BR A IN RE S E A RCH 1 3 18 ( 20 1 0 ) 8 7 9 5

available at www.sciencedirect.com

www.elsevier.com/locate/brainres

Research Report

Judgments of learning do not reduce to memory encoding

operations: Event-related potential evidence for distinct
metacognitive processes
Ida-Maria Skavhaug a,, Edward L. Wilding b , David I. Donaldson a
a

The Psychological Imaging Laboratory, Department of Psychology, University of Stirling, Stirling FK9 4LA, UK
Cardiff University, Cardiff University Brain Research Imaging Centre (CUBRIC), School of Psychology, Cardiff, Wales, UK

A R T I C LE I N FO

AB S T R A C T

Article history:

To examine how judgments of learning (JOLs) are made, we used event-related potentials

Accepted 19 November 2009

(ERPs) to compare neural correlates of JOLs and successful memory encoding. Participants

Available online 5 December 2009

saw word pairs, and for each made a JOL indicating how confident they were that they would
remember the pairing on a later cued recall task. ERPs were recorded while JOLs were made

Keywords:

and were separated according to whether items were: (i) remembered or forgotten on the

Memory Encoding

subsequent test, and (ii) rated likely or unlikely to be remembered. An early positive-going

Cued Recall

ERP effect was associated with both of these comparisons, whereas a later negative-going

Metacognition

effect was present only in the separation based upon JOL ratings. ERP data therefore indicate

Judgments of Learning

that JOLs do not reduce to encoding processes that predict the accuracy of memory

Event-Related Potentials

judgments.
2009 Elsevier B.V. All rights reserved.

1.

Introduction

Judgments of learning (JOLs) are metacognitive assessments

of memorability. They are operationalized as estimates of the
likelihood that studied material will be subsequently remembered (cf. Nelson and Narens, 1990). In most JOL studies,
behavioural manipulations have been employed to determine
their functional basis. In this experiment, we used brain
imaging (event-related potentials ERPs) to investigate the
similarities and differences between the cognitive processes
that support JOLs, and those that support successful memory
encoding.
Both neuropsychological (Vilkki et al., 1999; Kennedy and
Yorkston, 2000) and psychopharmacological data (Dunlosky
et al., 1998; Izaute and Bacon, 2005; Merritt et al., 2005) have
indicated dissociations between JOLs and memory. Together,

these findings suggest that some processes supporting JOLs

operate independently of memory encoding. Consistent with
this view, Kao, Davis and Gabrieli (2005) reported functional
Magnetic Resonance Imaging (fMRI) results that suggest JOLs
are based on a combination of shared and independent neural
circuitry. In keeping with previous memory findings (e.g.,
Wagner et al., 1998; Qin et al., 2007; for reviews see Fernandez
and Tendolkar, 2006; Diana et al., 2007), study items that were
subsequently remembered rather than forgotten were associated with increased activity in the medial temporal lobes (a
subsequent memory effect; Paller et al., 1987). More importantly,
whilst some brain regions (including left lateral prefrontal
cortex) were equally active for successful encoding and JOLs,
other regions (including left ventro-medial and dorso-medial
prefrontal cortex) were more active for JOLs than for successful memory encoding.

Corresponding author. Fax: + 44 1786 467641.

E-mail address: is10@stir.ac.uk (I. Skavhaug).
0006-8993/$ see front matter 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.brainres.2009.11.047

