Full text of DOI:10.3758/BF03195263

Memory & Cognition
2002, 30 (1), 34-45
Is the go/no-go lexical decision task an alternative
to the yes/no lexical decision task?
MANUEL PEREA, EVA ROSA, and CONSOLACIÓN GÓMEZ
Universitat de València, València, Spain
In the go/no-go lexical decision task (LDT), participants are instructed to respond as quickly as they
can when a word is presented and not to respond if a nonword is presented. By minimizing part of the
response selectionprocess in the experimentaltask, the impact of response decisiontime on the obtained
lexical decision time is probably reduced relative to the standard yes/no LDT (Gordon, 1983). Experi-
ments 1 and 2 show that the go/no-go LDT is sensitive to the effects of word frequency and associative
priming—the magnitude of these effects is similar with the two tasks. More important, the go/no-go
LDT has a number of advantages with respect to the “standard” yes/no LDT: It offers faster response
times, more accurateresponding, and fewer processing demands than does the yes/no task. Accordingly,
the go/no-go task appears to be an excellent alternativeto the standard yes/no task.
The lexical decision task (LDT for short) has been the the comparison between tasks similar to the comparison
most widely used task in the study of visual word recogni- betweenDonders’s (1868/1969;see alsoLuce, 1986)Type B
tion. In the standard yes/no LDT, participantsare to decide and Type C reaction time (RT) tasks. In the Type B reac-
as rapidly and as accurately as possible whether a string of tion task (as in the yes/no LDT), the participantmust make
word
letters is a word or not by pressing either the
or the a distinct response to each of several possible stimuli,
nonword
key on a response box. However, in the usual ex- whereasintheTypeC reactiontask(as inthego/no-goLDT),
perimental setup, after noting whether the letter string is a a response is required to only one of several possible stim-
word or not,theparticipantmust remember which response uli. Nonetheless, we must be assured that the go/no-go
to make for the words and which response to make for the procedureapplied to the LDT givesthe same—or better—
nonwords. Since this assignment is arbitrary, it may cause measures of the variables of interest and that this proce-
variability in the produced lexical decision times that is dure does not fundamentally modify the processes of in-
due to deciding what response to make, which is not rele- terest underlying the experimental task.
vant to the process of interest (see Pachella, 1974). In fact,
The processing in the two tasks would presumably in-
participants often make errors in the LDT, not because volve selecting the appropriate lexical unit (lexical selec-
they have misclassified the letter string, but because, once tion stage) and completing whatever decision-making
they have classified the letter string, they have executed processes are necessary to verify that it is the appropriate
the wrong response.
unit (responsedecisionstage). The basicdifference between
word
In order to avoid these problems, a go/no-go LDT has making a correct
response in the two tasks seems to
been proposedinstead (see Gordon,1983;Gordon & Cara- be the mental operationsinvolvedin the response selection
mazza, 1982). In the go/no-go LDT, the participant is in- stage. In the yes/no task, after deciding whether the letter
structed to respond as quicklyas he or she can when a word string is a legitimate word, participants are required to
word nonword
or ).
is presented but to withhold any response if the presented chooseoneof two alternativeresponses(
stimulus is a nonword. (Another form of the go/no-go In the go/no-go task, the response selection process is ap-
LDT would be to respond to a nonword and withhold re- parently simpler, because nonwords do not require an
sponse to a word; see Measso & Zaidel, 1990.)This makes overt response (Gibbs & Van Orden, 1998; Gordon, 1983;
Yelland, 1993). As a result, if a successful lexical selection
is diagnosticof the letter string’s beinga word, and if we as-
Parts of this research were reported at the 41st Annual Meeting of the sume a discrete stage model of performance in which the
Psychonomic Society, New Orleans, 2000. The first author was sup-
ported by a grant from the Secretaría de Estado de Educación y Univer-
any lexicaleffects (e.g., word frequency, associativeprim-
sidades of Spain. The second and third authors were recipients of pre-
subprocesses are identified as successive temporal stages,
ing)should be approximatelythe same size in the two tasks
doctoral grants from the Spanish Ministry of Science and Culture. We
(i.e., additive-factor method; Sternberg, 1969). (We discuss
an alternative account in the General Discussion section.)
The main goal of this study is to evaluate the possibil-
ity of usingthe go/no-go version of the lexicaldecisiontask
as a replacementfor the standardyes/no version. Although
the lexical decision task has had a controversial history
thank Steve Lupker, Ken Paap, Yasushi Hino, Gregory Ashby, and two
anonymous reviewers for helpful and constructive criticism on earlier
versions of this paper. Correspondence concerning this paper should be
sent to M. Perea, Departament de Metodologia, Facultat de Psicologia,
Av. Blasco Ibáñez, 21, 46010-València, Spain (e-mail: mperea@uv.es).
—Accepted by previous editorial team
Copyright 2002 Psychonomic Society, Inc.
34

THE YES/NO AND THE GO/NO-GO LDT
35
(see, e.g., Balota & Chumbley, 1984, 1990; Grainger & go/no-go task than in the yes/no task (Hino & Lupker,
Jacobs, 1996; Hino & Lupker, 1996, 1998; Monsell, 1998, 2000; but see Measso & Zaidel, 1990).
Task demands
4. . The simpler the task is, the less time-
Doyle, & Haggard, 1989; Paap, McDonald, Schvaneveldt,
& Noel, 1987; Pollatsek, Perea, & Binder, 1999), it is consumingdecisional,task-specificprocesseswill be nec-
word
likely that researchers will continue to use it, since com- essary to make a correct
response. In this light, the
peting tasks—such as the pronunciation task—have their go/no-go task appears to be simpler to perform than the
own problems (see, e.g., Forster & Shen, 1996;Paap et al., yes/no task (see above). For instance, Yelland (1993) indi-
1987). Thus, it seems reasonable to examine any modifi- cated that the go/no-go LDT (but not the yes/no LDT) can
cations in the procedure of the lexical decision that can easily be applied to experiments with special populations
make it more sensitiveand validin thescientificenterprise. (e.g., children; see also Davis, Castles, & Iakovidis,1998).
In order to compare the two techniques, we will focus on
In Experiment 1, we reexamine the word frequency ef-
fect with thetwo techniques.The two techniquesproduced
four important criteria.1
The number of errors
1.
. This is not a trivial issue, since similar RT versus word frequency functions in the Gor-
firm conclusionsabout the locus of the effect of a manip- don and Caramazza (1982) study. In an on-going project,
ulation must be restricted to cases in which error rates are one of us also found a similar word frequency advantage
very low (McClelland,1979). It is difficultto construe the intheyes/no LDT and in thego/no-goLDT (74 vs. 63 msec,
reduced errors to word stimuli as anythingbut positive.In respectively; Perea et al., 1998). However, Hino and Lup-
this light, several studieshave consistentlyfound that par- ker (1998; see also Hino & Lupker, 2000) found larger
ticipants commit substantially fewer errors in the re- word frequency effects with the go/no-go LDT than with
sponses to words in the go/no-go task than in the yes/no the yes/no task (see below), which casts doubt on the con-
task (Gibbs & Van Orden, 1998; Gordon & Caramazza, clusion that the two factors have additiveeffects on RT. In
1982; Hino & Lupker, 1998, 2000; Perea, Fernández, & Experiment 2, we examine the presence of an automatic
Carreiras, 1998).2 In addition,thedataconcerningfalse pos- associative priming effect with the two techniques. More
word
itive errors (i.e., nonword error rates:
nonword stimuli)needs to be taken into account,since this a second block of trials (with a yes/no task or a go/no-go
critical
responses to important, Experiment2 analyzeswhether the latenciesin
is
for an accurate evaluation of performance in task) are modulated by the task used in the first block
go/no-go tasks, as compared with binary decision tasks (go/no-go or yes/no task), that is, task is manipulatedboth
(i.e., it may alter the speed with which one responds to the between subjects and within subjects.
word stimuli).
The speedofthe RT
2.
