author's personal copy - university of leicester...accepted 6 october 2011 available online 20...

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

Event related potentials to digit learning: Tracking neurophysiologic changesaccompanying recall performance

Marijtje L.A. Jongsma a, Niels J.H.M. Gerrits a, Clementina M. van Rijn a,Rodrigo Quian Quiroga b, Joseph H.R. Maes a,⁎a Donders Institute for Brain, Cognition, and Behavior, DCC, Radboud University Nijmegen, Montessorilaan 3, 6525 HR Nijmegen, The Netherlandsb Department of Engineering, University of Leicester, Department of Engineering, University of Leicester, Leicester LE1 7RH, UK

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

Article history:Received 6 January 2011Received in revised form 22 September 2011Accepted 6 October 2011Available online 20 October 2011

Keywords:Event related potentialsDigit spanLearningMemoryP300

The aim of this study was to track recall performance and event-related potentials (ERPs) across multiple tri-als in a digit-learning task. When a sequence is practiced by repetition, the number of errors typically de-creases and a learning curve emerges. Until now, almost all ERP learning and memory research has focusedon effects after a single presentation and, therefore, fails to capture the dynamic changes that characterizea learning process. However, the current study used a free-recall task in which a sequence of ten auditorydigits was presented repeatedly.Auditory sequences of ten digits were presented in a logical order (control sequences) or in a random order(experimental sequences). Each sequence was presented six times. Participants had to reproduce the sequenceafter each presentation. EEG recordings weremade at the time of the digit presentations. Recall performance forthe control sequences was close to asymptote right after the first learning trial, whereas performance for theexperimental sequences initially displayed primacy and recency effects. However, these latter effects graduallydisappeared over the six repetitions, resulting in near-asymptotic recall performance for all digits. The perfor-mance improvement for the middle items of the list was accompanied by an increase in P300 amplitude, imply-ing a close correspondence between this ERP component and the behavioral data. These results, which werediscussed in the framework of theories on the functional significance of the P300 amplitude, add to the scarceempirical data on the dynamics of ERP responses in the process of intentional learning.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

1.1. ERPs in cognitive tasks

Event-related potentials (ERPs) are time-locked voltage fluctua-tions in the EEG, resulting from neuronal responses to sensory,motor, or cognitive events (Rugg and Coles, 1996). Because of theirhigh temporal resolution, neural correlates of the successive aspectsof information processing can be instantaneously measured. There-fore, ERPs are particularly useful for studying changes that areexpected to occur in information processing, for example, in explicitor implicit learning tasks. In a previous study (Jongsma et al., 2006),we focused on tracking neurophysiologic changes accompanyingrecall performance in an implicit learning task. Implicit learning, orprocedural learning, is defined as learning complex information

without the participant being able to verbalize the acquired knowl-edge. Instead, explicit learning, or declarative learning, results in aconscious, verbalized recollection of what and how specific knowl-edge is acquired (Seger, 1994; Shanks, 2010). The focus of the presentstudy was on ERP correlates of explicit learning.

One of the most widely investigated ERP components in informa-tion processing is the P300, also referred to as ‘P3’ or ‘P3b’ component(e.g., Polich, 2007). This positive wave reaches its highest amplitudebetween 300 and 600 ms after stimulus presentation and has its max-imal amplitude on the central-posterior part of the brain. Its ampli-tude is sensitive to a number of task parameters, such as stimulusprobability and target-to-target interval (Polich, 2007). In terms ofits underlying cognitive, functional significance, a dominant viewis that the P300 reflects a process that updates the hypotheses ormodels of the environment held by the participant (Donchin andColes, 1988; Duncan-Johnson and Donchin, 1982). A high P300 ampli-tude has also been associated with memory performance. For exam-ple, Karis et al. (1984) were the first to link recall probability toincreased P300 amplitude. Later studies (Bentin et al., 1992; Curran,2004; Curran and Clearly, 2003; Fabiani et al., 1986; Kim et al.,2001; Paller et al., 1988; Rushby et al., 2002) found a similar relation-ship. From these and other experiments, Polich (2007) concluded

International Journal of Psychophysiology 85 (2012) 41–48

⁎ Corresponding author at: Department of Biological Psychology, Radboud UniversityNijmegen, PO Box 9104, 6500HE Nijmegen, The Netherlands. Tel.: +31 24 3615992;fax: +31 24 3616066.

E-mail addresses: [email protected] (M.L.A. Jongsma),[email protected] (N.J.H.M. Gerrits), [email protected] (C.M. van Rijn),[email protected] (R.Q. Quiroga), [email protected] (J.H.R. Maes).

0167-8760/$ – see front matter © 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.ijpsycho.2011.10.004

Contents lists available at SciVerse ScienceDirect

International Journal of Psychophysiology

j ourna l homepage: www.e lsev ie r .com/ locate / i jpsycho


that stimuli that are well encoded and stored in memory produce alarger P300 compared to less successfully encoded and stored stimuli.