88

BR A IN RE S EA RCH 1 3 18 ( 20 1 0 ) 8 7 95

From a theoretical perspective, one interpretation of the

fMRI data is that JOLs engage additional metacognitive
assessments that operate downstream of successful encoding
operations. Given the poor temporal resolution of the data,
current fMRI results are unable to speak to this possibility.
This possibility can, however, be assessed by employing a
real-time measure of neural activity, the prediction being that
neural activity differentiating JOLs from successful memory
encoding will occur after (or extend beyond) activity that is
shared by these two kinds of information processing operation. Real-time measures of neural activity thus offer the
opportunity to contribute to theoretical accounts of JOLs by
delineating the temporal relationships between the different
classes of cognitive processes that may be engaged.
In this study, ERPs were acquired whilst JOLs were made.
Participants studied a series of word pairs and were asked how
likely they would be to remember the second word, if
presented with the first word, during a later test. ERPs were
acquired during the study phase and separated according to
whether; (i) the second word was or was not recalled
subsequently, and (ii) the study pair elicited a high or low
JOL. These contrasts permit assessment of the temporal and
functional correspondences between the neural signatures of
successful memory encoding and JOLs.

2.

Results

2.1.

Behavioral data

2.1.1.

Study

trials. Figs. 2 (subsequent memory) and 3 (JOLs) show the

waveforms for these response categories at representative
electrodes.
Fig. 4 shows the scalp distributions of the subsequent
memory and JOL effects (recalled minus missed, high minus
low JOL) during early (550 to 1000 ms) and late (1300 to
1900 ms) time windows. Both the subsequent memory and JOL
distributions have a similar widespread positivity in the
earlier time window, although the subsequent memory effect
extends to a greater degree to anterior locations than the JOL
effect. During the later time window, however, the two effects
differ; the JOL contrast reveals a strong left hemisphere
negative-going effect which is not present in the subsequent
memory contrast.
For each contrast, data were analyzed across 6 rows of
electrodes (Fig. 5), from frontal to occipital scalp, using ANOVA
with factors of condition (recalled versus missed, high versus

Participants had a preference for assigning intermediate JOLs

(Fig. 1a). ANOVA on response rates revealed a main effect of
JOL [F(4,72) = 7.0, p < 0.001], with the accompanying quadratic
trend [F(1,18) = 18.6, p < 0.001] reflecting the concentration of
responses towards the middle of the scale.

2.1.2.

Test

Mean numbers of Hits and False Alarms (old judgments to old

and new items) were 77% and 12%. In keeping with the
correspondence between response accuracy and study JOL
shown in Fig. 1b, ANOVA comparing recall rates revealed a
main effect of JOL [F(4,72) = 26.1, p < 0.001] and a linear trend
[F(1,18) = 52.8, p < 0.001] unmodified by any significant higher
order trends. Performance was also examined using GoodmanKruskal gamma, which is an ordinal correlation measure
of JOL accuracy (Nelson, 1984). Gamma coefficients range
between 1 and +1 (perfect negative and positive correlations). The mean gamma score of 0.29 (SD = 0.14) was
significantly above zero [t(19) = 9.38, p < 0.001].

2.2.

Brain imaging data

The initial analyses comprised separate assessments of the

study phase ERPs. First, subsequent memory effects: study
ERPs separated according to memory accuracy at test
(recalled versus missed). Second, JOL effects: ERPs associated
with high or low JOLs. It is not possible to contrast directly
these two effects as they contain overlapping subsets of

Fig. 1 Performance measures averaged across participants

(mean and standard error). (a) Bars represent the division of
responses across JOL level. (b) Bars represent the mean
percentage of correctly recalled items across JOL level.

BR A IN RE S E A RCH 1 3 18 ( 20 1 0 ) 8 7 9 5

89

Fig. 2 Grand average ERPs for subsequently missed items (bold lines) and subsequently recalled items (thin lines). ERP
waveforms are shown for midline frontal (FZ), midline parietal (PZ) and left and right central electrodes (C3 and C4).

low JOL), hemisphere (left, right) and electrode site (superior,

mid, inferior). The outcomes of these analyses for early and
late time windows are shown in Table 1.

2.2.1.

electrodes. No statistical support for this impression was

obtained; ANOVA with factors of condition (recall, miss), and
electrode site (FZ, FCZ, CZ, CPZ, PZ, POZ), revealed only a main
effect of condition [F(1.0,19.0) = 12.34, p < 0.05].