. Undoubtedly, it is desirableto use The Word Frequency Effect in the Yes/No and
a word identificationtask that minimizes time-consuming the Go/No-Go Lexical Decision Task
decisional processes that are unnecessary to recognize a
Althoughmost modelsofvisualword recognitionassume
given word, so that RTs will be closer to the total duration that the time taken to access the lexical entry of a word in
of eye fixations on a word when it is read in isolation— the mental lexicon is mainly governed by the frequency of
around 350–450 msec, (see, e.g., Grainger, O’Regan, Ja- occurrence of the word (e.g., search model, Forster, 1976;
generally
cobs, & Seguí, 1989). In this context, RTs are
resonancemodel, Gordon, 1983; multipleread-out model,
faster with the go/no-go task than with the yes/no task Grainger & Jacobs, 1996; interactive activation model,
(e.g., Chiarello,Nuding, & Pollock 1988; Gordon & Cara- McClelland & Rumelhart, 1981; activation-verification
mazza, 1982; Measso & Zaidel, 1990; Perea et al., 1998; model, Paap, Newsome, McDonald, & Schvaneveldt,
but see Gibbs & Van Orden, 1998; Hino & Lupker, 1998, 1982), otherauthors(e.g., Balota& Chumbley, 1984,1990)
2000). However, the fact that the participantsin most of the argue that word frequency also impacts the response deci-
above-cited experiments were not randomly assigned to sion stage. Nonetheless, if it is true that the difference be-
each level of task (except in the experiments by Chiarello tween the two tasks is in the response selection stage
et al., 1988, and Measso & Zaidel, 1990) makes it difficult (rather than in the lexical selection or the response deci-
to compare the main effect of task.
sion stage), we can still concludethat the yes/no LDT and
The variabilityin the data
3.
. There is littledoubtthat re- the go/no-go LDT should show a word frequency effect of
ducedvariability(noise) is a desirableproperty in the data similar size (see Gordon & Caramazza, 1982, and Perea
from any word identification task. Gordon (1983, 1985) et al., 1998, for a demonstration of the effect).
indicatedthat by minimizingpart of the response selection
However, Hino and Lupker (1998) found a much larger
process in the experimental task, the impact of response word frequency effect in the go/no-go LDT (163 msec)
decision time on the obtained lexical decision time in the than in the yes/no LDT (107.5 msec). Given that the word
go/no-go task would be reduced (relative to the standard frequency effect in the yes/no LDT might be inflated by
yes/no task), therebyreducingsubsequentvariability. How- the response decision stage (e.g., the word frequency ef-
ever, Gordon (1983, 1985; Gordon & Caramazza, 1982) fect is larger in the yes/no LDT than in the standard nam-
did not report data that can be used to assess this claim. In ing task or in a reading task; see Perea & Pollatsek, 1998;
fact, recent publishedstudieshave generally found higher Schilling, Rayner, & Chumbley, 1998), it does not seem
MS
values—if anything—for mean square errors ( es) in the reasonable to argue that lexical decisions with the stan-

36
PEREA, ROSA, AND GÓMEZ
word
dard yes/no procedure would underestimate the word fre- note that the nonword error rate (i.e.,
responses to
quency effect. Rather, the larger word frequency effect in nonwords) was higher in the yes/no task than in the go/no-
the go/no-go LDT would callinto questionthe role ofword go task in the Hino and Lupker experiments (13.0% vs.
frequency in the lexical selection process (see Hino & 8.7% in the 1998 experiment and 11.7% vs. 6.5% in the
Lupker, 1998, for a discussion of this issue).
2000 experiment), which suggests that the participants
word
Hino and Lupker (1998) proposed an ingenious expla- could have used different strategies—for example, a
nation for the different magnitude of the word frequency bias in the yes/no task—in the two tasks (the overall word
effect on the basis of the different demands of the two error rates for the yes/no and the go/no-go tasks were, re-
tasks: In the yes/no LDT, there is pressure to make a rapid spectively, 11.6% vs. 4.9% in the 1998 experiment and
response to all stimuli (words and nonwords). When an 6.7% vs. 4.1% in the 2000 experiment).
unfamiliar low-frequency word is encountered, partici-
Given the theoreticalimplicationsof the Hino and Lup-
pants might make the wrong answer, and these trials ker results with respect to the role of word frequencyin the
would end up counting as errors and not contributing to lexical selection process, it is of theoretical relevance to
the mean latency for low-frequency words. In contrast, in establish why it is enhanced in at least some implementa-
the go/no-go LDT, nonwords do not require a response. tions of the go/no-go task. We must bear in mind that Gor-
Thus, when an unfamiliar word is encountered, a trial- don and Caramazza (1982)found a similar RT versus word
terminatingnegativeresponse would not be made, and lex- frequency function with the two techniques(see also Perea
ical processing would continue. In this case, the partici- et al., 1998). One basic difference between the Hino and
pant may make a slow response (assuming that the word Lupker (1998) study and the other two studies would be
is in the individual’s vocabulary). Consistent with this in- the range of word frequency: The low-frequency words in
terpretation,Hino and Lupker (1998) found that error rate the Hino and Lupker (1998) study had a lower frequency
decreased for low-frequency words with the go/no-go of occurrence than did those employed by Gordon and
LDT (19.3% vs. 8.4%), whereas mean latency for low- Caramazza and Perea et al. Interestingly, in the Hino and
frequency words increased with the go/no-go LDT (684.5 Lupker(2000)paper, the low-frequencywords were higher
vs. 656 msec). In contrast,mean latencyfor high-frequency in frequency than those in their 1998 paper, and the inter-
words decreased in the go/no-go LDT (522 vs. 548 msec), actionbetween word frequencyand task was much weaker;
which is reasonable giventhat the go/no-go LDT seems to it was not tested, however. Perhaps the effects found by
be simpler than the yes/no LDT. However, Hino and Lup- Hino and Lupker (1998) are restricted to unfamiliar, very
ker (1998) did not provide any additionalstatisticalanaly- low frequency items. In order to examine this possibility,
ses in support of their explanation.Mean lexical decision besidesthehigh-frequency condition,we selectedtwo groups
time would be of limited use as a statistic in this case, be- of low-frequency words in Experiment 1: low-frequency
cause the main interest is the shape of the response time words and very low frequencywords (similar in frequency
distribution(see, e.g., Balota & Spieler, 1999).
to the low-frequency words in the Hino & Lupker, 1998,
More recently, Hino and Lupker (2000) found a some- study). According to the Hino and Lupker (1998) analy-
what larger word frequency effect in the go/no-go LDT sis, the more straightforward outcome would be to ob-
(120.5 msec) than in the yes/no LDT (96 msec; a differ- serve additivity when high-frequency words are paired
ence less than half in size of that in their 1998 paper). The with low-frequency words, and an interaction when they
reliability of the interaction between task and word fre- are paired with the very low frequency words.
quency was not tested, however. They concludedthat “the
task-specific nature of the frequency effects in the stan-
dard and go/no-go LDTs suggests that these results are
more consistent with a task-specific account” (Hino &
EXPERIMENT 1
Word Frequency Effect
Lupker,2000,pp. 179–180).We wouldliketo note thatRTs
much
Method
Participants. Forty-eight psychology students from the Univer-
sity of València took part in the experiment for course credit. All of
them had either normal or corrected-to-normal vision and were na-
tive speakers of Spanish.
Materials. One hundred and twenty Spanish words were col-
lected on the basis of the word frequency norms for Spanish
(Alameda & Cuetos, 1995). Forty of those words were of high fre-
quency, 40 of low frequency, and 40 of very low frequency. Fre-
quency counts for high-frequency words were greater than 125 per
two million (mean, 348; range, 140–998), frequency counts for the
low-frequency words were greater than 15 and less than 25 (mean,
20.5; range, 17–24), and frequency counts for the very low frequency
words were less than 4 (mean, 2.4; range, 2–3). The target words
were all 6 letters long. The 120 target nonwords were 80 stimuli con-
structed by changing an interior letter from a 6-letter Spanish word
in the go/no-go LDT were
greater than those in the
yes/no LDT, not only for low-frequency words (715 vs.
635 msec, respectively), but also for high-frequency
words (595 vs. 539 msec, respectively). This finding
seems rather surprising, since the go/no-go LDT is pre-
sumably simpler than the yes/no LDT. It is likely that this
difference was caused because the participants were not
randomlyassigned to each level of task (the two tasks were
run over a year apart; Lupker, personal communication,
January 20, 2000), and thereby it may be misleading to
compare the effect of task under these circumstances. In
fairness to Hino and Lupker, we must note that the com-
parison between these two tasks was not the main focus of
any of their two papers. Finally, it may be of interest to other than one from the experimental set (e.g., lógeca; the word

THE YES/NO AND THE GO/NO-GO LDT
37
F
MS
would be lógica, the Spanish for logic) and 40 stimuli constructed
by transposing two interior letters from a 6-letter Spanish word other
than one from the experimental set (e.g., tineda; the word would be
tienda, the Spanish for shop).