1.2. ERPs and the serial-position effect in learning and memory

The serial position effect is one of the most well-establishedphenomena in studies on learning and memory (Lewandowsky andFarrell, 2008; Murdock, 1962; Rundus, 1971). After presenting a listof to-be-learned words or numbers, recall is better for items at the be-ginning (primacy effect) and the end (recency effect) of the list, com-pared to items presented in the middle of the list. Recently, Azizianand Polich (2007) investigated serial position memory using both be-havioral and ERP measurements (see also Wiswede et al., 2007). Theyfound greater recall and larger P300 amplitudes for primacy andrecency items compared to middle words. It was suggested that theprimacy items were memorized better because more attentionalresources were available for the initial stimuli. Presumably, theseresources were reflected in increased P300 amplitude.

When a sequence is practiced by repetition, the number of errorstypically decreases and a learning curve emerges (Anderson, 2000).Until now, almost all ERP learning and memory research has focusedon effects after a single presentation and, therefore, fails to capturethe dynamic changes that characterize a learning process. In onestudy, Johnson et al. (1985) used a study-test recognition paradigmin which target words had to be memorized in a study conditionand were presented among distracter words in a test condition. Thisprocedure was repeated four times, using the same target words,but different distracters. They found an increase in P300 amplitudeover repetitions on the parietal (Pz) scalp as the number of correctlyclassified items increased over repetitions. The increase in amplitudewas interpreted as reflecting the increased discriminability of thetarget words as the memory trace increased in strength.

1.3. Main aim

The primary aim of this study was to contribute to the very scarceempirical findings on ERP correlates of repeated exposure to to-be-learned material. To this end, we used a recall task in which asequence of ten auditory digits was presented repeatedly. Based onthe notion of the serial-position effect, memory performance(remembering the items in the correct position in the sequence)was expected to gradually increase for middle items as a result ofthese repeated presentations. Instead, the ‘recency’, but especiallythe ‘primacy’, items were expected to yield a relatively high recallperformance right after their first presentation (see Fig. 1 for a graph-ical representation of the expected memory performance).

The main question addressed was of an empirical nature and con-cerned the changes in P300 amplitude that occur in the learningprocess that was expected to take place especially for middle items.Although mainly focusing on the P300 component, learning-inducedchanges in the N100, P200, and N200 ERP components were alsoinvestigated. Although not explicitly designed for this purpose, asecondary question of the experiment, addressed in the Discussion,was related to possible theoretical implications of the observed ERPchanges in terms of cognitive processes they might reflect.

2. Methods

2.1. Participants and apparatus

Ten students from Radboud University Nijmegen, The Netherlands(6 females, 4 males, mean age 24.7±3.3 years) participated forcourse credits. Only healthy participants, not using medication andwithout a neurological or psychiatric history, were accepted. Thestudy was approved by a local ethics committee and all participantssigned a written statement of informed consent. The students were

tested in a dimly lit, electrically shielded, sound-attenuating cubicle(inside dimensions: 2×2.2×2 m). Responses were collected viaa numerical keyboard. Stimuli consisted of spoken numbers (inDutch, male voice) and were presented through headphones with a21–18.000 Hz frequency band width. Each digitized auditory stimuluswas edited so that all stimuli were similar in frequency and amplitudeand had a duration of 200 ms. Stimuli were presented with an 800-msinter-stimulus interval (ISI) and were presented at a sound pressurelevel of 65 dB. The program E-Prime was used for stimulus presenta-tions and response collection.

2.2. Experimental design and procedure

Each participant was presented with control and experimentalstimulus sequences. Each sequence consisted of ten spoken numbers(0–9) and was repeated six times. Ten different control and ten differ-ent experimental sequences were presented alternately. Each partic-ipant first received a control sequence. Control sequences consistedof digits in a logical order. The first control sequence consisted of asequence from zero to nine, the second control sequence consistedof a sequence from one to zero, the third control sequence wentfrom two to one, and so forth. In this way, ten different control se-quences of ten digits were constructed. Experimental sequences con-sisted of the same ten digits, but put in a randomized order (see Fig. 2for a representation of the paradigm).

Each sequence was repeated six times, to invoke a learning pro-cess. The sequences were counterbalanced over the ten participantsby using a Latin square design. By doing so, each participant receivedthe same sequences, but at different times during the experiment.Because stimuli were presented in an ‘eyes closed’ condition, partici-pants were asked to open their eyes after each repetition and toreproduce the presented sequence in the order of presentation via anumerical keyboard. There was no time limit for responding and allresponses were recorded for later analyses. Before receiving a newsequence, participants were asked to close their eyes again. It tookapproximately 1 h to complete the experiment.

Fig. 1. The figure shows a schematic representation of the expected memory perfor-mance (y-axis). On the first presentation of the sequence (x-axis), a u-shaped curveis expected to arise due to primacy and recency effects. With repetition (z-axis) learn-ing curves are expected to emerge, especially with respect to the middle stimuli withina sequence.

42 M.L.A. Jongsma et al. / International Journal of Psychophysiology 85 (2012) 41–48


2.3. EEG recordings

EEG (band-pass: DC — 100 Hz, sampling rate 2000 Hz) wasrecorded with a SYNAMPS amplifier (Neuroscan, Herndon, VA) from19 Ag/AgCl electrodes (Fp1, Fp2, F7, F3, Fz, F4, F8, T3, C3, Cz, C4, T4,T5, P3, Pz, P4, T6, O1, O2) mounted in an elastic cap (EASICAP, FMS,Breitenbrunn, GER) and filled with Electrocap (ECI) Electrode gelfrom MedCaT B.V. at placements based on the International 10–20recording system (American Encephalographic Society, 1994), refer-enced to linked mastoids and stored on disk for offline processing.Vertical and horizontal eye movements were recorded by two addi-tional bipolar channels placed above and below the right eye and onthe outer canthi of each eye. The impedance of each electrode waskept below 5 kΩ.