Subsequent memory effects: Table 1a

In the 550- to 1000-ms time window the ANOVAs revealed

significant interactions between condition and site for all six
rows, reflecting the fact that the subsequent memory effect is
a broadly distributed greater relative positivity for remembered than for forgotten items that is largest at sites closest to
the midline. From 1300 to 1900 ms ANOVA revealed only main
effects of condition from fronto-central to parietal electrode
rows. As can be seen in Fig. 2, these effects reflect primarily a
(weakened) continuation of the effect in the preceding epoch.
Given the apparent posterior focus of the early subsequent
memory effect (see Fig. 4), additional analyses were carried out
to investigate whether it was significantly larger at posterior

2.2.2.

JOL effects: Table 1b

From 550 to 1000 ms the ANOVAs revealed interactions between

condition and site from fronto-central to parieto-occipital
electrode rows. As for the subsequent memory analyses in
this time window, these outcomes reflect a positive-going
modulation (high JOL > low JOL) that is largest at sites closest to
the midline. Importantly, the similarities between statistical
outcomes for the two effects do not continue in the later time
window. From 1300 to 1900 ms ANOVA of the JOL effects
revealed significant interactions between condition and hemisphere from frontal to centro-parietal electrode rows. As Figs. 3
and 4 illustrate, ERPs elicited by items attracting high and low

90

BR A IN RE S EA RCH 1 3 18 ( 20 1 0 ) 8 7 95

Fig. 3 Grand average ERPs for items assigned a low JOL (bold lines) and a high JOL (thin lines). Sites as for Fig. 2.
JOLs differ primarily at left hemisphere sites, where the high JOL
ERPs are markedly more negative-going.
As for the subsequent memory contrast, additional analyses were carried out to investigate whether the apparent
posterior maximum of the JOL effect was reliable during the
early time window, using ANOVA with factors of condition
(high JOL, low JOL) and electrode site (FZ, FCZ, CZ, CPZ, PZ,
POZ). A main effect of condition [F(1.0,19.0) = 10.79, p < 0.05]
was moderated by a condition and electrode site interaction
[F(1.73,32.89) = 7.34, p < 0.05], reflecting the centro-posterior
maximum of the early effect.

2.2.3.

Analyses of scalp distributions

Analyses of scalp distributions were conducted to determine

whether the subsequent memory and JOL ERP effects change
over time. Changes between the scalp distributions of ERP
effects license the claim that not entirely the same cognitive
processes were engaged. These analyses were performed on
difference scores (mean amplitudes for recalled minus

missed, and high minus low JOLs), and these were first
normalized using the Max/Min method (MacCarthy and
Wood, 1985) to avoid confounding changes between the
sizes of ERP effects with changes in their scalp distributions.
The scalp distribution analyses were conducted using
ANOVA with factors of time window (early, late), location
(Frontal, Fronto-Central, Central, Centro-Parietal, Parietal,
Parieto-Occipital), hemisphere (left, right) and site (superior,
mid, inferior). For the subsequent memory effects, the ANOVA
revealed no significant change in distributions over time (all
Fs < 2.4). For the JOL effects, ANOVA revealed significant
interactions between time and hemisphere [F(1.0,19.0) = 7.3,
p < 0.05], time and site [F(1.2,22.0) = 15.1, p < 0.001], as well as
between time, location and hemisphere [F(1.8,34.0) = 3.6,
p < 0.05], and time, location and site [F(3.7,70.5) = 2.9, p < 0.05].
Figs. 24 illustrate that only the scalp distributions of the JOL
effects change markedly over time, and these statistical
outcomes confirm this impression. The reliable interactions
that were revealed in the JOL analyses indicate that the early

BR A IN RE S E A RCH 1 3 18 ( 20 1 0 ) 8 7 9 5

Fig. 4 Topographic maps illustrating the scalp distributions

of the subsequent memory effects (subsequent recall minus
subsequent miss) and JOL effects (high JOL minus low JOL)
over the 5501000 ms and 13001900 ms time windows. The
front of the head is at the top of each map and the scale bar
represents the size of the effect in V.

and late JOL effects are generated by at least partially nonoverlapping sets of neural generators, and therefore index
distinct classes of cognitive operations.