Design. Task type (go/no-go LDT, yes/no LDT)was varied between
participants (24 participants were randomly assigned at each level of
task type), whereas word frequency (high frequency, low frequency,
and very low frequency) was varied within participants. Each par-
ticipant was given a total of 240 experimental trials: 120 word trials
and 120 nonword trials.
Procedure. The participants were tested in groups of 4 – 8 in a
quiet room. Presentation of the stimuli and recording of RTs were
controlled by Apple Macintosh Classic II microcomputers. The rou-
tines for controlling stimulus presentation and RT collection were
obtained from Lane and Ashby (1987) and from Westall, Perkey, and
Chute (1986), respectively. On each trial, the sequence “> <” was
presented for 200 msec on the center of the screen. After a 50-msec
blank, a lowercase letter string was presented. The stimulus remained
on the computer screen until the participant responded or until 2 sec
had elapsed (similar to the experiments of Gordon, 1983, and Hino
& Lupker, 1998, 2000). In the go/no-go task, the participants were
instructed to press the mouse with their dominant hand if the letter
string was a legitimate Spanish word. In the yes/no task, the partic-
ipants were instructed to press one of two buttons on the keyboard
to indicate whether the letter string was a Spanish word or not. The
participants used their dominant hand to make the word response. In
both tasks, this decision was to be made as rapidly and as accurately
as possible. The intertrial interval was set to 400 msec. Each partic-
ipant received a different random order of stimuli. Each participant
received a total of 24 practice trials prior to the experimental phase.
The session lasted approximately 15 min.
2(1,117)= 33.36, e = 3,797.6],andlow-frequency words
were responded to faster than very low frequency words
MS MS
F
[
F
1(1,46)= 205.34, e = 569.3; 2(1,117) = 55.86,
=
e
3,797.6].The main effect of task type was also significant
F
MS F
= 23,420.9; 2(1,117) = 345.28,
[
1(1,46) = 6.73,
= 790.97]: The eparticipants were faster with the
MSe
go/no-go LDT than with the yes/no LDT. More important,
the interaction between task type and word frequency was
F
MS
F
not significant[ 1(2,92) = 0.48,
MS
e = 579.5; 2(2,117) =
0.479,
= 790.97]. The magnitude of the word fre-
e
quencyeffect was similarin thetwo tasks. Furthermore, the
correlation between lexical decision time across items in
the two tasks was quitehigh ( = .85), which indicatesthat
the two tasks bear a high degree of similarity. Finally, we
should note that if we analyze separately the two tasks,
error variance was a bit higher with the go/no-go task than
with the yes/no task (the differences were not significant,
r
MS
however): The
370.91 versus 320.71 ( 1) and 2,475.33 versus 2,113.23
es for the word frequency effect were
F
F
(
2), respectively.
In the analyses of word error rates, the main effect of
F
word frequency was also significant [ 1(2,92) = 48.37,
MS MS
F
e = 7.87; 2(2,117) = 14.21,
the participantsmade fewer errors with the go/no-go LDT
MS
e = 44.63]. In addition,
F
than with the yes/no LDT [ 1(1,46) = 39.74,
e = 15.7;
e = 19.5, < .001]. The interaction
between task type and word frequency was significant
MS MS
F
MS
p
2(1,117) = 51.08,
F
F
[
1(1,102)= 375.31, e = 684; 2(1,162)= 20.56,
=
e
Results
19.5], which reflected the existence of a floor effect with
Lexicaldecision latenciesless than 250 msec or greater the go/no-go LDT (see Table 1).
than 1,500 msec were excluded from the latency analyses
(less than 0.4%). Mean lexical decision latencies for cor- Discussion
rect responses and mean error rates were calculatedacross
The present experiment has shown that the RTs were
individuals and across items. Participant and item analy- faster and more accurate with the go/no-go task than with
ses of variance (ANOVAs) based on the participants’ and the yes/no task. Furthermore, the word frequency advan-
items’ response latencies and error rates were conducted tage in the go/no-go LDT was the same magnitude as that
3
basedona 3 (word frequency, high,low, verylow) 2 (task in the yes/no LDT: Lexical decision times were about
type, go/no-go LDT vs. yes/no LDT) design. The .05 level 67 msec faster in the go/no-go LDT across the three word
of significance was adopted throughout. The mean RTs frequency conditions. This result is consistent with some
and percentageerror from the participantanalysisare pre- findings (Gordon & Caramazza, 1982; Perea et al., 1998)
sented in Table 1.
Not surprisingly, in theanalysesof responselatencies,the 2000). It is importantto stress that thisword frequency ad-
not
but contradictsotherrecent results (Hino & Lupker, 1998,
F
main effect of word frequency was significant, [ 1(2,92) = vantage in the go/no-go task was
accompanied by a
e = 3,797.6]: larger number of false positive responses [the percentage
High-frequency words were responded to faster than of false positive errors was similar in the two tasks;
MS
MS MS
F
329.69,
e = 579.5; 2(2,117) = 88.26,
F
F
(1,46) < 1]. Finally, we should note that contrary to Gor-
low-frequency words [ 1(1,46) = 218.78,
= 345.8;
e
Table 1
Mean Response Times (RTs, in Milliseconds) and Percentages of Error (PE) for the Word
Targets With the Go/No-Go Lexical Decision Task (LDT) and with the Yes/No LDT
Word Frequency
High
Low
Very Low
LDT
M
PE
M
PE
M
PE
Go/no-go
Yes/no
Difference
544
611
67
0.0
1.7
1.7
598
669
71
0.4
3.6
3.2
673
734
61
2.4
10.0
7.6
Note—Mean RTs for nonwords in the yes/no LDT were 779 msec. Mean error rates for nonwords were 5.8% and
5.4% for the go/no-go LDT and the yes/no LDT, respectively.

38
PEREA, ROSA, AND GÓMEZ
don’s (1983) claim, error variance was not lower with the could be argued that this analysis would be more infor-
go/no-go task than with the yes/no task. mativewhenconductedontheHinoandLupker(1998)data,
It seems necessary to explain why Hino and Lupker since our data showed additive effects of word frequency
(1998)found a strongerword frequencyeffect inthego/no- and task. Nevertheless, we believe that it is important to
go task than in the yes/no task. As we said in the intro- show that the RT distributionsin the two tasks have a sim-
duction, Hino and Lupker (1998) indicated that when an ilar shape.
unfamiliar low-frequency word is encountered, partici-
But there is another way to test Hino and Lupker’s
pants might make the wrong answer in the yes/no task, and (1998) interpretation. If the Hino and Lupker (1998) ac-
those trials would end up counting as errors and not con- count is correct, words with a high percentage of errors
tributing to the mean latency for that condition. In con- (i.e., the “difficult” words) should show faster lexical de-
trast, in the go/no-go LDT, nonwords do not require a re- cision times in the yes/no LDT than in the go/no-go LDT.
sponse and the participantcould eventuallyrealize that the Only 4 out of the 40 very low frequency words in our ex-
unfamiliar item is a word, producinga slow response and, periment showed a faster mean lexicaldecision time in the
thereby, a larger word frequency effect. Obviously, an yes/no LDT than in thego/no-goLDT. ConsistentwithHino
analysis based on the mean RTs per condition is not the and Lupker’s (1998) interpretation, two of these words
dehesa billón
and ; the Spanish for pasture and billion,re-
best way to test this hypothesis. Instead, we proceeded to
(
compare the shapes of the group RT distributionsby using spectively) showed, by far, the highest error rates in the
the procedure described extensively by Ratcliff (1979).3 experiment (25% of the errors each in the go/no-go LDT
If participantsmake a higher number of slow responses to and 41.7% of the errors each in the yes/no LDT).4 As a re-
the very low frequency words in the go/no-go LDT than sult, it could be argued that the 10% error rate observed
in the yes/no LDT, the right tail of the RT distribution with the very low frequency words in the yes/no LDT was
should be longer for the go/no-go LDT. However, the Vin- too small to trigger the Hino and Lupker (1998) pattern.
3
cent histograms with 5% quantiles based on all the cor- As we indicated earlier, the word frequency task inter-
rect RTs show that the RT distribution for the go/no-go action was relatively weak in the Hino and Lupker (2000)
task does not have a longer tail than the RT distributionfor paper, in which the low-frequency words were higher in
the yes/no task (if anything, the distribution is a bit more frequency than those in their 1998 paper.5 Thus, part of the
skewed in the yes/no task; see Figure 1). Of course, it word frequencyeffect reported by Hino and Lupker(1998)
Figure 1. Group reaction time distributionsfor the very low frequency words
in the go/no-go lexical decision task (LDT) and the yes/no LDT.

