by a sign test ( p � .80).
Output time was again analyzed with a 2 � 2 repeated measures ANOVA with word length (short vs. long) and list type (pure vs. mixed) as within subject factors. There was no difference in output time as a function of length, F(1, 31) � 1.500, MSE � 0.023, partial �2 � 0.046, p � .20, with short items taking 1.211 s to recall, compared with 1.178 s for long items. There was also no difference in output time as a function of list type (F � 1), with equivalent mean output times in pure and mixed lists (1.204 s vs. 1.185 s). The interaction was not reliable (F � 1).
Discussion
The only change between Experiment 1B and Experiment 3 was the set of words used, and the specific change was removing the confound of length and orthographic neighborhood size. The short words in both experiments were all monosyllabic, and the long words were all trisyllabic. However, all the words in Experiment 3 had an orthographic neighborhood of 1.0. Despite seeing robust effects of word length in Experiment 1B, no such effects were observed in Experiment 3.
It is not plausible to argue that we did not have a sufficiently powerful manipulation of length. First, the number of syllables in the short and long words was the same as in Experiment 1B. Second, although we did not measure pronunciation time, an informal examination showed that no matter what temporal mea-
sure was used (i.e., “normal” speaking, fast speaking, etc.), the long words were longer than the short. Third, we observed a word length effect in Experiment 1B with half the number of partici- pants as in Experiment 3. Even so, null results may obtain for a variety of reasons, and given the variability in results in word length effect experiments due to the particular stimulus set used, we deemed a replication necessary. To this end, Experiment 4 was designed as a replication of Experiment 3 but with a different set of stimuli.
Experiment 4
One possible concern with Experiment 3 is that the null results observed are due to some peculiarity of the particular stimulus set used. Experiment 4, therefore, was a replication of Experiment 3, but with a new set of short and long words that were also equated for orthographic neighborhood size.
Method
Participants. Thirty-two different undergraduate students from Memorial University of Newfoundland and the College of New Jersey participated in exchange for a small honorarium or course credit. All participants were native speakers of English, and none had participated in previous experiments.
Stimuli, design, and procedure. The only change from Ex- periment 3 was the set of stimuli. A new set of 14 short and 14 long words was created (see Appendix C) in which the short and long words were equated on the same dimensions as in Experiment 3. For these measures, the smallest p value associated with a t test was p � .56. The PSIMETRICA measure of phonological dissim- ilarity was .28 for the short words, compared with .28 for the long words. In addition, the short and long words were equated for orthographic neighborhood size (this time 2.0 rather than 1.0) as well as orthographic frequency.
Results and Discussion
As can be seen in the middle panel of Figure 2, Experiment 4 replicated Experiment 3: With short and long words equated for orthographic neighborhood size, there were no apparent effects of word length.
A 2 � 2 repeated measures ANOVA with word length (short vs. long) and list type (pure vs. mixed) as within subject factors confirmed these trends. There was no effect of word length, F(1, 31) � 1.132, MSE � 0.008, partial �2 � .035, with short and long words recalled equivalently (.684 vs. .701, respectively). Unlike in Experiment 3, the effect of list type failed to reach conventional levels of significance, F(1, 31) � 3.441, MSE � 0.008, partial �2 � .100, p � .05; the proportion of items recalled from mixed lists was .708, compared with .678 for pure lists. The interaction between length and list type was not significant (F � 1).
Because of the finding of no word length effect, we reanalyzed the data including shared first letter among the items as a covariate. A 2 � 2 ANCOVA was conducted with word length (short vs. long) and list type (pure vs. mixed) as factors and shared first letter as a covariate. The homogeneity of the regression assumption was respected. When shared first letter was controlled for, there was no effect of word length (F � 1), with recall approximately the same
346 JALBERT, NEATH, BIRETA, AND SURPRENANT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
for short and long words (0.728 vs. 0.713, respectively). The main effect of list type failed to reach significance (F � 1), with mixed lists being recalled as well as pure lists (.733 vs. .708, respec- tively). The interaction between length and list type was also not significant (F � 1).
As in Experiment 3, we again assessed whether additional measures of similarity yielded differences between the short and long words. As with the stimuli used in Experiment 3, the short and long words used here did not differ in bigram frequency, t(26) � 0.59, p � .50, and there was no neighbor overlap. The short and long items did differ in Levenshtein similarity, with a value of .22 for short and .15 for long items, t(178) � 11.4, p � .01. An ANCOVA was conducted on recall performance with word length (short vs. long) as a factor and Levenshtein similarity as a covariate. The homogeneity of the regression assumption was respected. When Levenshtein similarity was controlled for, there was still no difference in recall performance as a function of length, F(1, 25) � 1.57, MSE � 0.001, partial �2 � .059, p � .20.
In pure lists, 16 participants recalled more short than long words, 15 showed the reverse pattern, and there was one tie. This is not significant by a sign test ( p � .90). In mixed lists, nine participants recalled more short than long words, 19 showed the reverse, and there were four ties. Although the latter just fails to reach conventional levels of significance ( p � .08), the direction of the difference is in favor of the long words, not the short words.
