Examining ERP correlates of recognition memory: Evidence of accurate source recognition without...

(This is a sample cover image for this issue. The actual cover is not yet available at this time.)

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

http://www.elsevier.com/copyright

Author's personal copy

Examining ERP correlates of recognition memory: Evidence of accurate sourcerecognition without recollection

Richard J. Addante a,b,⁎, Charan Ranganath a,c, Andrew P. Yonelinas a,d

a University of California - Davis, Center for Neuroscience, USAb University of Texas at Dallas, Center for Vital Longevity, USAc University of California - Davis, Department of Psychology, USAd University of California - Davis, Center for Mind & Brain, USA

a b s t r a c ta r t i c l e i n f o

Article history:Accepted 12 April 2012Available online 25 April 2012

Keywords:RecollectionFamiliarityEvent-related potentialsSource memoryEpisodic memory

Recollection is typically associated with high recognition confidence and accurate source memory. However,subjects sometimes make accurate source memory judgments even for items that are not confidently recog-nized, and it is not known whether these responses are based on recollection or some other memory process.In the current study, we measured event related potentials (ERPs) while subjects made item and sourcememory confidence judgments in order to determine whether recollection supported accurate source recog-nition responses for items that were not confidently recognized. In line with previous studies, we found thatrecognition memory was associated with two ERP effects: an early on-setting FN400 effect, and a later pari-etal old–new effect [late positive component (LPC)], which have been associated with familiarity and recol-lection, respectively. The FN400 increased gradually with item recognition confidence, whereas the LPC wasonly observed for highly confident recognition responses. The LPC was also related to source accuracy, butonly for items that had received a high confidence item recognition response; accurate source judgmentsto items that were less confidently recognized did not exhibit the typical ERP correlate of recollection or fa-miliarity, but rather showed a late, broadly distributed negative ERP difference. The results indicate that ac-curate source judgments of episodic context can occur even when recollection fails.

© 2012 Elsevier Inc. All rights reserved.

Introduction

Recognition memory judgments can be based on recollection ofqualitative information about a previous event or on assessmentsof stimulus familiarity. The familiarity process is generally assumedto reflect the rapid assessment of global familiarity or memorystrength of an item, whereas recollection reflects a search of memo-ry similar to that used in free recall, whereby associative informationabout the study event is retrieved. Recollection and familiarity havebeen shown to be functionally dissociable and to rely on partiallyseparable brain regions (Diana et al., 2007; Eichenbaum et al.,2007; Yonelinas, 2001a, 2001b; Yonelinas and Parks, 2007;Yonelinas et al., 2002). In addition, recollection and familiarityhave been associated with distinct event related potential (ERP)modulations (Cansino and Trejo-Morales, 2008; Curran, 2000;Leynes et al., 2005; Rugg and Curran, 2007; Wilding and Rugg,1996; Yu and Rugg, 2010). That is, at time of retrieval, familiarity

is associated with modulations of the FN400, an enhanced positivityfor old items relative to new items observed from approximately400–600 ms post stimulus onset. The FN400 tends to have a mid-frontal scalp distribution which can extend to left and right frontalareas, plus central midline regions depending on which experimen-tal materials are used (Friedman, 2000; Friedman and Johnson,2000; Gruber and Otten, 2010; Rugg and Curran, 2007; Rugg et al.,1998a; Voss and Paller, 2007). In contrast, recollection has beenlinked with modulations of a positive going waveform that emergesapproximately 600 ms post stimulus and that is typically maximalover left parietal sites (Curran, 2000; Rugg et al., 1998a), and re-ferred to as the late positive component (LPC). Dissociations be-tween these ERP modulations have been reported based onsubjective reports of recollection and familiarity (Duzel et al.,1997), as well as correct and incorrect source memory discrimina-tions (Wilding and Rugg, 1996, for a review see Rugg and Curran,2007). In addition, the FN400 is found to increase gradually as afunction of item recognition confidence whereas the LPC is limitedto high confidence recognition responses (Woodruff et al., 2006;Yu and Rugg, 2010), consistent with cognitive models of familiarityand recollection (Yonelinas, 1999; Yonelinas et al., 2010).

Although the distinction between recollection and familiarity isrelatively well established, important debates remain about the

NeuroImage 62 (2012) 439–450

⁎ Corresponding author at: University of Texas at Dallas Center for Vital Longevity,1600 Viceroy Drive, Suite 800, Dallas, TX 75235, USA. Fax: +1 972 883 3250.

E-mail address: [email protected] (R.J. Addante).

1053-8119/$ – see front matter © 2012 Elsevier Inc. All rights reserved.doi:10.1016/j.neuroimage.2012.04.031

Contents lists available at SciVerse ScienceDirect

NeuroImage

j ourna l homepage: www.e lsev ie r .com/ locate /yn img


functional nature of these processes and about how to best separatethese processes (for a review see Yonelinas, 1999, 2001a; Yonelinasand Levy, 2002; Yonelinas et al., 2010). One approach is to use testsof source memory as measures of recollection (i.e., tasks that requiresubjects to retrieve where or how an item was studied), and comparethis to performance on tests of item recognition (i.e., tasks that requiresubjects to discriminate between old and new items). The idea is that,if accurate source discriminations rely exclusively on recollection thenperformance on these tests can be used as an index of recollection. Incontrast, if an item is recognized as old but leads to an incorrect sourcememory judgment then this can be used as a measure of familiarity. Analternative approach is to estimate recollection on the basis of recogni-tion confidence judgments. The idea is that the recollection of qualita-tive information about a study event should lead to confidentrecognition memory responses, whereas familiarity in the absence ofrecollection should support lower confidence recognition responses.Thus in standard item recognition tests recollection should be restrict-ed primarily to high confidence recognition responses whereas famil-iarity should increase gradually across levels of confidence, see (Parksand Yonelinas, 2007, 2009; Yonelinas and Parks, 2007).

Usually, correct source judgments are associated with the highestlevel of recognition confidence, so these two approaches often lead tothe same conclusions. However, subjects can often recognize itemswith less than the highest level of item confidence, and yet go on toaccurately determine the source of those items (Hicks et al., 2002;Levine et al., 1999; Quamme et al., 2002; Wais et al., 2010;Yonelinas, 2001a; Yonelinas et al., 2001). What memory processmight support these source discriminations is not yet clear. If theyare based on recollection, one might expect that items associatedwith lower recognition confidence but correct source should be asso-ciated with an LPC modulation, perhaps in a graded fashion (Leynesand Phillips, 2008). In contrast, if they are based on familiarity(Diana et al., 2011) or conceptual implicit memory (Paller et al.,2007), one might expect these trials to be associated with an ERPmodulation resembling the FN400.

We sought to test these alternatives by examining the ERPs re-lated to item recognition confidence and source memory judg-ments. In Experiment 1, we recorded ERPs during a recognitionmemory test in which subjects made old–new recognition judg-ments and source memory judgments (see Fig. 1). Based on priorwork, we expected to see an early fronto-central effect related tofamiliarity from 400 to 600 ms (FN400) and a later left hemisphereeffect related to recollection from 600 to 800 ms (LPC). The FN400effect was expected to increase gradually across item recognition

confidence with increases in familiarity strength, whereas the LPCwas expected to be restricted to the high confidence recognition re-sponses. Moreover, we hypothesized that items leading to correctsource memory judgments should produce a recollection effectcompared to incorrect source judgments. Most importantly, wethen examined the ERPs associated with correct source memory re-sponses that were not recognized with the highest levels of recog-nition confidence. No prior study that we are aware of hasexamined this critical condition. If these items reflect the operationof recollection then they should exhibit the late left hemisphereERP signature consistent with recollection. In contrast, if thesesource memory responses are reflecting the operation of familiarity,they should exhibit the FN400 ERP effect consistent with familiari-ty. Experiment 1 produced a pattern of results that was not consis-tent with either the recollection or the familiarity account, butrather suggested that source correct responses for low confidenceitem recognition might reflect the contribution of some other mem-ory process. To assess the extent to which this novel finding couldbe replicated, we examined the results from a similar EEG experi-ment (Addante et al., 2011) and found similar results, which arereported as Experiment 2.

Methods

Subjects and stimuli

Twenty-five right-handed undergraduate students (eight males)were recruited from the University of California–Davis PsychologyDepartment subject pool, and received credit for participation. Sub-jects were free from neurological, visual, motor, or other medical dis-orders, and the experiment was conducted as approved by theUniversity of California–Davis IRB protocol for research on humansubjects.

Word stimuli were selected from the Medical Research CouncilPsycholinguistics Database (http://www.psych.rl.ac.uk/MRC PsychDb.html). Word stimuli had an average rating of concreteness of589.50 (min=400, max=670), image-ability of 580.11 (min=424,max=667), Kucera–Francis Frequency of 30.38 (min=3,max=198), and an average number of 4.89 letters in each word(min=3, max=8). Stimuli were presented in uppercase letters in awhite font, centered on a black background screen (Fig. 1). Subjectswere seated approximately 44 in. away from the screen.