3.

Discussion

We investigated the relationship between judgments of

learning (JOLs) and successful memory encoding using
behavioral and brain imaging measures. The behavioral
results showed a clear relationship between memory encoding and JOLs and the brain imaging results provided novel
insights into this relationship not available via the behavior
alone. These insights follow from two critical contrasts
between ERPs acquired during the experiment study phase;

Fig. 5 Map (front of the head at the top) illustrating the

choice of electrodes.

91

ERPs elicited by studied items attracting correct or incorrect

judgments on the subsequent memory test and ERPs elicited
by items attracting high or low JOLs at study. The ERP data
were analyzed for two time windows: early (5501000 ms) and
late (13001900 ms). Findings for each window are discussed in
turn.
In both contrasts, reliable and markedly similar effects
were evident from 550 to 1000 ms. If this early positive going
ERP effect indexes successful memory encoding (Paller et al.,
1987), then the presence of this effect in the high/low JOL
contrast could be taken to suggest that JOLs can be based upon
operations that support successful encoding. The fact that
overlapping subsets of trials contributed to the subsequent
memory and JOL effects means, however, that the effect could
reflect an encoding process that is only found in the JOL
contrast because of trial overlap, or vice versa. If this was the
case, because trial overlap is not exact, the effect would be
considerably larger in the condition that was driving it. Thus,
the fact that the early effects are of equivalent magnitude for
encoding and JOLs therefore provides some evidence for the
engagement of common processes during successful encoding
and JOL decisions. By this view the presence of the effect in
both contrasts is not an artifact of trial overlap, but a real
reflection of shared processing operations.
Critically, from 1300 to 1900 ms, the subsequent memory
and JOL effects diverged markedly. During the later time
window a negative going ERP effect was evident for JOLs,
but not memory encoding. The later effect differs from the
earlier effect in terms of timing, location, polarity and
eliciting conditions, strongly suggesting that the two effects
are functionally distinct. In addition, the analyses of scalp
distributions revealed that the neural activity predicting
JOLs differed reliably across time windows, indicating
distinct neural origins (hence functional significances) for
the early shared and late JOL-specific processes. Our
findings therefore converge with fMRI, neuropsychological
and behavioural findings in neurologically intact individuals, all of which indicate that the processes supporting
JOLs do not reduce to the processes that support successful
encoding.
The new insight that is not available from preceding
studies is that the JOL-specific neural activity followed activity
shared by JOLs and successful encoding operations. This
observation is consistent with the view that JOLs are, at least
partly, a consequence of additional cognitive operations that
occur downstream of those supporting successful encoding.
The time course of the JOL-specific effects revealed here also
explains why no comparable neural correlates were found in a
previous ERP study of JOLs. Using faces as the stimuli for
which JOLs were required, Sommer et al. (1995) observed
comparable neural signatures for JOL and subsequent memory
effects. They only examined the ERPs up to 1000 ms poststimulus, however, thereby precluding identification of lateonsetting JOL effects comparable to the one revealed in this
study.
What is the likely functional significance of the JOL-specific
effect identified here? An important aspect of this finding is that
the JOL effect emerged from a contrast between items attracting
either high or low JOLs. One consequence of this finding is that
the effect cannot reflect a general assessment process that is

92

Table 1 Outcomes of the analysis of the ERP effects (F = Frontal; FC = Fronto-Central; C = Central; CP = Centro-Parietal; P = Parietal; PO = Parieto-Occipital). (a) Results of ANOVA
examining the subsequent memory effects for the 5501000 ms and 13001900 ms time windows. (b) Results of ANOVA examining the JOL effects for the 5501000 ms and
13001900 ms time windows.
F

