neural correlates of encoding predict infants’ memory in the paired-comparison procedure

30
Neural Correlates of Encoding Predict Infants’ Memory in the Paired-Comparison Procedure Kelly A. Snyder Department of Psychology University of Denver The present study used event-related potentials (ERPs) to monitor infant brain activity during the initial encoding of a previously novel visual stimulus, and examined whether ERP measures of encoding predicted infants’ subsequent performance on a visual memory task (i.e., the paired-comparison task). A late slow wave component of the ERP measured at encoding predicted infants’ immediate performance in the paired-comparison task: amplitude of the late slow wave at right-central and temporal leads decreased with stimulus repeti- tion, and greater decreases at right-anterior-temporal leads during encoding were associated with better memory performance at test. By contrast, neither the amplitude nor latency of the negative central (Nc) component predicted infants’ subsequent performance in the paired-comparison task. These findings are discussed with respect to a biased competition model of visual attention and memory. The ability to form enduring memories of new experience is of fundamental importance for human learning and development. Memory formation in young infants is typically studied in the context of behavioral tasks that rely on the duration of infant looking behavior at test to provide evidence that a memory was formed (e.g., habituation dishabituation, visual paired- comparison). Although these tasks have provided important insights into the rate at which infants encode visual stimuli at different ages (for a review, Correspondence should be sent to Kelly A. Snyder, Department of Psychology, University of Denver, 2155 S. Race St., Denver, CO 80208. E-mail: [email protected] Infancy, 15(3), 270–299, 2010 Copyright Ó International Society on Infant Studies (ISIS) ISSN: 1525-0008 print / 1532-7078 online DOI: 10.1111/j.1532-7078.2009.00015.x

Upload: kelly-a-snyder

Post on 20-Jul-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Neural Correlates of Encoding PredictInfants’ Memory in the Paired-Comparison

Procedure

Kelly A. SnyderDepartment of Psychology

University of Denver

The present study used event-related potentials (ERPs) to monitor infant brainactivity during the initial encoding of a previously novel visual stimulus, andexamined whether ERP measures of encoding predicted infants’ subsequentperformance on a visual memory task (i.e., the paired-comparison task). A late

slow wave component of the ERP measured at encoding predicted infants’immediate performance in the paired-comparison task: amplitude of the lateslow wave at right-central and temporal leads decreased with stimulus repeti-

tion, and greater decreases at right-anterior-temporal leads during encodingwere associated with better memory performance at test. By contrast, neitherthe amplitude nor latency of the negative central (Nc) component predicted

infants’ subsequent performance in the paired-comparison task. These findingsare discussed with respect to a biased competition model of visual attentionand memory.

The ability to form enduring memories of new experience is of fundamentalimportance for human learning and development. Memory formation inyoung infants is typically studied in the context of behavioral tasks that relyon the duration of infant looking behavior at test to provide evidence that amemory was formed (e.g., habituation ⁄dishabituation, visual paired-comparison). Although these tasks have provided important insights intothe rate at which infants encode visual stimuli at different ages (for a review,

Correspondence should be sent to Kelly A. Snyder, Department of Psychology, University

of Denver, 2155 S. Race St., Denver, CO 80208. E-mail: [email protected]

Infancy, 15(3), 270–299, 2010Copyright � International Society on Infant Studies (ISIS)ISSN: 1525-0008 print / 1532-7078 onlineDOI: 10.1111/j.1532-7078.2009.00015.x

see Rose, Feldman, & Jankowski, 2004), important questions remain aboutthe nature of the processes by which infants initially encode new experience.Insights into effective memory formation in adults have been gained bymonitoring adults’ brain activity during the initial encoding of a previouslynovel stimulus and relating this neural activity to behavioral evidence that amemory was formed (for a review, see Paller & Wagner, 2002). Although afew studies have examined relations between infant brain activity recordedshortly after encoding and behavioral measures of recall in the deferred imi-tation task (e.g., Bauer, Wiebe, Carver, Waters, & Nelson, 2003; Carver,Bauer, & Nelson, 2000; Lukowski et al., 2005), no study to date has moni-tored infant brain activity during the initial encoding of a novel stimulusand related this activity to behavioral evidence that a memory was formed.The goal of the research reported here was to investigate the neural mecha-nisms of encoding in young infants. To that end, we used event-relatedpotentials (ERPs) to monitor infant brain activity during the initial encodingof a previously novel visual stimulus, and examined whether ERP measuresof encoding predicted infants’ subsequent performance on a visual memorytask (i.e., the paired-comparison task).

Visual recognition memory in young infants has traditionally beenassessed using the paired-comparison procedure (Fagan, 1970; Fantz, 1964;for a recent review, see Rose , Feldman, & Jankowski 2007). In the encodingphase of this procedure, two identical stimuli are presented side by side (orone stimulus is presented centrally) for either a fixed period of time (e.g.,30 sec), or until the infant accumulates a fixed amount of looking. At test,the now ‘‘familiar’’ stimulus is presented alongside a novel stimulus, and thelength of time the infant fixates each stimulus is measured. The dependentmeasure is the novelty score, computed as the proportion of time the infantspends fixating the novel stimulus relative to both stimuli across the two testtrials (Fagan, 1974; Rose & Feldman, 1990). Memory for the familiar stimu-lus is typically inferred from longer looking to the novel stimulus.

Depending on the complexity of the stimulus and the age of the infant,encoding in the paired-comparison procedure can be a relatively protractedprocess, occurring over time and with prolonged (or repeated) exposure tothe familiarization stimulus. Novelty scores at test have been shown to pro-gress from familiarity preferences to chance responding to novelty prefer-ences (familiarity fi chance fi novelty) with increasing familiarization(Hunter, Ames, & Koopman, 1983; Hunter, Ross, & Ames, 1982; Richards,1997; Rose, Gottfried, Melloy-Carminar, & Bridger, 1982; Wagner & Sako-vits, 1986), although the intermediate stage of chance responding may reflectan artifact of averaging over groups of infants, some of whom have suffi-ciently encoded the stimulus to show a novelty preference and some who are‘‘slower’’ processors and still prefer the familiar stimulus (Roder, Bushnell,

NEURAL CORRELATES OF ENCODING IN INFANTS 271

& Sasseville, 2000). The opposite trend has been observed during the courseof forgetting, with infant preferences shifting from novelty preferences tochance responding to familiarity preferences (novelty fi chance fi familiar-ity) with increasing delays between study and test (Bahrick, Hernandez-Reif,& Pickens, 1997; Bahrick & Pickens, 1995; Courage & Howe, 1998).

In electrophysiological paradigms with infants, differing amounts ofexposure to a stimulus results in differences in the amplitude of a late slowwave component that is thought to reflect processes related to memoryupdating or encoding. In general, relatively little exposure elicits a positive-going deflection of the late slow wave, whereas relatively more exposureelicits smaller (less positive) amplitudes of the slow wave or a return to thebaseline response (for a review, see Nelson & Monk, 2001). Furthermore,this relation between stimulus exposure and amplitude of the slow waveappears to hold regardless of whether stimulus exposure is preexperimental,or occurs during the course of testing. For instance, Nelson and Collins(1991) familiarized 6-month-old infants to two faces (for 5 sec each) prior toERP testing. At test, infants were exposed to one of the familiar faces on60% of the trials (frequent–familiar), the other familiar face on 20% of thetrials (infrequent–familiar), and a trial-unique novel face on the remaining20% of trials (infrequent–novel) while ERPs were recorded. The late slowwave component was observed to be larger in amplitude to the infrequent–familiar face compared with the frequent–familiar face, which the authorsinterpreted as reflecting differences in the degree to which the two faces hadbeen encoded. As the frequent–familiar face was presented more often,resulting in relatively more exposure and presumably greater encoding, thegreater slow wave amplitude in response to the infrequent–familiar facewas inferred to reflect memory updating for a partially encoded stimulus.Similarly, the late slow wave has been observed to be larger in amplitude toa novel stimulus compared with an a priori familiar stimulus (e.g., a pictureof the mother’s face or the infant’s favorite toy; de Haan & Nelson, 1997,1999). In these experiments both the familiar and novel pictures wererepeated over multiple trials (as many as 50); so, the greater slow waveamplitude in response to the ‘‘novel’’ stimuli was again considered to reflectmemory updating for a partially encoded stimulus.

Taken together, these findings indicate an inverse relation between ampli-tude of the late slow wave and the amount of experience an infant has with astimulus. In general, more stimulus exposure is associated with smaller (lesspositive) amplitudes of the slow wave. This, in turn, suggests that the lateslow wave may provide a sensitive index of the degree to which a stimulushas been encoded. If so, amplitude of the late slow wave (measured atencoding) may predict infant’s subsequent performance on the paired-comparison test.