THE YES/NO AND THE GO/NO-GO LDT
39
Table 2
Summary of Word Frequency Studies (Mean Response Times and Percentages of Error)
That Have Used Both the Go/No-Go Task and the Yes/No Task
Word Frequency
High
Low
Very Low
Nonword
Task
M
PE
M
PE
M
PE
WF Effect
M
PE
Experiment 1
Go/no-go
Yes/no
544
611
0.0
1.7
598
669
0.4
3.6
673
734
2.4
10.0
129/54
123/68
–
779
5.8
5.4
Perea, Fernández, and Carreiras (1998)
Go/no-go
Yes/no
544
591
0.0
1.3
618
654
0.5
6.1
–
–
–
–
74
63
–
736
3.4
3.5
Hino and Lupker (1998)
Go/no-go
Yes/no
522
548
1.4
4.0
–
–
–
–
685
656
8.4
19.3
163
108
–
724
8.7
13.0
Hino and Lupker (2000)
Go/no-go
Yes/no
594
539
1.0
2.6
715
635
7.2
10.9
–
–
–
–
121
96
–
719
6.5
11.7
Note—WF effect refers to the difference between the low-frequency (LF) words (and/or very low frequency words) and the high-frequency (HF)
words.
in the go/no-go LDT could be attributed to participants’ mean RT of 734 msec would go higher, and again we
3
not knowing for sure the meaning of a particular word, would have a task word frequency interaction, but in
opposite
rather than to the frequency of occurrence of that word. the
direction from that reported by Hino and
Finally, it might be argued that the interaction between Lupker (1998, 2000). (We will discuss this issue in the
word frequency and task in the Hino and Lupker (1998) General Discussion section.)
paper was caused by different participants’ strategies in
the experiments.6 We must bear in mind that the partici-
pants in our experiment made fewer errors—for both word
and nonword stimuli (see Table 2)—than did those in the
Hino and Lupker (1998) experiments, which may be taken
EXPERIMENT 2
Associative Priming Experiment
CHAIR
It is well documentedthat a word (e.g., the target,
)
to suggest that stimulus processing was somehow deeper is responded to faster when it is preceded by an associa-
table
in our experiment. (In addition, it seems that there was a tively related word (the prime,
word
) than when it is pre-
biaswiththeyes/notask, relativeto thego/no-go task, ceded by an unrelated word: This is the associative prim-
in the Hino and Lupker, 1998, experiments.)If lexical de- ing effect (see Neely, 1991, for a review). In the classical
spreading activation
cisions are based on deep processing of the stimuli, ana-
lytic processing would be made before making lexical de- lexical/semantic network, so that the prime word automat-
preactivates
account, activation spreads along a
cisions in both the yes/no LDT and the go/no-go LDT, and ically
thereby a similar word frequency effect would be expected words (Collins & Loftus, 1975). Thus, the identification
preactivated
associatively/semantically related
in the two tasks. However, if lexicaldecisionsare based on time of the subsequently
words is sped up. As
shallow processing of the stimuli, the time pressure to we said earlier, if a successful lexical selection is diagnos-
make a rapid response in the yes/no LDT may provoke rel- tic of the letter string’s being a word, the associativeprim-
atively fast lexical decision times to unfamiliar words ing shouldbe approximatelythe same size in the two tech-
(with the cost of a relatively high proportion of errors). In niques. Even assuming that additional decisional and
contrast, there is no time pressure for nonwords in the response processing occurs after lexical access has al-
go/no-go LDT, which may provoke the presence of long ready occurred, but before the overt lexical decision has
lexical decision times for unfamiliar words and, thereby, a been executed (e.g., semantic-checking processes; see
stronger frequency effect for the go/no-go task.
In sum, differentparticipants’strategies—orthepresence these processes can affect differentiallythe
of unfamiliar low-frequency words—might provoke the in the go/no-go LDT and the yes/no LDT.
de Groot, 1984; Neely, 1991), it is difficult to see how
word
decisions
presenceofan interactionbetweenword frequencyand task,
To our knowledge, no previous associative priming ex-
such as that found by Hino and Lupker(1998, 2000).What periments have compared the two techniques.8 In order to
is more, onecan take a very differentperspective and imag- obtain strategy-free automatic activation, we used a stim-
ine what would occur if participants were told that 10% ulus onset asynchrony (SOA) between the prime and the
was not an acceptable word error rate and, instead, they target of 100 msec. The choiceof 100 msec as SOA was to
were asked to be as accurate in the yes/no task as they are make sure that the associativepriming effect would be siz-
in the go/no-go task (2.4%).7 The only cell in which par- able (e.g., McNamara & Altarriba, 1988, pointed out that
ticipantsneed to trade speed for more accuracy is the very automaticactivationreachesasymptoteinabout100 msec).
low frequency in the yes/no task. If they could do it, the One important feature of this experiment is that task

40
PEREA, ROSA, AND GÓMEZ
(go/no-go LDT, yes/no LDT) was manipulated not only Twenty-eight participants performed the go/no-go task in the first
block and the yes/no task in the second block, whereas the other 28
between subjects, but also within subjects. Specifically,
participants performed the yes/no task in the first block and the
half of the participantsperformed the go/no-go task in the
go/no-go task in the second block. Each participant was given a total
first blockand the yes/no task in the secondblock, whereas
of 100 experimental trials in each block: 50 word–word trials (25 re-
lated, 25 unrelated) and 50 word-nonword trials.
the other half performed the yes/no task in the first block
and the go/no-go task in the second block. By doing this,
we avoid the potential interpretive difficulties that may
arise when comparing results across two different sets of
participants.
Procedure. The participants were tested in groups of 4 –8 in a
quiet room. Presentation of the stimuli and recording of RTs were
controlled by Apple Macintosh Performa 6200 microcomputers. On
each trial, a forward mask composed of a row of 11 hash-marks
(###########) was presented for 500 msec on the center of the
screen. Next, a centered lowercase prime word was presented for
100 msec. Primes were immediately replaced by an uppercase tar-
get item. The target word (or nonword) remained on the computer
screen until the participant responded or until 2 sec had elapsed. RTs
were measured from target onset until the participant’s response. In
the go/no-go block, the participants were instructed to press a but-
ton on the keyboard with their dominant hand if the uppercase letter
string was a legitimate Spanish word. In the yes/no block, the par-
ticipants were instructed to press one of two buttons on the keyboard
to indicate whether the uppercase letter string was a Spanish word
or not. The participants used their dominant hand to make the word
response. In both tasks, this decision was to be made as rapidly and
as accurately as possible. The intertrial interval was set to 400 msec.
Each participant received a total of 20 practice trials prior to each ex-
perimental block. There was a short break between the first and the
second block. Stimulus presentation was randomized, with a differ-
ent order for each participant. The session lasted approximately
18 min.
Method
Participants. A total of 56 students from introductory psychol-
ogy courses at the University of València took part in the experiment
in exchange for course credit. All of them had either normal or
corrected-to-normal vision and were native speakers of Spanish.
None of them had participated in the previous experiment.
Materials. One hundred related pairs were selected from the free-
production norms in Spanish supplied by Algarabel, Sanmartín,
García, and Espert (1986). In all cases, the prime was the most fre-
quent associate of the target (e.g., julio–AGOSTO; July–AUGUST).
The mean associative strength was 35% (range, 17%–72%). Unre-
lated primes were also created for each target. Unrelated and related
primes were matched on number of letters and word frequency. The
mean length in letters for the word–word pairs was 5.0 (range, 4 –7)
for the primes and 5.4 (range, 3–12) for the targets. One hundred
word–nonword pairs were also created for the purposes of the lexi-
cal decision task. Nonwords were orthographically legal and had
been constructed by replacing a letter of a Spanish word other than
one of the experimental set.
Results
Two sets of stimuli were created (Set A and Set B). Fifty word–word
pairs were randomly assigned to Set A, whereas the other 50 word–
word pairs were assigned to Set B. In each set, two stimulus lists
were created for each set by matching each of those targets with ei-
ther its associated prime or an unrelated prime. Word–word pairs
were rotated across the priming condition across two groups of par-
ticipants (in each block) so that no participant was presented any sin-
gle prime or target word more than once but every participant re-
ceived both priming conditions in each block. Specifically, the
participants were randomly assigned to one of the following four se-
quences: Set A–List 1 Set B–List 2; Set A–List 2 Set B–List 1; Set
B–List 1 Set A–List 2; Set B–List 2 Set A–List 1. With respect to the
word–nonword trials, 50 pairs were randomly assigned to the first
block, whereas the other 50 pairs were assigned to the second block.