Output time was analyzed with a 2 � 2 repeated measures ANOVA with word length (short vs. long) and list type (pure vs. mixed) as within subject factors. There was no difference in output time as a function of length, F(1, 31) � 1.936, MSE � 0.031, partial �2 � .059, p � .15, with equivalent mean output time for short and long words (1.191 s vs. 1.234 s). Similarly, there was no difference in output time as a function of list type (F � 1), with equivalent mean output times in pure and mixed lists (1.220 s vs. 1.204 s). The interaction was not reliable (F � 1).
Experiment 4, with a different set of stimuli, replicated the null results from Experiment 3: When short and long words are equated for orthographic neighborhood size, there is no difference in recall of the short and long words.
Experiment 5
One possible concern is that the null results observed in Exper- iments 3 and 4 could be due to the recall method, even though there was no important difference between the results of Experi- ments 1A (strict written serial recall) and 1B (strict serial recon- struction of order). Nonetheless, Experiment 5 was a replication of Experiment 3, but with a spoken strict serial recall test.
Method
Participants. Sixteen undergraduate students from Memorial University of Newfoundland volunteered to participate in ex- change for a small honorarium. All participants were native speak- ers of English, and none had participated in previous experiments.
Stimuli, design, and procedure. The stimuli, design, and procedure were the same as in Experiment 3, except for the recall procedure. Following the presentation of the list, participants were asked to repeat out loud the words that were just presented. They were instructed to do so in the order of presentation.
Results and Discussion
As can be seen in the right panel of Figure 2, Experiment 5 replicated Experiments 3 and 4 in that there was no word length effect apparent in either the pure or mixed lists when the short and long words were equated for orthographic neighborhood size.
Spoken recall performance was analyzed with a 2 � 2 repeated measures ANOVA with word length (short vs. long) and list type (pure vs. mixed) as within subject factors. There was no difference in recall performance as a function of word length, F(1, 15) � 1.666, MSE � 0.003, partial �2 � .100, p � .20, with a similar recall performance for short and long words (.614 vs. .641). There was a significant main effect of list type, F(1, 15) � 4.840, MSE � 0.005, partial �2 � .244, with better recall in mixed lists than pure lists (.651 vs. .614). The interaction between word length and list type was significant, F(1, 15) � 14.785, MSE � 0.003, partial �2 � .496. A Tukey HSD test confirmed that there was a reliable reversed word length effect in the mixed lists but not in the pure lists.
This latter finding differs from that seen in Experiments 3 and 4, but we note that the change in test could lead to differences in processing. For example, preparing to say a list of items out loud may lead participants to rehearse or process the items differently than when no articulation is required. One consequence may be enhanced or more complete access of lexical information. How- ever, it could also be the case that this reverse effect is spurious; further empirical work is required before more can be made of this finding. It is important to note, however, that this difference is, if anything, even more problematic for traditional accounts of the word length effect. The major result is consistent with the idea that factors other than length per se may be behind the word length effect.
General Discussion
Experiment 1B demonstrated that strict reconstruction of order tests produce the same standard word length effect seen with strict written serial recall tests. It is important to note that the set of stimuli used deliberately confounded length and neighborhood size, such that the one-syllable short items had a larger neighbor- hood than the three-syllable long items. Experiment 2 replicated and extended the results of Roodenrys et al. (2002) and Allen and Hulme (2006) by showing that visually presented words with large orthographic neighborhoods were better recalled than words with smaller orthographic neighborhoods. Experiment 3 used a set of short (one-syllable) and long (three-syllable) words equated for orthographic neighborhood size and frequency, and the word length effect observed in Experiment 1B disappeared. Experiment 4 replicated the null results of Experiment 3 with a different set of short and long words equated for orthographic neighborhood and frequency. Finally, Experiment 5 extended the results of Experi- ments 3 and 4 by showing no word length effect with a spoken recall test.
The short and long words in Experiments 3, 4, and 5 were equated on all relevant dimensions thought to be important, but no effects of length were observed. Because it is possible that we have overlooked some other important dimension, and because of the past history of differing word length results as a function of specific stimulus sets (e.g., Neath et al., 2003; Bireta et al., 2006),
347LENGTH AND THE WORD LENGTH EFFECT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
we think it critically important that other researchers replicate these results using different stimulus sets. For any such sets, we recommend that in addition to controlling for output time, re- searchers ensure that their short and long words are equated on at least the following dimensions: concreteness, familiarity, image- ability, frequency (Kucera-Francis, Thorndike-Lorge, and CELEX), orthographic frequency, orthographic neighborhood size, bigram frequency, neighbor overlap, PSIMETRICA dissim- ilarity, and Levenshtein similarity.
Until more stimulus sets are tested, however, we suggest that the following conclusion is warranted: Neighborhood size—and pos- sibly other related lexical and linguistic factors correlated with it—rather than length per se is one of the critical factors underlying the syllable-based word length effect.