272 SNYDER

Current study

We examined whether amplitude of the late slow wave (measured at encod-ing) predicts performance on the paired-comparison test in 6-month-oldinfants. During an initial encoding phase, infants attended to repeated pre-sentations of a single novel object while we recorded ERPs. Immediately fol-lowing encoding, infants’ memory for the repeated object was assessed inthe paired-comparison procedure. To investigate the neural mechanisms ofmemory formation in infants, we examined the degree to which ERPs elic-ited during encoding predicted individual infants’ behavioral performance inthe paired-comparison. As partial encoding of a stimulus results in familiar-ity preferences (in the paired-comparison) and more positive amplitudes ofthe late slow wave component (in ERP paradigms), we predicted a negativecorrelation between amplitude of the late slow wave measured at encodingand infants’ novelty scores at test. That is, smaller (less positive) amplitudesof the late slow wave should predict larger novelty scores in the paired-com-parison. We examined these hypotheses in 6-month-old infants in order tobuild on previous research investigating memory-related modulation of thelate slow wave component. In order to evaluate the specificity of the lateslow wave for predicting infants’ performance in the paired-comparison, wealso examined possible associations between novelty scores and the ampli-tude and latency of a mid-latency negative component (Nc) that peaksbetween 400 and 800 msec following stimulus onset and is commonlyobserved over fronto-central scalp regions. Although the Nc is most com-monly thought to reflect processes associated with attention and orienting,Nc amplitude has been found to differ in magnitude for highly familiar andnovel stimuli (e.g., de Haan & Nelson, 1997, 1999).

METHODS

Participants

The final sample consisted of fifty-six 6-month-old infants (31 girls, 25 boys)born full term (i.e., 38–42 weeks gestation) with no known history of visualor neurological problems. Infants were tested within 1 week of turning6 months of age (mean age at test = 25.83 weeks, SD = .51). Only infantswho provided good data for both the encoding and test phases of the studywere included in the final sample. An additional 63 infants were tested andexcluded from the final sample for gestational diabetes (n = 4), refusal towear the electrode cap (n = 6), failure to complete the test phase (n = 10),problems with video recording during the test phase, such as the infantleaning out of view of the camera (n = 11), side bias in the looking data

NEURAL CORRELATES OF ENCODING IN INFANTS 273

(n = 1), low interrater reliability for the looking data (n = 1), or poor ERPdata quality (n = 30).1 Overall, 76% of eligible2 infants tested completedboth ERP and behavioral testing, and 64% of these infants (or 49% of alleligible infants) provided data for the final sample. Considering our require-ment that infants provide both ERP and behavioral data, and the subjectretention rates reported in previous infant ERP studies (e.g., 19% in deHaan, Pascalis, & Johnson, 2002) and behavioral studies (e.g., 47% in Rob-inson & Pascalis, 2004), the retention rate reported here is good. Infantswere recruited from a database of families maintained by the University ofMinnesota who were contacted at the birth of their baby and returned apostcard indicating their interest in participating in research.

Materials

Stimuli consisted of color pictures of Greebles as shown in Figure 1 (imagesare by courtesy of Michael J. Tarr, Brown University, Providence, RI).Greebles are a class of artificial three-dimensional objects that have beenused to study visual perception in adults (e.g., see Gauthier & Tarr, 1997). AGreeble consists of four horizontally oriented parts (similar to appendages)arranged on a vertically oriented central part (similar to a head ⁄ torso com-bination). All Greebles are of the same uniform color, and have the samenumber of horizontal and vertical parts arranged in the same spatial config-uration, but differ from one another in the shape of their horizontal andvertical parts. In addition, the Greebles are organized into categories suchthat each Greeble belongs to one of two ‘‘genders’’ (defined by whether the

Figure 1 Stimuli. For each infant, one Greeble from the set served as the familiar stim-

ulus and a different Greeble from the set served as the novel stimulus.

1‘‘Side bias’’ in the looking data was defined as spending more than 85% of the total time

during the test trials looking to one side of the screen; ‘‘Low interrater reliability’’ for looking

data during the test phase was defined as Pearson correlations below .85; ‘‘Poor ERP data qual-

ity’’ was defined as having an insufficient number of artifact-free trials for construction of ERPs

(i.e., <10 trials).2Does not include infants excluded for gestational diabetes.

274 SNYDER

horizontal parts are all pointing upward or all pointing downward), andeach Greeble belongs to one of five ‘‘families’’ (defined by the shape of thevertical part). Five different Greebles of the same gender but from differentfamilies were selected as stimuli for this study. By selecting Greebles fromdifferent families, we aimed to maximize shape differences between the Gree-bles. Pilot testing indicated that infants did not have a priori preferences forany of the five Greebles in the set.

We selected Greebles as stimuli for several reasons. First, infants areunlikely to have had experience with Greebles outside the laboratory, mak-ing them novel from the perspective of both perceptual memory andsemantic knowledge. This was critical for ensuring that the ERP data col-lected during the encoding phase reflected encoding-related neural process-ing, and not long-term memory for a particular object, pattern, or stimuluscategory (e.g., faces). Second, Greebles are relatively complex objects, mak-ing them well suited for collecting ERP data. In order to extract the ERPfrom the background electroencephalogram (EEG), a stimulus (or condi-tion) must be repeated across multiple trials and then averaged together. Ifinfants habituate to a stimulus too quickly, they may not produce enoughtrials for ERP construction. As Greebles are relatively complex objects,and infants need more time to encode complex stimuli (Caron & Caron,1969; Fagan, 1971), we expected to be able to collect a sufficient number oftrials for ERP construction. Finally, as Greebles are physically similar toone another, and the degree of physical similarity between the familiar andthe novel stimulus is known to affect infants’ visual preferences at test (e.g.,Bornstein, 1981; Miranda & Fantz, 1974), we expected that discriminatingtwo Greebles would be sufficiently difficult to produce variability in infants’performance at test.

Each infant saw two different Greebles. One Greeble served as the famil-iar stimulus and was presented during the encoding phase and the test phase,and a different Greeble served as the novel stimulus and was presented dur-ing the test phase only. The Greebles that served as the familiar and novelstimuli were rotated across infants such that each of the five Greebles servedequally often as the familiar stimulus or the novel stimulus.

Greebles were fuchsia colored against a white background and measured20 cm · 24 cm when presented on the computer screen. For the encodingphase, a single Greeble was centered in the subject screen. For the test phase,two Greebles were centered within the left and right halves of the subjectscreen. Stimuli were presented on a 24-inch-wide screen controlled by aMacintosh computer running EGIS experimental software (Electrical Geo-desics Inc., Eugene, OR). A videocassette recorder, camera, and time ⁄dategenerator were used to video-tape participants at 30 frames per second dur-ing the test phase.

NEURAL CORRELATES OF ENCODING IN INFANTS 275

Electroencephalogram recording procedure

EEG was recorded using a 64-channel system (Electrical Geodesics Inc.) con-sisting of 63 sintered silver ⁄ silver chloride (Ag ⁄AgCl) electrodes embedded insponges soaked in electrolyte solution and evenly distributed across the scalpsurface, including electrodes for recording the electrooculogram. All signalswere recorded referenced to a single vertex electrode, sampled at 200 Hz, andfiltered online using a bandpass of .1 to 80 Hz. A 60-Hz notch filter was alsoapplied. Impedances were checked online prior to recording and wereaccepted when they were below 50 KX.

Procedure

Infants were tested in a single session3 that consisted of an encoding phasefollowed immediately by a paired-comparison test. During both phases,infants were seated on their parent’s lap approximately 60 cm from themonitor, and parents were blindfolded and instructed not to interact with theinfant.

Encoding phase

EEGwas continuously recorded while the infant attended to a single Gree-ble that was repeated between 20 and 100 times. An experimenter monitoredthe infant’s corneal reflections via a video camera and pressed a button topresent the next picture when the infant was calm and attending to the screen.Each trial consisted of a 500-msec stimulus presentation, and a randominterstimulus interval (ISI) that varied between 1,800 and 2,800 msec at theminimum, and during which the screen was blue. The software did not allowthe experimenter to present a picture during the minimum computer-con-trolled ISI. This procedure allows the experimenter to synchronize stimuluspresentation with periods of calm attention on the part of the infant towardthe screen, while also ensuring that the timing of stimulus presentation is var-ied. If needed, a second experimenter tapped the monitor to attract theinfant’s attention to the screen prior to presentation of the images. Data col-lection ended when the infant indicated loss of interest in the picture by look-ing away from the screen for 3 sec or more on three separate occasions(Diamond, 1995), or started to become fussy. Overall, infants viewed thefamiliar stimulus between 21 and 70 times (M = 41.87, SD = 11.81).

3Infants in this study were also invited back at a later date to participate in tasks designed to

assess long-term memory for the familiar stimulus. These additional tests were intended to

address questions that differ from the focus of this paper (i.e., questions about memory retrie-

val) and are not reported here.