Design. Task (go/no-go LDT, yes/no LDT) and prime–target re-
Incorrect responses (3.2%) and RTs less than 250 msec
or greater than 1,500 msec (0.24% of the data) were ex-
cluded from the latency analysis. Mean lexical latencies
for correct responses and error rates were calculated
across individuals and across items. Separate ANOVAs
were performed for Block 1 and Block 2. Participant and
item ANOVAs based on the participants’ and items’ re-
sponse latencies and percentage error in each block were
conducted on the basis of a 2 (associative relatedness, re-
3
3
lated or unrelated) 2 (task, go/no-go or yes/no) 2 (set,
Set A or Set B) 2 (list, List 1 or List 2) design. The set
and list factors were included as dummy variables to ex-
3
latedness (related, unrelated) were varied within participants. tract the variance owing to the error associated with the
Table 3
Mean Response Times (RTs, in Milliseconds) and Percentages
of Errors on Target Words in Experiment 2
Condition
Block 1
Block 2
Go/No-Go Task
Yes/No Task
M
PE
M
PE
Go–Yes/No Group
Related
Unrelated
547
571
24
0.4
0.4
0.0
604
621
17
6.6
7.0
0.4
Priming effect
Yes/No Task
Go/No-Go Task
M
PE
M
PE
Yes/No–Go Group
Related
Unrelated
613
623
10
5.3
5.3
0.0
556
575
19
0.4
0.4
0.0
Priming effect
Note—Priming effect refers to the difference between the unrelated condition and the related
condition.Percentages of errors for nonwordswere 6.5% and 6.1% for the go/no-go lexical de-
cision task (LDT) and the yes/no LDT, respectively.

THE YES/NO AND THE GO/NO-GO LDT
41
lists (see Pollatsek & Well, 1995). The mean RTs and per- had performed the go/no-go task in the first block had
much longer
centage error from the participant analysis are presented
in Table 3.
latencies (and more errors) when they per-
formed the yes/no task in the second block [559 vs.
MS
F
Block 1. The ANOVA on the latency data showed that 612.5msec, respectively; 1(1,24)= 25.95, e = 3,645.9;
F
MS
the main effect of associative relatedness was significant
MS MS
2(1,96)= 318.9,
e = 1,179.2;the error rates were 0.4%
F
F
[
1(1,48) = 15.31,
e = 548.6; 2(1,96) = 11.72,
=
e
vs. 6.8%, respectively]. Taken together, these results con-
2,654.7],in which targets precededby related primes were firm the fact that the participants’responses are faster and
responded to faster than when preceded by unrelated more accurate in the go/no-go task. Second, the go/no-go
primes. The main effect of task was also significant task is clearly sensitive to the influence of automatic as-
F
MS MS
1(1,48)= 11.65, e = 8,528.8; 2(1,96)= 318.87, =
e
F
[
sociative priming. In fact, the priming effect was (if any-
1,179.2]: The participants in the go/no-go task had faster thing) a bit greater for the go/no-go task (21.5 msec) than
RTs than did those in the yes/no task. The interaction be- for the yes/no task (13.5 msec), although the interaction
F
tween the two factors was not significant[ 1(1,48) = 2.54, did not approach significancein a combinedanalysiswith
MS
p
F
MS
p
e = 548.5, > .10; 2(1,96) = 2.70,
.10].
e = 1,803.0, >
the two blocks.
It is important to note that the number of false positives
The ANOVA on the error data showed only that the (nonword error rates) was a bit higher in the go/no-go task
F
MS
main effect of task was significant[ 1(1,48)= 30.82,
MS
=
than in the yes/no task (6.5% vs. 6.1%), but this difference
e
F
F
28.81; 2(1,96)= 30.96,
e = 73.99]:The participantsin was nowhere near significance ( < 1). Finally, as in Ex-
theyes/notaskmademore errors thandidthoseinthego/no- periment 1, the error variance in the go/no-go task was not
go task.
Block 2. The ANOVA on the latency data showed that
the main effect of associative relatedness was significant
MS MS
consistently lower than that in the yes/no task.
GENERAL DISCUSSION
F
F
[
1(1,48) = 20.06,
= 413.7; 2(1,96) = 8.55,
=
e
e
4,493.7]: Targets preceded by related primes were re-
The present study has providedevidencethat the go/no-
spondedto faster than those preceded by unrelatedprimes. go LDT is sensitiveto the effects of word frequencyand as-
F
The main effect of task was also significant [ 1(1,48) = sociative priming: The magnitude of these effects was
MS
F
MS
= 8,365.0; 2(1,96) = 114.1, = 2,100.7]: similar to that found with the standard yes/no LDT. More
e
7.49,
The partiecipants in the go/no-go task had faster RTs than important,the go/no-go task appears to enjoy a number of
did those in the yes/no task. The interactionbetween these advantages over the yes/no task (in three out of the four
F
two factors did not approach significance (both s < 1). criteria discussed in the introduction):It offers faster RTs,
The ANOVA on the error data showed only that the more accurate responding, and fewer processing demands
F
main effect of task was significant [ 1(1,48) = 87.17, than the yes/no task. However, contrary to Gordon’s
MS MS
F
e = 13.57; 2(1,96) = 100.6,
e = 41.0]: The partici- (1983) claim, the variability in the data, as measured by
MS
pants in the yes/no task made more errors than did those the
in the go/no-go task.
e, does not seem to be smaller in the go/no-go task.
Finally, we should note that if we analyzeseparately the Speed of Response and Accuracy
two tasks, the error variance for the relatedness effect, in in the Go/No-Go Task
MS
terms of
es, did not differ significantlybetween the two
MS
RTs seem to be substantially faster with the go/no-go
task than with the yes/no task (for both high- and low-
F
tasks. Specifically, the
es were 560.5 versus 536.6 (
,
1
F
first block), 415.0versus 412.3( 1, secondblock), 2,061.5 frequency words), both across participants (Experi-
F
versus 2,396.5( 2, first block), and 3,917.7versus 2,971.0 ments 1– 2) and within participants(Experiment2). More
F
(
2, second block), in the go/no-go and the yes/no task, important, this reduction in RTs in the go/no-go task was
respectively.
accompaniedby a dramatic decrease in the word error rate
(while keeping a similar rate of nonword errors), which
error component
in the yes/no task
Discussion
implies that there is an
The resultsof this experiment are straightforward. First, that is minimized in the go/no-go task. It seems that the
the participants were faster and more accurate when per- standard yes/no task may require some time-consuming
forming thego/no-gotaskthanwhen performing the yes/no decisional processes that are unnecessary to make a cor-
word
task, replicatingthe data from Experiment 1. Interestingly, rect
we found the advantageof the go/no-go task not only with
response.
It is of obviousinterest to analyzewhy participantscom-
word
a between-subjectsmanipulation(as in Experiment1), but mit errors to
stimuli under unlimitedviewing condi-
also with a within-subjectsmanipulation:The participants tions in a lexical decision task. Perhaps the most cited and
who had performed the yes/no task in the first block had detailed framework for understanding the yes/no lexical
much faster latencies (and fewer errors) when they per- decision task is the one proposed by Balota and Chumb-
formed the go/no-go task in the second block [618 vs. ley (1984,1990;see also Balota& Spieler,1999). Although
F
MS
565.5msec, respectively; 1(1,24)= 32.79, e = 2,389.0; Balota and Chumbley apply their model primarily to RT
F
MS
2(1,96)= 114.1, e = 2,100.7;the error rates were 5.3% data, they also analyze the causes of word errors. Specif-
vs. 0.4%, respectively]. Conversely, the participants who ically, within the Balota and Chumbley model, errors to

42
PEREA, ROSA, AND GÓMEZ
word stimuli in the yes/no task can derive from multiple as successive temporal stages (i.e., lexical selection, re-
sources (see Balota & Chumbley, 1984; Balota & Spieler, sponse decision,and response selection).This assumption
1999), as follows.
of successive processing seems compelling, since the
1. Errors could occur because of misperceptions in processes are logically contingent upon each other.10 We
early perceptual analyses.
also assumed that the basic difference between making a
word
response in the two techniques(go/no-go and yes/no
fast guess
2. Errors could occur in terms of a
when a
word has an extremely low familiarity value (e.g., when LDT) seems to be the mental operations involved in the
the word has an unusual spelling: lilac, yacht).
response selectionstage. The response selectionstage was
lexicalselection
3. Errors could occurin the
stage when positedto be independentoffactors suchas word frequency,
the participant has established a time criterion for a since the decision has already been taken in the response
positive/negativeresponse,afterwhich a guesswill be made decision process. As a result, it was predictedthat task and
because he or she is still unsure about whether the letter word frequency should show additive effects, as actually
string is spelledcorrectly or not (see Forster & Veres, 1998; occurred in Experiment 1. However, the data from Hino
Lesch & Pollatsek, 1998).
