Backward Recall and the Word Length Effect
AIMÉE M. SURPRENANT, MARK A. BROWN, ANNIE JALBERT, and IAN NEATH
Memorial University of Newfoundland
TAMRA J. BIRETA
The College of New Jersey
GERALD TEHAN
The University of Southern Queensland
The word length effect, the finding that words that have fewer syllables are recalled better
than otherwise comparable words that have more syllables, is one of the benchmark effects
that must be accounted for in any model of serial recall, and simulation models of immediate
memory rely heavily on the finding. However, previous research has shown that the effect disappears when participants are asked to recall the items in strict backward order. The present 2 experiments replicate and extend that finding by manipulating the participant’s foreknowledge of
recall direction (Experiment 1) and by giving the participant repeated practice with one direction
by blocking recall direction (Experiment 2). In both experiments, a word length effect obtained
with forward but not backward recall. The results are problematic for all models that currently
have an a priori explanation for word length effects. The finding can be accounted for but is not
predicted by Scale-Independent Memory, Perception, and Learning (SIMPLE), a model in which
item and order information are differentially attended to in the 2 recall directions.
Recently, we reported four experiments showing that
the four benchmark effects of working memory—the
word length effect, the irrelevant speech effect, the
acoustic confusion effect, and the concurrent articulation effect—were observed in forward recall but
XFSFHSFBUMZBUUFOVBUFEDPODVSSFOUBSUJDVMBUJPO
PS
absent (word length, irrelevant speech, and acoustic
DPOGVTJPO
XJUICBDLXBSESFDBMM#JSFUBFUBMöôõô
In those experiments, participants did not know
whether they would be asked to recall in forward or
backward order until after the list had been presented.
The current studies replicate and extend that finding by manipulating the participant’s foreknowledge
PGSFDBMMEJSFDUJPO&YQFSJNFOUõ
BOECZHJWJOHUIF
participant repeated practice with one direction by
CMPDLJOHSFDBMMEJSFDUJPO&YQFSJNFOUö
Surprisingly, there are few studies examining how
people recall information in reverse order and even
fewer models of memory that try to account for backXBSESFDBMM0OFSFBTPOXFDPOTJEFSUIJTTVSQSJTJOH
is the prominent role that backward span measures
play in certain areas of research. For example, investigators interested in working memory capacity or
individual differences often assess both forward and
backward span, and the tasks are often found to differ substantially. For example, unlike forward span,
backward span commonly correlates with measures
quite removed from simple serial order tasks such as
American Journal of Psychology
4QSJOHöôõõ7PMõöø/PõQQûùmüúr¥öôõõCZUIF#PBSEPG5SVTUFFTPGUIF6OJWFSTJUZPG*MMJOPJT
AJP 124_1 text.indd 75
1/27/11 1:16:34 PM
memory updating and verbal and spatial coordinaUJPOUBTLTFH0CFSBVFS4ۓ4DIVM[F8JMIFMN
8JUUNBOOöôôô
In the typical experiment, then, backward recall is
used mainly as part of a working memory construct,
and its characteristics and relationship to forward recall are rarely examined. Indeed, although almost all
computational models of short-term memory address
GPSXBSESFDBMM-FXBOEPXTLZBOE'BSSFMMöôôü
DPOcluded that only one such model, the primacy model
1BHF/PSSJTõýýü
PďFSTBEFUBJMFEBOEJOUFHSBUFE
account of backward recall.ı As described in more detail later, the primacy model predicts that in general,
effects observed with forward serial recall should also
be observable with backward serial recall. Similarly, a
TJNQMFFYUFOTJPOPGUIFGFBUVSFNPEFM/BJSOFõýýô
/FBUI/BJSOFõýýù/FBUIöôôô
SFTVMUTJOUIFTBNF
prediction: no appreciable difference between forward
and backward recall. Are these predictions correct?
5IFJOJUJBMTFUPGTUVEJFTCZ#JSFUBFUBMöôõô
DPOcluded that they were not. Here, we follow up on those
previous results, focusing on the word length effect.
The word length effect is the finding that lists of
short words (e.g., cat, boat, pear
BSFSFDBMMFECFUUFS
than lists of long words (e.g., gorilla, bicycle, strawberry
5IJTTFFNJOHMZVODPNQMJDBUFESFTVMUIBT
played such a significant role in the development of
theories of memory that it is now regarded as a finding that current theories of short-term or working
NFNPSZNVTUBEESFTTDG-FXBOEPXTLZ'BSSFMM
öôôü
/PUPOMZBSFUIFSFBMBSHFOVNCFSPGDPNQVtational models that try to account for the effect (e.g.,
#JSFUBFUBMöôõô#SPXO)VMNFõýýù#VSHFTT
)JUDIõýýý/FBUI/BJSOFõýýù1BHF/PSSJT
õýýü
CVUUIFFďFDUJUTFMGIBTCFFODBMMFEUIFiCFTU
remaining solid evidence” for the existence of decay
JOTIPSUUFSNNFNPSZ$PXBOõýýùQøö
The standard paradigm used to investigate the
word length effect is forward immediate serial recall, but the effect is also observable with a variety
of other memory tests including reconstruction of
PSEFS/BJSOF/FBUI4FSSBõýýû
TFSJBMSFDPHOJUJPO#BEEFMFZ$IJODPUUB4UBďPSE5VSLöôôö
GSFFSFDBMM8BULJOTõýûö
TJOHMFJUFNQSPCFSFDBMM
"WPOT8SJHIU1BNNFSõýýø
BOEDPNQMFYTQBO
5FIBO)FOESZ,PDJOTLJöôôõ
Researchers distinguish between two word
length effects. In order to demonstrate a time-based
word length effect, one uses two sets of items that
differ solely in terms of pronunciation time. For example, Baddeley, Thomson, and Buchanan (1975,
&YQFSJNFOUø
DPOTUSVDUFEUXPTFUTPGXPSETUIBU
were equated for word frequency, the number of syllables, and the number of phonemes (given Scottish
QSPOVODJBUJPO
CVUEJďFSFEJOQSPOVODJBUJPOUJNF
The short words were bishop, pectin, pewter, phallic,
and wicket, and the long words were coerce, cyclone,
Friday, harpoon, and zygote.PSFTIPSUXPSETXFSF
recalled than long words, a finding consistent with the
phonological loop hypothesis: Items in the phonological store decay unless revived by covert articulation. Because short words take less time to rehearse,
more short words can be kept from decaying away
completely relative to long items. This effect has been
SFQMJDBUFEOVNFSPVTUJNFTFH$PXBOFUBMõýýö
-POHPOJ3JDIBSETPO"JFMMPõýý÷-PWBUU"WPOT
.BTUFSTPOöôôô/BJSOFFUBMõýýû
CVUPOMZXJUI
the stimuli created by Baddeley et al. Five other sets
of words that vary only in pronunciation time have
failed to yield a time-based word length effect (Lovatt
FUBMöôôô/FBUI#JSFUB4VSQSFOBOUöôô÷
*U
has been suggested that the time-based word length
effect is really an artifact of some unknown property
PGUIFPSJHJOBMTUJNVMVTTFU+BMCFSU/FBUI#JSFUB
4VSQSFOBOUJOQSFTT
BOETPXFGPDVTPOUIFTFDPOE
type of word length effect.