276 SNYDER

The encoding procedure used here was a hybrid between electrophysio-logical recording methods and infant-controlled habituation. The habitua-tion criterion was modeled after a procedure used by Diamond (1995), andwas intended to provide a compromise between electrophysiological record-ing methods and the infant control habituation procedure. In a typicalinfant control procedure, look duration is used as an index of attention andhabituation of the infants’ attention is considered to have occurred once theduration of the infants’ last two or three looks toward the repeated stimulusis less than 50% of the average of the infants’ first (or longest) two or threelooks (Horowitz, Paden, Bhana, & Self, 1972). This is known as the 50%decrement criterion (Dannemiller, 1984). The 50% decrement criterion is anarbitrary criterion, and is meant primarily to equate infants’ level of habitu-ation or familiarity with the repeated stimulus prior to testing responserecovery, rather than to provide an absolute index of habituation (Gilmore& Thomas, 2002; Thomas & Gilmore, 2004). In the present context, wecould not use look duration toward the stimulus as an index of attentionbecause ERP designs necessitate the use of discrete (and relativelyshort—i.e., 500 msec) stimulus presentation durations. Importantly, Dia-mond showed that using a criterion of looking away from a stimulus for3 sec or more on three separate occasions was sufficient for obtaining nov-elty preferences in infants. Thus, the ‘‘3 sec on three occasions’’ criterionaccomplishes the same goals as the infant control procedure by: (a) allowingthe amount of stimulus exposure (or number of trials) to vary as a functionof individual differences in infant information processing, thereby (b) givinginfants sufficient experience with the stimulus to show enhanced respondingto a novel stimulus at test (i.e., equating infants’ level of familiarity with therepeated stimulus prior to testing response recovery). Thus, instead of usinglook duration toward the stimulus as an index of infant attention andresponse decrement, we used look duration away from the location wherethe stimulus reliably appeared. In this way, stimulus duration on each trialwas fixed to allow for ERP construction, whereas the number of trials pre-sented was infant controlled to allow for individual differences in the rate atwhich infants encode information, similar (but not equivalent) to an infant-controlled habituation procedure (Horowitz et al., 1972). Based on previousinfant ERP work, we expected that differences in the number of trials thatinfants completed would reflect, in part, individual differences in speed ofencoding (Snyder, Webb, & Nelson, 2002).

Paired-comparison

Infants participated in a paired-comparison test immediately followingencoding. The paired-comparison test consisted of two 5-sec trials, during

NEURAL CORRELATES OF ENCODING IN INFANTS 277

which the familiar and a novel Greeble were presented side by side. The left–right position of the familiar stimulus was counterbalanced across test trialsand across subjects. To begin each trial, an experimenter flashed a red lightlocated at the top, center of the subject screen to attract the infant’s atten-tion. The experimenter viewed the infant via a video camera, pressed abutton to begin the trial when the infant was attending to the light, andheld down a button when the infant was attending to either of the stimuli.Each trial ended when the infant accumulated 5 sec of looking at the stimuli.

Scoring of infant looking behavior

Encoding phase

Infants’ visual fixations during the encoding phase were coded off-linefrom videotape by two independent raters. Raters conducted a frame-by-frame analysis of the videotapes and recorded the number of frames thatinfants’ spent looking toward and away from the screen. Raters alsorecorded behavioral indicators of ‘‘fussiness,’’ defined as (a) distress orprotest vocalizations (i.e., vocalizations other than babbling), (b) pulling onthe net, (c) rubbing the eyes, (d) arching the back, (e) trying to stand on theparent’s lap, or (f) crying. Interrater reliability was assessed for lookdurations by computing Pearson correlations for the duration of individuallooks toward and away from the screen across raters. Mean rater agreementfor look durations was .91 (SD = .03).

Paired-comparison

Infants’ visual fixations in the paired-comparison task were coded off-linefrom videotape by two independent raters who were blinded to the left–rightlocation of the familiar stimulus across trials. Raters conducted a frame-by-frame analysis of the videotapes and recorded the number of frames theinfant fixated the left and right sides of the screen. Interrater reliability wasassessed by computing Pearson correlations for the duration of individuallooks to the left and right across raters. As we were interested in examiningpossible associations between ERP data collected at encoding and looking-time data collected at test, we excluded test data with interrater reliabilitybelow .85 from further analysis (n = 1). Mean rater agreement for retaineddata was .96 (SD = .03). Novelty scores were the dependent measure, andwere computed as the proportion of time the infant fixated the novel stimu-lus with respect to both stimuli across the two test trials. Novelty scores arethe traditional dependent measure examined in the paired-comparison test,and are typically tested against a chance level of 50%.

278 SNYDER

EEG artifact elimination and data reduction

Epochs consisting of 1,600 msec (100 msec prestimulus, 500 msec stimuluspresentation, and 1,000 msec poststimulus) were obtained off-line from thecontinuous EEG data. The procedure proposed by Junghofer, Elbert,Tucker, and Rockstroh (2000) for artifact identification and control of high-density electrophysiological data was used. This method excludes and inter-polates bad channels and time points when statistical criteria indicate thepresence of artifact, including rejecting trials containing eye blinks or move-ments. In addition, epochs were rejected that contained: (a) more than 12bad channels, (b) a cluster of bad channels grouped at one region of thescalp (which would result in invalid interpolation), or (c) signals exceeding100 lV. This last criterion was necessary due to the high frequency of move-ment artifact obtained when testing infants. Data were mathematically con-verted to the average reference, and this reference was used for allsubsequent steps of analysis. The mean number of artifact-free trials avail-able for analysis was 27.23 (SD = 10.55; range = 11–57).

Analysis of ERP data

A single cross-average was constructed for each infant by averaging theinfant’s EEG data across all available artifact-free trials collected duringthe encoding period. A grand mean ERP was then constructed by aver-aging together the cross-averages. The cross-averages reflect average ERPresponses of individual infants during encoding, and were used to exam-ine relations between the timing and magnitude of ERP componentselicited during encoding and novelty scores at test. The grand meanreflects the average ERP response elicited during encoding across infants,and is used to illustrate the average timing and topography of compo-nents in figures.

Based on our hypotheses and previous infant ERP studies of visualrecognition memory (for a review, see Nelson & Monk, 2001), analyseswere focused on two components: (a) the Nc from 300 to 700 msecfollowing stimulus onset, and (b) the late slow wave from 1,000 to1,500 msec.

RESULTS

In the following and in all subsequent analyses, an alpha level of .05 isused for all statistical tests, t-tests are all two-tailed, and effect-size statis-tics (including Cohen’s d for paired-sample t-tests) are reported for all

NEURAL CORRELATES OF ENCODING IN INFANTS 279

significant effects. The data reported for all behavioral analyses are forrater 1. Data for raters 1 and 2 were analyzed separately, and all effectswere significant in both data sets. To test our hypotheses, we examinedpossible associations between the timing and magnitude of ERP compo-nents elicited during encoding and novelty scores at test for two overlap-ping groups of infants: (a) infants who met criteria for habituation duringencoding (n = 39), and (b) all infants who provided sufficient artifact-freeERP data during encoding and reliable looking-time data during the testphase (n = 56). We reasoned that the neural correlates of preferencescores in infants who habituated, and subsequently showed a novelty pref-erence, would be most closely related to the underlying mechanisms thatgenerate infants’ preferential-looking patterns in typical behavioral tasks.We also reasoned that the neural correlates of preference scores in infantswho became too fussy to continue encoding, and subsequently did nothabituate, would be more ambiguous and difficult to interpret, as brainactivity in these infants may reflect a combination of encoding-relatedprocesses and emotional responses (related to fussiness). On the otherhand, inclusion of these infants might enhance our ability to examineencoding-related processes by including infants in whom encoding wasincomplete, as stopping familiarization prior to habituation typicallyresults in familiarity preferences. We therefore report associations betweenERP and preference scores for both groups of infants (a and b above, seeTable 1) but focus our discussion on effects that are significant in thegroup of infants who habituated.