Accounting for Neighborhood Effects
Why does a large neighborhood size benefit immediate recall? This result is surprising, as large neighborhood size has previously been associated with detrimental effects. In particular, there is a large literature that shows that spoken word recognition is better facilitated for words with smaller neighborhoods than those with larger neighborhoods (e.g., Luce & Pisoni, 1998). However, facil- itative effects for words with large neighborhoods have been shown on certain production—as opposed to perception—tasks. For example, Vitevitch (2002) showed that more errors were elicited for words with fewer similar sounding words (i.e., small neighborhood) than words with more similar sounding words (i.e., large neighborhood). Similarly, in a picture naming task, words from small neighborhoods were identified more slowly than words from large neighborhoods (see also Vitevitch & Sommers, 2003).
Roodenrys (2009; see also Roodenrys & Miller, 2008) sug- gested one way in which both phonological and orthographical neighborhood size could have a beneficial effect on recall. Many models of memory posit that at retrieval, one major task facing the rememberer is the interpretation of degraded items. Typically, a redintegrative process is invoked which recruits additional infor- mation to help interpret the ambiguous remnants of the to-be- remembered items. If one were to assume that the degraded string serves as input to an interactive network, such as might be en- countered in speech production, then the slight activation in the network accruing from the commonalities of the neighbors could readily lead to more successful redintegration of a target. Such a process could also be extended to account for other beneficial effects of linguistic or lexical factors, and this could be added to those models that already include a redintegrative component.
Implications for Theories
We noted earlier that only one set of English words reliably produces a time-based word length effect, whereas all other sets tested so far do not (e.g., Lovatt et al., 2000; Neath et al., 2003). To this, we now add evidence that previous demonstrations of the syllable-based word length effect may be due to a confound in the stimulus sets, and when this confound is removed, so too are the effects of length.
To the extent that additional sets of stimuli can be found in which short and long words are appropriately equated and no word length effect emerges, models and theories based on the phono-
logical loop (e.g., Baddeley, 1986, 1992; Burgess & Hitch, 1999; Page & Norris, 1998) are critically compromised. The basic archi- tecture of these models requires that a word length effect be observed; if no such effects are observable then the processes and architecture that predict the word length effect would need to be removed. Doing so, however, would also remove the model’s ability to account for many other aspects of immediate memory.
The implications for accounts based on item properties are different. The account offered by the feature model (Neath & Nairne, 1995) does not require a time-based word length effect, so the lack of one is not a problem, but it does make an incorrect prediction about the syllable-based word length effect in mixed lists (see Hulme et al., 2004; Hulme et al., 2006), and the results of Experiment 1 compound this problem. However, if length per se is no longer a factor that needs to be explained, the processes that produce a word length effect can be removed. Unlike the case for models based on the phonological loop, removing these word- length specific processes does not affect the feature model’s ability to account for the other core phenomena. Indeed, because a rudi- mentary redintegrative process was included in the original version of the model (Nairne, 1990), it may be possible to add the bene- ficial redintegrative effects of a large neighborhood.
The Brown and Hulme (1995) model also explained the effects of length based on item-specific factors and made incorrect pre- dictions about recall of short and long words in mixed lists. As the model was intended to demonstrate that rehearsal was not neces- sary for the word length effect, its scope and purpose was limited. With the demonstration that length effects are not always ob- served, the fundamental assumption of this account, differential decay rates, is also questioned. This does not, of course, make the model meaningless; rather, it continues to serve as an existence proof that rehearsal is not necessary in order to explain certain immediate memory effects.
The final model we consider (Hulme et al. (2004; see also Neath & Brown, 2006) is based on the framework of SIMPLE (scale invariant memory and perceptual learning; Brown, Neath, & Chater, 2007). SIMPLE is a relative distinctiveness model and assumes that items are represented on one or more dimensions. An item that “stands out” on its dimension (or position in multidimen- sional space) will be better recalled than one that has lots of neighbors. The word length effect was explained by noting that short words are typically more distinctive (i.e., easier to appre- hend) than long items. In mixed lists, long words benefit from emergent distinctiveness; that is, compared with the short items, they now stand out more than when presented in a pure list. Indeed, when only one long item appears in a list of short items, it is in fact recalled better than the short items (Hulme et al., 2006). The challenge for SIMPLE is resolve the paradox that items with fewer close neighbors are seen as more distinct but items with more orthographic (or phonological) neighbors are recalled better. SIMPLE does not yet include a redintegration stage.
Time and Memory
As Nairne (2002) noted in his comprehensive review, the so-called “standard model” of short-term/working memory posits that items decay unless offset with rehearsal, and rehearsal speed is assumed to be related to pronunciation time. If items take longer to rehearse, less of them can be refreshed, so fewer can be recalled, compared with
348 JALBERT, NEATH, BIRETA, AND SURPRENANT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
shorter items. The syllable-based word length effect is a highly robust phenomenon demonstrated in numerous studies. However, those stud- ies have confounded length with orthographic neighborhood size (see Table 1). When short and long words are equated for neighborhood size, no word length effect is observed. This result is devastating for any model that incorporates the idea of time-based decay offset by rehearsal: It is simply not possible to explain why three-syllable words are recalled equally as well as one-syllable words when three-syllable words take longer to rehearse and so should be more prone to decay.