In contrast to the time-based word length effect,
the syllable-based word length effect is robust and
has been observed with many different stimulus sets.
With this effect, the short and long items differ in
both pronunciation time and the number of syllables
and phonemes. In many such studies, the experimenters use spoken or written serial recall. This can
add a confound in that it takes longer to say or write a
three-syllable word (e.g., bicycle
UIBOBPOFTZMMBCMF
word (e.g., car
#FDBVTFJUJTXFMMLOPXOUIBUEJGferent output times can affect recall independently
PGUIFUZQFPGJUFN%PTIFS.Bõýýü
SFTFBSDIers have developed different ways of removing this
confound. Some continue to use written recall but
ask the participants to write down only the first few
MFUUFSTPGFBDIXPSE#BEEFMFZFUBMöôôö
"MUIPVHI
it presumably takes the same amount of time to write
down the three letters bic and car, there may be differences between writing part of a word and stopping
and writing a complete word, even if the two written
t SURPRENANT ET AL .
AJP 124_1 text.indd 76
1/27/11 1:16:34 PM
forms are of the same length. An alternative strategy
is to use a strict serial reconstruction of order task.
With this task, all list items are presented at test,
either in a new order or in alphabetical order, and
the participant’s task is to click on the words to recSFBUFUIFQSFTFOUBUJPOPSEFS0OFDBOUIFONFBTVSF
output time and empirically determine whether the
potential confound is present. It should be noted
that this type of test produces the same benchmark
working memory effects observable with strict writUFOTFSJBMSFDBMMUFTUUIBUJTPOFPCUBJOTXPSEMFOHUI
FďFDUT/BJSOFFUBMõýýû
BDPVTUJDDPOGVTJPOFďFDUT
4VSQSFOBOU/FBUI-F$PNQUFõýýý
JSSFMFWBOU
TQFFDIFďFDUT/FBUI'BSMFZ4VSQSFOBOUöôô÷
and concurrent articulation effects (Surprenant,
-F$PNQUF/FBUIöôôô
0OMZBIBOEGVMPGTUVEJFTIBWFJOWFTUJHBUFEXPSE
length effects with backward recall, but until recently
none compared forward and backward recall of short
and long words within a single experiment (see Bireta
FUBMöôõôGPSBNPSFUIPSPVHISFWJFX
'PSFYBNQMF
$PXBO8PPEBOE#PSOFõýýø
PCTFSWFEBOFďFDU
of word length with spoken backward recall, but because there was no forward condition, it is difficult to
assess the magnitude of the effect. Walker and Hulme
õýýý
BMTPVTFETQPLFOCBDLXBSESFDBMMPGTIPSUBOE
long words, but they also included a manipulation
of concreteness. Although they found better recall
of short abstract words than long abstract words,
there was little evidence of a word length effect for
DPODSFUFXPSETXJUICBDLXBSESFDBMM.PSFPWFSOP
group performed forward recall with these stimuli.
'JOBMMZ5FIBOBOE.JMMTöôôû
SFQPSUFEOPXPSE
length effect with backward recall of four-item lists,
but they also had no forward control group. Given
this small literature, it is difficult to assess whether
word length effects are observable in backward recall
and, if they are, whether they are of the same magnitude as in forward recall.
Although the evidence is at best equivocal, the
predictions of two computational models are crystal
clear: Two computational models predict that a word
length effect will be observed with backward recall.2
5IFQSJNBDZNPEFM1BHF/PSSJTõýýüöôô÷
JTBO
instantiation of the phonological loop hypothesis and
explains the word length effect by invoking the tradeoff between decay and rehearsal, despite the problems
noted earlier concerning the lack of a pure time-based
effect. The model assumes that the strength of the
activation of successive list items decreases over list
positions, with the first-presented item having the
most activation, followed by the second-presented
item, and so on. This results in a primacy gradient
in which activation is the mechanism retaining order
information. The most active item is output first at
SFDBMMBOEUIFOJTTVQQSFTTFEUIFOUIFOFYUNPTU
active item is output, and so on. This produces the
basic serial position function. Rehearsal offsets the
decay of activation, and because long words take longer to rehearse than do short words, their memory
representations have lower levels of activation and
thus worse recall.
The primacy model implements serial recall by
assuming that people perform a series of forward
recalls. The participant recalls each item in the list,
beginning with the first item, until the most active
item is reached, in which case it is output as the response. Then, the participant again recalls the list
from the first item but now stops one item earlier and
outputs that item. The repeated forward recalls act
like rehearsal within the model, and so the primacy
gradient is maintained at a high level. This allows the
model to predict high levels of recall performance
in backward recall conditions (see, for example, the
TJNVMBUJPOSFQPSUFECZ1BHF/PSSJTõýýüQûûö
Because backward recall is basically multiple forward recalls with activation levels maintained, the
model predicts that word length effects should be
observed with backward recall and that the magnitude of the effect should be comparable to that seen
with forward recall.
In contrast to the primacy model, the feature
NPEFM/FBUI/BJSOFõýýù
EPFTOPUJOWPLFUIF
concepts of decay or rehearsal. Rather, items are represented as vectors of features, and memory depends
on successfully matching a degraded cue of an item
in primary memory with its intact representation in
secondary memory. Forgetting is due mainly to interference. Word length effects are seen as item-level
effects: Long words have more segments than short
words. If one assumes a fixed probability of an error
in assembling segments, then overall, longer words
will be recalled worse than short words. Although
published versions of the feature model do not include an account of backward recall, a straightforward
extension is possible. In the model, forward serial
WORD LENGTH AND RECALL DIRECTION t AJP 124_1 text.indd 77
1/27/11 1:16:34 PM
recall proceeds by using the cue for the first item and
determining the best match. There is no reason why
recall could not begin with the cue for the last item,
however. We made this change and ran the model to
determine its predictions for the word length effect
with both forward and backward recall. As predicted,
the model produces a word length effect with backward recall that is comparable to the one observed
with forward recall. The reason is that because most
of the computations that drive the word length effect
within the model occur before retrieval, the effect is
established before any part of the retrieval process.