FC

CP

PO

F(1.0,19.0) = 9.2; p < 0.01

F(1.0,19.0) = 9.7; p < 0.01

F(1.0,19.0) = 7.5; p < 0.05

F(1.5,28.7) = 19.1; p < 0.01

F(1.2,23.1) = 5.7; p < 0.05

F(1.1,20.6) = 6.4; p < 0.05

F(1.1,21.0) = 4.6; p < 0.05

F(1.0,19.0) = 5.4; p < 0.05

F(1.0,19.0) = 9.5; p < 0.01

F(1.0,19.0) = 12.2; p < 0.01

F(1.0,19.0) = 6.2; p < 0.05

F(1.0,19.0) = 9.4; p < 0.01

F(1.0,19.0) = 17.5; p < 0.01

F(1.0,19.0) = 12.7; p < 0.01

F(1.1,21.5) = 11.0; p < 0.01

F(1.3,23.0) = 12.1; p < 0.01

(a) Recall/Miss

F(1.5,29.1) = 10.4; p < 0.01

F(1.1,21.7) = 12.1; p < 0.05

(b) High JOL/Low JOL

5501000 ms
Condition
Condition Hemisphere
Condition Site
Condition Hemisphere Site
1300 1900 ms
Condition
Condition Hemisphere
Condition Site
Condition Hemisphere Site

F(1.0,19.0) = 8.6; p < 0.01

F(1.3,24.9) = 7.9; p < 0.01

F(1.2,22.0) = 4.8; p < 0.05

F(1.1,20.7) = 8.5; p < 0.01

F(1.0,19.0) = 5.8; p < 0.05

F(1.0,19.0) = 10.0; p < 0.01

F(1.0,19.0) = 8.0; p < 0.05

BR A IN RE S EA RCH 1 3 18 ( 20 1 0 ) 8 7 95

5501000 ms
Condition
Condition Hemisphere
Condition Site
Condition Hemisphere Site
13001900 ms
Condition
Condition Hemisphere
Condition Site
Condition Hemisphere Site

BR A IN RE S E A RCH 1 3 18 ( 20 1 0 ) 8 7 9 5

engaged equally whenever a JOL is requiredbecause a JOL was

required in all cases any processing associated with the
requirement to make a JOL would be invisible in this comparison. The effect more likely reflects some aspect of the
assessment process that is sensitive to the amount or kind of
information that is accessed. For example, low JOL rating may be
produced when no information is accessed, and high JOL ratings
when some information is accessed. By this view the electrical
record indexes a general assessment process that is differentially engaged by certain contents, giving rise to the differences
between activities across conditions.
Strong claims about the functional significance of the late
effect are constrained by the fact that the only electrophysiological information comes from the result of a binary
(high vs. low JOL) contrast. As a result, whether the effect is
driven primarily by high or low JOLs, or reflects a process that
varies parametrically with JOLs, remains to be seen. The
presence of the effect, however, indicates that participants
have an imperfect understanding of some of the factors that
influence memorability. That is, at least on some trials,
participants assigned importance to factors that do not in
fact contribute substantively to effective encoding. One
possibility is that the JOL-specific effect reflects the earlier
recovery of a certain kind of content for word pairs that were
subsequently given either high or low JOLs. This account can
explain the current data if the fluency with which information
becomes available influences JOLs but is a poor predictor of
the subsequent memorability of the critical stimuli.
Irrespective of the accuracy of the functional considerations,
however, the behavioral and ERP findings in this study (i) support
claims that the processes differentiating high and low JOLs do
not reduce to those that support successful memory encoding,
and (ii) demonstrate for the first time that JOL-specific processes
operate downstream of those that are shared between encoding
operations and judgments about the subsequent memorability
of studied material. In short, using the high-temporal resolution
of real-time imaging, we provide direct evidence of a distinct
neural correlate that may reflect metacognitive processing
which is engaged when judgments of learning are made.

4.

Experimental procedures

4.1.