Procedure

During study, subjects encoded 200 words (presented in 4 lists of50 words each) during an incidental encoding task. Two separateencoding tasks (i.e., “Animacy” and “Manmade” judgments), wereused, which served as the basis for source memory decisions duringretrieval (i.e., subjects made a yes/no responses to indicate if theitem was alive, or to indicate if the item was manmade). Theseencoding tasks were selected to lead to comparable levels of sourcememory performance while allowing for reasonable levels of sourcediscriminability (Ranganath et al., 2004). The two encoding taskswere presented in a blocked ABBA design, counterbalanced betweensubjects for the order of the two tasks. Prior to each stimulus presen-tation, a fixation cross appeared in the middle of the screen for750 ms. The encoding probe then appeared on the screen for1500 ms, during which time the subject viewed the stimuli, butwere not yet cued to respond. After the 1500 ms trial, subjects werethen asked to decide either “yes” or “no” to either the animacy ormanmade task by responding with a button press (i.e.: a subject-paced response). Prior to each study block, subjects heard the instruc-tions and there was a practice session of 10 stimuli that the experi-menter and subject performed together in order to be sure that thesubject understood the task. None of the practice stimuli appeared

OK TO BLINK

TREE1 2 3 4 5

New Old

TREE

Fixation (650ms)

Stimuli (1500ms) (ERP epoch)

Item Recogntion (subject paced)

Source Memory (subject paced)

Blink (1000ms)

Fixation (650ms)

TREE54321

Alive ? Manmade

Fig. 1. Schematic depiction of events during test trials. For each test item, subjects firstmade an item memory confidence judgment, followed by a source memory confidencejudgment. ERPs were recorded during the presentation of the test word, and classifiedaccording to the ensuing item and source memory responses.

440 R.J. Addante et al. / NeuroImage 62 (2012) 439–450


in the test phase. After the 4 study blocks were presented there was abreak, during which the ERP cap was affixed to the subjects, beforethe retrieval phase of the experiment commenced.

During retrieval, the 200 stimuli presented during the encodingphase were randomly intermixed with 100 new words (lures), for atotal of 300 test stimuli (Fig. 1). The test stimuli were presented in6 test blocks of 50 stimuli each. Prior to each stimulus presentation,a fixation cross appeared in the middle of the screen for 650 ms, dur-ing which subjects were instructed not to blink, so as to minimize oc-ular artifacts in the EEG. The retrieval probe then appeared on thescreen for 1500 ms, during which time the subject viewed the stimuli,but were not yet cued to respond (Fig. 1). Subjects were alsoinstructed not to blink while each stimulus was on the screen. Afterthe 1500 ms probe, subjects were asked first to make an item recog-nition judgment followed by a source recognition judgment; each ofthese responses was subject paced and therefore provided a variabletemporal jittering of each subsequent stimuli's presentation. Prior tocommencement of the testing phase, subjects practiced on 10 sampletrials with the experimenter present to make sure they understoodinstructions and used the scale correctly.

For the item recognition judgment, subjects responded on a 5-point confidence scale, with 5 indicating that they were sure it wasold, 4 indicating that it was probably old, 3 indicating they wereguessing, 2 indicating it was probably new, and 1 indicating theywere sure it was new. For the source recognition judgment, subjectsalso responded on a 5-point scale with 5 indicating that they weresure it was from the animacy encoding task, 4 indicating that theythought it was from the animacy task but were not sure, 3 indicatingthey were guessing (or that the item was rated “new”), 2 indicatingthat they thought it was from the manmade task but were not sure,and 1 indicating they were sure it was from the manmade task.

During later analysis, source memory responses were analyzedonly for previously studied items (e.g. “old” items), and not for sourceresponses of new items (i.e.: source ratings of “3”), so as to avoid in-clusion of random response trials or item false alarms. For ease of in-terpretation during data analysis, source memory responses werecollapsed into a modified scale of source accuracy (1–5) that cor-responded to the same direction of the item confidence ratings:source memory responses reported as “5” reflect correct source judg-ments reported as highly confident/sure, “4” representing correctsource judgments reported as not being completely sure, “3” beingsource “unknown/forgotten”, “2” responses being incorrect sourcejudgments (misattributions) though not completely sure, and “1”being incorrect source judgments (misattributions) that subjectsreported as being highly confident/sure of.

EEG acquisition and analysis

EEGwas recorded using a BioSemi ActiveTwo Recording Systemwitha 32 channel electrode cap conforming to the standard International10–20 System of electrode locations (Klem et al., 1999). Each subjectwas tested individually inside a sound-attenuating chamber. Stimuluspresentation and behavioral response monitoring were controlledusing Presentation software on a Windows PC. EEG was sampled at arate of 1024 Hz. Subjects were instructed to minimize jaw and muscletension, eye movements, and blinking. EOG was monitored in the hori-zontal direction and vertical direction, and this data was used to elimi-nate trials contaminated by blink, eye-movement, or other artifacts.

All EEG analyses were performed using custom Matlab code andfunctions from the EEGLAB Toolbox for Matlab (Delorme andMakeig, 2004). Raw EEG data was re-referenced to averaged mas-toids, downsampled to 256 Hz, and high-pass filtered at .1 Hz inorder to optimize independent component analysis (ICA) decomposi-tion for artifact correction. These data were epoched from 200 ms be-fore the onset of the retrieval item to 1.5 s following the retrievalitem, and was baseline subtracted from −200 to 0 ms. Epochs

containing single channel data which exceeded 4 standard deviationsof the channel's mean across epochs were removed to optimize ICAdecomposition, as were epochs containing data 6 standard deviationsfrom the pooled channel mean. This procedure was designed to re-move primarily non-biological noise, while allowing stereotypical ar-tifacts (such as eye-blinks) to remain. Data were then decomposedinto temporally independent components using Infomax ICA (Belland Sejnowski, 1995). Artifactual components (eye-blinks, muscletension, etc.) were manually identified and subtracted from the dataand the artifact-corrected data were manually screened a secondtime to reject any remaining epochs with artifacts.

ERPs were averaged from the EEG data using ERPLAB software(http://erpinfo.org/erplab), a plug-in toolbox of Matlab functions forthe EEGLAB software (Delorme and Makeig, 2004). ERPs were grandaveraged to a baseline of the 200 ms preceding stimulus onset,using the un-weighted average of individual subjects' trials. Meanamplitudes of latencies of interest for each condition were obtained,and analyzed in separate statistical software.

We used a priori defined latencies of interest, based upon theestablished ERP literature of familiarity and recollection-related ef-fects. Therefore, we focused our analysis on the time periods of400–600 and 600–800 ms, and our old–new analysis (Fig. 2) con-firmed the detection of FN400 and LPC effects during these epochsin the current study. For our initial analysis (old–new effects, i.e.:Figs. 2A, B, C), we used a priori-selected mid-frontal and left parietalelectrode sites of Fz and P3 to explore old–new ERP differencesfrom 400 to 600 ms and 600 to 800 ms for the FN400 and LPC, respec-tively. These old–new effects were highly significant (each pb .001),but closer inspection of the topographies of these effects (Fig. 2A)revealed that these sites (Fz and P3) were not the sites where effectswere actually focused upon, but that they were insteadmost apparentat the adjacent mid-frontal and left parietal electrode sites of Cz andCp5, respectively. The same old–new effects were also highly signifi-cant at these adjacent sites, Cz and Cp5 (each pb .001) (Figs. 2B, C),consistent with findings of other labs (Wolk et al., 2004). Therefore,in order to more fully characterize these old–new effects with the ad-ditional contrasts of item and source memory responses that were ofparticular interest for this experiment, we focused the remaininganalyses upon these representative mid-frontal (Cz) and left parietal(Cp5) sites for all ensuing analyses. When conducting ANOVA tests,results are reported after Greenhouse–Geisser corrections wereperformed.

For most of our contrasts, all subjects were included (N=25), andthe average number of trials per condition was quite high (e.g., in theitem confidence analysis an average number of trials per confidencelevel was of 37, 48, 26, 54, and 100 trials per subject). However, insome conditions, trial counts were quite low for a few subjects,which is to be expected given that each memory response is dividedup between five different levels of confidence (as long as perfor-mance is above chance subjects cannot produce equally distributedold and new responses across levels of confidence). However, incases in which it was necessary to exclude subjects, we report thesample size in the relevant result sections. For each ERP contrast,we followed similar criteria used by prior studies (Gruber andOtten, 2010; Kim et al., 2009; Otten et al., 2006), and included sub-jects only if they had at least 13 artifact-free trials per condition. Im-portantly, for these particular comparisons (i.e.: Fig. 5), the sameoverall pattern of results were still evident (and statistically signifi-cant) when every subject was included in the analysis.