Reversing the order of retrieval, then, has essentially
no influence on the word length effect.
#JSFUBFUBMöôõô
DPNQBSFEGPSXBSEBOECBDLXBSESFDBMMPGTIPSUPOFTZMMBCMF
BOEMPOHUISFF
TZMMBCMF
JUFNTVTJOHBTUSJDUTFSJBMSFDPOTUSVDUJPOPG
order task. The participants did not know the recall
direction until after list presentation. This was done
to make sure that whatever processing strategies were
adopted, they would necessarily be equivalent up until the time of recall. Thus, on each trial, the list was
presented and then the participant was told whether
to recall in strict forward order or strict backward order. Contrary to the predictions of both the primacy
model and the feature model, there was a robust word
length effect with forward recall but no effect with
backward recall.
0OFMJNJUBUJPOPGUIFSFTVMUTPG#JSFUBFUBMöôõô
is that they did not examine forward and backward
recall when people knew the recall direction ahead
of time. It is possible that in the absence of information about recall direction, people act as if they
are expecting forward recall. This could be because
people regularly use forward serial recall in everyday
activities (e.g., recalling a phone number, processing
TQFFDI
CVUBMNPTUOFWFSVTFCBDLXBSETFSJBMSFDBMM
In this case, one would expect disruption in the backward recall condition, possibly enough to remove the
effect of word length. A second possibility is that the
design used by Bireta et al. may have prevented participants from developing appropriate processing and
retrieval strategies for backward recall. In their study,
recall direction was randomly determined on each
USJBMUIBUJT5SJBMn might be forward recall, Trial n+1
might be backward recall, and so on, preventing the
participants from becoming sufficiently comfortable
with backward recall. If participants were to receive a
block of trials all of which used the same recall direction, then it might be more conducive to the development of appropriate processing that would yield a
word length effect.
The two experiments reported here examine
UIFTFUXPJTTVFT*O&YQFSJNFOUõXFDPNQBSFEUXP
groups of participants, one of which knew ahead of
time the recall direction and the other of which did
not. This latter group also served as replication of
&YQFSJNFOUõPG#JSFUBFUBMöôõô
&YQFSJNFOUö
addressed the question of whether a word length effect is observed with backward recall if the lists are
presented in blocks, thus removing any uncertainty
on a trial-to-trial basis about the nature of the test.
EXPERIMENT 1
&YQFSJNFOUõEJSFDUMZUFTUFEXIFUIFSGPSFLOPXMFEHF
of the recall direction affects whether a word length
FďFDUJTPCTFSWFE0OFHSPVQPGQBSUJDJQBOUTXBT
UFTUFEJOBDPOEJUJPOEFTJHOFEUPSFQMJDBUF&YQFSJNFOUõPG#JSFUBFUBMöôõô
5IFZTBXTFWFOJUFN
MJTUTPGTIPSUPOFTZMMBCMF
PSMPOHUISFFTZMMBCMF
words, and after list presentation was over, they were
instructed whether to recall the words in forward or
backward order. The only change for the second
group of participants was that they were informed
of the recall direction before the presentation of the
to-be-remembered items. Both groups received a
strict serial reconstruction of order test, because this
permits output time to be equated.
METHOD
Participants
Fifty undergraduates from the College of New Jersey volunteered to participate in exchange for course
credit. All identified themselves as native speakers of
"NFSJDBO&OHMJTI
Stimuli
The to-be-remembered stimuli were 15 short (oneTZMMBCMF
BOEõùMPOHUISFFTZMMBCMF
XPSETTFFUIF
"QQFOEJY
FRVBUFE GPS GSFRVFODZ DPODSFUFOFTT
JNBHFBCJMJUZBOEGBNJMJBSJUZBDDPSEJOHUPUIF.3$
1TZDIPMJOHVJTUJD%BUBCBTF$PMUIFBSUõýüõ
'PSFBDI
trial, seven words were randomly sampled from the
BQQSPQSJBUFQPPMTIPSUPSMPOH
BOETIPXOJOSBOdom order.
t SURPRENANT ET AL .
AJP 124_1 text.indd 78
1/27/11 1:16:34 PM
to click on the buttons to recreate either the original
presentation order if they saw the word “Forward” or
The design was a 2 (advance knowledge vs. no adthe reverse order if they saw the word “Backward.”
vance knowledge of recall direction, manipulated
That is, with forward recall, they clicked on the first
CFUXFFO QBSUJDJQBOUT
r2 (short or long words,
word first, the second word second, and so on. With
NBOJQVMBUFEXJUIJOQBSUJDJQBOUT
r2 (forward or
backward recall, they clicked on the last word first,
CBDLXBSESFDBMMNBOJQVMBUFEXJUIJOQBSUJDJQBOUT
UIFQFOVMUJNBUFXPSETFDPOEBOETPPO0ODFBSFmixed design.
sponse had been made, it could not be changed. No
There were 40 experimental trials, 10 in each
feedback was given (i.e., a clicked button remained
condition. Assignment of trials to condition was
VODIBOHFEBGUFSUIFDMJDL
SBOEPNJ[FEGPSFBDIQBSUJDJQBOU&BDIQBSUJDJQBOU
was tested individually, and the experimenter remained in the room to ensure that the instructions RESULTS
were followed.