Historically, decay as a cause of forgetting has been vigorously and repeatedly rejected (e.g., McGeoch, 1932; Osgood, 1953), and it was not until the so-called cognitive revolution that theories started in- cluding decay and de-emphasizing other causes of forgetting (see Neath, 1998, for a review). Now, it appears as though the tide is turning once again away from time-based decay as an explanatory construct. Indeed, there are an increasing number of empirical (e.g., Berman, Jonides, & Lewis, 2009) and theoretical (e.g., Le- wandowsky, Oberauer, & Brown, 2009) articles which suggest that time-based decay simply does not exist; instead, forgetting is attrib- uted to a number of different causes, including interference, changed cues, inappropriate processing, relative distinctiveness, and the like. Our results add to this growing consensus.
Working Memory and Language
A final speculation concerns the relation between memory and language development. One dominant view is that as the phono- logical loop develops, it aids vocabulary acquisition (see Badde- ley, 2003, for a review). However, if the word length effect is due to neighborhood size, one possible implication is the reverse: that as vocabulary develops, it aids the development of the phonolog- ical loop or, more precisely, the immediate recall of items in order. The data are largely correlational, so the direction of causation is open to interpretation. Indeed, this suggestion is not new (see, for example, Brown & Hulme, 1996; Snowling, Chiat, & Hulme, 1991) and fits nicely into a recent emphasis of speech production mechanisms on immediate memory (e.g., Woodward, Macken, & Jones, 2008). The basic idea is that as vocabulary is acquired, memory as measured by span or immediate serial recall increases due to the introduction of neighborhood and other lexical effects.
Summary
The word length effect has been termed one of the benchmark findings that any theory of short-term memory must account for. Indeed, the effect was one that led directly to the development of working memory and the phonological loop. Experiments 1A and 1B replicated the typical effects of length when short and long words were equated on all relevant dimensions previously identified in the literature. However, previous studies investigating the effect of word length have confounded length with orthographic neighborhood size. In English, short words are more likely to have a larger neighborhood size than long words, and Experiment 2 replicated the finding that words with a large neighborhood are recalled better than words with a small neighborhood. When a new set of short and long items were also equated for neighborhood size, the word length effect disap- peared. These findings add to the growing literature showing that performance in many memory tasks is affected by particular proper- ties of the stimulus set used and compounds the problems for theories
of memory, such as working memory, that include decay offset by rehearsal as a central feature.
References
Allen, R., & Hulme, C. (2006). Speech and language processing mecha- nisms in verbal serial recall. Journal of Memory and Language, 55, 64 – 88. doi:10.1016/j.jml.2006.02.002
Baddeley, A. (1992). Is working memory working? The Fifteenth Bartlett Lecture. Quarterly Journal of Experimental Psychology, 44A, 1–31.
Baddeley, A. D. (1986). Working memory. Oxford, England: Oxford Uni- versity Press.
Baddeley, A. (2003). Working memory and language: An overview. Jour- nal of Communication Disorders, 36, 189 –208. doi:10.1016/S0021- 9924(03)00019-4
Baddeley, A. D., Chincotta, D., Stafford, L., & Turk, D. (2002). Is the word length effect in STM entirely attributable to output delay? Evidence from serial recognition. The Quarterly Journal of Experimental Psychol- ogy, 55A, 353–369.
Baddeley, A. D., Thomson, N., & Buchanan, M. (1975). Word length and the structure of short-term memory. Journal of Verbal Learning and Verbal Behavior, 14, 575–589. doi:10.1016/S0022-5371(75)80045-4
Berman, M. G., Jonides, J., & Lewis, R. L. (2009). In search of decay in verbal short-term memory. Journal of Experimental Psychology: Learn- ing, Memory, and Cognition, 35, 317–333. doi:10.1037/a0014873
Bireta, T. J., Fry, S. E., Jalbert, A., Neath, I., Surprenant, A. M., Tehan, G., & Tolan, G. A. (2010). Backward serial recall and benchmark effects of working memory. Memory & Cognition, 38, 279 –291. doi:10.3758/ MC.38.3.279
Bireta, T. J., Neath, I., & Surprenant, A. M. (2006). The syllable-based word length effect and stimulus set specificity. Psychonomic Bulletin & Review, 13, 434 – 438.
Brown, G. D. A., & Hulme, C. (1995). Modeling item length effects in memory span: No rehearsal needed? Journal of Memory and Language, 34, 594 – 621. doi:10.1006/jmla.1995.1027
Brown, G. D. A., & Hulme, C. (1996). Nonword repetition, STM, and age-of-acquisition versus pronunciation-time limits in immediate recall for forgetting-matched acquisition: A computational model. In S. E. Gathercole (Ed.), Models of short-term memory (pp. 129 –148). Hove, England: Psychology Press.