Results

Behavioral results

Memory performance for each encoding task is shown in Table 1.As expected, item recognition performance was similar for the two

441R.J. Addante et al. / NeuroImage 62 (2012) 439–450


encoding conditions, so for all ensuing analyses we collapsed acrossencoding condition. On average, recognition confidence ratingswere higher for old items than for new items, t(24)=21.58,pb .001, indicating that subjects were able to discriminate betweenold and new items. Table 2 shows the proportions of confidence re-sponses for the source memory judgments. An examination of sourcememory responses indicated that .35 of the studied items were recog-nized with the highest level of confidence and received a correctsource response (.15 and .20 led to low and high levels of source con-fidence, respectively), whereas .10 of the studied items were recog-nized with the second highest level of confidence and received acorrect source response (.09 and .01 led to low and high levels ofsource confidence, respectively). Table 3 shows that subjects per-formed at comparable levels of source accuracy for the two encodingtasks (no significant difference in source accuracy among the twoencoding tasks, t(24)=1.29, p=.207), so our source memory analy-sis proceeded by collapsing across tasks. Source memory accuracy(collapsing across low- and high-confidence recognition) was abovechance (i.e., >50%) for both the low confidence item recognition tri-als (M=.58, SD=.15, t(24)=2.50, p=.02); (reflecting the propor-tion of source correct divided by the sum of items receiving asource correct and source incorrect response) and the high confi-dence item recognition trials (M=.74, SD=.08, t(24)=14.35,pb .001). Thus subjects made accurate source memory responses foritems that were recognized with both high and low levels of item rec-ognition confidence.

Electrophysiological results

Recognition memoryItem recognition was first examined by contrasting the ERPs asso-

ciated with hits (old items receiving a 4 or 5 response) and correct-rejections (new items receiving a 1 or 2 response). Topographicmaps of item recognition difference waves (hits−correct rejections)for each 200 ms time window of the recording epoch are shown inFig. 2A. ERPs for hits were more positive going than those for correctrejections (i.e., warmer colors on the topographical maps). This effectwas apparent by approximately 400 ms after stimulus onset and wascentrally distributed with a maximum over central midline electrodeCz (i.e.: FN400 effect), t(24)=5.35, pb .001, though broadly evidentat many adjacent electrode sites. However, between 600 and800 ms the FN400 effect diminished, and an LPC effect was observedthat was maximal over the left parietal electrode Cp5, t(24)=3.73,pb .001, though this too was also quite widespread in its spatial distri-bution across adjacent electrode sites. The ERPs observed at the Cz

and Cp5 electrodes for the hits and correct rejections are plotted inFigs. 2B and C, respectively. The early and late ERP differences corre-spond temporally and topographically to those identified in previousstudies with the FN400 and the LPC, respectively (for reviews seeCurran, 2000; Friedman and Johnson, 2000; Rugg and Curran, 2007).In addition, to control for potential differences in the confidencelevels of the hits and correct rejections we examined performancefor only the highest confidence responses (i.e., 1 s and 5 s), and thisalso led to the same pattern of FN400 and LPC effects. Thus, in oursubsequent analyses, we used mean voltage amplitudes at electrodeCz from 400 to 600 ms as a measure of the FN400, and at Cp5 from600 to 800 ms as a measure of the LPC.

Item memory confidenceTo examine the relationship between the FN400 and LPC effects

with the processes of familiarity and recollection, we plotted the av-erage amplitude of the ERPs at each level of recognition confidence(collapsed across study status). Figs. 2D and E present results froman analysis which included the 14 subjects who had more than 13 tri-als in each of the 5 response confidence bins (the same pattern wasobserved, however, when we included all subjects regardless of num-ber of trials in each cell). As shown in Fig. 2D (see also Fig. 3 for ERPsand topographic maps of each condition), the FN400 increased in alinear manner across confidence levels (R2=.927, pb .001), and in-troducing a quadratic component did not lead to a significant increasein variance accounted for (Fb1). In contrast, LPC amplitudes showeda non-linear U-shaped function across confidence levels (Fig. 2E, ERPsshown in Fig. 3) and introducing a quadratic component led to a sig-nificant increase in fit compared to a linear model (R2=.994, F(1,2)=272.56, pb .005). Subsequent analysis indicated that anLPC effect was evident for confidently recognized items (i.e., “5”)compared to confidently rejected items (i.e., “1”) (t(13)=4.83,pb .001), whereas the low confidence recognition responses (i.e.:“4”) were not different from the confidently rejected items(“1's”), t(13)=−.293, p=.77. In contrast, FN400 amplitudeswere more positive for both the high and low confidence recogni-tion responses (see Figs. 2D, 3) than for confidently rejecteditems (“1's”), (i.e. t(13)=6.25, pb .001, and t(13)=4.02, p=.001for high and low confidence responses, respectively).1

Overall, the item recognition results are consistent with previousstudies showing that recollection and familiarity are associated withtopographically and temporally distinct ERP correlates (Rugg andCurran, 2007; Woodruff et al., 2006; Yu and Rugg, 2010), and withother studies indicating that familiarity increases gradually acrossitem recognition confidence, whereas recollection contributes pri-marily to high confidence recognition responses (Yonelinas, 2001b).

Table 1Distribution of responses for each level of item recognition confidence, as a proportionof all responses (see response schematic Fig. 1).

Item recognition confidence 1 2 3 4 5

All old items .045 .099 .077 .245 .534New items .327 .349 .146 .139 .039Animacy task .054 .116 .092 .256 .483Manmade task .036 .082 .062 .234 .586

Table 2Proportion of responses for each level of source memory confidence.

Sourcememoryconfidence

Incorrecthighconfidence

Incorrectlowconfidence

Unknown Correct lowconfidence

Correcthighconfidence

All old items .046 .154 .344 .248 .207All new items .010 .073 .851 .060 .006Animacy task .035 .140 .393 .244 .188Manmadetask

.057 .168 .296 .252 .226

Table 3Source memory accuracy. Proportions of source correct and source incorrect are givenfor all old items, as well as for each encoding task. Accuracy values collapse across low-and high-confidence responses, and reflect the proportion of source correct divided bythe sum of items receiving a source correct and source incorrect response.

Memory accuracy Source correct Source incorrect

All old items .694 .305Animacy task .711 .288Manmade task .679 .320

1 Note that a further examination of Fig. 2E indicated that the confident new re-sponses (i.e., “1” responses) were associated with a more positive going memory effectat CP5 during the 600–800 ms time window than the low confidence rejected items(i.e., “2“ responses). Although this difference was not statistically significant(pb0.05), we examined it further and found that it had a more posterior and later scalpdistribution than the recollection-related response. This pattern of results has beenreported before as dissociable from recollection, and has been attributed to novelty-related processing and confidence (Curran, 2004; Woodruff et al., 2006).



Source recognitionERPs were then examined to identify the correlates of recollection

and familiarity on the basis of source discrimination. Thus, ERPs wereexamined for “source correct” trials irrespective of the preceding itemjudgments (i.e., old items receiving a low or high confidence correctsource judgment), “source incorrect” trials (i.e., old items receivingan incorrect source judgment or a “source unknown” response) and“correct rejection” trials (i.e., new items receiving a low or high con-fident new response). Compared to correct rejections, source correcttrials should exhibit an LPC effect indicative of recollection, whereasthe source incorrect trials should exhibit a reduced or non-evidentLPC because recollection has presumably failed. In contrast, bothsource correct and source incorrect trials should be familiar andthus should exhibit the FN400 effect indicative of familiarity.

As expected, FN400 amplitudesweremore positive for source incor-rect trials than for correct rejections, t(24)=3.05, p=.005, and alsomore positive going for correct source trials than for correct rejections,t(24)=5.72, pb .001. FN400 amplitudes were higher for source correcttrials than for the source incorrect trials as well, t(24)=5.14, pb .001.The latter effect might reflect some degree of familiarity-based sourcerecognition (Diana et al., 2011). An examination of the scalp topographyof these effects indicated that they were maximal over the central mid-line electrode Cz (Fig. 4B).

The amplitude of the LPC was significantly higher for source cor-rect trials compared to incorrect source trials, t(24)=5.603, pb .001and compared to correct rejections, t(24)=5.07, pb .001, whereasLPC amplitudes did not differ between the source incorrect trialsand the correct rejections t(24)=.673, p=.50. Topographic maps

Fig. 2. Recognition memory ERP effects. (A) Topographic maps of mean amplitude differences between hits and correct rejections for each 200 ms latency window. (B and C) MeanERPs for hits and correct rejections plotted for electrodes Cz (B) and Cp5 (C). (D and E) Mean ERP amplitude plotted as a function of item recognition confidence (collapsed acrossold and new items) for Cz (D) and Cp5 (E) effects. (F and G) Mean ERP amplitudes for correct rejections, source incorrect and source correct trials for Cz (F) and Cp5 (G). (For in-terpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)



contrasting the source correct and source incorrect trials to correctrejections showed that this effect exhibited a left lateralized effect(Fig. 4D) for correct source memories, but no differences across thescalp for incorrect source judgments from 600 to 800 ms.