Design
Proportion Correct
Items were scored as correct only if they were placed
Participants were informed that the purpose of the in their correct serial position. The main results (Figstudy was to see how accurately they could remember VSFõ
JMMVTUSBUFUIBUBTJHOJêDBOUXPSEMFOHUIFďFDU
UIFPSEFSJOXIJDIUIFZTBXBMJTUPGJUFNT0OFBDI was observed for forward but not backward recall
trial, each word was displayed one at a time for 1 s in and that advance knowledge of the recall direction
black 28-point Helvetica against a white background had no discernible effect.
on a computer screen. In the advance knowledge con- " ö XPSE MFOHUI
rö SFDBMM EJSFDUJPO
r2
dition, the word “Forward” or “Backward” appeared LOPXMFEHFUZQF
BOBMZTJTPGWBSJBODFanova
XBT
at the top of the window and remained there for 2 conducted.
s, disappearing before the beginning of each list. In
There was a significant main effect of length,
the no advance knowledge condition, this step was
Fõøü
õüúùöMSEôôôýQBSUJBMH2öüô
omitted. For both conditions, the test was identical:
Seven buttons labeled with the seven words just seen p < .001, with better recall of short than long words
BQQFBSFEPOUIFTDSFFOUIFMBCFMTXFSFBSSBOHFEPO ùùùWTøýû
5IFNBJOFďFDUPGSFDBMMEJSFDUJPO
the buttons in alphabetical order. Simultaneously, was not significant, F < 1, with the proportion correct
the word “Forward” or “Backward” appeared above BQQSPYJNBUFMZUIFTBNFGPSGPSXBSEùöý
BOECBDLthe response buttons. The participants were asked XBSEùöø
SFDBMM5IFNBJOFďFDUPGLOPXMFEHFUZQF
Procedure
FIGURE 1. Proportion of short and long words recalled as a function of recall direction and whether recall direction was known before list
presentation. Error bars show the standard error of the mean
WORD LENGTH AND RECALL DIRECTION t AJP 124_1 text.indd 79
1/27/11 1:16:35 PM
failed to reach conventional levels of significance, F(1,
øü
÷õúôMSEôôúöQBSUJBMH2ôúöpôü
although numerically performance was slightly better when there was advance knowledge of the recall
direction than when there was not (.557 vs. .495, reTQFDUJWFMZ
No interaction involving type of knowledge was
significant. Indeed, for all such interactions, F < 1. Crucially, the interaction between length and recall direction was significant, Fõøü
õõúúøMSEôôôú
partial H2õýùp < .001, with a word length effect
BQQBSFOUXJUIGPSXBSESFDBMMùûúWTøüõ
CVUOPFGGFDUXJUICBDLXBSESFDBMMù÷ùWTùõ÷
The critical interaction between word length and
recall direction was further assessed using a Tukey
HSD test. When recall direction was known, short
words were recalled significantly more accurately than
MPOHXPSETXJUIGPSXBSESFDBMMúôýWTùõý
8JUI
backward recall, recall of short and long words did not
EJďFSùúöWTùøô
5IFQBUUFSOXBTJEFOUJDBMXIFO
recall direction was not known: There was a significant
FďFDUPGXPSEMFOHUIXJUIGPSXBSESFDBMMùø÷WTøøø
CVUOPUXJUICBDLXBSESFDBMMùôûWTøüù
Output Time
During recall, the computer recorded how long the
participant took to make each response. The purpose was to ensure that the reconstruction of order
test removed the confound of output time. A 2 (word
MFOHUI
röSFDBMMEJSFDUJPO
röLOPXMFEHFUZQF
mixed-design anova was performed on mean output
times for correct responses.
As can be seen in Figure 2, there was no main effect of word length, Fõøü
öú÷öMSEõôûüùô
partial H2ôùöpõô0VUQVUUJNFGPSTIPSUXPSET
was similar to output time for long words (1,711 ms
WTõûüûNT
5IFSFXBTBTJHOJêDBOUFďFDUPGSFDBMM
direction, Fõøü
ùýøõüMSE÷ûôøöüQBSUJBM
H2ùù÷p < .001, with longer output time for backXBSEUIBOGPSXBSESFDBMMöôøôNTWTõøùüNT
5IF
main effect of knowledge type was also significant,
Fõøü
úôõ÷ MSEùõøûûýQBSUJBM H2õõõ
p < .05, with shorter output times when recall direction was known in advance than when recall direcUJPOXBTOPULOPXOõúöùNTWTõüû÷NT
/POFPG
UIFJOUFSBDUJPOTXFSFTJHOJêDBOUUIFPOMZJOUFSBDUJPO
with an F > 1 was group by direction, Fõøü
õ÷ôõ
MSEöüøüôüQBSUJBMH2ôöúp > .25.
DISCUSSION
&YQFSJNFOUõSFQMJDBUFEUIFSFTVMUTPG#JSFUBFUBM
öôõô
CZTIPXJOHUIBUBXPSEMFOHUIFďFDUJTPCserved with forward but not backward recall when the
recall direction is not known before list presentation.
&YQFSJNFOUõBMTPGPVOEUIFTBNFQBUUFSOXIFOSFDBMM
direction was identified before list presentation. Advance knowledge of recall direction had a marginally
beneficial effect on accuracy and also significantly re-
FIGURE 2. Mean output time for items correctly recalled as a function of recall direction and whether recall direction was known before
list presentation. Error bars show the standard error of the mean
t SURPRENANT ET AL .
AJP 124_1 text.indd 80
1/27/11 1:16:35 PM
duced output time. However, this advance knowledge
of recall direction did not interact with word length.
Backward recall took longer than forward recall, but
there was no significant difference in output times
as a function of word length, replicating the finding
that word length effects can be observed when the
confound of differential output time is not present
DG#JSFUBFUBMöôõô+BMCFSUFUBMJOQSFTT
EXPERIMENT 2
identified themselves as native speakers of American
&OHMJTI
Stimuli
5IFTUJNVMJXFSFUIFTBNFBTJO&YQFSJNFOUõ
Design
The design was a 2 (forward recall first or backward
SFDBMMêSTUNBOJQVMBUFECFUXFFOQBSUJDJQBOUT
r2
(short or long words, manipulated within particiQBOUT
r2 (forward or backward recall, manipulated
XJUIJOQBSUJDJQBOUT
NJYFEEFTJHO
*O&YQFSJNFOUõSFDBMMEJSFDUJPOXBTEFUFSNJOFE
randomly on each trial. A second way to manipulate recall direction is to have the first half of the Procedure
5IFQSPDFEVSFXBTBMNPTUUIFTBNFBTJO&YQFSJNFOU
experiment consist of just one recall direction and
1. For the forward-first group, on the first 20 trials the
then change direction for the second half. It is posparticipant was asked to recall the items in forward
sible that with consistent, repeated practice with
order, and no mention was made of backward recall
backward recall, a word length effect may emerge. In
until the second half of the experiment, when new
&YQFSJNFOUöPOFHSPVQPGQBSUJDJQBOUTXFSFUPMEUP
instructions were provided asking the participant to
SFDBMMUIFJUFNTJOGPSXBSEPSEFSIBMGXBZUISPVHI
recall the items in backward order. For the backwardthey were told to recall the items in backward orfirst group, the reverse was true.
der. A second group of participants were given the
reverse instructions.