Habituation

We first examined whether infants included in the final sample habituated tothe repeated stimulus during the encoding phase or became fussy, causingthe encoding phase to end prior to habituation. Infants were considered tomeet criteria for habituation if the encoding phase ended after the third lookaway from the screen lasting 3 sec or more, and they did not exhibit signs offussiness at the end of the phase. Look durations away from the screen wereexamined to assess whether infants met criteria for habituation. Encodingphases that ended as a result of infant ‘‘fussiness’’ were determined by thepresence of two or more behavioral indicators of fussiness at the end of theencoding period (i.e., within the last four observations), as coded by bothobservers, regardless of the duration of looking away from the screen. Thelatter exception is important because infants who are fussy may look awayfrom the screen out of fussiness (e.g., while trying to stand or arching theirbacks) rather than because they have habituated to the stimulus. Of the 56infants who provided sufficient artifact-free ERP data during encoding and

280 SNYDER

TA

BLE

1

First-

ord

er

part

ialc

orr

ela

tions

pre

dic

ting

novelty

score

from

the

Nc

and

late

slo

ww

ave

com

ponents

elic

ited

during

encodin

g,contr

olli

ng

for

num

ber

ofart

ifact-

free

tria

lsin

clu

ded

inth

eE

RP

Measure

12

cFz

cCz

cPz

cC3

cC4

cT3

cT4

Nc(Peakamplitude)

Trialscompleted

—).05

().01)

).12

().02)

).03

(.07)

.15

(.07)

.01

(.13)

).25

().20)

——

Novelty

score

—.24

(.06)

).03

().08)

).18

().10)

).02

().10)

).06

().13)

——

Nc(Latency)

Trialscompleted

—).05

().01)

).17

().06)

).21

().10)

.22

(.18)

).28�

().19)

).15

().18)

——

Novelty

score

—).03

().04)

).16

().13)

.07

(.09)

.24

(.02)

).28�

().11)

——

Late

SlowWave(M

eanamplitude)

Trialscompleted

—).05

().01)

—).49**

().29)*

—).19

().04)

).35*

().17)

.24

(.09)

.19

(.08)

Novelty

score

——

.14

(.08)

—.13

().01)

).11

().08)

.14

(.11)

).41**

().38)**

Note.Values

inthetoprow

representinfants

whomet

criteria

forhabituationduringencoding(n

=39);values

inthebottom

row

inpar-

entheses

representallinfants

whoprovided

sufficientartifact-freeERPdata

duringencodingandreliable

looking-tim

edata

duringthetest

phase

(n=

56);

�p

£.10(2-tailed).*p

£.05(2-tailed).**p

£.01(2-tailed).

NEURAL CORRELATES OF ENCODING IN INFANTS 281

reliable looking-time data during the test phase (i.e., included in the finalsample), 39 met criteria for habituation and 17 met criteria for fussiness atthe end of the encoding phase.

Novelty scores

As shown in Figure 2, infants included in the final sample who met criteriafor habituation spent significantly more time looking at the novel stimulusat test, t(38) = 2.14, p < .05. By contrast, infants for whom the encodingphase was ended due to fussiness spent significantly more time looking atthe familiar stimulus at test, t(16) = )3.34, p < .01. These results are con-sistent with previous work showing that infants typically show a familiaritypreference if tested before reaching criterion for habituation, and a noveltypreference if tested after reaching criterion for habituation (e.g., Hunteret al., 1982, 1983). The number of times an infant saw the familiar stimulusduring encoding did not correlate with novelty score at test, r(54) = ).01,p > .9.

Late slow wave

Figure 3a plots the grand average waveform for encoding trials at a repre-sentative anterior-temporal electrode. The late slow wave is evident as apositive deflection in the waveform that begins approximately 1,000 msecfollowing the onset of the stimulus. Figure 3b shows the scalp topography ofthe late slow wave at the time of its peak (1,100–1,400 msec). Consistent

Figure 2 Mean duration of looking to the familiar and novel objects during the paired-

comparison test as a function of whether infants met criterion for habituation (Habitu-

ated) or for whom the encoding phase was ended due to fussiness (Fussy). Error bars

indicate SE. Asterisks indicate significant differences in looking at the familiar and novel

objects for each group.

282 SNYDER

with previous reports, the late slow wave can be observed to have a centraland anterior-temporal distribution.

Our primary hypothesis was that stimulus exposure would lead toencoding-related decreases in the amplitude of the slow wave that would, in

(a) (b)

(c) (d)

(e) (f)

Figure 3 The late slow wave component. (a) Grand average ERP waveform for encod-

ing trials at a representative electrode in cT4. The late slow wave component is the large

positive deflection that begins at approximately 1,000 msec and continues until the end

of the trial. (b) Top-down view of the head showing the scalp topography of the late slow

wave. The late slow wave is depicted as large positive values (shown in red), and can be

seen to cover a region of scalp extending from central to right-anterior-temporal leads.

White dots indicate electrodes included in right-central and right-anterior-temporal clus-

ters. (c) Scatter plot depicting the negative correlation between number of trials an infant

completed during the encoding phase and mean amplitude of the late slow wave compo-

nent measured during encoding (at cCz). (d) Scatter plot depicting the negative correla-

tion between mean amplitude of the late slow wave component measured during

encoding (at cT4) and novelty scores at test. (e) Change across blocks in mean amplitude

of the late slow wave component during encoding (at cT4). Error bars indicate SE. (f)

Scatter plot depicting the positive correlation between change scores (Block 5 ) Block 1)

for amplitude of the late slow wave component measured during encoding (at cT4) and

novelty scores at test.

NEURAL CORRELATES OF ENCODING IN INFANTS 283

turn, predict subsequent memory performance. Previous studies of 6-month-old infants have most consistently found exposure-related amplitude reduc-tions and memory-related effects for the late slow wave at central leads(Cz, C3, C4; de Haan & Nelson, 1999; Nelson & Collins, 1991; Snyder et al.,2002) and anterior-temporal leads (T3, T4; de Haan & Nelson, 1997, 1999;Snyder et al., 2002). To facilitate comparison of our results with previouswork, we constructed electrode groups corresponding to the location ofthese leads as shown in Figure 4. Mean amplitude of the late slow wave,

Figure 4 Electrode groups defined for analyses of variance of the Nc and late slow

wave. Electrodes within each group were averaged together to approximate the activity

typically recorded from a single electrode of the 10 ⁄ 20 system (Jasper, 1958). Dark circles

indicate the location of relevant electrodes from the 10 ⁄ 20 system. Boxes surround the

electrodes included in each group. Each electrode group is assigned a label consisting of

the letter ‘‘c’’ (for cluster), followed by the label of the corresponding electrode from the

10 ⁄ 20 system. For instance, ‘‘cCz’’ is used to refer to the cluster of electrodes (5, 55, and

30) that were averaged together to approximate electrode Cz of the 10 ⁄ 20 system.

284 SNYDER

defined as the average response within the 1,000- to 1,500-msec time win-dow, was averaged across electrodes within each group. To reduce the num-ber of comparisons, thereby maintaining the familywise error rate at .05, welimited correlations to these five groups. Bonferroni correction was used toadjust the p-value for multiple comparisons (five comparisons, p = .01).

Relation to stimulus exposure at encoding

We hypothesized that more stimulus exposure would be associatedwith smaller (less positive) amplitudes of the slow wave. To test thishypothesis, we examined partial correlations between mean amplitude ofthe slow wave measured at encoding and the number of trials an infantcompleted during the encoding phase, controlling for the number of arti-fact-free trials included in the ERP. We examined partial correlations inorder to factor out the influence of the number of trials in the ERP onslow wave amplitude, as averaging over more trials diminishes the ampli-tude of ERP components (Thomas, Grice, Najm-Briscoe, & Miller,2004). As shown in Table 1, there was a negative correlation betweenamplitude of the slow wave over mid- and right-central leads and thenumber of trials an infant completed during the encoding phase. Specifi-cally, increased exposure to the familiar stimulus during the encodingphase was associated with reduced amplitude of the slow wave as shownin Figure 3c.

Relation to memory performance

We also hypothesized that smaller (less positive) amplitudes of thelate slow wave would predict larger novelty scores in the paired-comparison.To test this hypothesis, we examined partial correlations between meanamplitude of the slow wave measured at encoding and novelty scores at test,again controlling for the number of artifact-free trials included in the ERP.As predicted, smaller amplitudes of the slow wave predicted betterperformance (i.e., larger novelty scores) as shown in Figure 3d. This effectwas observed over right-anterior-temporal leads.

As our hypothesis that smaller late slow wave amplitudes would predictlarger novelty scores was predicated on the hypothesis that the amplitude ofthe late slow wave decreases with increasing stimulus exposure (and is there-fore an index of encoding), we examined whether late slow wave amplitudedecreased with stimulus exposure over right-anterior-temporal sites. Toexamine this, we divided the number of artifact-free trials for each partici-pant at right-anterior-temporal leads by five (to create five blocks of trialsfor each participant) and then averaged artifact-free trials within each block

NEURAL CORRELATES OF ENCODING IN INFANTS 285

for each participant. The first and last block reflect the average ERPresponse during initial and final stages of encoding, and blocks 2–4 reflectintermediate stages of encoding. As shown in Figure 3e, late slow waveamplitude first increased and then decreased across blocks, although therewere no significant differences between blocks.

If the correlation between late slow wave amplitude at right-anterior-temporal leads and novelty score reflects an encoding-related decrease inslow wave amplitude associated with better encoding of the familiar stimu-lus, greater decreases in slow wave amplitude during encoding should pre-dict a stronger preference for the novel stimulus at test. To test thishypothesis, we examined the prediction of novelty scores from ERP changescores over right-anterior-temporal leads, calculated by subtracting meanslow wave amplitude for the last block (final stage of encoding) from slowwave amplitude for the first block (initial stage of encoding; first block ) lastblock), for infants who met criterion for habituation, again controlling forthe number of artifact-free trials included in the ERP. Change in slow waveamplitude during encoding positively correlated with novelty scores at test,pr(36) = .41, p < .01, indicating that larger changes (greater decreases)during encoding predicted better memory performance (i.e., larger noveltyscores).