Source memory for low and high-confidence item recognition hitsThe preceding analyses suggested that the LPC was enhanced for

items that received the highest confidence rating (i.e.: “Item 5”)and/or items that were associated with accurate source decisions,which is consistent with a large body of evidence linking the LPC torecollection (Allan and Rugg, 1998; Allan et al., 1998; Cansino andTrejo-Morales, 2008; Curran, 2000, 2004; Curran and Doyle, 2011;Curran et al., 2006, 2007; Duarte et al., 2004; Duzel et al., 1997; Fried-man, 2000; Friedman and Johnson, 2000; Leynes and Phillips, 2008;Leynes et al., 2005; Rugg and Curran, 2007; Rugg and Wilding,2000; Rugg et al., 1998a, 1998b; Wilding, 2000; Wilding and Rugg,1996; Woodruff et al., 2006; Yu and Rugg, 2010). No significant LPCwas seen for low confidence item hits (i.e.: “Item 4”), which might in-dicate that these items were not recollected. Our behavioral analyses,however, indicated that source memory accuracy was above chancefor item hits that were recognized with low confidence (“Item 4” re-sponses). If recollection supported accurate source decisions for lowconfidence item recognition hits, then it is possible that an LPC en-hancement might be observed specifically for “Item 4” responsesthat were also associated with correct source decisions (i.e.: “Item 4+Source Correct”). Alternatively, if familiarity solely supportedthese types of responses (i.e.: “Item 4+Source Correct”), we wouldexpect these trials to exhibit an FN400 modulation as well. Thus, forthis particular analysis, we assessed correct source memory ERPs foritems that were recognized with high confidence (“Item 5+SourceCorrect”), as well as items recognized with low confidence (“Item4+Source Correct”), as compared to correct rejections (new itemsrated “1” and “2”). In order to avoid confounding this comparisonwith differences in source confidence between item recognitionconditions, we limited this analysis to including only source judg-ments with low confidence (i.e.: “Item 5+Source 4”, and “Item4+Source 4”), so that confidence for accurate source was held

constant between conditions of high and low item recognitionconfidence.

Fig. 5 shows a series of topographic maps illustrating the timecourse for the amplitude difference between trials of “Item 4+Source

Fig. 4. Source memory effects. (A) ERPs at electrode Cz for source correct, source incor-rect and correct rejections. (B) Scalp topographies of the difference wave comparisonsof “source correct” vs. “correct rejection” trials, and the “source incorrect” vs. "correctrejections" (CR) during the early period (400–600 ms). (C) ERPs at electrode Cp5 for“source correct”, “source incorrect” and “correct rejections”. (D) Scalp topographiesof difference wave comparisons of “source correct” vs. “correct rejections”, and “sourceincorrect” vs. “correct rejections” at Cp5 during the later period (600–800 ms).

Fig. 3. Electrophysiological correlates of itemmemory confidence. Topographic maps of the item recognition confidence (N=25). Difference waves of ERPs at each confidence level(“2”, “3”, “4”, and “5”) compared to the ERP for the items rated “new” (i.e.: “1” responses) during the early (A) and late (B) latency windows of the putative correlates of familiarityand recollection, respectively. C) ERPs for each level of item recognition confidence at electrodes Cz and Cp5.



4” responses of accurate sourcememory for low confidence recognitionand correct rejections, as well as ERPs at a representative left parietalelectrode site (for ERPs at each electrode see Supplementary Fig. 1).For comparison purposes, the time course of the difference betweenhigh confidence hits with correct source (“Item 5+Source 4”) and cor-rect rejections is also shown. Unlike the high confidence hits with cor-rect source, which were associated with a significant increase of theFN400 and LPC (t(12)=.4.23, p=.001; t(12)=2.297, p=.04, respec-tively) (Fig. 5A), “Item 4+Source 4” judgments were not significantlymore positive than the correct rejections, during either of the FN400or LPC time windows (both t(12)'sb .10). Instead, these trials of “Item4+Source 4” memories were associated with a broadly distributednegative going ERP that peaked during the 800–1000 ms time window(Fig. 5B).

An exploratory 2×2×2 (hemisphere [left/right]×region [anteri-or/posterior]×memory condition [“Item 4+Source 4” /correct rejec-tion]) ANOVA was performed on representative electrode sites fromthe four scalp quadrants during the 600–800 ms and the 800–1000 ms window (i.e., left frontal (F3), right frontal (F4), left parietal(P3), and right parietal (P4); c.f. Curran et al., 2006; Woodruff et al.,2006; Yu and Rugg, 2010). From 600 to 800 ms there was a significanteffect of memory condition, F(1,12)=6.443, p=.026) and a main ef-fect of region (F,12)=4.788, p=049, indicating that during the LPCtime window of 600–800 ms these “Item 4+Source 4” ERPs for accu-rate source judgments were significantly more negative than ERPs forcorrect rejections and that this emerged as a frontally-based effect. Inaddition, from 800 to 1000 ms this effect became more pronounced,with a significant effect of condition (F(1,12)=14.309, p=.003)and hemisphere (F(1,12)=6.16, p=.029), as well as a marginallysignificant hemisphere by condition interaction, F(1,12)=3.772,p=.076, suggesting that during this period the memory effect be-came robust across the scalp, and was slightly larger over the rightthan left hemisphere. Thus, contrary to our initial predictions, lowconfidence item hits that were associated with correct source deci-sions exhibited neither the FN400 nor the LPC modulations that

have been previously associated with familiarity and recollection, re-spectively. Instead, these trials were associated with a late, broadlydistributed, negative-going ERP.

One possible concern raised by a reviewer was that the more neg-ative going ERP for “Item 4+Source 4” trials relative to correct rejec-tions may have been due to the fact that the correct rejections wereassociated with higher confidence than the “Item 4+Source 4” re-sponses. That is, the correct rejections included both high confidence(i.e., “1” responses) and low confidence trials (i.e., “2” responses), andour initial item confidence analysis had suggested that the high con-fidence correct rejections were associated with a large positivity(see Fig. 2E). To control for potential confidence effects we contrastedthe ERPS for “Item 4+Source 4” memories to ERPs of the low confi-dent correct rejections (“2” responses). There were fifteen subjectswith a sufficient number of ERP trials for inclusion in this analysis.This contrast revealed that the 800–1000 ms negative ERP for correctsource memory related to low confidence recognition was still ob-served (see Fig. 6). As in the original analysis, “Item 4+Source 4” tri-als were more negative going than the low confidence correctrejections (F=(1,14)=5.052, p=.041), and there was a marginalmemory condition by hemisphere interaction (F=(1,14)=3.945,p=.067) suggesting the effect was slightly larger over the righthemisphere.

Another possible concern was the extent to which this effect wasspecific to source memory accuracy, so we also contrasted the sourcememory ERPs of “correct source” vs. “incorrect source” for “Item 4”responses, utilizing the same approach used to assess general sourcememory effects (i.e.: Source recognition section, see Fig. 4), and thistoo resulted in a similar late negative going ERP (see Fig. 7). Therewas a marginally significant effect of condition (F=(1,14)=4.407,p=.054), a significant effect of hemisphere (F(1,14)=6.324,p=.025) and a significant condition by hemisphere interaction (F(1,14)=5.04, p=.041), indicating that the memory effect was largerover the right hemisphere. Taken together the results indicate thatthe late negative going source memory ERP effect for Item 4 memo-ries was not due to differences in overall response confidence, andthat it was specific for accurate source memory.

Experiment 2

The results of Experiment 1 were unexpected in the sense that thelow confidence recognition items that were associated with correctsource memory (“Item 4+Source Correct”) did not exhibit ERP sig-natures indicative of either recollection or familiarity. Before dis-cussing the implications of this finding, we briefly describe areanalysis of an additional study from the lab that used a similar ex-perimental design to that used in Experiment 1, in order to verifythat this new ERP effect was also observed in an additional sampleof subjects (we thank an anonymous reviewer for this suggestion).The data was from Addante et al. (2011), in which we examined oscil-latory EEG activity related to item and source recognition confidence.The results were re-analyzed by examining ERP data in the same waythat we analyzed the results of Experiment 1, reported above. Ouraim was to determine if the late negative ERP difference related tolow confidence recognition with accurate source memory in thefirst experiment would also be observed in a separate sample ofsubjects.

Methods and procedure

Themethods used for Experiment 2were the same aswas describedfor Experiment 1, with the following differences: seventeen subjects,different from Experiment 1 were tested (from Addante et al., 2011),the encoding tasks were a “pleasantness rating task” and an “animacyjudgment task”, and there was a 15 min delay between encoding andtest. In our data analysis we focused on the conditions yielding the

Fig. 5. Correct source memory ERPs for items receiving high and low item recognitionconfidence. Time course of topographic maps of (A) mean ERP differences between highconfidence item hits for which the source was correctly recognized (“Item 5+Source4”) and correctly rejected new items, and (B) mean ERP differences between low confi-dence item hits for which the source was correctly recognized (“Item 4+Source 4”) andcorrectly rejected new items. (C) ERPs at left parietal electrode Cp5.



0-200 200-400 400-600 600-800 800-1000 1000-1200 1200-1400(ms)

1

2

µV

Item 4 + Source Correct vs. Low Confidence Correct RejectionA

Item 4 + Source Correct (low confidence)

Correct Rejection (low Confidence)

B

0

4

8

µV 0 200 400 600 800 1000 1200 1400ms µV 0 200 400 600 800 1000 1200 1400ms

µV 0 200 400 600 800 1000 1200 1400ms µV 0 200 400 600 800 1000 1200 1400ms

0

4

8

0

4

8

0

4

8

F4F3

P4P3

Fig. 6. Source ERPs when controlling for item confidence. In this analysis, confidence of new and old responses are held constant, comparing old items recognized with low con-fidence of “4” and low confidence source correct (i.e.: “Item 4+ Source 4”, the same ERP condition plotted in Fig. 5 and Supplementary Fig. 1) vs. low confidence correct rejections(new items rated “2”). A) Topographic maps throughout the 1500 ms epoch. B) ERPs from representative electrodes from left frontal (F3), right frontal (F4), left parietal (P3), andright parietal (P4) are shown. The topographic locations of each site on the scalp are shown in red dots on the electrode map inset on site F4. (For interpretation of the references tocolor in this figure legend, the reader is referred to the web version of this article.)