RESULTS
The main results, displayed in Figure 3, illustrate that
a significant word length effect was observed for forward but not backward recall and that this was unafParticipants
Fifty undergraduates from the College of New Jer- fected by which block was performed first.
sey volunteered to participate in exchange for course "öHSPVQ
röMJTUUZQF
röSFDBMMEJSFDUJPO
DSFEJU/POFIBEQBSUJDJQBUFEJO&YQFSJNFOUõBOEBMM mixed-design anova was conducted. The main efMETHOD
'*(63& Proportion of short and long words recalled as a function of recall direction and whether the order of blocks was forward recall
followed by backward recall (forward first) or the reverse (backward first). Error bars show the standard error of the mean
WORD LENGTH AND RECALL DIRECTION t AJP 124_1 text.indd 81
1/27/11 1:16:35 PM
fect of group was not significant, Fõøü
ôýýú
MSEôôøôQBSUJBMH2ôöôp > .30, with the proportion of words correctly recalled in order approximately the same in both groups, .538 for forward first
and .510 for backward first. There was a significant
main effect of length, Fõøü
õüûüûMSEôôôû
partial H2öüõp < .001, with better recall of short
UIBOMPOHXPSETùùôWTøýý
5IFNBJOFďFDUPG
recall direction was also significant, Fõøü
ùøøú
MSEôôõ÷QBSUJBMH2õôöp < .05, with more
XPSETCFJOHSFDBMMFEJOGPSXBSEùø÷
UIBOGPSCBDLXBSEùôù
SFDBMM
The direction by group interaction was significant, Fõøü
õøüõøMSEôôõ÷QBSUJBMH2ö÷ú
p < .01, due mainly to better performance on the second block for the backward-first group compared with
the forward-first group. The latter group improved
from .526 for forward recall to .551 for backward
recall, whereas the backward-first group improved
from .459 to .561. Neither the interaction between
length and group nor the three-way interaction was
TJHOJêDBOUGPSCPUIFõøü
õ
"TJO&YQFSJNFOUõUIFDSVDJBMJOUFSBDUJPOCFtween length and recall direction was significant, F(1,
øü
õõôöõMSEôôô÷QBSUJBMH2õüûp < .01,
with a word length effect apparent with forward recall
ùüúWTùôõ
CVUOPFďFDUXJUICBDLXBSESFDBMMùõø
WTøýú
The critical interaction between word length and
recall direction was further assessed using a Tukey
HSD test. In the forward-first group, short words
were recalled significantly more accurately than long
XPSETXJUIGPSXBSESFDBMMùùýWTøý÷
8JUICBDLward recall, recall of short and long words did not
EJďFSùùýWTùøö
5IFQBUUFSOXBTJEFOUJDBMGPSUIF
backward-first group: There was a significant effect
PGXPSEMFOHUIXJUIGPSXBSESFDBMMúõ÷WTùôý
CVU
OPUXJUICBDLXBSESFDBMMøûôWTøùô
Output Times
As can be seen in Figure 4, backward recall took longer than forward recall, but the time taken to output
short items was equivalent to the time needed to outQVUMPOHJUFNT"öHSPVQ
röMJTUUZQF
r2 (recall
EJSFDUJPO
NJYFEEFTJHOanova was performed on the
output time data.
There was no significant difference in output times between the two groups, Fõøü
õJO
the forward-first and backward-first groups (1,547
NTWTõùü÷NT
5IFSFXBTBMTPOPNBJOFďFDUPG
length, Fõøü
õXJUINFBOPVUQVUUJNFTFRVJWBMFOUGPSTIPSUõùøúNT
BOEMPOHJUFNTõùüøNT
There was a significant main effect of recall direction,
Fõøü
øøùûúMSEöûöôùõQBSUJBMH2øüö
p < .001, with longer times to recall backward comQBSFEUPGPSXBSEõüõõNTWTõ÷õýNT
The only significant interaction was between direction and group, Fõøü
õõúýøMSEöûöôùõ
FIGURE 4. Mean output time for items correctly recalled as a function of recall direction and whether the order of blocks was forward
recall followed by backward recall (forward first) or the reverse (backward first). Error bars show the standard error of the mean
t SURPRENANT ET AL .
AJP 124_1 text.indd 82
1/27/11 1:16:35 PM
partial H2õýúp < .001. This was due to a larger
difference between forward and backward recall for
UIFCBDLXBSEêSTUHSPVQõöõôNTWTõýùùNT
UIBO
GPSUIFGPSXBSEêSTUHSPVQõøöûNTWTõúúüNT
For all the remaining interactions, F < 1.
Explaining Backward Recall
Both the primacy model and the feature model predict that word length effects will be observable with
backward recall. The results observed here provide
an additional challenge to those accounts. No other
current models address all four benchmark working
memory phenomena and backward recall, so there
DISCUSSION
is currently no general explanation that can account
In the blocked design, performance was better in
for the data from backward recall (see also LewanBlock 2 than Block 1 for the group that had backward
EPXTLZ'BSSFMMöôôüGPSBTJNJMBSDPODMVTJPO
recall first, an effect we attribute to practice and greatBCPVUCBDLXBSESFDBMM
#JSFUBFUBMöôõô
EFTDSJCFE
er familiarity with the procedure and to the change
two possible accounts, both of which focused in genfrom a very difficult condition to an easier one. Howeral terms on how backward recall might differ from
ever, significant word length effects emerged only
forward recall and both of which remain viable given
with forward recall and not with backward recall.
the current results.
"TJO&YQFSJNFOUõXFXFSFTVDDFTTGVMJOSFNPWJOH
In their discussion of the theory of distributed
the confound of output time, and we still observed
BTTPDJBUJWFNFNPSZ-JBOE-FXBOEPXTLZõýý÷
a word length effect with forward recall but not with
noted that in addition to other factors, forward and
backward recall.
backward recall differ in terms of output interference.
With forward recall, every item is followed by the
GENERAL DISCUSSION
same number of intervening events. For example, if
In two experiments, participants recalled significantly the to-be-remembered sequence is A B C D E F, it
more short words than long words with forward re- must be recalled as A B C D E F. Therefore, A is folcall, but there was no effect of word length with back- lowed by five events before recall, B is also followed
ward recall. This pattern was obtained regardless of by five events (presentation of C D E F and recall of
whether the participants knew the recall direction A
BOETPPO*ODPOUSBTUXJUICBDLXBSESFDBMMJUFNT
before list presentation and regardless of whether re- differ greatly in the number of intervening events.
call direction was blocked or varied from trial to trial. With the same sequence, F is followed by 0 events,
These results replicate and extend those reported whereas A is followed by 10 events. Li and LewanCZ#JSFUBFUBMöôõô
5IFEJďFSFODFCFUXFFOSFDBMM dowsky further noted that when this observation is
of short and long words is larger with forward recall incorporated into the theory of distributed associathan with backward recall. In our two studies, the UJWFNFNPSZ-FXBOEPXTLZ.VSEPDLõýüý
UIF
latter difference was so small that recall of short and model predicts, in general, that variables that affect
forward recall should have little or no effect on backlong items did not differ statistically.