Participants

These were 24 students at the University of Stirling, all of

whom were right-handed native English speakers with no
known neurological impairments. Participants provided informed consent, as approved by the Department of Psychology
Ethics Committee at the University of Stirling. They were
given financial compensation and course credits where
applicable. Three participants were excluded due to equipment failure or excessive EEG artifacts, and one due to poor
performance. The remaining 20 participants (12 female) had a
mean age of 22 (range: 1730).

4.2.

Stimuli

Four hundred and twenty word pairs were used as stimuli

during ERP recording (see Appendix 1 for a sample). The pairs

93

had a mean forward associative strength of 0.42 (sd = 0.16) and

a mean backward associative strength of 0.02 (sd = 0.02)
(according to the norms of Nelson et al., 1998). Twelve
additional word pairs were used for practice.

4.3.

Procedure

The experiment was conducted on a PC, using Eprime

(Psychology Software Tools; www.pstnet.com) and a compatible response box (Psychology Software Tools). There was
one study session during which JOLs were made, followed by
a memory test session (thereby avoiding confounds associated with JOLs made during multiple study-test blocks; see
Koriat et al., 2002; Kelemen et al., 2007). The study phase
comprised 280 trials, each involving a word pair selected
randomly from the initial 420 pairs. The first word of each of
the 280 pairs was re-presented at test, along with 140 new
words. Word presentation order was determined randomly
for each participant. Breaks were at 70 trial intervals, and an
initial practice session familiarized participants with the
procedures.
Each study trial began with a white fixation cross
presented in the centre of a blue screen for 1000 ms. A
word pair was then presented, one word above and one
below the central fixation point. Words were displayed in
white against a blue background, in uppercase 18-point
Courier New font. After 3000 ms a blue screen appeared,
replaced after 500 ms by the prompt PROBABILITY TO
RECALL. This was the instruction for participants to
indicate via button press how likely they would be to recall
the second word successfully if presented with the first word
on a subsequent test. Participants were asked to respond on
a 5-point scale: 1 (definitely forget), 2 (probably forget), 3
(unsure), 4 (probably remember), 5 (definitely remember).
The need to make use of the full scale throughout the
experiment was emphasized. Participants were asked to try
to remember the word pairs, but no specific memorization
instructions were given. After each JOL was made, a blue
screen was presented for 1000 ms before the next trial
started.
Each test trial also began with presentation of a white
fixation cross in the centre of a blue screen for 1000 ms. A
single word was then presented centrally on the blue screen,
displayed in white uppercase 18-point Courier New font. The
word remained on the screen for 1500 ms and was followed
by a blue screen for 2500 ms, providing a 4 s response
window. Participants were instructed to press buttons 1 or 5
depending on whether the word was old (presented at study)
or new (not presented). If a new response was made (or no
response occurred within the 4 s window) the trial
terminated. Following an old response the prompt CAN
RECALL? was presented. Participants were asked to press
buttons 1 or 5 to indicate whether they could or could not
remember the word's partner at study. The prompt
remained visible until a response was made. If the
participant responded no the current trial was terminated.
Following a yes response the prompt RECALL WORD
appeared, and participants were instructed to verbally
complete the word pair. After recording the response, the
experimenter initiated the next trial.

94
4.4.