2.4. Data-processing

The behavioral responses were analyzed by comparing the partici-pant's responses with the correct responses. For example, if the pre-sented sequence had been ‘0826539417’ and the participant hadreproduced the sequence ‘0826590731’, the response at each of thepositions 1–5 was considered ‘correct’, whereas the response at posi-tions 6–10 was ‘incorrect’. In case the participant had produced thesequence ‘0826951347’, the response at each of the positions 1, 2, 3,4, and 10 was correct and that at positions 5–9 incorrect. If a partici-pant produced fewer than 10 responses, the maximum number ofcorrect responses was equal to the total number of responses made.For example, given the sample sequence above, a ‘0826’ responsewould yield a correct response for positions 1–4, and an incorrectresponse for positions 5–10. However, in the case of fewer than 10responses, one ‘empty’ position was tolerated; for example, with aresponse consisting of the sequence ‘08263’, the response at stimuluspositions 1, 2, 3, 4, and 6 would be considered correct.

In this way, using the 10 control and 10 experimental sequences,mean proportion correct responses were computed for each stimulusposition at each repetition, separately for each sequence type.

EEG data were analyzed in BrainVision Analyser. For each partici-pant, the entire signal was rereferenced with linked mastoids as areference, ocular correction was applied (Gratton et al., 1983),and segmented into epochs from −512 ms to 512 ms around all

(n=1200) presented auditory stimuli. Next, each epoch wasbaseline-corrected at 512 ms before stimulus presentation and fil-tered (0.5–40 Hz). Then, ERPs elicited by the auditory stimuli wereaveraged per type of sequence (control or experimental) over theten sequences. Accordingly, the averaged ERPs were based on 10(stimuli)×6 (repetitions)×2 (sequence types)=120 responses foreach participant.

The average ERPs contained a clearly distinguishable N100 (meanlatency: 138 ms), P200 (mean latency: 220 ms), N200 (mean latency:280 ms), and P300 (mean latency: 335 ms) component. After exam-ining scalp distributions, components appeared to be most pro-nounced on the midline electrodes Fz, Cz, and Pz. Thus, for the sakeof data reduction, these electrodes were used for further analyses.For each electrode, the amplitude of the four components was deter-mined by peak-picking.

2.5. Statistical analysis

2.5.1. Behavioral analysisTo examine the effect of sequence type, repetition, and stimulus

position on correct recall (in proportion), a three-within subjectsANOVA was performed with the factors: sequence type (two levels:control vs. experimental), repetition (six levels: 1, 2, 3, 4, 5, 6) andstimulus position (ten levels: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).

2.5.2. EEG analysisThe effect of sequence type, repetition, and stimulus position on

the amplitude (μV) of each of the four ERP components (N100,P200, N200, P300) was investigated with a three-within subjectsANOVA, using the same levels as in the behavioral analysis. The ana-lyses were performed for each electrode-site (Fz, Cz, Pz) separately.

Given that our main research question concerned the relationshipbetween learning and the P300 component, in our analyses we fo-cussed on the effect of repetition for each stimulus position, and onthe effect of stimulus position for each level of the repetition factor,for both the proportion correct recall and the P300 amplitude. Be-cause of heterogeneous covariances for both the behavioral and EEGdata, we used the Greenhouse–Geisser correction in all statistical an-alyses. When reporting F-statistics, we present the original degrees offreedom (dfs) and the value of the Epsilon parameter (ε) that was

Fig. 2. The figure shows a schematic example of the used paradigm. A sequence of ten spoken numbers (in Dutch) was presented (x-axis). This sequence was repeated six times(z-axis). Both experimental sequences (depicted on the left) followed by control sequences (depicted on the right) were presented.

43M.L.A. Jongsma et al. / International Journal of Psychophysiology 85 (2012) 41–48


used to adjust these dfs to correct for sphericity violation. The level ofsignificance was set at pb .05 throughout.

3. Results

3.1. Behavioral data

Fig. 3 shows a graphic representation of the behavioral resultsfor both the experimental and control sequences.

A sequence type×repetition×stimulus position ANOVA on theproportion correct recall revealed a significant effect for each of thethree main factors, as well as for all possible interactions amongthem (Fs>12.82, psb .001). The significant sequence type×repeti-tion×stimulus position interaction (F(45, 405)=13.75, pb .001,ε=.13) was examined further by performing a repetition×stimulusposition ANOVA for each sequence type separately.