Summary

Amplitude of the slow wave predicted immediate performance in thepaired-comparison test. Specifically, smaller amplitudes of the slow wavewere associated with larger novelty scores at right-anterior-temporal leads.Furthermore, amplitude of the slow wave at mid-central, right-central, andright-anterior-temporal leads decreased during the encoding phase(although measurable decreases at right-anterior-temporal leads did notreach statistical significance), and larger changes (greater decreases) overright-anterior-temporal leads during encoding predicted better memoryperformance at test (i.e., larger novelty scores).

Nc

Figure 5a plots the grand average waveform for encoding trials at the vertexelectrode (i.e., Cz). The Nc is clearly evident as a large negative deflection inthe waveform that peaks approximately 500 msec following the onset of thestimulus. Figure 5b shows the scalp topography of the Nc at the time of itspeak (550 msec). Consistent with previous reports, the Nc can be observedto have a fronto-central distribution, and a slight right hemisphere bias (seealso Nelson, Thomas, de Haan, & Wewerka, 1998).

286 SNYDER

Previous studies of 6-month-old infants have most consistently foundmemory-related effects for peak amplitude of the Nc at midline leads (Fz,Cz, Pz; de Haan & Nelson, 1997; Reynolds & Richards, 2005; Webb, Long,& Nelson, 2005), and central leads (C3, Cz, and C4; de Haan & Nelson,1999). To facilitate comparison of our results with previous work, and main-tain the familywise error rate at .05, the Nc was analyzed at these five sitesas shown in Figure 4. Peak amplitude, defined as the minimum point of theresponse relative to baseline within the 300- to 700-msec time window, andlatency to peak amplitude, defined as the time relative to stimulus onset atwhich peak amplitude occurs, were averaged across electrodes within eachgroup. To examine possible associations between the Nc and subsequentmemory performance, we examined partial correlations between peakamplitude and latency of the Nc measured at encoding and novelty scores attest, controlling for the number of artifact-free trials included in the ERP.Bonferroni correction was used to adjust the p-value for multiple compari-sons (five comparisons, p = .01).

Relation to stimulus exposure at encoding

Number of trials completed was not associated with statistically signifi-cant differences in Nc amplitude or latency (see Table 1).

Relation to memory performance

Neither Nc amplitude nor latency predicted infants’ memory perfor-mance (see Table 1).

(a) (b)

Figure 5 The Nc component. (a) Grand average ERP waveform for encoding trials at

the vertex electrode. The Nc component is the large negative deflection peaking at

approximately 500 msec. (b) Top-down view of the head showing the scalp topography

of the Nc at the time of its peak (550 msec). The Nc is depicted as large negative values

(shown in blue), and can be seen to cover a large fronto-central region of the scalp.

NEURAL CORRELATES OF ENCODING IN INFANTS 287

DISCUSSION

The aim of this study was to investigate whether ERPs measured duringthe initial encoding of a previously novel object would predict infants’subsequent memory for that object in the paired-comparison test. Wehypothesized that smaller (less positive) amplitudes of the late slow wavewould predict better memory performance (i.e., larger novelty scores). Aspredicted, amplitude of the slow wave at mid-central, right-central,and right-anterior-temporal leads decreased with stimulus exposure(although measurable decreases at right-anterior-temporal leads did notreach statistical significance), and smaller amplitudes of the slow wave atright-anterior-temporal leads were associated with larger novelty scores.Furthermore, larger changes (greater decreases) in the amplitude of theslow wave over right-anterior-temporal leads during encoding predictedbetter memory performance at test (i.e., larger novelty scores). By contrast,neither the amplitude nor latency of the Nc was significantly associatedwith infants’ subsequent performance in the paired-comparison task. Thesefindings have implications for understanding the neural mechanisms ofinfant encoding in passive viewing tasks, and for understanding memoryperformance in the paired-comparison procedure.

Encoding

We used a hybrid between electrophysiological recording methods andinfant-controlled habituation to habituate infants to a previously novelobject while we recorded ERPs. A single Greeble was presented for500 msec across multiple trials until the infant indicated loss of interest inthe picture by looking away from the screen for 3 sec or more on threeseparate occasions, or started to become fussy. In this way, stimulusduration on each trial was fixed to allow for ERP construction, whereasthe number of trials presented was infant controlled to allow for individ-ual differences in the rate at which infants encode information. For someinfants, it was necessary to end the encoding phase at the first signs offussiness as a crying infant cannot participate in the test phase of theexperiment.

Consistent with findings from other infant-controlled habituation proce-dures (e.g., Hunter et al., 1982, 1983), infants who met our criteria for habit-uation showed a novelty preference at test, whereas infants who did nothabituate (and for whom the encoding phase was ended due to fussiness)showed a familiarity preference at test. These results provide an importantvalidation of our encoding procedure, and suggest that infants who met ourcriteria for habituation had fully encoded the stimulus, whereas infants who

288 SNYDER

became fussy before reaching criterion for habituation may have only par-tially encoded the familiar stimulus.

We also found that infants’ performance in the paired-comparison testwas not strictly a function of stimulus exposure or repetition, as indicatedby the absence of associations between the number of times an infant sawthe familiar stimulus during encoding and novelty scores. Although thisreflects a null finding, and should therefore be interpreted with caution, it isconsistent with findings from behavioral studies of habituation, as well as arecent electrophysiological study of habituation in 6-month-olds that alsofound that the number of times an infant saw the familiar stimulus duringhabituation did not predict novelty scores at test (Snyder & Keil, 2008).Behavioral studies have shown that some infants require more exposure ortrials than others to fully encode a stimulus (i.e., habituate). The number oftrials per se would therefore not be expected to predict novelty scores.Instead, the relevant criterion is whether or not an infant habituates, ashabituation has been shown to reliably predict novelty preferences (e.g.,Hunter et al., 1982, 1983). As some infants habituate in relatively few trials,and others require a lot more exposure, number of trials is not related tohabituation in an absolute sense, and would therefore not be expected topredict novelty scores. This could help explain why stimulus exposure wasnot related to amplitude of the slow wave at right-anterior-temporal leads,where slow wave amplitude predicted memory performance.

Associations between encoding-related neural activity and memory perfor-mance

Consistent with previous findings, we found that amplitude of the slow waveat mid-central, right-central, and right-anterior-temporal leads decreasedwith stimulus exposure, although measurable decreases at right-anterior-temporal leads did not reach statistical significance. The inverse relationbetween amplitude of the late slow wave and stimulus exposure observedhere is consistent with previous ERP work in infants. For instance, Wiebeet al. (2006) reported that slow wave amplitude was larger in magnitude fora condition in which different novel pictures were each presented once (norepetition) compared with a condition in which the same initially novel pic-ture was presented across multiple trials (repetition). In addition, Snyderet al. (2002) examined the effects of repetition on amplitude of the late slowwave in 6-month-olds using a block design, and found that the amplitude ofthe slow wave decreased from early to later blocks of trials, regardless ofwhether the repeated stimulus was initially novel or familiar. Furthermore,these effects were observed over anterior-temporal and central leads, consis-tent with the topography of repetition effects found in the present study.

NEURAL CORRELATES OF ENCODING IN INFANTS 289

Most recently, Snyder, Garza, Zolot, and Kresse (in press) found that evena single repetition of a visual stimulus results in reduced amplitudes of thelate slow wave amplitude over right-central leads, and that stimuli for whichthe infant has a great deal of exposure (i.e., highly familiar stimuli) elicitreduced amplitudes of the late slow wave over anterior-temporal leads com-pared with novel stimuli. Taken together, findings from this and previouswork suggest that the amplitude of the late slow wave component is sensitiveto stimulus exposure, and decreases in magnitude with more exposure, par-ticularly over central and anterior-temporal leads.

One question is whether these exposure-related amplitude modulations ofthe late slow wave reflect encoding, or some other process such as decreasesin cortical arousal or attention? There are several reasons to think that thereduction in amplitude of the late slow wave with stimulus repetition, atleast over right-anterior-temporal leads, reflects encoding. First, infants whomet the criterion for habituation exhibited a novelty preference at test, indi-cating that they had fully encoded the stimulus. Second, among theseinfants, the amplitude of the slow wave decreased across blocks of trials,and smaller amplitudes of the slow wave predicted better memory perfor-mance at test (i.e., larger novelty scores). Third, greater decreases in slowwave amplitude at right-anterior-temporal leads during encoding predictedbetter memory performance at test, suggesting that exposure-relateddecreases in slow wave amplitude at right-anterior-temporal leads reflectbetter encoding of the familiar stimulus. The right hemisphere focus ofencoding-related amplitude modulations of the late slow wave observedhere is consistent with a right hemisphere bias in the encoding of nonverbal,complex visual stimuli (Brewer, Zhao, Desmond, Glover, & Gabrieli, 1998).