Item 4 + Source CorrectItem 4 + Source Incorrect

Correct Rejections

B0-200 200-400 400-600 600-800 800-1000 1000-1200 1200-1400(ms)

1

2

µV

Item 4 + Source Correct vs.Item 4 + Source IncorrectA

0

4

8

µV 0 200 400 600 800 1000 1200 1400ms

0

4

8

0

4

8

0

4

8

µV 0 200 400 600 800 1000 1200 1400ms

µV 0 200 400 600 800 1000 1200 1400ms µV 0 200 400 600 800 1000 1200 1400ms

F4F3

P4P3

Fig. 7. Source correct vs. source incorrect ERPs. A) Topographic maps contrasting trials of “Item4+Source 4” vs. “Item 4+Source 2” trials throughout the recording epoch. B) ERPsfrom representative electrodes from left frontal (F3), right frontal (F4), left parietal (P3), and right parietal (P4) are shown for trials in which subjects indicated accurate and in-accurate source memory for items recognized with low confidence. The topographic locations of each site on the scalp are shown in red dots on the electrode map inset on siteF4. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)



unexpected results in Experiment 1: correct source memory responsesfor high and low confidence recognition judgments (“Item 4+SourceCorrect”). Of the seventeen subjects studied, one did not have any trialsin the condition of accurate sourcememory for low confidence recogni-tion (the specific bin of interest for this experiment), so this subject wasexcluded from the analysis. Unlike Experiment 1 and the oscillatoryanalysis used in Addante et al. (2011), artifact rejection for this ERPanalysis was performed by individual inspection of the data unbiasedto trial type, without reliance upon ICA decomposition for artifact cor-rection (i.e.: eye blinks)—a minor change which we did not expect tosignificantly alter the outcome of the analysis. In addition, becausethere were fewer “Item 4+Source 4” responses than in Experiment 1,source confidence ratings were collapsed into “Item 4+Source 4 and5” in order to obtain sufficient ERP trials for analysis. This difference isnot critical because if recollection was supporting accurate sourcememory for low confidence recognition, including the high confidentsource judgments (i.e.: “Source 5”) in the ERP analysis, then thiswould increase the likelihood of detecting an LPC, and thus would de-crease the likelihood of observing the same negative ERP differenceseen for this condition in Experiment 1.

Results

BehaviorOverall performance was similar to that observed in Experiment 1.

Most importantly, source accuracy for the low confidence recognition(“Item 4”) was significantly greater than chance (M=.65), t(15)=3.53,p=.003, andwas numerically but not significantly higher than the .58 ac-curacy observed in Experiment 1 (t(39)=1.37, p=.177). Overall itemrecognition and source recognition accuracy (i.e., high and lowconfidencehits minus high and low confidence false alarms) was .78 and .18 for Ex-periment 1, and .80 and .18 for Experiment 2, respectively.

ElectrophysiologyERPs were specifically analyzed for conditions of correct source

memory for low confidence recognition, “Item 4+Source 4 and 5”,and for high confidence recognition “Item 5+Source 4 and 5), ascompared to correct rejections. The analysis was motivated theoreti-cally to focus on the potential roles of item familiarity and recollec-tion in source memory via the 400–600 ms and 600–800 mslatencies of the FN400 and LPC, respectively, as well as the role of anegative-going ERP difference seen in Experiment 1 that extendedbeyond this latency and into the 800–1000 ms epochs. As such, andtowards the goal of assessing the replicability of the observed pat-terns in Experiment 1, initial analyses were constrained to the sameCz electrode reported for FN400 effects, and left parietal electrodeCp5 where effects were evident in Experiment 1. Fig. 8 shows the re-sults of this analysis in topographic maps as well as ERPs at Cp5.

A comparison of Fig. 8 (Experiment 2) and Fig. 5 (Experiment 1)shows that the main ERP results of Experiment 1 replicated quite well.As in Experiment 1, an FN400 was evident at the Cz site from 400 to600 ms for “Item 5+Source 4 and 5” trials, t(15)=2.67, p=.017, butnot for “Item4+Source 4 and 5” trials of accurate sourcememory asso-ciated with low recognition confidence (t(15)=−.210, p=.836). Sim-ilarly, in comparison to the correct rejections, an LPC was observedduring 600–800 ms at electrodeCp5 for the high confidence recognitionresponses with accurate source memory (“Item 5+Source 4 and 5”)t(15)=2.19, p=.044 (Fig. 8), whereas, for source correct trials associ-ated with low item confidence (Item 4+Source 4 and 5), once againthere was no evidence of an LPC (t(15)=−.580, p=.570). Moreover,as in Experiment 1, the ERPwas significantlymore negative whenmea-sured at left parietal electrode P3 (t(15)=−3.15, p=.006) (Fig. 8). Inaddition, as in Experiment 1, from 800 to 1000 ms this negative ERPwas broadly distributed across the scalp, with a significant effect of con-dition (F(1,15)=30.777, pb .001) that was evident throughout the restof the epochs.

We also performed the same follow-up analysis as was done in Ex-periment 1 in order to rule out the possibility that these effects couldbe due to confidence differences from the correct rejections. For thisanalysis we contrasted low confident accurate source judgments foritems recognized with low confidence (i.e.: “Item 4+Source 4”) tolow confidence correct rejections (“new” items rated “2”) on thetwelve subjects who contained a sufficient number of trials in theseconditions. As with Experiment 1 (see Supplementary Fig. 2), “Item4+Source 4” responses were associated with a late effect of negativeERP differences. The effect of condition was not quite significant (F(1,11)=1.856, p=.20), likely due to the substantially reduced sam-ple size in this data set, however, there was a significant conditionby region interaction, (F(1,11)=4.69, p=.05), indicating that the ef-fects were again larger over the frontal regions than they were overposterior regions, the same pattern that was initially observed in Ex-periment 1. Because there were only 8 subjects with sufficient num-bers of low confidence incorrect source judgments, we were unableto contrast the source correct and source incorrect trials for low con-fidence as was done in Experiment 1 (i.e.: Fig. 7).

General discussion

The current study examined ERPs associated with item and sourcememory retrieval. In line with previous studies we found evidence tosuggest that two well-known ERP modulations, the FN400 and theLPC, were differentially related to recollection and familiarity, as mea-sured by recognition confidence and source memory accuracy. Final-ly, we found that accurate source memory for items that wererecognized with low confidence was associated with a topographical-ly widespread, late on-setting negative ERP modulation. Below, weconsider the implications of each finding in more detail.

The amplitude of the FN400 increased linearly with item recogni-tion confidence, whereas LPC amplitudes were enhanced specificallyfor items that were recognized at the highest confidence level.

0-200 200-400 400-600 600-800 800-1000 1000-1200 1200-1400

A

BTime (ms)

C

1

2

µV

0 200 400 600 800 1000 1200 1400

0

1

2

3

4

5

6

Item 5 + Source CorrectItem 4 + Source CorrectCorrect Rejection

CP5

µVms

Item 5 + Source Correct vs. Correct Rejections

Item 4 + Source Correct vs. Correct Rejections

Fig. 8. ERP results of correct source memory in Experiment 2. (A) Topographic maps ofmean ERP differences between high confidence item hits for which the source was cor-rectly recognized (“Item 5+Source 4 and 5”) and correctly rejected new items, and (B)between low confidence item hits for which the source was correctly recognized(“Item 4+Source 4 and 5”) and correctly rejected new items. (C) ERPs at representa-tive left parietal site Cp5. Compare to similar results found in Experiment 1 (shown inFig. 5).



Moreover, the LPC was enhanced specifically for old items that wereassociated with correct source memory judgments, whereas theFN400 was observed even for recognized items that were associatedwith incorrect source judgments. The FN400 and LPC results are con-sistent with previous ERP studies, neuroimaging studies (Kirwan etal., 2008; Montaldi et al., 2006; Ranganath et al., 2004), as well asneuropsychological and behavioral studies (Bowles et al., 2007;Yonelinas et al., 2002) that have shown that recollection supportshigh confidence item recognition responses, whereas familiaritystrength increases gradually across levels of item confidence(Woodruff et al., 2006; Yu and Rugg, 2010). The results are also con-sistent with previous studies showing that recollection supports theretrieval of qualitative source information (Duarte et al., 2004; Guoet al., 2006; Johnson et al., 2008; Leynes and Phillips, 2008; Mitchelland Johnson, 2009; Rugg and Wilding, 2000; Vilberg et al., 2006;Wilding, 2000; Wilding and Rugg, 1996) and that familiarity can beobserved when the retrieval of source information fails (Curran etal., 2006; Johnson et al., 1997, 1998; Mecklinger et al., 2011; Rugg etal., 1998a; Senkfor and Van Petten, 1998; Wilding et al., 1995; Yoveland Paller, 2004).