For control sequences, the ANOVA revealed no significant effects(maximum F(5, 45)=4.29, p=.058, ε=.24, for the repetition factor),suggesting the absence of learning processes. For experimentalsequences, the repetition×stimulus position was highly significant(F(45, 405)=13.96, pb .001, ε=.13), as were each of the main factors(Fs>35.76, psb .001). Subsequent analysis of the effect of the repeti-tion factor for each stimulus position separately revealed a significanteffect for each stimulus position (Fs>9.56, psb .001) except for thefirst position (F(5, 45)=2.74, p=.10, ε=.38). For each of stimuluspositions 2–10, the repetition effect reflected significant linear andquadratic trends. Specifically, polynomial contrasts revealed thatthe repetition effect reflected a linear increase for each of stimuluspositions 2–10 (Fs(1, 9)>16.0, psb .01), and a significant quadratictrend for stimulus position 2 and each of positions 5–10 (Fs(1, 9)>6.37, psb .05). The absence of a repetition effect for the first stimulusposition reflects near-asymptotic recall performance for items at thisposition from the very first presentation on. Analysis of the effect ofstimulus position at each level of the repetition factor revealed a sig-nificant effect for each of repetitions 1–3 (Fs>9.40, psb .001), but notfor each of the repetitions 4–6 (Fsb2.82, ps>.08). The significantstimulus position effect for the initial three repetitions reflected sig-nificant linear (Fs(1, 9)>29.05, psb .001) and cubic (Fs(1, 9)>7.62,psb .05) trends. Looking at the proportion of correct data, the cubictrends mainly reflected an increase in proportion correct recall forthe last item position (position 10), relative to the preceding two

item positions (positions 8 and 9). Specifically, the mean proportionsfor positions 8, 9, and 10 were 0.13, 0.15, and 0.23 for repetition1, 0.43, 0.44, and 0.51 for repetition 2, and 0.71, 0.71, and 0.75, forrepetition 3. The absence of a significant stimulus position effect forthe final three repetitions reflects (near-) asymptotic recall perfor-mance for the digits at each stimulus position.

3.2. EEG data

Fig. 4 displays the grand average ERPs.The amplitudes of all four ERP components in response to the first

stimulus were relatively large compared to subsequent stimuli. Be-cause it is possible that effects are due solely to this initial reaction,two analyses were performed: one using all ten stimuli and onewith the first stimulus excluded.

A sequence type×repetition×stimulus position ANOVA was per-formed for each of the N100, P200, N200, and P300 ERP componentsfor each of the Fz, Cz, and Pz electrode positions. Moreover, each anal-ysis was performed with and without the first stimulus included, im-plying 24 ANOVAs. Below, we only present the significant main andinteraction effects for each component and electrode position; effectsnot explicitly mentioned were all non-significant.

The N100 component had a decreasing amplitude across stimulus po-sitions on the Fz (F(9, 81)=4.57, p=.01, ε=.32), Cz (F(9, 81)=11.35,pb .001, ε=.27), and Pz electrodes (F(9, 81)=12.94, pb .001, ε=.33;main effect of stimulus position). Furthermore, a decreasing amplitudeacross repetitions was present on the Pz electrode (main effect of repeti-tion, F(5, 45)=3.18, p=.04, ε=.58). All effects failed to reach signifi-cance when excluding the first stimulus.

The amplitude of the P200 increased over repetitions on the Pzelectrode (main effect of repetition, F(5, 45)=3.75, p=.016,ε=.71) and decreased across stimulus positions on the Fz (maineffect of stimulus position, F(9, 81)=23.58 pb .001, ε=.35), Cz (F(9,81)=34.50, pb .001, ε=.33), and Pz electrodes (F(9, 81)=16.65,pb .001, ε=.43). After exclusion of the first stimulus, all these effectsremained significant.

The N200 amplitude was larger for control sequences than for ex-perimental sequences on the Fz (main effect of sequence type, F(1,9)=11.21, p=.01, ε=1) and Cz electrodes (F(1, 9)=7.72, p=.02,ε=1). Furthermore, its amplitude increased over stimulus positionon the Fz (main effect of stimulus position, F(9, 81)=4.59, p=.01,

Fig. 3. The figure depicts the total correct recall proportion (y-axis) for experimental sequences (depicted on the left) and control sequences (depicted on the right). Stimulusposition is depicted on the x-axes, repetitions on the y-axes. The correct recall proportion increased over repetitions for the experimental sequences, but this increase differedfor consecutive stimulus positions. The correct recall proportion for the control sequences was immediately close to 1. No repetition nor stimulus effects were found for thesesequences.



ε=.42), Cz (F(9, 81)=3.84, p=.02, ε=.34) and Pz electrodes (F(9,81)=3.58, p=.03, ε=.33). After excluding the first stimulus, onlythe larger amplitude for the control sequences on Fz and Cz remainedsignificant.

The P300 amplitude increased across stimulus positions on theFz (main effect of stimulus position, F(9, 81)=4.28, p=.01, ε=.40)and Cz electrodes (F(9, 81)=3.09, p=.03, ε=.41). In addition, a se-quence type×repetition interaction was present for the Pz electrode(F(5, 45)=4.31, p=.02, ε=.49), reflecting that the increase overrepetitions was not the same for the sequence types. The amplitudeincreased over repetitions for the experimental sequences (maineffect of repetition, F(5, 45)=7.13, p=.001, ε=.59), but remainedconstant for control sequences (Fb1, p=.423). After excluding thefirst stimulus, only the increasing amplitude over repetitions for theexperimental sequences on Pz remained significant. See also Fig. 5.