In addition, the pattern of findings for the late slow wave observed here isconsistent with a large body of literature indicating that neural responses inthe occipital-temporal visual processing pathway and portions of the med-ial-temporal lobe are reduced when stimuli are repeated. Electrophysiologi-cal recordings from the temporal lobe of monkeys during recognition tasksand neuroimaging work in adults has shown that a common mechanismunderlying visual recognition memory is decreased neural firing to stimulithat were previously encountered, a phenomenon known as repetition sup-pression (Begleiter, Porjesz, & Wang, 1993; Fahy, Riches, & Brown, 1993;Li, Miller, & Desimone, 1993; Wan, Aggleton, & Brown, 1999; Wiggs &Martin, 1998; Xiang & Brown, 1998; Zhu, Brown, McCabe, & Aggleton,1995). Both animal and human work has demonstrated that repeated exposureto the same visual stimulus leads to short- and long-term suppression of neuro-nal responses in certain medial-temporal lobe structures (i.e., entorhinal andperirhinal cortex), as well as adjacent high-level visual perceptual areas (e.g.,visual area TE; for a review, see Aggleton & Brown, 2005). It has been pro-

290 SNYDER

posed that repetition suppression reflects learning about the critical features ofa stimulus (Desimone, 1996; Wiggs & Martin, 1998). As a stimulus is repeated,the response of neurons that were initially activated but were not highly selec-tive for the features of the stimulus decreases (Li et al., 1993). As a result, repeti-tion suppression results in an overall reduction in neural activity that reflectsgreater specificity in the representation of stimulus features. The constructionof this type of stimulus memory is thought to reflect cortical learningmechanisms, independent of the hippocampus, and has been found to belong lasting (Fahy et al., 1993; Li et al., 1993). In the present context, wefound that greater decreases in slow wave amplitude at right-anterior-temporal leads during encoding predicted better memory performance at test.The pattern and topography of these findings are consistent with observationsof repetition suppression in the occipital-temporal visual processing pathway.

Our main hypothesis was that exposure-related decreases in slow waveamplitude during encoding would predict better memory performance at test(i.e., larger novelty scores). Although we found support for this hypothesisat right-anterior-temporal leads, the findings at mid- and right-central leadsare less straightforward. That is, although slow wave amplitude decreasedwith stimulus repetition at both mid- and right-central leads, amplitude ofthe slow wave at these leads did not predict memory performance at test.One possibility is that repetition-related decreases in slow wave amplitude atthese leads reflect decreases in cortical arousal or general attention (i.e., non-specific effects of stimulus repetition). Another possibility is that repetitionsuppression at these leads reflects recency effects (i.e., whether a particularstimulus has been encountered recently) rather than familiarity effects (i.e.,whether a particular stimulus is familiar or not).

Characterizations of item-specific neural activity in the temporal lobehave identified three distinct patterns of neuronal activity associated withrecognition judgments (Baylis & Rolls, 1987; Fahy et al., 1993; Li et al.,1993; Miller, Li, & Desimone, 1991; Xiang & Brown, 1998). Most relevantto the present study, neurons that signal familiarity exhibit reduced activ-ity in response to familiar compared with novel stimuli (e.g., Fahy et al.,1993; Xiang & Brown, 1998), whereas neurons that signal recency exhibitreduced activity in response to stimulus repetition, regardless of whetherthe repeated stimulus is initially familiar or novel. Given that familiarityand recency neurons are simultaneously active in the temporal lobe, it ispossible that the late slow wave reflects a mixture of signals, with differ-ent signals being more detectable at some scalp regions than others. Inline with this possibility, and the results reported here, Snyder et al.(in press) recently found different effects of stimulus repetition over centraland anterior-temporal leads. Specifically, single repetitions of individualstimuli resulted in suppression of slow wave amplitude for both highly

NEURAL CORRELATES OF ENCODING IN INFANTS 291

familiar and novel stimuli over right-central leads, most consistent withrecency effects, whereas familiar compared with novel stimuli elicitedreduced amplitudes of the late slow wave over anterior-temporal leads,most consistent with familiarity effects. In the present context, stimulusrepetition was associated with statistically significant decreases in slowwave amplitude at mid- and right-central leads that was not predictive ofinfants’ subsequent ability to distinguish familiar and novel stimuli,whereas smaller (and nonsignificant) reductions in slow wave amplitude atright-anterior-temporal leads did predict memory performance. These find-ings are consistent with the possibility that slow wave activity at centralleads may reflect recency signals, whereas slow wave activity at anterior-temporal leads may reflect familiarity signals. As familiarity signals distin-guish between familiar and novel stimuli and recency signals do not,familiarity signals (at temporal leads) should predict memory performancein the paired-comparison, whereas recency signals (at central leads) shouldnot.

In contrast to the slow wave, amplitude of the Nc did not predict infants’memory performance at test. This lack of association between Nc amplitudeand memory performance is consistent with previous studies in 6-month-oldinfants that have reported no differences in Nc amplitude between familiarand novel stimuli when the infant has received only brief preexposure to thefamiliar stimulus prior to testing (Nelson & Collins, 1991; Richards, 2003).By contrast, studies comparing highly familiar stimuli (e.g., a picture of themother’s face or the infant’s favorite toy) to novel stimuli in 6-month-oldshave reported amplitude differences for the Nc (de Haan & Nelson, 1997,1999; Snyder et al., in press). One possibility is that Nc amplitude may bemodulated by memory in 6-month-olds only when the infant has a signifi-cant amount of previous exposure to a stimulus, such as experience inter-acting with a stimulus repeatedly over a period of days or weeks. If so, Ncamplitude may be modulated by long-term but not short-term memory in6-month-olds. In studies employing brief preexposure, as well as the presentstudy, infants may not have had sufficient time to consolidate representa-tions of familiar stimuli into long-term memory. This possibility is consistentwith the idea that the ERPs measured in the present study reflect processesassociated with the encoding of a new stimulus more so than retrieval offamiliar representations from long-term memory stores.

Implications for memory performance in the paired-comparison procedure

Our main finding was that encoding-related decreases in slow wave ampli-tude at right-anterior-temporal leads predicted better memory performanceat test. Desimone and Duncan (1995) proposed a model of selective visual

292 SNYDER

attention, the biased competition model, that can account for the relationbetween repetition suppression and the bias to look longer at novel stimuli.A critical assumption of this model is that the nervous system has a limitedcapacity for processing visual information such that objects in the visualenvironment must compete for processing resources. In theory, reduced acti-vation to a repeated stimulus results in a smaller neural signal for familiarstimuli, biasing the competition for visual processing resources (and, hence,visual attention) toward novel stimuli. Empirically, repetition suppressionhas been shown to produce orienting to a novel stimulus in monkeys (Desi-mone, Miller, Chelazzi, & Lueschow, 1994; Li et al., 1993). In the presentstudy, we found that greater decreases in slow wave amplitude over right-anterior-temporal leads were associated with enhanced orientation to anovel stimulus at test. The pattern and topography of these findings areconsistent with observations of repetition suppression, and suggest that nov-elty preferences obtained immediately after encoding in the paired-compari-son test may reflect a stimulus-driven bias toward novelty in visual selectiveattention. In line with this possibility, recent work in 6-month-old infantshas shown that greater repetition-related decreases in stimulus-inducedgamma activity are associated with greater orienting toward a novel stimu-lus in a habituation–dishabituation paradigm (Snyder & Keil, 2008).

The biased competition model represents an alternative to the morewidely accepted view that novelty preferences in the paired-comparisondepend on the hippocampus and reflect a form of explicit or declarativememory (McKee & Squire, 1993; Nelson, 1995; Rose et al., 2004; but seeSchacter & Moscovitch, 1984; Haith, 1998; Snyder, 2007; Mandler, 2007).This latter view is viable, in part, because amnesic patients with impairedexplicit memory and nonhuman primates with medial-temporal lobe lesionsare impaired in exhibiting novelty preferences, although there are alternativeinterpretations of these results (Mandler, 2007; Snyder, 2007). Importantly,however, evidence from amnesic patients and experimental animals withhippocampal or other medial-temporal lobe damage indicates that noveltypreferences are not impaired when subjects are tested at short delays or imme-diately after encoding (for reviews, see Snyder, 2007; Snyder & Torrence,2008). By contrast, monkeys with lesions of visual perceptual areas, but notthe medial-temporal lobe, are impaired in the paired-comparison testat all delays (Buffalo et al., 1999; Haggar, Brickson, & Bachevalier, 1985).These data suggest that immediate performance in the paired-comparisonmay not require the hippocampus or medial-temporal lobe, and could besupported by short-term forms of memory in high-level visual perceptualareas (such as inferior-temporal cortex). Future work utilizing structuralmagnetic resonance imaging in combination with source analysis techniquesmay provide better spatial information needed to evaluate this hypothesis.