Brown, G. D. A., Neath, I., & Chater, N. (2007). A temporal ratio model of memory. Psychological Review, 114, 539 –576. doi:10.1037/0033- 295X.114.3.539
Burgess, N., & Hitch, G. J. (1999). Memory for serial order: A network model of the phonological loop and its timing. Psychological Review, 106, 551–581. doi:10.1037/0033-295X.106.3.551
Chomsky, N., & Halle, M. (1968). The sound pattern of English. New York, NY: Harper & Row.
Coltheart, M., Davelaar, E., Jonasson, J. T., & Besner, D. (1977). Access to the internal lexicon. In S. Dornic (Ed.), Attention and performance VI (pp. 535–555). San Diego, CA: Academic Press.
Coltheart, V., Mondy, S., Dux, P. E., & Stephenson, L. (2004). Effects of orthographic and phonological word length on memory for lists shown at RSVP and STM rates. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30, 815– 826. doi:10.1037/0278-7393.30.4.815
Cowan, N. (1995). Attention and memory: An integrated framework. New York, NY: Oxford University Press.
Cowan, N., Baddeley, A. D., Elliot, E. M., & Norris, J. (2003). List composition and the word length effect in immediate recall: A compar- ison of localist and globalist assumptions. Psychonomic Bulletin & Review, 10, 74 –79.
Cowan, N., Day, L., Saults, J. S., Kellar, T. A., Johnson, T., & Flores, L. (1992). The role of verbal output time in the effect s of word-length on immediate memory. Journal of Memory and Language, 31, 1–17. doi: 10.1016/0749-596X(92)90002-F
349LENGTH AND THE WORD LENGTH EFFECT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
Cowan, N., Wood, N. L., & Borne, D. N. (1994). Reconfirmation of the short-term storage concept. Psychological Science, 5, 103–106. doi: 10.1111/j.1467-9280.1994.tb00639.x
Cowan, N., Wood, N. L., & Nugent, L. D., & Treisman, M. (1997). There are two word-length effects in verbal short-term memory: Opposed effects of duration and complexity. Psychological Science, 8, 290 –295. doi:10.1111/j.1467-9280.1997.tb00440.x
Dosher, B. A., & Ma, J.-J. (1998). Output loss or rehearsal loop? Output- time versus pronunciation-time limits in immediate recall for forgetting- matched materials. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24, 316 –335. doi:10.1037/0278-7393.24.2.316
Freeman, E., Heathcote, A., Chalmers, K., & Hockley, W. (2010). Item effects in recognition memory for words. Journal of Memory and Lan- guage, 62, 1–18. doi:10.1016/j.jml.2009.09.004
Goh, W. D., & Pisoni, D. B. (2003). Effects of lexical competition on immediate memory span for spoken words. The Quarterly Journal of Experimental Psychology, 56A, 929 –954.
Hulme, C., Neath, I., Stuart, G., Shostak, L., Surprenant, A. M., & Brown, G. D. A. (2006). The distinctiveness of the word-length effect. Journal of Experimental Psychology: Learning, Memory, and Cognition, 32, 586 –594. doi:10.1037/0278-7393.32.3.586
Hulme, C., Surprenant, A. M., Bireta, T. J., Stuart, G., & Neath, I. (2004). Abolishing the word-length effect. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30, 98 –106. doi:10.1037/0278- 7393.30.1.98
Hulme, C., & Tordoff, V. (1989). Working memory development: The effects of speech rate, word length, and acoustic similarity on serial recall. Journal of Experimental Child Psychology, 47, 72– 87. doi: 10.1016/0022-0965(89)90063-5
La Pointe, L. B., & Engle, R. W. (1990). Simple and complex word spans as measures of working memory capacity. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 1118 –1133. doi: 10.1037/0278-7393.16.6.1118
Lewandowsky, S., & Farrell, S. (2008). Short-term memory: New data and a model. In B. H. Ross (Ed.), The psychology of learning and motivation (Vol. 49, pp. 1– 48). New York, NY: Elsevier.
Lewandowsky, S., & Oberauer, K. (2008). The word-length effect provides no evidence for decay in short-term memory. Psychonomic Bulletin & Review, 15, 875– 888. doi:10.3758/PBR.15.5.875
Lewandowsky, S., Oberauer, K., & Brown, G. D. A. (2009). No temporal decay in verbal short-term memory. Trends in Cognitive Sciences, 13, 120 –126. doi:10.1016/j.tics.2008.12.003
Longoni, A. M., Richardson, J. T. E., & Aiello, A. (1993). Articulatory rehearsal and phonological storage in working memory. Memory & Cognition, 21, 11–22.
Lovatt, P., Avons, S. E., & Masterson, J. (2000). The word-length effect and disyllabic words. The Quarterly Journal of Experimental Psychol- ogy, 53A, 1–22. doi:10.1080/027249800390646
Luce, P. A., & Pisoni, D. B. (1998). Recognizing spoken words: The neighborhood activation model. Ear & Hearing, 19, 1–36. doi:10.1097/ 00003446-199802000-00001
McGeoch, J. A. (1932). Forgetting and the law of disuse. Psychological Review, 39, 352–370. doi:10.1037/h0069819
McNeil, A. M., & Johnston, R. S. (2004). Word length, phonemic, and visual similarity effects in poor and normal readers. Memory & Cogni- tion, 32, 687– 695.