Because of our specific interest in the link between EEG and recalldata, and the fact that the P300 on Pz displayed a differential repeti-tion effect as a function of sequence type, for the experimentalsequences, we further examined the effects on the Pz P300 ampli-tudes of the repetition and stimulus position factors separately. Theeffect of repetition was (near) significant for items at position 5(F=2.66, p=.057), 6 (F=3.18, p=.031), 7 (F=3.09, p=.032), and8 (F=2.66, p=.058), in each case reflecting a significant linearincrease (Fs(1, 9)>6.10, psb .05). However, the repetition effect wasclearly insignificant for items at positions 1–4 (Fsb2.26, ps>.15),and 9–10 (Fsb1, ps>.60). These results implicate that the mostmarked amplitude changes (increase) occurred for the middleitems. The effect of stimulus position was not significant for any ofthe repetitions (Fsb1.58, ps>.20). However, on the very first repeti-tion, the P300 amplitude differences across stimulus positions dis-played a significant quadratic trend (p=.047). These results reflecta relatively high P300 amplitude for items at the beginning and endof the list right from the first repetition on (relative to middle

items), and a decreased difference between items at these differentpositions from the second repetition on.

4. Discussion

4.1. Effects of digit learning on ERP components and recall performance

The present study investigated the relationship between the am-plitude of four ERP components, N100, P200, N200, P300, and memo-ry performance across repeated presentations of to-be-learned digits.The proportion of correct recall for the control sequences was close to1.0 from the first presentation on. Experimental sequences initiallydisplayed a primacy and a recency effect, which disappeared in thecourse of the repetitions, when the proportions approached 1.0 forall stimulus positions. A primacy effect was reflected in the fact that,for the first item, recall performance was already close to the asymp-tote from the first repetition on, with a corresponding absence of arepetition effect for that item. At the same time, performance forthe items at the other list positions improved with increasing repeti-tions, yielding significant linear trends. Some evidence for a recencyeffect was obtained from the recall data in the form of a cubic trendwhich was observed in each of the first three repetitions, where theitem in the last position was recalled better than was the case forthe preceding two items. Notably, stronger evidence for a primacythan a recency effect was also reported by Azizian and Polich(2007), who used a single-trial learning experiment rather than thepresent multi-trial set-up. The present experimental sequences con-sisted of digits presented in a random order, whereas the digitsin the control sequences were presented in a logical order. It washypothesized that a learning process would only be induced inthe experimental sequences. This hypothesis was confirmed by thebehavioral data. Because all other variables were kept constant,

Fz

1

2

3

4

5

6N

100

P20

0N

200

P30

0

Cz

1

2

3

4

5

6

N10

0

P20

0N

200

P30

0

Pz

1

2

3

4

5

6

N10

0

P20

0N

200

P30

0

repe

titi

on

repe

titi

on

repe

titi

on

+

-

5µV

Fig. 4. The figure depicts grand average ERPs. ERPs were averages over participants and over stimuli within a trial, emphasizing the repetition effects of experimental compared tocontrol trials. Solid lines depict grand average ERPs elicited by experimental sequences whereas dotted lines depict grand average ERPs elicited by experimental sequences. ERPs aredepicted for Fz (left), Cz (middle), and Pz (right) electrodes. Moreover, ERPs to all 6 repetitions are depicted from bottom to top. Note the increment of the P300 on consecutiverepetitions for experimental trial only.



differences between control and experimental sequences can be at-tributed to learning and memory processes.

With respect to the N100 and N200 components, no clear evi-dence was found that these two components were influenced bylearning, at least not in the present task. Most effects were due tothe response to the first stimulus and disappeared when this re-sponse was excluded from the analysis.

The P200 amplitude at the Pz electrode increased over repetitions.However, no condition×repetition interaction was present, indicat-ing that this increase in amplitude was equal for both control and ex-perimental sequences. This suggests that the P200 amplitude increasewas not related to learning and memory in the present paradigm.

The increase in proportion correct recall for the experimentalsequences was accompanied by an increase in the amplitude of theP300 component. Two pieces of evidence supported this. First, asequence type×repetition interaction on the signal from the Pz elec-trode was found. The amplitude only increased for experimental se-quences, when learning and memory were involved. Second,focusing on the increase per stimulus position, the P300 amplitudeonly significantly (linearly) increased for items presented in the mid-dle of the list. Moreover, during the first repetition, the differences inP300 amplitude across stimulus positions revealed a significantquadratic trend, suggesting a larger P300 amplitude for items thatwere relatively well recalled (primacy and recency items) thanfor items less well recalled (middle items). Summarized, there wasa fairly close correspondence between P300 amplitude and memoryperformance.

4.2. The ERP N100, P200 and N200

Two studies (Conley et al., 1999; Golob and Starr, 2004) examinedthe amplitude of the N100 component in relation to memory proces-sing. It was concluded that the amplitude decreased with an increasein memory load. In terms of our study, this assumed relationshipwould have implied a significant sequence type×repetition interac-tion, which was not found. Although repetition and stimulus positioneffects were present, these were due to the relatively large amplitudeof the first stimulus. This observed difference could be (partly)explained by an effect of a preceding ERP on the baseline correctionof the following ERP. Since for the first stimulus within a trial therewere no preceding stimuli, the baseline correction of the first-stimulus ERPs might have resulted in a slightly different ERP

morphology than with the baseline corrections of following stimuliwithin a trial. When excluding the first stimulus, both effects failedto reach significance. Therefore, our study does not support a linkbetween N100 amplitude and memory.