The novel finding of the current study was related to the examina-tion of the ERPs for low confidence item hits that were associatedwith correct source judgments. Unlike items that were recognizedwith high confidence, these trials did not exhibit modulations of theLPC. To the extent that the LPC effect indexes the recollection process(Allan and Rugg, 1998; Allan et al., 1998; Leynes et al., 2005; Rugg andWilding, 2000; Wilding, 2000; for reviews see Curran, 2000;Eichenbaum et al., 2007; Friedman and Johnson, 2000; Mecklinger,2006; Rugg and Curran, 2007) the results indicate that accuratesource recognition can occur even in the absence of recollection.One might argue that the lower confidence source correct items didnot show an LPC simply because those responses included randomguesses. However, subjects were given the opportunity to indicatethat they did not know the study source (i.e., a source confidence of“3”), and these trials were not included in the analysis. Moreover,source accuracy for low confident item hits was above chance, indi-cating that some memory process must have supported these accu-rate source decisions. The lack of an LPC modulation for lowconfidence hits that were associated with correct source decisionsalso cannot be attributed to insufficient statistical power, becausethese trials were associated with a statistically significant negativegoing ERP effect, which is the opposite of the LPC that was evidentin the recollection contrasts. More specifically, these trials were asso-ciated with a topographically widespread negative ERP differencethat peaked between 800 and 1000 ms post-stimulus. This suggeststhat source recognition for low confidence item hits was supportedby a neurocognitive process distinct from recollection.

What processes support accurate source memory in the absence ofrecognition confidence? The current experiment cannot conclusivelyanswer this question, but there are several possibilities. One possibil-ity is that accurate source memory for lower confidence recognitiontrials relied on familiarity. Several studies have shown that familiaritycan support accurate source memory discriminations when recollec-tion fails (Diana et al., 2010; Quamme et al., 2002), and that familiar-ity can be sensitive to contextual or source information (Tsivilis et al.,2001). For example, unitizing item and source information can lead toan increase in behavioral and ERP measures of familiarity (Bader etal., 2010; Diana et al., 2010; Wiegand et al., 2010). Although it is pos-sible that familiarity supported accurate source decisions for low con-fidence hits, no significant FN400 modulation was observed for theseitems, so there is little evidence to support this hypothesis.

A second, more speculative account of the late negativity that wasrelated to source memory for low confidence recognition is that it re-flects a form of “contextual familiarity”, consistent with the "Conver-gence, Recollection, and Familiarity Theory" (CRAFT) (Montaldi &Mayes, 2010), which proposed that item familiarity, context familiarity,

and recollection are each processed differently in the medial temporallobe memory system. The background for this idea is that episodicmemories may reflect the binding of neural representations of itemand context information (Diana et al., 2007; Eichenbaum et al., 2007).According to this "Binding Item in Context" (BIC) view, item familiaritymay reflect the strength of the item representation (e.g., a word) thatsupports recognition memory. Context information might include in-formation about the place and time it was encountered, and how itwas processed (see Ranganath, 2010 and also Montaldi & Mayes,2010, for review). Recollection can be thought of as the recall of contextinformation associated with an item, but it is possible that processingof a studied item could elicit weak activation of the associated contextrepresentation even when recollection fails. This type of contextual fa-miliarity signal, in turn, could support source discrimination. The cur-rent data offers physiological support for the proposed role of`context familiarity' (Montalid & Mayes, 2010) by providing the firstevidence which we are aware of that accurate contextual informationcan be retrieved without associated indications of either recollection-or item familiarity-based processing, consistent with both the BIC andCRAFT models. This 'context familiarity' hypothesis is admittedly posthoc though some additional ERP evidence exists which distinguishesbetween effects of context and familiarity (Ecker et al., 2007a; Tsiviliset al., 2001)—so further studies that test this possibility will benecessary.

Another possible account of the negative ERP effect of accuratesource for low confidence hits is that it might reflect a controlled sea-rch process, possibly via a top-down executive control mechanism.Consistent with this idea, some source memory studies have reporteda similar late negative going ERP (the Late Posterior Negativity,“LPN”), which is considered a functionally heterogeneous effectwith interpretations ranging from a controlled search process to re-sponse fluency (Friedman et al., 2005; Herron, 2007; Johansson andMecklinger, 2003). There are several key distinctions though, whichsuggest that the ERPs found in the current study are not likely to bereflecting the same type of processing as the LPN. For example,while the negative ERP observed in the current study emerged at an-terior sites from 600 to 800 ms, its topography was generally muchmore widespread, becoming insensitive to anterior/posterior factorsfrom 800 to 1000 ms. Additionally, the LPN is not generally observedin studies such as ours that utilize item judgments followed by sourcediscriminations (Duarte et al., 2004; Trott et al., 1999;Wilding, 2000).Finally, when the LPN has been observed in source memory tasks, ithas been found to be either insensitive to source accuracy (Herron,2007) or to be even larger for inaccurate source judgments(Johansson and Mecklinger, 2003; Wilding, 1999). Together, thesefactors make it difficult to attribute the observed effects as an LPN.

The current results join a growing number of ERP studies that pro-vide support for dual process models (Curran, 2000; Curran et al.,2002; Diana et al., 2006; Duzel et al., 1999, 2001; Leynes et al.,2005; Rugg and Curran, 2007; Woodruff et al., 2006; Yu and Rugg,2010) which propose that recognition is supported by separable sig-nals for recollection and familiarity, and provides initial evidence tosuggest that there may be different forms of familiarity processing(item and context), consistent with recent theories (Diana et al.,2007; Eichenbaum et al., 2007; Montaldi and Mayes, 2010;Ranganath, 2010). The finding of temporally and topographically dis-tinct recognition memory ERP effects is consistent with the idea thatrecognition cannot be accounted for by a single underlying neuralprocess. Moreover, these ERP effects were functionally dissociablewith respect to how they varied across confidence, even when sourceaccuracy was held constant. We know of no single process model ofrecognition that can account for these types of dissociations withoutadditional post-hoc assumptions.

The current results have implications for how one goes aboutassessing the neural correlates of recollection and familiarity, andfor current theories of recognition. For example, ERP and fMRI studies



of recollection and familiarity are often conducted either by assessingrecognition confidence or by contrasting source and item recognitionjudgments. The extent to which they have been successful at dissoci-ating these processes suggests that they can provide a rough index ofthese processes, and these methods have largely led to converging re-sults (Spaniol et al., 2009). However, a few studies that have usedthese methods have failed to find evidence for significant dissocia-tions (Gold et al., 2006; Wais et al., 2006). One potential account ofthese latter results then is that by assessing only item confidence orsource recognition they did not successfully isolate recollectionfrom familiarity. Given the relative simplicity of collecting item confi-dence and source confidence responses as we did in the currentstudy, it would seem important to do so in future studies examiningthese processes (Wais et al., 2010).

The extent to which the current results generalize to other testprocedures is currently unknown. However, the procedures thatwere used in the current experiment were chosen to reflect standarditem confidence paradigms and standard source memory paradigms,so we expect the results to be quite general. It is possible that sourceidentification tasks involving the identification of physical propertiessuch as color and location, modalities (auditory, visual), or the time ofpresentation at encoding, may have yielded different findings, andsource tasks with greater than two sources could also confer differentprocessing demands. There is also evidence that distinctions betweenintrinsic and extrinsic source features can differentially influencesource memory (Aly et al., 2010; Ecker et al., 2007a, 2007b; Nyhusand Curran, 2009), and the extent to which the current results aresensitive to these factors remains to be tested.

Moreover, although recollection was found to support only thehighest confidence recognition responses, other conditions wouldlikely lead recollection to support a wider range of responses. For ex-ample, conditions in which subjects are forced to respond using aconfidence scale with many more levels of confidence should leadsubjects to use a wider range of confidence responses for recollecteditems. Indeed, some evidence exists for graded ERP correlates of rec-ollection (Leynes and Phillips, 2008; Vilberg et al., 2006), consistentwith the Source Monitoring Framework (SMF) and various dual pro-cess models of recognition, which contend that source monitoringcan be supported by varying degrees of recollection (Hicks et al.,2002; Johnson et al., 1993).

Conclusions

Recollection appears to support relatively high confidence itemrecognition responses, whereas familiarity based responses vary di-rectly with item confidence. In addition, recollection supports accu-rate source recognition, whereas familiarity can be observed evenwhen recollection of source information fails. However, results alsoshow that accurate source recognition can occur even when recogni-tion confidence is not high, and reveal that episodic context can be re-trieved independent of recollection. Under these conditions, itappears that accurate source recognition occurs in the absence ofrecollection.

Supplementary data related to this article can be found online atdoi:10.1016/j.neuroimage.2012.04.031.

Acknowledgments

The authors would like to thank Steve Luck, John Olichney, JoshuaKoen, and Kathleen Baynes for helpful comments on earlier drafts, aswell as Javier Lopez-Calderon and Steve Luck for guidance in the de-velopment of ERPLAB analysis software.

This work was supported by T32 MH18882-22 to RJA, R01MH59352-01 to APY, R01MH068721 to CR, and R01MH083734 toCR and APY.

References

Addante, R.J., Watrous, A.J., Yonelinas, A.P., Ekstrom, A.D., Ranganath, C., 2011. Pre-stimulus theta activity predicts correct source memory retrieval [Research Sup-port, N.I.H., Extramural] Proc. Natl. Acad. Sci. U. S. A. 108 (26), 10702–10707.