NEURAL CORRELATES OF ENCODING IN INFANTS 293

Limitations of the current study

There are differences between our presentation technique and typicalinfant-controlled habituation and paired-comparison procedures that maylimit the comparison of our study to typical looking procedures used withinfants. For instance, we presented the repeated stimulus for a fixed,relatively brief period of time in order to collect ERPs during encoding.By contrast, the familiar stimulus is presented for a longer, continuousperiod of time during a typical habituation procedure or during theencoding phase of the paired-comparison procedure. In addition, althoughthe familiar stimulus is sometimes presented centrally during the encodingphase of the paired-comparison, it is often presented simultaneously onthe left and right sides of the screen. Thus, one question is whetherinfants were encoding stimuli in the same way in our study as in otherstudies using habituation or the paired-comparison procedure (althoughnote that the same question might be raised for variants of the paired-comparison involving central versus side-by-side presentations of thefamiliar stimulus during the familiarization phase). Although there may bedifferences in the way infants encode stimuli presented for short versus longerperiods of time, we would emphasize that encoding in a typical looking para-digm and the present study is the result of repeated, passive visual exposure.That is, the infant forms a memory of the stimulus simply by looking at itrepeatedly. For this reason, we suggest that examining infants’ brain activityduring repeated presentations of a visual stimulus, and relating this neuralactivity to behavioral evidence that a memory was formed, can provideimportant insights into memory formation in typical infant looking para-digms.

In addition, it is interesting to note that although infants who met criteriafor habituation looked significantly longer at the novel stimulus at test,the size of the novelty preference was small. This could reflect the fact thatthe Greebles were sufficiently complex and similar in appearance to makethe discrimination task relatively difficult for this age group. Indeed, previ-ous work has shown that both stimulus complexity and the degree of physi-cal similarity between two stimuli affect infants’ visual preferences at test(e.g., Bornstein, 1981; Caron & Caron, 1969; Fagan, 1971; Miranda &Fantz, 1974). We selected Greebles specifically because they were complex,so that we would be able to collect a sufficient number of trials for ERP con-struction before habituation occurred, and also because they are physicallysimilar to one another, allowing for more variability in infants’ visual prefer-ences at test (as we intended to examine correlations). Thus, it appears thatthe selection of stimuli for optimization of ERP recording is somewhat atodds with the selection of stimuli for optimization of novelty preferences,

294 SNYDER

and this will likely present unique challenges for the use of ERPs for examin-ing the neural basis of preferential-looking in infants.

Alternatively, the small size of the novelty preference in infants whohabituated could reflect, in part, a violation of infants’ expectations aboutthe number and duration of stimuli presented. During the encoding phase, asingle object is presented in the center of the screen for 500 msec. During thetest phase, however, two objects are presented on the left and right sides ofthe screen for 5 sec or longer. Thus, there are two types of ‘‘novelty’’ presentduring the test phase: stimulus novelty (i.e., the novel object) and presenta-tion novelty (i.e., number, location, and duration of stimuli presented). Ifinfants’ had developed an expectation for how many stimuli would bepresented on each trial, and for how long, infants’ looking behavior duringthe test phase might reflect, in part, a violation of this expectation. It isimportant to note that novelty scores were calculated as the percentage oftime infants’ spent looking at the novel object, with respect to both objects,and therefore reflect discrimination of stimulus novelty and not discrimina-tion of the number, location, or duration of stimulus presentation betweenencoding and test. If infant looking during the test phase were driven purelyby surprise at the change in stimulus presentation, we would have foundequivalent looking at the familiar and novel stimulus at test. Yet, infantswho habituated showed a significant novelty preference and infants who didnot habituate showed a familiarity preference, indicating that infants wereable to discriminate between the familiar and novel stimuli despite differ-ences in presentation between encoding and test. It is possible, however, thatdifferences in stimulus presentation between encoding and test may haveaffected infants’ looking behavior, possibly interfering with infants’ discrimi-nation of the familiar and novel stimulus and reducing novelty scores. Thisissue reflects another unique challenge for the use of ERPs in examining theneural basis of preferential-looking in infants, as it is not possible to recordERPs in response to two stimuli presented simultaneously on the left andright sides of the screen.

CONCLUSION

The paired-comparison procedure has great utility as a nonverbal measureof memory, yet important questions remain about the nature of the pro-cesses that underlie infants’ performance in the task. The results reportedhere are consistent with a large body of evidence from amnesic patients andexperimental animals suggesting that infants’ immediate performance in thepaired-comparison test may reflect a stimulus-driven bias toward novelty invisual selective attention mediated by repetition suppression in the occipital-

NEURAL CORRELATES OF ENCODING IN INFANTS 295

temporal visual processing pathway. Given the important implications ofour results and conclusion, additional research should be aimed at testingthe biased competition model.

ACKNOWLEDGMENTS

Research and manuscript preparation were made possible in part by grantsfrom the National Institutes of Mental Health (MH12132) and Child Healthand Human Development (R03-HD049366) to Kelly Snyder, and by fundsprovided by the NIH to the laboratory of Dr. Charles A. Nelson (NS32976).I am indebted to Charles Nelson, Marshall Haith, and Sarah Watamura forvery helpful comments on previous versions of this manuscript. I also thankDana Keufner, Emily Schunk, Neil Wilczek, and Liza Zolot for assistancetesting participants and coding videotapes, and am especially grateful to theparents and infants who participated in this research.

REFERENCES

Aggleton, J. P., & Brown, M. W. (2005). Contrasting hippocampal and perirhinal cortex func-

tion using immediate early gene imaging. The Quarterly Journal of Experimental Psychology,

58B, 218–233.

Bahrick, L. E., Hernandez-Reif, M., & Pickens, J. N. (1997). The effect of retrieval cues on

visual preferences and memory in infancy: Evidence for a four-phase attention function. Jour-

nal of Experimental Child Psychology, 67, 1–20.

Bahrick, L. E., & Pickens, J. N. (1995). Infant memory for object motion across a period of

three months: Implications for a four-phase attention function. Journal of Experimental Child

Psychology, 59, 343–371.

Bauer, P. J., Wiebe, S. A., Carver, L. J., Waters, J. M., & Nelson, C. A. (2003). Developments

in long-term explicit memory late in the first year of life: Behavioral and electrophysiological

indices. Psychological Science, 14, 629–635.

Baylis, G. C., & Rolls, E. T. (1987). Responses of neurons in the inferior temporal cortex

in short term and serial recognition memory tasks. Experimental Brain Research, 65,

614–622.

Begleiter, H., Porjesz, B., & Wang, W. (1993). A neurophysiologic correlate of visual short-term

memory in humans. Electroencephalography and Clinical Neurophysiology, 87, 46–53.

Bornstein, M. H. (1981). Psychological studies of color perception in human infants: Habitua-

tion, discrimination and categorization, recognition, and conceptualization. Advances in

Infancy Research, 1, 1–40.

Brewer, J. B., Zhao, Z., Desmond, J. E., Glover, G. H., & Gabrieli, J. D. E. (1998). Making

memories: Brain activity that predicts how well visual experience will be remembered.

Science, 281, 1185–1187.

Buffalo, E. A., Ramus, S. J., Clark, R. E., Teng, E., Squire, L. R., & Zola, S. M. (1999). Dissoci-

ation between the effects of damage to perirhinal cortex and area TE. Learning & Memory, 6,

572–599.

296 SNYDER

Caron, A. J., & Caron, R. F. (1969). Degree of stimulus complexity and habituation of visual

fixation in infants. Psychonomic Science, 14, 78–79.

Carver, L. J., Bauer, P. J., & Nelson, C. A. (2000). Associations between infant brain activity

and recall memory. Developmental Science, 3, 234–246.

Courage, M. L., & Howe, M. L. (1998). The ebb and flow of infant attentional preferences:

Evidence for long-term recognition memory in 3-month-olds. Journal of Experimental Child

Psychology, 70, 26–53.

Dannemiller, J. L. (1984). Infant habituation criteria. I. A Monte Carlo study of the 50%

decrement criterion. Infant Behavior & Development, 7, 147–166.

de Haan, M., & Nelson, C. A. (1997). Recognition of the mother’s face by six-month-old

infants: A neurobehavioral study. Child Development, 68, 187–210.

de Haan, M., & Nelson, C. A. (1999). Brain activity differentiates face and object processing in

6-month-old infants. Developmental Psychology, 35, 1113–1121.

de Haan, M., Pascalis, O., & Johnson, M. H. (2002). Specialization of neural mechanisms

underlying face recognition in human infants. Journal of Cognitive Neuroscience, 14, 199–209.