Medler, D. A., & Binder, J. R. (2005). MCWord: An on-line orthographic database of the English language. Retrieved from http://www.neuro .mcw.edu/mcword/
Mueller, S. T., Seymour, T. L., Kieras, D. E., & Meyer, D. E. (2003). Theoretical implications of articulatory duration, phonological similar- ity, and phonological complexity in verbal working memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29, 1353– 1380. doi:10.1037/0278-7393.29.6.1353
Nairne, J. S. (1990). A feature model of immediate memory. Memory & Cognition, 18, 251–269.
Nairne, J. S. (2002). Remembering over the short-term: The case against the standard model. Annual Review of Psychology, 53, 53– 81. doi: 10.1146/annurev.psych.53.100901.135131
Nairne, J. S., Neath, I., & Serra, M. (1997). Proactive interference plays a role in the word-length effect. Psychonomic Bulletin & Review, 4, 541–545.
Neath, I. (1998). Human memory: An introduction to research, data, and theory. Pacific Grove, CA: Brooks/Cole.
Neath, I., Bireta, T. J., & Surprenant, A. M. (2003). The time-based word length effect and stimulus set specificity. Psychonomic Bulletin & Re- view, 10, 430 – 434.
Neath, I., & Brown, G. D. A. (2006). SIMPLE: Further applications of a local distinctiveness model of memory. In B. H. Ross (Ed.), The psy- chology of learning and motivation (pp. 201–243). San Diego, CA: Academic Press.
Neath, I., & Nairne, J. S. (1995). Word-Length effects in immediate memory: Overwriting trace decay theory. Psychonomic Bulletin & Re- view, 2, 429 – 441.
Osgood, C. E. (1953). Method and theory in experimental psychology. New York, NY: Oxford University Press.
Page, M. P. A., & Norris, D. (1998). The primacy model: A new model of immediate serial recall. Psychological Review, 105, 761–781. doi: 10.1037/0033-295X.105.4.761-781
Romani, C., McAlpine, S., Olsen, A., Tsouknida, E., & Martin, R. C. (2005). Length, lexicality, and articulatory suppression in immediate recall: Evidence against the articulatory loop. Journal of Memory and Language, 52, 398 – 415. doi:10.1016/j.jml.2005.01.005
Roodenrys, S. (2009). Explaining phonological neighbourhood effects in short-term memory. In A. Thorn & M. Page (Eds.), Interactions between short- and long-term memory in the verbal domain (pp. 177–197). London, England: Psychology Press.
Roodenrys, S., & Hinton, M. (2002). Sublexical or lexical effects on serial recall of nonwords? Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, 29 –33. doi:10.1037/0278-7393.28.1.29
Roodenrys, S., Hulme, C., Lethbridge, A., Hinton, M., & Nimmo, L. M. (2002). Word-frequency and phonological-neighborhood effects on verbal short-term memory. Journal of Experimental Psychology: Learning, Mem- ory, and Cognition, 28, 1019 –1034. doi:10.1037/0278-7393.28.6.1019
Roodenrys, S., & Miller, L. M. (2008). A constrained Rasch model of trace redintegration in serial recall. Memory & Cognition, 36, 578 –587. doi:10.3758/MC.36.3.578
Russo, R., & Grammatopoulou, N. (2003). Word length and articulatory suppression affect short-term and long-term recall tasks. Memory & Cognition, 31, 728 –737.
Service, E. (1998). The effect of word length on immediate serial recall depends on phonological complexity, not articulatory duration. The Quarterly Journal of Experimental Psychology, 51A, 283–304. doi: 10.1080/027249898391639
Snowling, M., Chiat, S., & Hulme, C. (1991). Words, nonwords, and phonological processes: Some comments on Gathercole, Willis, Emslie, and Baddeley. Applied Psycholinguistics, 12, 369 –373. doi:10.1017/ S0142716400009279
Surprenant, A. M., Bireta, T. J., Brown, M. A., Jalbert, A., Tehan, G., & Neath, I. (in press). Backward recall and the word length effect. Amer- ican Journal of Psychology.
Surprenant, A. M., LeCompte, D. C., & Neath, I. (2000). Manipulations of irrelevant information: Suffix effects with articulatory suppression and irrelevant speech. The Quarterly Journal of Experimental Psychology, 53A, 325–348. doi:10.1080/027249800390510
Surprenant, A. M., Neath, I., & Brown, G. D. A. (2006). Modeling age-related differences in immediate memory using SIMPLE. Journal of Memory and Language, 55, 572–586. doi:10.1016/j.jml.2006.08.001
Surprenant, A. M., Neath, I., & LeCompte, D. C. (1999). Irrelevant speech,
350 JALBERT, NEATH, BIRETA, AND SURPRENANT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
phonological similarity, and presentation modality. Memory, 7, 405– 420. doi:10.1080/741944920
Tehan, G., & Turcotte, J. (2002). Word length effects are not due to proactive interference. Memory, 10, 139 –149. doi:10.1080/ 09658210143000272
Vitevitch, M. S. (2002). The influence of phonological similarity neigh- borhood on speech production. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, 735–747. doi:10.1037/0278- 7393.28.4.735
Vitevitch, M. S., & Sommers, M. S. (2003). The facilitative influence of phonological similarity and neighborhood frequency in speech produc- tion in younger and older adults. Memory & Cognition, 31, 491–504.