The N100 amplitude is predominantly determined by the physicalfeatures of the evoking stimulus, such as its intensity and frequency,and the interval between stimuli (Näätänen and Picton, 1987). Fur-thermore, when the physical dimensions of the first stimulus arekept constant, the amplitude decreases over subsequent presenta-tions (e.g. Budd et al., 1998). In our study, the N100 amplitude de-creased in amplitude with increasing nominal stimulus positionwhich is in line with previous findings. No stimulus effect wasfound when the first stimulus was excluded, reflecting a constant am-plitude after the presentation of the first stimulus. This finding is inaccordance with Rosburg et al. (2004), who found that the amplitudeof the first stimulus is larger compared to later stimuli.

Several studies have also related the P200 amplitude to memoryprocessing (Chapman et al., 1978; Rushby et al., 2002; Smith, 1993;Voss and Paller, 2009). However, such a relationship appeared notto be present in our study, as reflected in the absence of a significantsequence type×repetition interaction effect for this component. Thisdifference between studies is difficult to interpret, given that the ex-perimental designs used differ in many respects, for example, asses-sing recognition rather than free recall. Other studies mention P200in relation to feature detection processes (Luck and Hillyard, 1994)or attentional mechanisms (Lijffijt et al., 2009; McDonough et al.,1992; Noldy et al., 1990).

The N200 amplitude component is also often related to novelty ormismatch detection (Folstein and Van Petten, 2008; Zheng et al.,2010). Moreover, the N200 has been found to be modulated in work-ing memory tasks (Morgan et al., 2010). No clear explanation can begiven for the fact that in our study the amplitude of the N200 of thecontrol sequenceswas larger than the amplitude of the N200 of exper-imental sequences. The control sequences were already known to theparticipants and, therefore, should not be novel or require workingmemory. When considering our results with previous studies, aN200 modulation was to be expected for the experimental sequences.However, Ferdinand et al. (2008) performed an experiment in whichparticipants had to respond to letters which were presented in eithera regular or an irregular sequence. The irregular sequences containedmore deviant stimuli, and larger N200 amplitudes were expected inthese sequences. However, the N200 amplitude was larger for regular

Fig. 5. Fig. 5 depicts the P300 component amplitudes at Pz. The x-axes depict the consecutive stimuli, the z-axes depict the 6 repetitions and the y-axes the amplitudes in μV. On theleft are P300 component amplitudes elicited within the experimental sequences whereas on the right P300 components elicited within control sequences are depicted.



than for irregular sequences. This also cannot be explained by theabove mentioned theories, but is in line with our results. How tointerpret these findings in a theoretical framework remains to beelucidated.

4.3. P300 and memory

A dominant view on the P300 is the context updating hypothesis(Donchin and Coles, 1988; Duncan-Johnson and Donchin, 1982),which states that this component reflects a process that is invokedwhenever a stimulus demands a revision of hypotheses or modelsof the environment held by the subject. The P300 is also associatedwith attentional resource allocation and cognitive demands (Kok,2001; Polich, 2007). When studying a stimulus with full attention, itis remembered better (Yonelinas, 2001), and has a larger P300 ampli-tude (Hillyard, 1985) than when studied with divided attention.Azizian and Polich (2007) used a free recall task in which a sequenceof fifteen stimuli was presented, while recording the EEG. A serialposition memory curve emerged and they found that an enlargedP300 amplitude predicted later retrieval of the primacy items. Thisfinding was explained by the primacy gradient model, whichstates that stimuli presented in the beginning of a list receive moreattention during the encoding stage than later stimuli (Farrell andLewandowsky, 2002; Page and Norris, 1998) and, therefore, are bet-ter encoded and retrieved. The authors concluded that this enhancedengagement of attentional resources was reflected in increased P300amplitude. This is in line with the data from our current study.

The increasing P300 amplitude for items being learned is also inline with findings from Johnson et al. (1985). They concluded, afterrepetitively presenting the same stimulus in a recognition task, thatthe increase in P300 amplitude reflects an increase in memory con-solidation. In addition, Vincent et al. (2006) found in an fMRI studythat specific parietal areas are sensitive to successful recollectionand that these areas are linked to the hippocampal formation. It wassuggested that these structures form a hippocampal–parietal memorynetwork. Because P300 often has the largest amplitude at posteriorsites, this would further support the assumed relation betweenP300 and memory storage.

However, the P300 amplitude has been associated with many dif-ferent, often strongly inter-related, cognitive processes, such as atten-tion allocation (also related to item distinctiveness), updating ofworking memory, and the process of memory access or retrieval(e.g., Polich, 2007). For example, the role of attention in relation tomemory is well established (Curran, 2004; Gardiner and Parkin,1990; Reinitz et al., 1994; Yonelinas, 2001). When stimuli receivefull attention, P300 amplitudes are larger (Hillyard, 1985), and mem-ory performance is better when compared to stimuli that are studiedwith divided attention. In our study, we observed relatively constantand high P300 amplitudes for primacy and recency items, in combi-nation with relatively low amplitudes for middle items that increasedin the course of learning trials. These observations might have impli-cations for theories about serial position effects (SPEs) in free recall.There are multiple competing theories explaining SPEs, which canbe roughly divided into dual-component or one-component models.The most prominent dual-component model (e.g., Atkinson and Shif-frin, 1968) states that primacy effects reflect enhanced rehearsal forinitial items, and a corresponding better transfer from short-termmemory to long-term memory for these items, relative to otheritems. Recency effects are explained by assuming that the items atthe end of the list are still being maintained in short-term memory.One-component models, of which there are many variants (e.g., seeLewandowsky and Farrell, 2008, for an overview), instead assumethat there is only one component or process underlying both primacyand recency effects. For example, Brown et al. (2007) proposed thatSPEs are due to the enhanced positional or temporal distinctivenessof items at the beginning of the list (which are not preceded by