4. Errors could occurin the
there is a lack of knowledgeabout the appropriatespelling
of a word.
and Lupker (1998, 2000) strongly suggest that word fre-
lexicalselection
decisionalprocess-
stage when quencycould also affect differentially
ing
in the yes/no and the go/no-go tasks; if not, the higher
word error rates in the yes/no task would be left unex-
5. Errors could occur on the basis of momentarily for- plained. Moreover, the long RTs in the yes/no task after a
getting which keys correspond to word and nonword. block with the go/no-go task (Experiment 2)—but not the
In contrast, in the go/no-go task, errors to word stimuli other way around—suggest that the difference in the two
will be caused primarily because of a lack of knowledge tasks could also rely on the response decision process (or
about the appropriate spelling of a word (keep in mind evenearlier), and notjust on the response selectionprocess.
that, unlike the yes/no task, there is no emphasis on gen-
An important source of difference between the yes/no
erating a fast response to the all stimuli). It is worth noting LDT and the go/no-go LDT can be the presence of re-
that the percentage of errors for low-frequency words in sponse competition in the yes/no task (see Grice & Reed,
Experiment 1 is negligible in the go/no-go task (0.4%; it 1992, for evidence of response competitionin a two-letter
was 0.0%for high-frequencywords), muchsmallerthanthat classificationexperimentwith the two procedures). In this
for high-frequency words in the yes/no task (1.7%). The way, it couldbe argued that a factor such as word frequency
1.7% error rate for high-frequencywords in theyes/no task has an effect at a response conflict locus in the yes/no
is probably due to some early perceptualerrors or because LDT, which would delay the response to the target item,
the participantexecuted the wrong response (once the let- since low-frequency words may produce evidence favor-
word
nonword
responses simultaneously
ter string was correctly classified as a word). (Of course, ing the
and the
it is possible that a very small percentage of errors to word (see Abrams & Balota, 1991). In contrast, response-
stimulicouldalsobe causedbytheparticipants’inattention.) conflictis probably absent—or at least minimized—in the
The presence of a 3%–6% false positiveerrors (relative go/no-go task, since nonwords do not require an overt re-
to a negligibleword error rate) in the go/no-go task might sponse.11 Apparently, there is a problem with this account:
word
suggestthat there is a substantial
biasin this task that It may actually predict a larger word frequency effect in
does not exist in the yes/no task (i.e., word and nonword the yes/no task than in the go/no-go task. But appearances
error rates are about equal in the yes/no task, around may be deceiving.This larger word frequency effect in the
equivalent
3%– 6%). However, as we stated earlier, well-motivated yes/no task would occur if the error rates were
participants should not make an error to a word item in a in the two tasks (see the Discussion section of Experi-
go/no-go LDT unless they do not know for sure the ment 1). However, in the usual experimental setup, partic-
spelling of a given word; after all, they have 2 sec to iden- ipants make substantially more word errors—especially
tify the word. As for the presence of occasional nonword for the low-frequency words—in the yes/no LDT than in
similar
errors in the go/no-go task—a rate that was
to that the go/no-go LDT. These trials in the yes/no task would
obtained with the yes/no task—we believe that these er- end up countingas errors and not contributingto the mean
rors are inherent whenever there is pressure for rapid re- latency for low-frequency words. In contrast, in the go/no-
sponding(see Luce, 1986). As Pachella(1974)has pointed go LDT, nonwords do not require a response, and the par-
out, if a participant never makes a false positive (i.e., a ticipantcouldeventuallyrealize that the unfamiliar item is
nonword error), he orshe will neverknowif he or she could a word, producinga slow response (Hino & Lupker, 1998).
be a little faster still without making errors. In any event, In other words, the relatively high error rates for the low-
this is an aspect of the experiment that is, to some degree, frequency words in the yes/no task may have introduced
outside the direct control of the researcher.
“potential distortion of the observed mean reaction time
from the mean reaction time that would be observed had
errors not removed trials that might otherwise have pro-
What is the Locus of the Go/No-Go Advantage?
In the introduction,we assumed a discrete stage model duced long reaction times” (McClelland, 1979, p. 319). If
3
of performance in which the subprocesseswere identified the previous reasoning is correct, the task word fre-

THE YES/NO AND THE GO/NO-GO LDT
43
quency interaction could show not only a larger word fre- Orden model with respect to the additivity/interaction be-
quency effect in the yes/no task (assuming a similar rate tween word frequencyand task? Unfortunately, as we said
of errors in the two tasks), but also additive effects (as in earlier, the Stone and Van Orden model does not make
Experiment1; see also Gordon & Caramazza, 1982;Perea quantitativepredictions,althoughit is likely that the net ef-
et al., 1998) or even a larger word frequency effect in the fect could depend on such factors as the accuracy of the
go/no-go task (as in Hino & Lupker, 1998, 2000). As Luce responses.
(1986) pointed out, Sternberg’s (1969) additive-factor
method focuses exclusively on RTs (i.e., assuming error- Task Demands and Variability
free performance), and it is not simple to generalize this in the Go/No-Go Task
method to the analysis of error data.
The fact that participants have longer latencies in the
One alternative explanation of the present results is in yes/no task after a block with the go/no-go task, whereas
terms of the random-walk framework proposed by Stone participants have much shorter latencies in the go/no-go
and Van Orden (1993; see also Gordon, 1983; Ratcliff, task after a block with the yes/no task (see Experiment 2),
1978; Ratcliff, Gómez, & McKoon, 2001) for the yes/no suggeststhat the go/no-go task makes fewer task demands;
LDT. From a modeler’s perspective (e.g., Smith, 2000),12 after all, the participantswere using the
word
key for
re- responses in the two blocks. Of course, this can also be
sponse would be emitted when the accumulatedactivation viewed as simply a restatementthat the go/no-go task pro-
same
word
the yes/no task has two absorbing barriers (i.e., a
a
of evidence reaches a given positive criterion 1, whereas duces faster RTs and fewer errors than does the yes/no
a
nonword
response would be emitted when the accumu- task. More direct evidenceon this issue would probably be
a
lated activation falls below a given negative criterion 2 ), obtainedwith a dual-taskexperiment.In any event, it is im-
whereas the go/no-go task has only one absorbing barrier portant to note that the go/no-go LDT, but not the yes/no
word
(i.e., a
response would be emitted when the accu- LDT, can be easily applied to experimentswith children or
a
mulated activation exceeds a given positive criterion ). special populations(Yelland, 1993).
There are several basic parameters of a random-walk
However, there is one importantpropertyin which there
processwhen appliedto a yes/no LDT (Stone & Van Orden, is not such an advantage of the go/no-go task relative to
word
nonword
1993): (1) the criterion level of evidence for a
sponse, (2) the criterion level of evidence for a
re- the yes/no task: the variability(noise) in the data. If we use
MS
error variance, as measured by
es, as an estimate of the
response, and (3) the parameters responsiblefor the rate of sensitivityoftheprocedure,thevariabilitytendstobe some-
word (and nonword) accumulation. In a go/no-go LDT, how higher with the go/no-go task than with the yes/no
nonword
the criterion for
responses (or the parameters for task (see also Hino & Lupker, 1998, 2000), although the
nonword accumulation) would be superfluous. Interest- pattern is not entirely consistent. In any case, these results
ingly, the Stone and Van Orden model can be used for pre- show that the claim that the go/no-go LDT produces data
cise (albeit qualitative) predictions concerning responses that are less noisy than those in the yes/no task (Gordon,
to both word and nonword stimuli. At first glance, the fact 1983, 1985) is probably wrong.13 One reason why the
word
that participantsmake faster
go task than in the yes/no task suggests that participants the yes/no task is probably related to the fact that some of
word
responses in the go/no- go/no-go task might not produce less error variance than
might use a lower criterionfor
responsesin thego/no- the variability corresponding to the word stimuli is re-
word
go task. However, moving the
criterion toward ori- moved by the false negativeerrors in the yes/no technique.