In addition, analysis of the output time data re- ward recall (for further details and simulations, see
vealed that participants took the same amount of time -J-FXBOEPXTLZõýý÷
5IJTQSFEJDUJPODPVMECF
to recall short and long items, confirming that the tested by devising a test that requires forward and
reconstruction of order test successfully removed a backward seriation but somehow equates for output
confound often present in the word length effect lit- interference. Given the results of the current two exerature. This finding replicates previous work (Bireta periments, this explanation remains viable.
A second possible account is based on the idea of a
FUBMöôõô+BMCFSUFUBMJOQSFTT
BOEGVSUIFSEFNonstrates that situations in which different outcomes trade-off between item and order information (Hendry
occur are not a necessary condition for observing a 5FIBOöôôù5FIBO5PMBOöôôû
8JUIGPSXBSE
word length effect. We suspect that testing proce- recall, a well-practiced task, less attention is devoted to
dures in which it takes longer to recall long words retaining the order of the words and more is focused on
UIBOTIPSUXPSETFHTQPLFOSFDBMMXSJUUFOSFDBMM
item differences. With backward recall, however, more
will find larger effects of word length than those re- attention is needed for the order task, and so some
attention is withdrawn from the item differences.
ported here.
WORD LENGTH AND RECALL DIRECTION t AJP 124_1 text.indd 83
1/27/11 1:16:35 PM
#JSFUBFUBMöôõô
TIPXFEIPXUIJTJEFBDBOCF
made more specific by instantiating it within Scale*OEFQFOEFOU.FNPSZ1FSDFQUJPOBOE-FBSOJOH
4*.1-&
#SPXO/FBUI$IBUFSöôôû/FBUI
#SPXOöôôú
*OUIJTNPEFMUXPEJNFOTJPOTBSF
used to represent the information, one corresponding
to relative time and one representing item differences.
Within the model, relative time is used to determine
the order in which the items were presented. WhenFWFS4*.1-&VTFTUXPPSNPSF
EJNFOTJPOTFBDI
dimension is weighted to determine its relative influence, with the constraint that the sum of the weights
NVTUCFõôTFF4VSQSFOBOU/FBUI#SPXOöôôú
5IFSFGPSFBOBQSJPSJQSFEJDUJPOPG4*.1-&JTUIBUJG
more attention is paid to one dimension, such as the
order dimension, less attention is paid to the second,
such as item information.
'PMMPXJOH)FOESZBOE5FIBOöôôùTFFBMTP5FIBO5PMBOöôôû
#JSFUBFUBMöôõô
GPDVTFEPO
a trade-off between item and order information as a
possible difference between forward and backward
recall. The logic was that forward recall is a highly
practiced task, and this enables people to devote more
attention to processing the items than to having to attend to the particular order. In contrast, backward
recall is a more effortful task, because it is not often
QSBDUJDFEJOPSEFSUPQFSGPSNUIJTUBTLNPSFBUUFOtion has to be paid to information conveying order
information, and so there is less attention available to
focus on differences between the items.
#JSFUBFUBMöôõô
JODMVEFEBTJNVMBUJPOJOXIJDI
4*.1-&êUUIFGPSXBSEBOECBDLXBSESFDBMMEBUBXFMM
producing a word length effect for forward but not
backward recall (R2 was .907 for forward recall and
ýøýGPSCBDLXBSESFDBMM
*OBEEJUJPOCBDLXBSESFDBMM
levels were roughly equivalent to forward recall levels.
Crucially, the only free parameter that varied for fitting forward and backward recall was the dimensional
XFJHIUXJUINPSFXFJHIUPOUIFUFNQPSBMPSEFS
EJmension for backward recall relative to forward recall.
The key assumption, then, is that with backward recall
participants increase the degree to which they attend
to the temporal dimension at the time of retrieval.
Some evidence consistent with finding this comes
from recent studies examining immediate recall of
verbal materials in which the emphasis is on the important role that language production plays in such
BTDFOBSJPFH"DIFTPO.BD%POBMEöôôý1BHF
$VNNJOH.BEHF/PSSJTöôôû
*ONPEFMTPG
speech production, ordering mechanisms often act at
a different level of representation than lexical or grammatical levels. Thus, forward recall would rely heavily
on the ordering level, whereas backward recall would
be less reliant on that level. If this speculation is well
founded, then other effects that rely on the linguistic
ordering level of language production (see Acheson
.BD%POBMEöôôýGPSBEFTDSJQUJPO
TIPVMEBMTP
disappear with backward recall. In contrast, effects
that rely on long-term linguistic knowledge (e.g., conDSFUFOFTTJNBHFBCJMJUZ
BSFNPSFMJLFMZUPCFPCTFSWFE
in backward recall.
Summary
Forward recall is characterized by a robust influence
of word length, whereas backward recall is not. None
of the existing theoretical accounts predicted this result, and so far, no model seems able to accommodate
UIFêOEJOH"MUIPVHI4*.1-&JTBCMFUPBDDPVOUGPS
the results, it did not predict them, and the account
rests on an untested assumption. However, the findings reported here that foreknowledge of the recall
direction does not restore the word length effect with
CBDLXBSESFDBMMBSFDPOTJTUFOUXJUI4*.1-&TBDDPVOU.PSFCSPBEMZUIFSFTVMUTBSFDPOTJTUFOUXJUI
a small but growing literature that suggests that backward recall is very different from forward recall, and
accounts based on assuming either the same process
JFNVMUJQMFGPSXBSESFDBMMT
PSBSFWFSTBMPGGPSXBSE
SFDBMMBSFVOMJLFMZUPCFTVDDFTTGVM0OFDPOTFRVFODF
is that these data, along with those reported by Bireta
FUBMöôõô
BSFJODPOTJTUFOUXJUIUIFTVQQPTFEDPOtribution of the phonological loop within the working
memory model.