other items) and at the end of the list (which are not followed byfurther items). The present finding of high P300 amplitudes for bothprimacy and recency items could be interpreted as evidence fora more parsimonious one-component instead of a two-componentapproach. Accordingly, assuming that item distinctiveness andattention are interrelated, with the latter being reflected in P300amplitude, one possible scenario might be that, because of theirenhanced distinctiveness, primacy and recency items received fullattention from the first learning trial on. Attention allocated toprimacy and recency items in the first repetition(s) implicatedrelatively little attention for middle items, with correspondinglow P300 amplitudes. In the course of learning trials, moreattentional resources became available for these middle items, withaccompanying increased P300 amplitudes, because they were nolonger used for the primacy and recency items. One problem withthis scenario is that it cannot explain the observation that the P300amplitude for primacy and recency items remained high throughoutthe learning trials. At least this observation seems to demand theassumption of an additional process that contributes to the P300amplitude, such as successful memory access or retrieval.

5. Conclusions

Whatever the merits of these theoretical speculations, to ourknowledge, this study is the first one to systematically track ERPchanges that are associated with multiple digit-learning trials, andcorresponding memory performance improvement. We observed amatch between the well-known serial position (behavioral) memoryeffect, and its disappearance in the process of learning, on the onehand, and the changes in the amplitude of especially the P300 ERPcomponent on the other. Future research should be directed at a clos-er examination of the cognitive components that underlie these ob-served learning-induced ERP changes.

References

American Encephalographic Society, 1994. Guidelines for standard electrode positionnomenclature. Journal of Clinical Neurophysiology 11, 40–73.

Anderson, J.R., 2000. Learning and Memory. Wiley, New York.Atkinson, R.C., Shiffrin, R.M., 1968. Human memory: a proposed system and its control

processes. In: Spence, K.W., Spence, J.T. (Eds.), The Psychology of Learning and Mo-tivation, Vol. 2. Academic Press, New York, pp. 89–195.

Azizian, A., Polich, J., 2007. Evidence for attentional gradient in the serial positionmemory curve from event-related potentials. Journal of Cognitive Neuroscience19, 2071–2081.

Bentin, S., Moscovitch, M., Heth, I., 1992. Memory with and without awareness: perfor-mance and electrophysiological evidence of savings. Journal of ExperimentalPsychology. Learning, Memory, and Cognition 18, 1270–1283.

Brown, G.D.A., Neath, I., Chater, N., 2007. A temporal ratio model of memory. Psychol-ogy Review 114, 539–576.

Budd, T.W., Barry, R.J., Gordon, E., Rennie, C., Michie, P.T., 1998. Decrement of the N1auditory event-related potential with stimulus repetition: habituation vs. refracto-riness. International Journal of Psychophysiology 31, 51–68.

Chapman, R.M., McCrary, J.W., Chapman, J.A., 1978. Short-term memory: the “storage”component of human brain responses predicts recall. Science 202, 1211–1214.

Conley, E.M., Michalewski, H.J., Starr, A., 1999. The N100 auditory cortical evokedpotential indexes scanning of auditory short-term memory. Clinical Neurophysiol-ogy 110, 2086–2093.

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

Curran, T., Clearly, A.M., 2003. Using ERPs to dissociate recollection from familiarity inpicture recognition. Cognitive Brain Research 15, 191–205.

Donchin, E., Coles, M.G.H., 1988. Is the P300 component a manifestation of contextupdating? The Behavioral and Brain Sciences 21, 149–168.

Duncan-Johnson, C.C., Donchin, E., 1982. The P300 component of the event-relatedbrain potential as an index of information processing. Biological Psychology 14,1–52.

Fabiani, M., Karis, D., Donchin, E., 1986. P300 and recall in an incidental memory para-digm. Psychophysiology 23, 298–308.

Farrell, S., Lewandowsky, S., 2002. An endogenous distributed model of ordering in se-rial recall. Psychonomic Bulletin and Review 9, 59–79.

Ferdinand, N.K., Mecklinger, A., Kray, J., 2008. Error and deviance processing in implicitand explicit sequence learning. Journal of Cognitive Neuroscience 20, 629–642.

Folstein, J.R., Van Petten, C., 2008. Influence of cognitive control and mismatch on theN2 component of the ERP: a review. Psychophysiology 45, 152–170.



Gardiner, J.M., Parkin, A.J., 1990. Attention and recollective experience in recognitionmemory. Memory and Cognition 18, 579–583.

Golob, E.J., Starr, A., 2004. Serial position effects in auditory event-related potentialsduring working memory retrieval. Journal of Cognitive Neuroscience 16, 40–52.

Gratton, G., Coles, M.G.H., Donchin, E., 1983. New method for off-line removal of ocularartifact. Electroencephalography and Clinical Neurophysiology 55, 468–484.

Hillyard, S.A., 1985. Electrophysiology of human selective attention. Trends in Neuro-sciences 8, 400–405.