gin would also increase the probabilityof a false positive In other words, what would be the (true) error variance if
(i.e., thenonworderror rate; see Stone& Van Orden, 1993). participantswere as accurate in the yes/no task as they are
In the present experiments, the faster responding in the in the go/no-go task? Bear in mind that errors in early per-
go/no-go task than in the yes/no task was also accompa- ceptual analysis would correspond to a false negative re-
nied by a similar nonword error rate in the two techniques, sponse in the yes/no task, whereas they would correspond
which seems to rule out this option. Instead, what would to a relatively long latency in the go/no-go task (Hino &
happen if the participants speed the word accumulation Lupker, 1998), thereby increasing variability in the data.
process in the simpler task (i.e., the go/no-go task, since it Undoubtedly, the fact that errors take out trials that would
may require less cognitive load)? The effect of speeding otherwise contribute to mean RTs is troublesome (Mc-
the word accumulationrate would be a decrease in the RTs Clelland, 1979).Thus, the(possible)reductioninerror vari-
and in the word error rate, relative to the more complex ance in the yes/no LDT (relative to the go/no-go task)
yes/no task, as actually occurred. More important, speed- might just reflect an artifact of the task.
not
ing the word accumulation rate does
modify the prob-
ability of a false positive (see Stone & Van Orden, 1993, Implications for Future Research
Table 6). Interestingly, recent research suggests that par-
The message we want toconveyis thatthe go/no-goLDT
ticipants can strategically increase/decrease the word ac- seems to be a simple, promisingtask—at least for devotees
input gain
cumulation rate (or the so-called
in connection- of the LDT—that offers more accurate responding, faster
ist models)intheprocessofword reading(seeKello& Plaut, RTs, and fewer processing resources than does the yes/no
2000). What are the predictions of the Stone and Van task. It may also provide “more time for word dynamicsto

44
PEREA, ROSA, AND GÓMEZ
run toward coherent states” (Gibbs & Van Orden, 1998, Forster, K. I., & Veres, C. (1998). The prime lexicality effect: Form-
priming as a function of prime awareness, lexical status, and discrim-
ination difficulty. Journal of Experimental Psychology: Learning,
Memory, & Cognition, 24, 498-514.
Gibbs, P., & Van Orden, G. C. (1998). Pathway selection’s utility for
p. 1180). Accordingly, unless one intends to examine the
RT for nonwords, the go/no-go task appears to be an ex-
cellentalternativeto the standardyes/no task. Furthermore,
as McClelland(1979) pointedout, firm conclusionsabout
the locus of the effect of a manipulationmust be restricted
to cases in which error rates are very low, as seems to be
the case with the go/no-go LDT. In fact, more attention
should be devoted to the analysis of word/nonword errors
in the yes/no LDT, since it might compromise the pattern
of results in the RT analysis for a given effect (see Luce,
1986). Undoubtedly, as Grainger and Jacobs (1996;Grain-
ger, Carreiras, & Perea, 2000) have pointed out, theoreti-
cal progress to study word recognitionis highly dependent
on the study of cross-task comparisons. Time will tell
whether the go/no-go procedure applied to the lexical de-
cision task is as fruitful as it seems.
control of word recognition. Journal of Experimental Psychology:
Human Perception & Performance, 24, 1162-1187.
Gordon, B. (1983).Lexical access and lexical decision:Mechanisms of
frequency sensitivity. Journal of Verbal Learning & Verbal Behavior,
22, 24-44.
Gordon, B. (1985). Subjective frequency and the lexical decision la-
tency function: Implications for mechanisms of lexical access. Jour-
nal of Memory & Language, 24, 631-645.
Gordon, B., & Caramazza,A. (1982). Lexical decision for open- and
closed-class words: Failure to replicate differential frequency sensi-
tivity. Brain & Language, 15, 143-160.
Grainger,J., Carreiras,M., & Perea, M. (2000).A new techniquefor
visual word recognitionresearch: The luminance increment paradigm.
Current Psychology Letters, 1, 107-116.
Grainger, J., & Jacobs, A. M. (1996). Orthographic processing in vi-
sual word recognition: A multiple read-out model. Psychological Re-
view, 103, 518-565.
REFERENCES
Grainger, J., O’Regan, J. K., Jacobs, A. M., & Seguí, J. (1989). On
therole of competingword unitsin visual word recognition:The neigh-
borhood frequency effect. Perception & Psychophysics, 45, 189-195.
Grice,G. R., & Reed, J. M. (1992).What makes targets redundant? Per-
ception & Psychophysics, 51, 437-442.
Hino, Y., & Lupker, S. J. (1996). Effects of polysemy in lexical deci-
sion and naming: An alternative to lexical access accounts. Journal of
Experimental Psychology: Human Perception & Performance, 22,
1331-1356.
Hino, Y., & Lupker, S. J. (1998). The effects of word frequency for
Japanese Kana and Kanji words in naming and lexical decision: Can
the dual-route model save the lexical-selection account? Journal of
Experimental Psychology: Human Perception & Performance, 24,
1431-1453.
Hino, Y., & Lupker, S. J. (2000). The effects of word frequency and
spelling-to-soundregularity in naming with and without lexical deci-
sion. Journal of Experimental Psychology: Human Perception & Per-
formance, 26, 166-183.
Abrams, R. A., & Balota, D. A. (1991). Mental chronometry: Beyond
reaction time. Psychological Science, 2, 153-157.
Alameda, J. R., & Cuetos, F. (1995). Diccionario de frecuencia de las
unidadeslingüísticas del castellano [Dictionary of word frequency in
Spanish]. Oviedo: Servicio de publicaciones de la Universidad de
Oviedo.
Algarabel,S., Sanmartín, J., García, J., & Espert, R. (1986). Nor-
mas de asociación libre de 400 sustantivos pertenecientes a BASPAL
[Norms of free association for 400 nouns in the BASPAL database].
Informes del Departamento de Psicología Experimental, Universitat
de València, Spain.
Balota, D. A., & Chumbley, J. I. (1984). Are lexical decisions a good
measure of lexical access? The role of word frequency in the neglected
decision stage. Journal of Experimental Psychology: Human Percep-
tion & Performance, 10, 340-357.
Balota, D. A., & Chumbley, J. I. (1990). Where are the effects of fre-
quency in visual word recognition tasks? Right where we said they
were! Comment on Monsell, Doyle, and Haggard (1989). Journal of
Experimental Psychology: General, 119, 231-237.
Kello, C. T., & Plaut, D. C. (2000). Strategic control in word naming:
Evidence from speeded responding in the tempo naming task. Jour-
nal of Experimental Psychology: Learning,Memory, & Cognition, 26,
719-750.
Balota, D. A., & Spieler, D. H. (1999). Word frequency, repetition,
and lexicality effects in word recognition. Journal of Experimental Lane, D. M., & Ashby, B. (1987). PsychLib: A library of machine lan-
Psychology: General, 128, 32-55.
guage routines for controlling psychology experiments on the Apple
Macintosh computer. Behavior Research Methods, Instruments, &
Computers, 19, 246-248.
Lesch, M. F., & Pollatsek,A. (1998).Evidence fortheuse of assembled
phonologyin accessing the meaning of printed words. Journal of Ex-
perimental Psychology:Learning, Memory, & Cognition, 24, 573-592.
Chiarello, C., Nuding, S., & Pollock, A. (1988). Lexical decision
and naming asymmetries: Influence of response selection and re-
sponse bias. Brain & Language, 34, 302-314.
Collins, A. M., & Loftus, E. F. A. (1975). A spreading-activationthe-
ory of semantic processing. Psychological Review, 82, 407-428.
Davis, C., Castles, A., & Iakovidis, E. (1998). Masked homophone Luce, R. D. (1986). Response times. New York:Oxford UniversityPress.
and pseudohomophone priming in children and adults. Language &
Cognitive Processes, 13, 625-651.
de Groot, A. M. B. (1984). Primed lexical decisions: Combined effects
McClelland, J. L. (1979). On the time relations of mental processes:
An examination of systems of processes in cascade. PsychologicalRe-
view, 86, 287-330.
of the proportion of related prime–target pairs and the stimulus-onset McClelland, J. L., & Rumelhart, D. E. (1981). An interactive acti-
asynchrony of prime and target. Quarterly Journal of Experimental
Psychology, 36A, 253-280.
vation model of context effects in letter perception: 1. An account of
basic findings. Psychological Review, 88, 375-407.
den Heyer, K., Sullivan, A., & McPherson, C. (1987). Mediated McNamara, T. P., & Altarriba, J. (1988). Depth of spreading activa-
priming in a single-response lexical decision task. Unpublished man-
uscript.
tion revisited: Semantic mediated priming occurs in lexical decisions.