Watkins, M. J. (1972). Locus of the modality effect in free recall. Journal
of Verbal Learning and Verbal Behavior, 11, 644 – 648. doi:10.1016/ S0022-5371(72)80048-3
Watkins, M. J., LeCompte, D. C., & Kim, K. (2000). Role of study strategy in recall of mixed lists of common and rare words. Journal of Experi- mental Psychology: Learning, Memory, and Cognition, 26, 239 –245. doi:10.1037/0278-7393.26.1.239
Woodward, A. J., Macken, W. J., & Jones, D. M. (2008). Linguistic familiarity in short-term memory: A role for (co-)articulatory fluency? Journal of Memory and Language, 58, 48 – 65. doi:10.1016/ j.jml.2007.07.002
Yates, M., Locker, L., & Simpson, G. B. (2004). The influence of phono- logical neighborhood on visual word perception. Psychonomic Bulletin & Review, 11, 452– 457.
Appendix A
The Short and Long Words Used in Experiments 1A and 1B
Word Conc. Fam. Imag. KF freq. TL freq. C freq. Orth. Orth. freq.
aisle 509 503 528 6 72 7.6 1 0.2 beam 502 476 539 21 127 9.2 9 23.7 draw 442 542 435 56 428 58.7 6 14.4 flood 553 523 598 19 325 15.6 2 158.3 howl 434 447 536 4 72 2.6 5 6.8 joke 388 580 483 22 230 34.6 6 6.2 lice 543 397 532 2 4 1.9 13 251.5 mink 589 524 604 5 27 3.3 16 42.2 pain 426 569 502 88 541 77.7 11 47.6 peal 402 451 433 1 13 1.1 15 39.5 pint 483 536 487 13 92 10.1 12 8.4 rose 608 556 623 86 801 80.1 16 30.5 sale 364 555 422 44 403 33.7 20 55.6 threat 335 524 408 42 108 64.3 2 28.2 wrath 304 466 377 9 51 7.5 1 0.1
M 458.8 509.9 500.5 27.9 219.6 27.2 9.0 47.5 SD 92.1 52.2 75.1 29.2 233.8 29.1 6.2 68.5
abundant 351 524 443 9 50 9.8 0 0.0 accident 419 564 518 33 399 50.3 1 0.1 approval 267 526 375 51 108 29.7 0 0.0 article 479 533 421 68 550 41.0 0 0.0 avenue 539 529 564 46 320 24.5 1 1.7 foreigner 492 499 516 4 92 7.4 0 0.0 hexagon 559 387 527 1 4 0.8 0 0.0 musician 564 558 585 23 72 5.3 0 0.0 occasion 346 566 305 58 424 64.8 0 0.0 paragraph 493 559 482 12 72 10.0 0 0.0 recital 476 468 495 8 27 3.5 0 0.0 sedative 459 423 459 1 13 1.3 0 0.0 sympathy 278 501 402 36 228 31.8 0 0.0 telegraph 547 460 518 21 126 3.0 0 0.0 telephone 619 605 655 76 800 102.9 1 0.1
M 459.2 513.5 484.3 29.8 219.0 25.7 0.2 0.1 SD 106.7 58.6 87.9 25.1 233.9 29.1 0.4 0.4
Note. The first four measures are from the Medical Research Council Psycholinguistics database (http:// www.psy.uwa.edu.au/mrcdatabase/uwa_mrc.htm), and the remaining measures are from the MCWord database of Medler and Binder (2005; http://www.neuro.mcw.edu/mcword/). Conc. � concreteness; Fam. � familiarity; Imag. � imageability; KF freq. � Kucera-Francis frequency; TL freq. � Thorndike-Lorge frequency; C. freq. � CELEX frequency; Orth. � number of orthographic neighbors; Orth. freq. � CELEX frequency of orthographic neighbors.