BR A IN RE S EA RCH 1 3 18 ( 20 1 0 ) 8 7 95

EEG recording

EEG was recorded from 62 silver/silver chloride electrodes

located in an elastic cap (Quick-Cap, Neuromedical Supplies)
in accordance with an extended version of Jasper's (1958)
international 10/20 system (FP1, FPZ, FP2, AF3, AF4, F7, F5, F3,
F1, FZ, F2, F4, F6, F8, FT7, FC5, FC3, FC1, FCZ, FC2, FC4, FC6, FT8,
T7, C5, C3, C1, CZ, C2, C4, C6, T8, TP7, CP5, CP3, CP1, CPZ, CP2,
CP4, CP6, TP8, P7, P5, P3, P1, PZ, P2, P4, P6, P8, PO7, PO5, PO3,
POZ, PO4, PO6, PO8, CB1, O1, OZ, O2, CB2). Additional
electrodes were placed on the mastoids, above and below
the left eye (Vertical EOG), and on the outer canthi (Horizontal
EOG) to monitor eye movements. Electrodes were referenced
to one positioned between CZ and CPZ during recording, then
re-referenced off-line to create an averaged mastoid reference.
Electrode impedances were below 5 k at the start of the
recording session. Recordings were made using a Synamps2
amplifier and Neuroscan 4.3 Acquire software (Quick-Cap,
Neuromedical Supplies). Signals were amplified with a gain of
2010, bandpass filtered (0.140 Hz) and digitized at 250 Hz.

4.5.

identify the novel JOL effects seen in the current data. In the
results section only effects involving condition are reported,
as the sites where ERP modulations are largest are interesting
only when they vary with condition.

Acknowledgments
Thanks to Mrs. Catriona Bruce for assistance with data
collection.
DID is a member of the SINAPSE Collaboration (www.sinapse.
ac.uk), a Pooling Initiative funded by the Scottish Funding
Council and the Chief Scientist Office of the Scottish Executive.

Appendix 1
Typical word pairs included in the experiment.
WORD1

WORD2

Forward
association

Backward
association

ACRE
PERFORM
PRINCIPAL
LUMBER
MOP

LAND
ACT
SCHOOL
WOOD
FLOOR

0.68
0.29
0.31
0.59
0.24

0.02
0.02
0.00
0.00
0.04

EEG data processing

EEG data were analyzed off-line using Neuroscan 4.3 Edit

software (Quick-Cap, Neuromedical Supplies: www.neuro.
com). Segments of data were rejected if they were excessively
noisy or clearly saturated (based on visual inspection of waveforms) to prevent the introduction of artifacts by algorithms
used in later processing stages. Eye blink artifacts were
attenuated using the Neuroscan Ocular Artifact Reduction
procedure (minimum of 32 blinks per participant). EEG study
data were epoched time-locked to stimulus presentation, using
a 104 ms to +2000 ms time window. Epochs were excluded
where drift (defined as the difference in amplitude between the
beginning and end of each recording epoch) greater than 50 V
was present, or where signal change exceeded 100 V. Data
were smoothed over 5 successive points and baseline corrected
with respect to the 104 ms pre-stimulus period.
Study phase ERPs were formed for four response categories: recalled (items subsequently recognized as old for which
the study partner was recalled), missed (items judged
incorrectly as being new), high JOL (study pairs assigned a
JOL of 4 or 5) and low JOL (JOL of 1 or 2). Study items attracting
an unsure JOL (3 responses) were discarded to separate the
high and low JOL categories. Individual epochs were combined
to produce averaged waveforms for each condition and
participant. Mean numbers of trials (with a criterion of
minimum of 16 per participant per condition) were 123, 49,
87 and 73 for the recalled, missed, high JOL and low JOL
categories.
The ERP data were analyzed using repeated measures
ANOVA, applying the GreenhouseGeisser correction (Greenhouse and Geisser, 1959) where appropriate. The analyses
were restricted to two post-stimulus time-windows: 5501000
and 13001900 ms. These time windows, as well as the
electrode locations submitted to analysis, were selected as
they capture the principal divergences between the critical
ERP conditions. The earlier time window is one in which
subsequent memory effects are typically observed (Paller
et al., 1987), and the later time window is designed to best