Johnson, R., Pfefferbaum, A., Kopell, B.S., 1985. P300 and long-term memory: latencypredicts recognition performance. Psychophysiology 22, 447–507.

Jongsma, M.L.A., Eichele, T., Van Rijn, C.M., Coenen, A.M.L., Hugdahl, K., Nordby, H.,Quian Quiroga, R., 2006. Tracking pattern learning with single-trial event-relatedpotentials. Clinical Neurophysiology 117, 1957–1973.

Karis, D., Fabiani, M., Donchin, E., 1984. “P300” and memory: individual differences inthe Von Restorff effect. Cognitive Psychology 16, 117–216.

Kim, M.S., Kim, J.J., Kwon, J.S., 2001. The effect of immediate and delayed word repeti-tion on event-related potential in a continuous recognition task. Cognitive BrainResearch 11, 387–396.

Kok, A., 2001. On the utility of P3 amplitude as a measure of processing capacity. Psy-chophysiology 38, 557–577.

Lewandowsky, S., Farrell, S., 2008. Short-term memory: new data and a model.Psychology of Learning and Motivation 49, 1–48.

Lijffijt, M., Lane, S.D., Meier, S.L., Boutros, N.N., Burroughs, S., Steinberg, J.L., Moeller,G.F., Swann, A.C., 2009. P50, N100, and P200 sensory gating: relationships with be-havioral inhibition, attention, and working memory. Psychophysiology 46,1059–1079.

Luck, S.J., Hillyard, S.A., 1994. Electrophysiological correlates of feature analysis duringvisual search. Psychophysiology 31, 291–308.

McDonough, B.E., Warren, C.A., Don, N.S., 1992. Event-related potentials in a guessingtask: the gleam in the eye effect. International Journal of Neuroscience 65,209–219.

Morgan, H.M., Jackson, M.C., Klein, C., Mohr, H., Shapiro, K.L., Linden, D.E.J., 2010.Neural signatures of stimulus features in visual working memory: a spatiotemporalapproach. Cerebral Cortex 20, 187–197.

Murdock, B.B., 1962. The serial position effect in free recall. Journal of ExperimentalPsychology 64, 482–488.

Näätänen, R., Picton, T., 1987. The N1wave of the human electric andmagnetic responseto sound: a review and analysis of the component structure. Psychophysiology 24,375–425.

Noldy, N.E., Stelmack, R.M., Campbell, K.B., 1990. Event-related potentials and recogni-tion memory for pictures and words: the effects of intentional and incidentallearning. Psychophysiology 27, 417–428.

Page, M.P.A., Norris, D., 1998. The primacy model: a new model of immediate serial re-call. Psychology Review 105, 761–781.

Paller, K.A., McCarthy, G., Wood, C.C., 1988. ERPs predictive of subsequent recall andrecognition performance. Biological Psychology 26, 269–276.

Polich, J., 2007. Updating P300: an integrative theory of P3a and P3b. Clinical Neuro-physiology 118, 2128–2148.

Reinitz, M.T., Morrissey, J., Demb, J., 1994. Role of attention in face encoding. Journal ofExperimental Psychology 20, 161–168.

Rosburg, T., Trautner, P., Korzyukov, O.A., Boutros, N.N., Schaller, C., Elger, C.E., Kurthen,M., 2004. Short-term habituation of the intracranially recorded auditory evokedpotentials P50 and N100. Neuroscience Letters 372, 245–249.

Rugg, M.D., Coles, M.G.H., 1996. Electrophysiology of Mind: Event-related Brain Poten-tials and Cognition. Oxford University Press, Oxford.

Rundus, D., 1971. Analysis of rehearsal processes in free recall. Journal of ExperimentalPsychology 89, 63–77.

Rushby, J.A., Barry, R.J., Johnstone, S.J., 2002. Event-related potential correlates of serial-position effects during an elaborative memory test. International Journal of Psy-chophysiology 46, 13–27.

Seger, C.A., 1994. Implicit learning. Psychological Bulletin 115, 163–196.Shanks, D.R., 2010. Learning: from association to cognition. Annual Review of Psychol-

ogy 61, 273–301.Smith, M.E., 1993. Neurophysiological manifestations of recollective experience during

recognition memory judgments. Journal of Cognitive Neuroscience 5, 1–13.Vincent, J.L., Snyder, A.Z., Fox, M.D., Shannon, B.J., Andrews, J.R., Raichle, M.E., Buckner,

R.L., 2006. Coherent spontaneous activity identifies a hippocampal–parietal mem-ory network. Journal of Neurophysiology 96, 3517–3531.

Voss, J.L., Paller, K.A., 2009. An electrophysiological signature of unconscious recogni-tion memory. Nature Neuroscience 12, 349–355.

Wiswede, D., Rüsseler, J., Münte, T.F., 2007. Serial position effects in free memory recall— an ERP-study. Biological Psychology 75, 185–193.

Yonelinas, A.P., 2001. Consciousness, control, and confidence: the 3 Cs of recognitionmemory. Journal of Experimental Psychology 130, 361–379.

Zheng, Y., Xu, J., Jin, Y., Sheng, W., Ma, Y., Zhang, X., Shen, H., 2010. The time courseof novelty processing in sensation seeking: an ERP study. International Journal ofPsychophysiology 76, 57–63.


author's personal copy - university of leicester...accepted 6 october 2011 available online 20...

Documents