Journal of Memory & Language, 27, 545-559.
Donders, F. C. (1969). Over de snelheid van psychische processen [On
the speed of psychological processes]. Acta Psychologica, 30, 412-
431. (Original work published 1868)
Forster, K. I. (1976). Accessing the mental lexicon. In R. J. Wales &
E. W. Walker (Eds.), New approachestolanguagemechanisms(pp. 257-
287). Amsterdam: North-Holland.
Measso, G., & Zaidel, E. (1990). Effect of response programming on
hemispheric differences in lexical decision. Neuropsychologia, 28,
635-646.
Monsell, S., Doyle, M. C., & Haggard, P. N. (1989). Effects of fre-
quency on visual word recognition tasks: Where are they? Journal of
Experimental Psychology: General, 118, 43-71.
Forster, K. I., & Shen, D. (1996). No enemies in the neighborhood: Neely, J. H. (1991). Semantic priming effects in visual word recogni-
Absence ofinhibitoryeffects in lexical decision and categorization.Jour-
nalof Experimental Psychology:Learning,Memory, & Cognition, 22,
696-713.
tion: A selective review of current findingsand theories. In D. Besner
and G. W. Humphreys (Eds.), Basic processes in reading:Visual word
recognition (pp. 264-336). Hillsdale, NJ: Erlbaum.

THE YES/NO AND THE GO/NO-GO LDT
45
Paap, K. R., McDonald, J. E., Schvaneveldt, R. W., & Noel, R. W.
(1987).Frequencyand pronounceabilityin visuallypresented naming and
lexical decision tasks. In M.Coltheart (Ed.),AttentionandPerformance
3. The first step is to compute the quantiles for each participant and
condition.In our study, we used 5% quantiles (i.e., 19 quantiles for each
participant and condition).Second, each quantile is averaged across par-
XII: The psychology of reading (pp. 221-243). Hove, U.K.: Erlbaum. ticipants to give group quantiles. Third, group distribution histograms
Paap, K. R.,Newsome, S. L., McDonald, J. E., & Schvaneveldt,R. W.
(Vincent histograms) can be easily constructed by plotting quantiles on
(1982). An activation-verification model for letter and word recogni- the abscissa and then constructingrectangles between adjacent quantiles
tion: The word superiority effect. Psychological Review, 89, 573-594. so that all the rectangles have equal areas (each rectangle has an area of
Pachella, R. G. (1974). The interpretation of reaction time in informa-
1/q, where q is the number of quantiles). Interestingly, Vincent his-
tion processing research. In B. Kantowitz (Ed.), Human information tograms produce average distributions with the same shape as compo-
processing: Tutorials in performance and cognition (pp. 41-81). Po- nent distributions (see Ratcliff, 1979).
tomac, MD: Erlbaum.
4. Although these words are not very unfamiliar, perhaps the fact that
billón has a similarly spelled word of much higher frequency (millón,
the Spanish for million; see, e.g., Perea & Pollatsek, 1998) and that de-
hesa has an unusual syllabic structure (i.e., in Spanish it is very unusual
to find a mute “h” in the middle of a word), combined with the time pres-
sure of the LDT, caused participants to frequently consider them as non-
words.
Perea, M., Fernández, L., & Carreiras, M. (1998, December). Efec-
tos secuenciales en la tarea de decisión léxica [Sequential effects in
the lexical decision task]. Paper presented at the II Conference of the
Spanish Experimental Psychology Society, Granada, Spain.
Perea, M., & Pollatsek, A. (1998). The effects of neighborhoodfre-
quency in reading and lexical decision. Journal of Experimental Psy-
chology: Human Perception & Performance, 24, 767-779.
5. We should note that the Hino and Lupker experiments were carried
Pollatsek, A., Perea,M., & Binder, K. (1999).The effects of “neigh- out in different languages (i.e., Japanese in the 1998 report and English
borhoodsize” in reading and lexical decision. Journal of Experimental in the 2000 report) than was the present experiment (i.e., Spanish). Per-
Psychology: Human Perception & Performance, 25, 1142-1158.
Pollatsek, A., & Well, A. (1995). On the use of counterbalanced de-
haps this fact might have created some differences between experiments
(especially the experiments in Japanese with respect to the experiments
signs in cognitive research: A suggestion for a better and more pow- with alphabetical languages), although we cannot see how this differ-
erful analysis. Journal of Experimental Psychology: Learning, Mem-
ory, & Cognition, 21, 785-794.
ence can account for the seemingly contradictory results.
6. We thank Yasushi Hino for this suggestion.
Ratcliff, R. (1978). A theory of memory retrieval. Psychological Re-
view, 85, 59-108.
Ratcliff, R. (1979). Group reaction time distributions and an analysis
of distribution statistics. Psychological Bulletin, 86, 446-461.
7. We thank Ken Paap for this sharp observation.
8. We should note that Neely (1991) cited an unpublished study of
den Heyer, Sullivan, and McPherson (1987), in which they found medi-
ated priming (e.g., lion–STRIPES; via tiger) with the go/no-go LDT, but
Ratcliff, R.,Gómez, P.,& McKoon,G. (2001).Adiffusionmodel account not with the standard yes/no LDT.
of the lexical decision task. Manuscript submitted for publication.
Schilling, H. E. H., Rayner, K., & Chumbley, J. I. (1998). Compar-
ing naming, lexical decision, and eye fixation times: Word frequency
9. We thank an anonymous reviewer for this suggestion.
10. Bear in mind that it is not simple to make predictionsfrom a model
in which the transfer of information between processes is taking place all
effects and individualdifferences. Memory & Cognition, 26, 1270-1281. the time (a cascade model; see, e.g., McClelland, 1979),since it assumes
Smith, P. L. (2000).Stochastic dynamic models of response time and ac-
curacy: A foundational primer. Journal of Mathematical Psychology,
44, 408-463.
very low error rates, which is not usually the case with low-frequency
words in the yes/no task. Furthermore, as Ulrich, Mattes, and Miller
(1999) pointed out, the serial processing assumption may also accom-
Sternberg, S. (1969). The discovery of processing stages: Extensions modate top-down processes, as well as bottom-up processes.
of Donders’ method. Acta Psychologica, 30, 276-315.
11. As Yasushi Hino has suggested, an interference may also occur at
Stone, G. O., & Van Orden, G. C. (1993). Strategic control of pro- the motor system, independent of any internal processing. In the yes/no
cessing in visual word recognition. Journal of Experimental Psychol-
ogy: Human Perception & Performance, 19, 744-774.
Ulrich, R., Mattes, S., & Miller,J. (1999). Donders’s assumption of
pure insertion: An evaluation on the basis of response dynamics. Acta
Psychologica, 102, 43-75.
Westall, R., Perkey, M. N., & Chute, D. L. (1986). Accurate mil-
lisecond timing on Apple’s Macintosh usingDrexler’s MilliTimer. Be-
havior Research Methods, Instruments, & Computers, 18, 307-311.
Yelland, G. (1993, November). Is there a place for phonological
processes in reading development? Paper presented at the 9th Aus-
tralian Language and Speech Conference, Sydney, Australia.
task, participants are required to press one of two keys, and thereby two
distinct motors programs have to be executed, depending on the lexical
status of the stimulus. In fact, it is quite usual for participants to say, after
a yes/no lexical decision experiment, that they have made some mistakes
because they automatically pressed a wrong key, especially after several
word (or nonword) trials in a row.
12. We thank Gregory Ashby for this suggestion.
13. In order to examine whether the go/no-go task could produce less
noisy data than the yes/no task, we also analyzed how the two tasks dif-
fered with respect to the within-cell standard deviation (averaged across
participants). Note that the task differences might be more directly re-
lated to within-subjects (trial-by-trial) variability than the MSes. How-
ever, the statistical analyses on these variability data also failed to find
any differences across tasks.
NOTES
1. We thank an anonymous reviewer for suggesting this scheme.
2. However, Chiarello, Nuding, and Pollock (1988) and Measso and
Zaidel (1990) did not find a reliable effect of task in the error data when
the stimuli were presented very briefly (100 msec in the Chiarello et al.
experiment and 80 msec in the Measso & Zaidel experiment) in the
left/right visual field.
(Manuscript received January 23, 2001;
revision accepted for publication October 29, 2001.)