NOTES
Portions of this work were supported by grants from the NatVSBM4DJFODFTBOE&OHJOFFSJOH3FTFBSDI$PVODJMPG$BOBEB
UP"+*/BOE".44PNFPGUIJTXPSLXBTQSFTFOUFEBU
UIFùôUI"OOVBM.FFUJOHPGUIF1TZDIPOPNJD4PDJFUZ#PTton, November 2009.
Address correspondence about this article to Aimée
4VSQSFOBOU%FQBSUNFOUPG1TZDIPMPHZ.FNPSJBM6OJWFSTJUZ
PG/FXGPVOEMBOE4U+PIOT/-"õ#÷9ý$BOBEBFNBJM
BTVSQSFO!NVODB
1. It should be noted that whereas Lewandowsky and
'BSSFMMöôôü
DPODMVEFEUIBU0TDJMMBUPS#BTFE"TTPDJBUJWF
3FDBMM04$"3
QSPWJEFTBOBDDPVOUPGCBDLXBSESFDBMMUIF
t SURPRENANT ET AL .
AJP 124_1 text.indd 84
1/27/11 1:16:35 PM
EFWFMPQFSTPG04$"3TUBUFEUIBUiXFSFHBSECBDLXBSESFDBMM
BTCFJOHPVUTJEFUIFTDPQFPG04$"3u#SPXO1SFFDF
)VMNFöôôôQõûø
2. Two other models, the start–FOENPEFM)FOTPOõýýü
BOEUIF#VSHFTTBOE)JUDIõýýý
NPEFMQSPCBCMZNBLFUIF
same prediction: equivalent word length effects with forward
and backward recall. However, the start–end model has not
been fit to word length effect data, and the Burgess and Hitch
model does not include an account of backward recall.
3. The figures show data collapsed over serial position,
both for clarity and because there were no theoretically significant interactions involving serial position. Supplemental
data showing recall and output time as a function of serial position for both experiments are available from the last author
PSGSPNIUUQNFNPSZQTZDINVODBQVCTUôüTIUNM
REFERENCES
"DIFTPO%+.BD%POBME.$öôôý
7FSCBMXPSLJOH
memory and language production: Common approaches
to the serial ordering of verbal information. Psychological
Bulletin, 135, 50–68.
"WPOT4&8SJHIU,-1BNNFS,õýýø
5IFXPSE
length effect in probed and serial recall. Quarterly Journal of Experimental Psychology, 47A, 207–231.
#BEEFMFZ"%$IJODPUUB%4UBďPSE-5VSL%öôôö
*TUIFXPSEMFOHUIFďFDUJO45.FOUJSFMZBUUSJCVUBCMFUP
PVUQVUEFMBZ &WJEFODFGSPNTFSJBMSFDPHOJUJPOQuarterly
Journal of Experimental Psychology, 55A, 353–369.
#BEEFMFZ"%5IPNTPO/#VDIBOBO.õýûù
8PSE
length and the structure of short-term memory. Journal of
Verbal Learning and Verbal Behavior, 14, 575–589.
#JSFUB5+'SZ4&+BMCFSU"/FBUI*4VSQSFOBOU"
.5FIBO(5PMBO("öôõô
#BDLXBSESFDBMM
and benchmark effects of working memory. Memory &
Cognition, 38, 279–291.
#SPXO(%")VMNF$õýýù
.PEFMJOHJUFNMFOHUI
effects in memory span: No rehearsal needed? Journal of
Memory and Language, 34, 594–621.
#SPXO(%"/FBUI*$IBUFS/öôôû
"UFNQPral ratio model of memory. Psychological Review, 114,
539–576.
#SPXO(%"1SFFDF5)VMNF$öôôô
0TDJMMBUPS
based memory for serial order. Psychological Review, 107,
127–181.
#VSHFTT/)JUDI(+õýýý
.FNPSZGPSTFSJBMPSEFS
A network model of the phonological loop and its timing.
Psychological Review, 106, 551–581.
$PMUIFBSU.õýüõ
5IF.3$1TZDIPMJOHVJTUJD%BUBCBTF
Quarterly Journal of Experimental Psychology, 33A,
497–505.
$PXBO/õýýù
Attention and memory: An integrated
framework/FX:PSL/:0YGPSE6OJWFSTJUZ1SFTT
Cowan, N., Day, L. J., Saults, J. S., Keller, T. A., Johnson, T.,
'MPSFT-õýýö
5IFSPMFPGWFSCBMPVUQVUUJNFJOUIF
effects of word length on immediate memory. Journal of
Memory and Language, 31, 1–17.
$PXBO/8PPE/-#PSOF%/õýýø
3FDPOêSNBtion of the short-term storage concept. Psychological Science, 5, 103–106.
%PTIFS#".B++õýýü
0VUQVUMPTTPSSFIFBSTBM
MPPQ 0VUQVUUJNFWFSTVTQSPOVODJBUJPOUJNFMJNJUTJO
immediate recall for forgetting-matched materials. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 24, 316–335.
)FOESZ-5FIBO(öôôù
"OJUFNPSEFSUSBEFPďFYplanation of word length and generation effects. Memory,
13, 364–371.
)FOTPO3/"õýýü
4IPSUUFSNNFNPSZGPSTFSJBMPSEFS
The start–end model. Cognitive Psychology, 36, 73–137.
+BMCFSU"/FBUI*#JSFUB5+4VSQSFOBOU".JO
QSFTT
iWhen does length cause the word length effect?”
Journal of Experimental Psychology: Learning, Memory,
and Cognition.
-FXBOEPXTLZ4'BSSFMM4öôôü
4IPSUUFSNNFNPSZ
/FXEBUBBOEBNPEFM*O#)3PTT&E
The psychology of learning and motivation7PMýQQõmøü
4BO
%JFHP$"&MTFWJFS
-FXBOEPXTLZ4.VSEPDL##õýüý
.FNPSZGPSTFrial order. Psychological Review, 96, 25–57.
-J4$-FXBOEPXTLZ4õýý÷
*OUSBMJTUEJTUSBDUPSTBOE
recall direction: Constraints on models of memory for serial order. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 19, 895–908.
-POHPOJ"3JDIBSETPO+5&"JFMMP"õýý÷
"Sticulatory rehearsal and phonological storage in working
memory. Memory & Cognition, 21, 11–22.
-PWBUU1"WPOT4&.BTUFSTPO+öôôô
5IFXPSE
length effect and disyllabic words. Quarterly Journal of
Experimental Psychology, 53A, 1–22.
/BJSOF+4õýýô
"GFBUVSFNPEFMPGJNNFEJBUFNFNPSZ
Memory & Cognition, 18, 251–269.