Allan, K., Rugg, M.D., 1998. Neural correlates of cued recall with and without retrievalof source memory. Neuroreport 9 (15), 3463–3466.

Allan, K., Wilding, E.L., Rugg, M.D., 1998. Electrophysiological evidence for dissociableprocesses contributing to recollection. Acta Psychol. (Amst.) 98 (2–3), 231–252.

Aly, M., Knight, R.T., Yonelinas, A.P., 2010. Faces are special but not too special: sparedface recognition in amnesia is based on familiarity. Neuropsychologia 48 (13),3941–3948.

Bader, R., Mecklinger, A., Hoppstadter, M., Meyer, P., 2010. Recognition memory forone-trial-unitized word pairs: evidence from event-related potentials. NeuroImage50 (2), 772–781.

Bell, A.J., Sejnowski, T.J., 1995. An information-maximization approach to blind separa-tion and blind deconvolution. Neural Comput. 7 (6), 1129–1159.

Bowles, B., Crupi, C., Mirsattari, S.M., Pigott, S.E., Parrent, A.G., Pruessner, J.C., et al.,2007. Impaired familiarity with preserved recollection after anterior temporal-lobe resection that spares the hippocampus. Proc. Natl. Acad. Sci. U. S. A. 104(41), 16382–16387.

Cansino, S., Trejo-Morales, P., 2008. Neurophysiology of successful encoding and re-trieval of source memory. Cogn. Affect. Behav. Neurosci. 8 (1), 85–98.

Curran, T., 2000. Brain potentials of recollection and familiarity. Mem. Cognit. 28 (6),923–938.

Curran, T., 2004. Effects of attention and confidence on the hypothesized ERP correlatesof recollection and familiarity. Neuropsychologia 42 (8), 1088–1106.

Curran, T., Doyle, J., 2011. Picture superiority doubly dissociates the ERP correlates ofrecollection and familiarity. J. Cogn. Neurosci. 23 (5), 1247–1262.

Curran, T., Tanaka, J.W., Weiskopf, D.M., 2002. An electrophysiological comparison ofvisual categorization and recognition memory. Cogn. Affect. Behav. Neurosci. 2(1), 1–18.

Curran, T., DeBuse, C., Woroch, B., Hirshman, E., 2006. Combined pharmacological andelectrophysiological dissociation of familiarity and recollection. J. Neurosci. 26 (7),1979–1985.

Curran, T., DeBuse, C., Leynes, P.A., 2007. Conflict and criterion setting in recognitionmemory. J. Exp. Psychol. Learn. Mem. Cogn. 33 (1), 2–17.

Delorme, A., Makeig, S., 2004. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J. Neurosci.Methods 134 (1), 9–21.

Diana, R.A., Reder, L.M., Arndt, J., Park, H., 2006. Models of recognition: a review of ar-guments in favor of a dual-process account. Psychon. Bull. Rev. 13 (1), 1–21.

Diana, R.A., Yonelinas, A.P., Ranganath, C., 2007. Imaging recollection and familiarity inthe medial temporal lobe: a three-component model. Trends Cogn. Sci. 11 (9),379–386.

Diana, R.A., Yonelinas, A.P., Ranganath, C., 2010. Medial temporal lobe activity duringsource retrieval reflects information type, not memory strength. J. Cogn. Neurosci.22 (8), 1808–1818.

Diana, R.A., Van den Boom, W., Yonelinas, A.P., Ranganath, C., 2011. ERP correlates ofsource memory: unitized source information increases familiarity-based retrieval.Brain Res. 1367, 278–286.

Duarte, A., Ranganath, C., Winward, L., Hayward, D., Knight, R.T., 2004. Dissociable neu-ral correlates for familiarity and recollection during the encoding and retrieval ofpictures. Brain Res. Cogn. Brain Res. 18 (3), 255–272.

Duzel, E., Yonelinas, A.P., Mangun, G.R., Heinze, H.J., Tulving, E., 1997. Event-relatedbrain potential correlates of two states of conscious awareness in memory. Proc.Natl. Acad. Sci. U. S. A. 94 (11), 5973–5978.

Duzel, E., Cabeza, R., Picton, T.W., Yonelinas, A.P., Scheich, H., Heinze, H.J., et al., 1999.Task-related and item-related brain processes of memory retrieval. Proc. Natl.Acad. Sci. U. S. A. 96 (4), 1794–1799.

Duzel, E., Vargha-Khadem, F., Heinze, H.J., Mishkin, M., 2001. Brain activity evidence forrecognition without recollection after early hippocampal damage. Proc. Natl. Acad.Sci. U. S. A. 98 (14), 8101–8106.

Ecker, U.K., Zimmer, H.D., Groh-Bordin, C., 2007a. Color and context: an ERP study onintrinsic and extrinsic feature binding in episodic memory [Research Support,Non-U.S. Gov't] Mem. Cognit. 35 (6), 1483–1501.

Ecker, U.K., Zimmer, H.D., Groh-Bordin, C., 2007b. The influence of object and back-ground color manipulations on the electrophysiological indices of recognitionmemory [Research Support, Non-U.S. Gov't] Brain Res. 1185, 221–230.

Eichenbaum, H., Yonelinas, A.P., Ranganath, C., 2007. The medial temporal lobe andrecognition memory. Annu. Rev. Neurosci. 30, 123–152.

Friedman, D., 2000. Event-related brain potential investigations of memory and aging.Biol. Psychol. 54 (1–3), 175–206.

Friedman, D., Johnson Jr., R., 2000. Event-related potential (ERP) studies of memoryencoding and retrieval: a selective review. Microsc. Res. Tech. 51 (1), 6–28.

Friedman, D., Cycowicz, Y.M., Bersick, M., 2005. The late negative episodic memory ef-fect: the effect of recapitulating study details at test. Brain Res. Cogn. Brain Res. 23(2–3), 185–198.

Gold, J.J., Smith, C.N., Bayley, P.J., Shrager, Y., Brewer, J.B., Stark, C.E., et al., 2006. Itemmemory, source memory, and the medial temporal lobe: concordant findingsfrom fMRI and memory-impaired patients. Proc. Natl. Acad. Sci. U. S. A. 103 (24),9351–9356.

Gruber, M.J., Otten, L.J., 2010. Voluntary control over prestimulus activity related toencoding. J. Neurosci. 30 (29), 9793–9800.

Guo, C., Duan, L., Li, W., Paller, K.A., 2006. Distinguishing source memory and itemmemory: brain potentials at encoding and retrieval. Brain Res. 1118 (1), 142–154.



Herron, J.E., 2007. Decomposition of the ERP late posterior negativity: effects of retriev-al and response fluency. Psychophysiology 44 (2), 233–244.

Hicks, J.L., Marsh, R.L., Ritschel, L., 2002. The role of recollection and partial informationin source monitoring. J. Exp. Psychol. Learn. Mem. Cogn. 28 (3), 503–508.

Johansson, M., Mecklinger, A., 2003. The late posterior negativity in ERP studies of ep-isodic memory: action monitoring and retrieval of attribute conjunctions. Biol.Psychol. 64 (1–2), 91–117.

Johnson, M.K., Hashtroudi, S., Lindsay, D.S., 1993. Source monitoring. Psychol. Bull. 114(1), 3–28.

Johnson, M.K., Kounios, J., Nolde, S.F., 1997. Electrophysiological brain activity andmemory source monitoring. Neuroreport 8 (5), 1317–1320.

Johnson Jr., R., Kreiter, K., Zhu, J., Russo, B., 1998. A spatio-temporal comparison of se-mantic and episodic cued recall and recognition using event-related brain poten-tials. Brain Res. Cogn. Brain Res. 7 (2), 119–136.

Johnson, M.K., Verfaellie, M., Dunlosky, J., 2008. Introduction to the special section onintegrative approaches to source memory. J. Exp. Psychol. Learn. Mem. Cogn. 34(4), 727–729.

Kim, A.S., Vallesi, A., Picton, T.W., Tulving, E., 2009. Cognitive association formation inepisodic memory: evidence from event-related potentials [Research Support,Non-U.S. Gov't] Neuropsychologia 47 (14), 3162–3173.

Kirwan, C.B., Wixted, J.T., Squire, L., 2008. Activity in the medial temporal lobe predictsmemory strength, whereas activity in the prefrontal cortex predicts recollection. J.Neurosci. 28, 10541–10548.

Klem, G.H., Luders, H.O., Jasper, H.H., Elger, C., 1999. The ten-twenty electrode systemof the International Federation. The International Federation of Clinical Neuro-physiology. Electroencephalogr. Clin. Neurophysiol. Suppl. 52, 3–6.

Levine, B., Freedman, M., Dawson, D., Black, S., Stuss, D., 1999. Ventral frontal contribu-tion to self-regulation: convergence of episodic memory and inhibition. Neurocase5 (3), 263–275.

Leynes, P.A., Phillips, M.C., 2008. Event-related potential (ERP) evidence for varied rec-ollection during source monitoring. J. Exp. Psychol. Learn. Mem. Cogn. 34 (4),741–751.

Leynes, P.A., Landau, J., Walker, J., Addante, R.J., 2005. Event-related potential evidencefor multiple causes of the revelation effect. Conscious. Cogn. 14 (2), 327–350.