REFERENCES

Diana, R.A., Yonelinas, A.P., Ranganath, C., 2007. Imaging

recollection and familiarity in the medial temporal lobe: a
three-component model. Trends Cogn. Sci. 11,
379386.
Dunlosky, J., Domoto, P.K., Wang, M.L., Ishikawa, T., Roberson, I.,
Nelson, T.O., Ramsay, D.S., 1998. Inhalation of 30% nitrous
oxide impairs people's learning without impairing people's
judgments of what will be remembered. Exp. Clin.
Psychopharmacol. 6, 7786.
Fernandez, G., Tendolkar, I., 2006. The rhinal cortex: gatekeeper
of the declarative memory system. Trends Cogn. Sci. 10,
358362.
Greenhouse, S.W., Geisser, S., 1959. On methods in the analysis of
profile data. Psychometrika 24, 95112.
Izaute, M., Bacon, E., 2005. Specific effects of an amnesic drug:
effect of Lorazepam on study time allocation and on judgment
of learning. Neuropsychopharmacology 30, 196204.
Jasper, H.A., 1958. The ten-twenty system of the international
federation. Electroencephologr. Clin. Neuropsychol. 10, 371375.
Kao, Y.-C., Davis, E.S., Gabrieli, J.D.E., 2005. Neural correlates of
actual and predicted memory formation. Nat. Neurosci. 8,
17761783.
Kelemen, W.L., Winningham, R.G., Weaver, C.A., 2007. Repeated
testing sessions and scholastic aptitude in college students'
metacognitive accuracy. Eur. J. Cogn. Psychol. 19, 689717.
Kennedy, M.R., Yorkston, K.M., 2000. Accuracy of metamemory
after traumatic brain injury: predictions during verbal learning.
J. Speech Lang. Hear. Res. 43, 10721086.
Koriat, A., Sheffer, L., Ma'ayan, H., 2002. Comparing objective and
subjective learning curves: judgments of learning exhibit
increased underconfidence with practice. J. Exp. Psychol.:
General 131, 147162.

BR A IN RE S E A RCH 1 3 18 ( 20 1 0 ) 8 7 9 5

MacCarthy, G., Wood, C.C., 1985. Scalp distribution of eventrelated potentials: an ambiguity associated with analysis of
variance models. Electroencephalogr. Clin. Neurophysiol.,
Evoked Potentials 62, 203208.
Merritt, P., Hirshman, E., Hsu, J., Berrigan, M., 2005. Metamemory
without the memory: are people aware of midazolam-induced
amnesia? Psychopharmacology 177, 336343.
Nelson, D.L., McEvoy, C.L., Schreiber, T.A., 1998. The University of
South Florida word association, rhyme and word fragment
norms. http://www.usf.edu/FreeAssociation/.
Nelson, T.O., 1984. A comparison of current measures of the
accuracy of feeling-of-knowing predictions. Psychol. Bull. 95,
109133.
Nelson, T.O., Narens, L., 1990. Metamemory: a theoretical
framework and new findings. In: Metcalfe, J., Shimamura, A.P.
(Eds.), 1994. Metacognition: Knowing about knowing. MIT
Press, Cambridge, MA.

95

Paller, K.A., Kutas, M., Mayes, A.R., 1987. Neural correlates of

encoding in an incidental learning paradigm.
Electroencephalogr. Clin. Neurophysiol. 67, 360371.
Qin, S., Piekeman, C., Petersson, K.M., Han, B., Luo, J., Fernandez,
G., 2007. Probing the transformation of discontinuous
associations into episodic memory: an event-related fMRI
study. NeuroImage 38, 212222.
Sommer, W., Heinz, A., Leuthold, H., Matt, J., Schweinberger, S.R.,
1995. Metamemory, distinctiveness, and event-related
potentials in recognition memory for faces. Mem. Cogn. 23, 111.
Vilkki, J., Surma-aho, O., Servo, A., 1999. Inaccurate prediction of
retrieval in a face matrix learning task after right frontal lobe
lesions. Neuropsychology 13, 298305.
Wagner, A.D., Schacter, D.L., Rotte, M., Koutstaal, W., Maril, A.,
Dale, A.M., Rosen, B.R., Buckner, R.L., 1998. Building memories:
remembering and forgetting of verbal experiences as predicted
by brain activity. Science 281, 11881191.