/BJSOF+4/FBUI*4FSSB.õýýû
1SPBDUJWFJOUFSGFSence plays a role in the word-length effect. Psychonomic
Bulletin & Review, 4, 541–545.
/FBUI*öôôô
.PEFMJOHUIFFďFDUTPGJSSFMFWBOUTQFFDIPO
memory. Psychonomic Bulletin & Review, 7, 403–423.
/FBUI*#JSFUB5+4VSQSFOBOU".öôô÷
5IF
time-based word length effect and stimulus set specificity.
Psychonomic Bulletin & Review, 10, 430–434.
/FBUI*#SPXO(%"öôôú
4*.1-&'VSUIFSBQQMJcations of a local distinctiveness model of memory. In B.
)3PTT&E
The psychology of learning and motivation
QQöôõmöø÷
4BO%JFHP$""DBEFNJD1SFTT
/FBUI*'BSMFZ-"4VSQSFOBOU".öôô÷
%JSFDUMZ
assessing the relationship between irrelevant speech and
articulatory suppression. Quarterly Journal of Experimental Psychology, 56A, 1269–1278.
/FBUI*/BJSOF+4õýýù
8PSEMFOHUIFďFDUTJOJN-
WORD LENGTH AND RECALL DIRECTION t AJP 124_1 text.indd 85
1/27/11 1:16:36 PM
NFEJBUFNFNPSZ0WFSXSJUJOHUSBDFEFDBZUIFPSZPsychonomic Bulletin & Review, 2, 429–441.
0CFSBVFS,4ۓ).4DIVM[F38JMIFMN08JUUNBOO88öôôô
8PSLJOHNFNPSZDBQBDJUZ'BDFUTPG
a cognitive ability construct. Personality and Individual
Differences, 29, 1017–1045.
1BHF.1"$VNNJOH/.BEHF"/PSSJT%(
öôôû
4QFFDIFSSPSTBOEUIFQIPOPMPHJDBMTJNJMBSJUZ
FďFDUJOTIPSUUFSNNFNPSZ&WJEFODFTVHHFTUJOHBDPNmon locus. Journal of Memory and Language, 56, 49–64.
1BHF.1"/PSSJT%õýýü
5IFQSJNBDZNPEFM"
new model of immediate serial recall. Psychological Review, 105, 761–781.
1BHF.1"/PSSJT%(öôô÷
5IFJSSFMFWBOUTPVOE
effect: What needs modelling, and a tentative model.
Quarterly Journal of Experimental Psychology, 56A,
1289–1300.
4VSQSFOBOU".-F$PNQUF%$/FBUI*öôôô
.Bnipulations of irrelevant information: Suffix effects with
articulatory suppression and irrelevant speech. Quarterly
Journal of Experimental Psychology, 53A, 325–348.
4VSQSFOBOU"./FBUI*#SPXO(%"öôôú
.PEFMJOHBHFSFMBUFEEJďFSFODFTJOJNNFEJBUFNFNPSZ
VTJOH4*.1-&Journal of Memory and Language, 55,
572–586.
4VSQSFOBOU"./FBUI*-F$PNQUF%$õýýý
*Srelevant speech, phonological similarity, and presentation
modality. Memory, 7, 405–420.
5FIBO()FOESZ-,PDJOTLJ%öôôõ
8PSEMFOHUI
and phonological similarity effects in simple, complex
and delayed serial recall task: Implications for working
memory. Memory, 9, 333–348.
5FIBO(.JMMT,öôôû
8PSLJOHNFNPSZBOETIPSU
term memory storage: What does backward recall tell us?
*O/0TBLB3-PHJF.%&TQPTJUP&ET
Working
memory: Behavioural and neural correlatesQQõù÷mõú÷
0YGPSE&OHMBOE0YGPSE6OJWFSTJUZ1SFTT
5FIBO(5PMBO("öôôû
8PSEMFOHUIFďFDUTJO
long-term memory. Journal of Memory and Language, 56,
35–48.
8BMLFS*)VMNF$õýýý
$PODSFUFXPSETBSFFBTJFS
UPSFDBMMUIBOBCTUSBDUXPSET&WJEFODFGPSBTFNBOUJD
contribution to short-term serial recall. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25,
1256–1271.
8BULJOT.+õýûö
-PDVTPGUIFNPEBMJUZFďFDUJOGSFF
recall. Journal of Verbal Learning and Verbal Behavior,
11, 644–648.
APPENDIX
STIMULI USED IN EXPERIMENT 1: 15 long (3-syllable) and 15 short (1-syllable) words equated for frequency,
concreteness (CNC), familiarity (FAM), and imageability (IMG)
Word
CNC
FAM
IMG
KFFRQ
TLFRQ
AISLE
509
503
528
6
BEAM
502
476
539
DRAW
442
542
435
FLOOD
553
523
HOWL
434
JOKE
LICE
Word
CNC
FAM
IMG
KFFRQ
TLFRQ
72
ABUNDANT
351
524
443
9
50
21
127
ACCIDENT
419
564
518
33
399
56
428
APPROVAL
267
526
375
51
108
598
19
325
ARTICLE
479
533
421
68
550
447
536
4
72
AVENUE
539
529
564
46
320
388
580
483
22
230
FOREIGNER
492
499
516
4
92
543
397
532
2
4
HEXAGON
559
387
527
1
4
MINK
589
524
604
5
27
MUSICIAN
564
558
585
23
72
PAIN
426
569
502
88
541
OCCASION
346
566
305
58
424
PEAL
402
451
433
1
13
PARAGRAPH
493
559
482
12
72
PINT
483
536
487
13
92
RECITAL
476
468
495
8
27
ROSE
608
556
623
86
801
SEDATIVE
459
423
459
1
13
SALE
364
555
422
44
403
SYMPATHY
278
501
402
36
228
THREAT
335
524
408
42
108
TELEGRAPH
547
460
518
21
126
WRATH
304
466
377
9
51
TELEPHONE
619
605
655
76
800
Mean
458.8
509.93
500.47
Mean
459.2
513.47
484.33
29.8
219
27.87
219.6
Note. All values obtained from the MRC Psycholinguistic Database (Coltheart, 1981; http://www.psy.uwa.edu.au/mrcdatabase/uwa_mrc.htm). KFFRQ = Kučera
& Francis frequency; TLFRQ = Thorndike–Lorge written frequency.
t SURPRENANT ET AL .
AJP 124_1 text.indd 86
1/27/11 1:16:36 PM