Mecklinger, A., 2006. Electrophysiological measures of familiarity memory. Clin. EEGNeurosci. 37 (4), 292–299.

Mecklinger, A., Brunnemann, N., Kipp, K., 2011. Two processes for recognition memoryin children of early school age: an event-related potential study. J. Cogn. Neurosci.23 (2), 435–446.

Mitchell, K.J., Johnson, M.K., 2009. Source monitoring 15 years later: what have welearned from fMRI about the neural mechanisms of source memory? Psychol.Bull. 135 (4), 638–677.

Montaldi, D., Mayes, A.R., 2010. The role of recollection and familiarity in the functionaldifferentiation of the medial temporal lobes. Hippocampus 20, 1291–1314, http://dx.doi.org/10.1002/hipo.20853.

Montaldi, D., Spencer, T.J., Roberts, N., Mayes, A.R., 2006. The neural system that medi-ates familiarity memory. Hippocampus 16 (5), 504–520.

Nyhus, E., Curran, T., 2009. Semantic and perceptual effects on recognition memory:evidence from ERP [Research Support, N.I.H., Extramural Research Support, U.S.Gov't, Non-P.H.S.] Brain Res. 1283, 102–114.

Otten, L.J., Quayle, A.H., Akram, S., Ditewig, T.A., Rugg, M.D., 2006. Brain activity beforean event predicts later recollection. Nat. Neurosci. 9 (4), 489–491.

Paller, K.A., Voss, J.L., Boehm, S.G., 2007. Validating neural correlates of familiarity.Trends Cogn. Sci. 11 (6), 243–250.

Parks, C.M., Yonelinas, A.P., 2007. Moving beyond pure signal-detection models: com-ment on Wixted (2007). Psychol. Rev. 114 (1), 188–202 discussion 203–189.

Parks, C.M., Yonelinas, A.P., 2009. Evidence for a memory threshold in second-choice rec-ognition memory responses. Proc. Natl. Acad. Sci. U. S. A. 106 (28), 11515–11519.

Quamme, J.R., Frederick, C., Kroll, N.E., Yonelinas, A.P., Dobbins, I.G., 2002. Recognitionmemory for source and occurrence: the importance of recollection. Mem. Cognit.30 (6), 893–907.

Ranganath, C., 2010. A unified framework for the functional organization of the medialtemporal lobes and the phenomenology of episodic memory. Hippocampus 20(11), 1263–1290.

Ranganath, C., Yonelinas, A.P., Cohen, M.X., Dy, C.J., Tom, S.M., D'Esposito, M., 2004. Dis-sociable correlates of recollection and familiarity within the medial temporal lobes.Neuropsychologia 42 (1), 2–13.

Rugg, M.D., Curran, T., 2007. Event-related potentials and recognition memory. TrendsCogn. Sci. 11 (6), 251–257.

Rugg, M.D., Wilding, E.L., 2000. Retrieval processing and episodic memory. TrendsCogn. Sci. 4 (3), 108–115.

Rugg, M.D., Mark, R.E., Walla, P., Schloerscheidt, A.M., Birch, C.S., Allan, K., 1998a. Dis-sociation of the neural correlates of implicit and explicit memory. Nature 392(6676), 595–598.

Rugg, M.D., Walla, P., Schloerscheidt, A.M., Fletcher, P.C., Frith, C.D., Dolan, R.J., 1998b. Neuralcorrelates of depth of processing effects on recollection: evidence from brain potentialsand positron emission tomography. Exp. Brain Res. 123 (1–2), 18–23.

Senkfor, A.J., Van Petten, C., 1998. Who said what? An event-related potential investigationof source and item memory. J. Exp. Psychol. Learn. Mem. Cogn. 24 (4), 1005–1025.

Spaniol, J., Davidson, P.S., Kim, A.S., Han, H., Moscovitch, M., Grady, C.L., 2009. Event-re-lated fMRI studies of episodic encoding and retrieval: meta-analyses using activa-tion likelihood estimation [Meta-Analysis Research Support, Non-U.S. Gov't]Neuropsychologia 47 (8–9), 1765–1779.

Trott, C.T., Friedman, D., Ritter, W., Fabiani, M., Snodgrass, J.G., 1999. Episodic primingand memory for temporal source: event-related potentials reveal age-related dif-ferences in prefrontal functioning. Psychol. Aging 14 (3), 390–413.

Tsivilis, D., Otten, L.J., Rugg, M.D., 2001. Context effects on the neural correlates of rec-ognition memory: an electrophysiological study. Neuron 31 (3), 497–505.

Vilberg, K.L., Moosavi, R.F., Rugg, M.D., 2006. The relationship between electrophysio-logical correlates of recollection and amount of information retrieved. Brain Res.1122 (1), 161–170.

Voss, J.L., Paller, K.A., 2007. Neural correlates of conceptual implicit memory and theircontamination of putative neural correlates of explicit memory. Learn. Mem. 14(4), 259–267.

Wais, P.E., Wixted, J.T., Hopkins, R.O., Squire, L.R., 2006. The hippocampus supportsboth the recollection and the familiarity components of recognition memory. Neu-ron 49 (3), 459–466.

Wais, P.E., Squire, L.R., Wixted, J.T., 2010. In search of recollection and familiarity sig-nals in the hippocampus. J. Cogn. Neurosci. 22 (1), 109–123.

Wiegand, I., Bader, R., Mecklinger, A., 2010. Multiple ways to the prior occurrence of anevent: an electrophysiological dissociation of experimental and conceptually driv-en familiarity in recognition memory. Brain Res. 1360, 106–118.

Wilding, E.L., 1999. Separating retrieval strategies from retrieval success: an event-re-lated potential study of source memory. Neuropsychologia 37 (4), 441–454.

Wilding, E.L., 2000. In what way does the parietal ERP old/new effect index recollec-tion? [Research Support, Non-U.S. Gov't] Int. J. Psychophysiol. 35 (1), 81–87.

Wilding, E.L., Rugg, M.D., 1996. An event-related potential study of recognition memo-ry with and without retrieval of source. Brain 119 (Pt 3), 889–905.

Wilding, E.L., Doyle, M.C., Rugg, M.D., 1995. Recognition memory with and without re-trieval of context: an event-related potential study. Neuropsychologia 33 (6),743–767.

Wolk, D.A., Schacter, D.L., Berman, A.R., Holcomb, P.J., Daffner, K.R., Budson, A.E., 2004.An electrophysiological investigation of the relationship between conceptual flu-ency and familiarity [Comparative Study Research Support, Non-U.S. Gov't Re-search Support, U.S. Gov't, P.H.S.] Neurosci. Lett. 369 (2), 150–155.

Woodruff, C.C., Hayama, H.R., Rugg, M.D., 2006. Electrophysiological dissociation of theneural correlates of recollection and familiarity. Brain Res. 1100 (1), 125–135.

Yonelinas, A.P., 1999. The contribution of recollection and familiarity to recognition andsource–memory judgments: a formal dual-process model and an analysis of receiveroperating characteristics. J. Exp. Psychol. Learn. Mem. Cogn. 25 (6), 1415–1434.

Yonelinas, A.P., 2001a. Components of episodic memory: the contribution of recollec-tion and familiarity. Philos. Trans. R. Soc. Lond. B Biol. Sci. 356 (1413), 1363–1374.

Yonelinas, A.P., 2001b. Consciousness, control, and confidence: the 3 Cs of recognitionmemory. J. Exp. Psychol. Gen. 130 (3), 361–379.

Yonelinas, A.P., Levy, B.J., 2002. Dissociating familiarity from recollection in human rec-ognition memory: different rates of forgetting over short retention intervals. Psy-chon. Bull. Rev. 9 (3), 575–582.

Yonelinas, A.P., Parks, C.M., 2007. Receiver operating characteristics (ROCs) in recogni-tion memory: a review. Psychol. Bull. 133 (5), 800–832.

Yonelinas, A.P., Hopfinger, J.B., Buonocore, M.H., Kroll, N.E., Baynes, K., 2001. Hippo-campal, parahippocampal and occipital–temporal contributions to associativeand item recognition memory: an fMRI study. Neuroreport 12 (2), 359–363.

Yonelinas, A.P., Kroll, N.E., Quamme, J.R., Lazzara, M.M., Sauve, M.J., Widaman, K.F., etal., 2002. Effects of extensive temporal lobe damage or mild hypoxia on recollec-tion and familiarity. Nat. Neurosci. 5 (11), 1236–1241.

Yonelinas, A.P., Aly, M., Wang, W.C., Koen, J.D., 2010. Recollection and familiarity: ex-amining controversial assumptions and new directions. Hippocampus 20 (11),1178–1194.

Yovel, G., Paller, K.A., 2004. The neural basis of the butcher-on-the-bus phenomenon:when a face seems familiar but is not remembered. NeuroImage 21 (2), 789–800.

Yu, S.S., Rugg, M.D., 2010. Dissociation of the electrophysiological correlates of familiar-ity strength and item repetition. Brain Res. 1320, 74–84.


Examining ERP correlates of recognition memory: Evidence of accurate source recognition without...

Documents

Transcript of Examining ERP correlates of recognition memory: Evidence of accurate source recognition without...