(Appendices continue)
351LENGTH AND THE WORD LENGTH EFFECT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
Appendix B
The Short and Long Words Used in Experiments 3 and 5
Word Conc. Fam. Imag. KF freq. TL freq. C. freq. Orth. Orth. freq.
disc 553 466 575 6 5 8.0 1 23.08 grief 303 505 480 10 137 16.2 1 47.00 guess 247 585 330 56 933 60.3 1 25.46 numb 379 487 477 4 55 4.2 1 11.72 phase 360 516 319 72 91 31.2 1 9.16 rogue 424 378 478 1 11 2.6 1 3.45 shriek 481 458 515 5 101 3.9 1 6.19 sponge 597 538 577 7 51 7.5 1 1.43 square 516 576 610 143 573 92.0 1 3.69 squeak 461 506 492 1 22 2.9 1 1.84 teeth 618 593 611 103 405 82.1 1 10.53 throng 400 377 452 3 60 3.6 1 9.76 wheat 594 510 577 9 158 30.3 1 4.22
M 456.4 499.6 499.5 32.3 200.2 26.5 1.0 12.1 SD 117.3 68.7 94.7 46.7 276.0 31.7 0.0 12.9
assemble 394 482 413 9 98 5.5 1 26.95 avenue 539 529 564 46 320 24.5 1 1.67 depression 303 541 453 24 244 24.9 1 11.36 destroyer 513 448 508 2 14 3.9 1 45.81 emission 397 446 416 32 1 1.5 1 3.39 fisherman 567 471 610 5 70 3.6 1 6.19 gentleman 516 537 559 28 580 24.4 1 29.09 insolent 311 388 357 2 25 2.3 1 1.43 minister 563 500 584 61 228 101.0 1 13.68 officer 550 549 593 101 585 79.3 1 33.43 photograph 590 551 618 18 342 28.7 1 1.49 primary 326 497 367 96 58 40.9 1 1.96 socialist 443 480 352 21 17 56.7 1 33.26
M 462.5 493.8 491.8 34.2 198.6 30.6 1.0 16.1 SD 105.1 48.4 102.2 33.3 207.3 31.5 0.0 15.5
Note. Conc. � concreteness; Fam. � familiarity; Imag. � imageability; KF freq. � Kucera-Francis frequency; TL freq. � Thorndike-Lorge frequency; C. freq. � CELEX frequency; Orth. � number of orthographic neighbors; Orth. freq. � CELEX frequency of orthographic neighbors.
(Appendices continue)
352 JALBERT, NEATH, BIRETA, AND SURPRENANT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
Appendix C
The Short and Long Words Used in Experiment 4
Word Conc. Fam. Imag. KF freq. TL freq. C. freq. Orth. Orth. freq.
birch 620 518 561 2 34 2.5 2 38.4 broad 399 523 463 84 282 42.3 2 40.3 cloud 554 553 595 28 367 32.5 2 10.5 flask 595 401 614 5 16 4.3 2 14.1 gloom 399 475 429 14 74 11.3 2 5.5 itch 488 526 486 5 20 2.5 2 12.6 myth 334 514 359 35 22 19.9 2 2.0 pledge 360 442 408 3 70 5.5 2 0.6 prune 611 444 578 1 104 2.0 2 3.8 split 417 514 445 30 119 38.7 2 0.3 swarm 406 463 488 3 76 3.3 2 0.4 trend 328 503 373 46 75 22.7 2 3.2 tweed 570 429 540 5 76 5.1 2 0.2 vault 550 445 550 2 35 3.6 2 19.6
M 473.6 482.1 492.1 18.8 97.9 14.0 2.0 10.8 SD 107.2 45.3 82.9 23.9 102.4 14.5 0.0 13.5
altitude 373 420 472 4 53 4.4 2 41.8 charity 373 518 445 8 158 14.3 2 5.6 convention 488 466 502 28 251 16.1 2 2.4 deduction 327 492 316 12 20 5.9 2 12.1 invader 485 402 419 1 15 1.0 2 3.6 lecturer 561 574 551 6 24 7.3 2 9.1 observer 505 469 489 16 82 12.7 2 20.5 opening 455 542 462 83 124 62.6 2 0.0 procession 500 462 534 5 89 12.8 2 11.9 radio 615 644 613 120 393 73.6 2 6.8 retailer 521 429 445 1 27 1.0 2 0.6 scavenger 486 474 501 1 10 0.6 2 0.2 treasurer 557 511 493 14 34 4.5 2 4.1 vocation 349 458 404 3 19 2.7 2 14.6
M 471.1 490.1 474.7 21.6 92.8 15.7 2.0 9.5 SD 86.2 64.5 71.1 35.4 110.3 22.9 0.0 11.1
Note. The first four measures are from the Medical Research Council Psycholinguistics database (http:// www.psy.uwa.edu.au/mrcdatabase/uwa_mrc.htm), and the remaining measures are from the MCWord database of Medler and Binder (2005; http://www.neuro.mcw.edu/mcword/). Conc. � concreteness; Fam. � familiarity; Imag. � imageability; KF freq. � Kucera-Francis frequency; TL freq. � Thorndike-Lorge frequency; C. freq. � CELEX frequency; Orth. � number of orthographic neighbors; Orth. freq. � CELEX frequency of orthographic neighbors.
Received June 15, 2009 Revision received August 18, 2010
Accepted August 27, 2010 �
353LENGTH AND THE WORD LENGTH EFFECT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly
GET SOLUTION FOR THIS ASSIGNMENT, Get Impressive Scores in Your Class
NO PLAGIARISM (100% Sure)
CLICK HERE TO MAKE YOUR ORDER When does Length Cause the Word Length Effect?
TO BE RE-WRITTEN FROM THE SCRATCH