Desimone, R. (1996). Neural mechanisms for visual memory and their role in attention. PNAS,

93, 13494–13499.

Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual attention. Annual

Reviews in Neuroscience, 18, 193–222.

Desimone, R., Miller, E. K., Chelazzi, L., & Lueschow, A. (1994). Multiple memory systems in

the visual cortex. In M. Gazzaniga (Ed.), Cognitive neurosciences (pp. 475–486). Cambridge,

MA: MIT Press.

Diamond, A. (1995). Evidence of robust recognition memory early in life even when assessed by

reaching behavior. Journal of Experimental Child Psychology, 59, 419–456.

Fagan, J. F. (1970). Memory in the infant. Journal of Experimental Child Psychology, 9, 217–

226.

Fagan, J. F. (1971). Infants’ recognition memory for a series of visual stimuli. Journal of Experi-

mental Child Psychology, 11, 244–250.

Fagan, J. F. (1974). Infant recognition memory: The effects of length of familiarization and type

of discrimination task. Child Development, 45, 351–356.

Fahy, F. L., Riches, I. P., & Brown, M. W. (1993). Neuronal activity related to visual recogni-

tion memory: Long-term memory and the encoding of recency and familiarity information in

the primate anterior and medial inferior rhinal cortex. Experimental Brain Research, 96, 457–

472.

Fantz, R. L. (1964). Visual experience in infants: Decreased attention to familiar patterns rela-

tive to novel ones. Science, 146, 668–670.

Gauthier, I., & Tarr, M. J. (1997). Becoming a ‘‘Greeble’’ expert: Exploring mechanisms for

face recognition. Vision Research, 37, 1673–1682.

Gilmore, R. O., & Thomas, H. (2002). Examining individual differences in infants’ habituation

patterns using objective quantitative techniques. Infant Behavior & Development, 25, 399–412.

Haggar, C., Brickson, M., & Bachevalier, J. (1985). Sparing of visual recognition after neonatal

lesions of inferior temporal cortex in infant monkeys. Journal of the Society for Neuroscience

Abstracts, 11, #831.

Haith, M. M. (1998). Who put the cog in infant cognition? Is rich interpretation too costly?

Infant Behavior and Development, 21, 167–179.

Horowitz, F. D., Paden, L., Bhana, K., & Self, P. (1972). An infant-control procedure for

studying infant visual fixations. Developmental Psychology, 7, 90.

Hunter, M. A., Ames, E. W., & Koopman, R. (1983). Effects of stimulus complexity and

familiarization time on infant preferences for novel and familiar stimuli. Developmental

Psychology, 19, 338–352.

NEURAL CORRELATES OF ENCODING IN INFANTS 297

Hunter, M. A., Ross, H. S., & Ames, E. W. (1982). Preferences for familiar or novel toys: Effect

of familiarization time in 1-year-olds. Developmental Psychology, 18, 519–529.

Jasper, H. H. (1958). The ten-twenty electrode system of the international federation. Electroen-

cephalography and Clinical Neurophysiology, 10, 371–375.

Junghofer, M., Elbert, T., Tucker, D. M., & Rockstroh, B. (2000). Statistical control of artifacts

in dense array EEG ⁄MEG studies. Psychophysiology, 37, 1–10.

Li, L., Miller, E. K., & Desimone, R. (1993). The representation of stimulus familiarity in

anterior inferior temporal cortex. Journal of Neurophysiology, 69, 1918–1929.

Lukowski, A. F., Wiebe, S. A., Haight, J. C., DeBoer, T., Nelson, C. A., & Bauer, P. J. (2005).

Forming a stable memory representation in the first year of life: Why imitation is more than

child’s play. Developmental Science, 8, 279–298.

Mandler, J. M. (2007). How do we remember? Let me count the ways. In L. M. Oakes & P. J.

Bauer (Eds.), Short- and long-term memory in infancy and early childhood: Taking the first

steps toward remembering (pp. 271–290). New York: Oxford University Press.

McKee, R. D., & Squire, L. R. (1993). On the development of declarative memory. Journal of

Experimental Psychology: Learning, Memory, and Cognition, 19, 397–404.

Miller, E. K., Li, L., & Desimone, R. (1991). A verbal mechanism for working and recognition

memory in inferior temporal cortex. Neuroscience Abstracts, 2, 1377–1379.

Miranda, S. B., & Fantz, R. L. (1974). Recognition memory in Down’s syndrome and normal

infants. Child Development, 45, 651–660.

Nelson, C. A. (1995). The ontogeny of human memory: A cognitive neuroscience perspective.

Developmental Psychology, 31, 723–738.

Nelson, C. A., & Collins, P. F. (1991). Event-related potential and looking-time analysis of

infants’ responses to familiar and novel events: Implications for visual recognition memory.

Developmental Psychology, 27, 50–58.

Nelson, C. A., & Monk, C. S. (2001). The use of event-related potentials in the study of cogni-

tive development. In C. A. Nelson & M. Luciana (Eds.), Handbook of developmental cognitive

neuroscience (pp. 125–136). Cambridge, MA: MIT Press.

Nelson, C. A., Thomas, K. M., de Haan, M., & Wewerka, S. S. (1998). Delayed recognition

memory in infants and adults as revealed by event-related potentials. International Journal

of Psychophysiology, 29, 145–165.

Paller, K. A., & Wagner, A. D. (2002). Observing the transformation of experience into

memory. Trends in Cognitive Science, 6, 93–102.

Reynolds, G. D., & Richards, J. E. (2005). Familiarization, attention, and recognition memory

in infants: An event-related potential and cortical source localization study. Developmental

Psychology, 41, 598–615.

Richards, J. E. (1997). Effects of attention on infants’ preference for briefly exposed visual

stimuli in the paired-comparison recognition-memory paradigm. Developmental Psychology,

33, 22–31.

Richards, J. E. (2003). Attention affects the recognition of briefly presented visual stimuli in

infants: An ERP study. Developmental Science, 6, 312–328.

Robinson, A. J., & Pascalis, O. (2004). Development of flexible visual recognition memory in

human infants. Developmental Science, 7(5), 527–533.

Roder, B. J., Bushnell, E. W., & Sasseville, A. M. (2000). Infants’ preferences for familiarity

and novelty during the course of visual processing. Infancy, 1, 491–507.

Rose, S. A., & Feldman, J. F. (1990). Infant cognition: Individual differences and

developmental continuities. In J. Colombo & J. W. Fagen (Eds.), Individual differences in

infancy: Reliability, stability, prediction (pp. 229–245). Hillsdale, NJ: Lawrence Erlbaum.

Rose, S. A., Feldman, J. F., & Jankowski, J. J. (2004). Infant visual recognition memory. Devel-

opmental Review, 24, 74–100.

298 SNYDER

Rose, S. A., Feldman, J. F., & Jankowski, J. J. (2007). Developmental aspects of visual recogni-

tion memory in infancy. In L. M. Oakes & P. J. Bauer (Eds.), Short- and long-term memory in

infancy and early childhood: Taking the first steps toward remembering (pp. 153–178). New

York: Oxford University Press.

Rose, S. A., Gottfried, A. W., Melloy-Carminar, P., & Bridger, W. H. (1982). Familiarity and

novelty preferences in infant recognition memory: Implications for information processing.

Developmental Psychology, 18, 704–713.

Schacter, D. L., & Moscovitch, M. (1984). Infants, amnesics, and dissociable memory systems.

In M. Moscovitch (Ed.), Infant memory (pp. 173–216). New York: Plenum Press.

Snyder, K. A. (2007). Neural mechanisms of attention and memory in preferential-looking

tasks. In L. M. Oakes & P. J. Bauer (Eds.), Short- and long-term memory in infancy and early

childhood: Taking the first steps toward remembering (pp. 179–208). New York: Oxford

University Press.

Snyder, K. A., Garza, J., Zolot, L., & Kresse, A. (in press). Electrophysiological signals of

familiarity and recency in the infant brain. Infancy.

Snyder, K. A., & Keil, A. (2008). Repetition suppression of induced gamma activity predicts

enhanced orienting toward a novel stimulus in 6-month-old infants. Journal of Cognitive Neu-

roscience, 20, 2137–2152.

Snyder, K. A., & Torrence, C. M. (2008). Habituation and novelty. In M. M. Haith & J. B. Ben-

son (Eds.), Encyclopedia of infant and early childhood development (Vol. 2, pp. 51–63). Oxford:

Elsevier Press.

Snyder, K. A., Webb, S. J., & Nelson, C. A. (2002). Theoretical and methodological implica-

tions of variability in infant brain response during a recognition memory paradigm. Infant

Behavior & Development, 25, 466–494.

Thomas, H., & Gilmore, R. O. (2004). Habituation assessment in infancy. Psychological Meth-

ods, 9, 70–92.

Thomas, D. G., Grice, J. W., Najm-Briscoe, R. G., & Miller, J. W. (2004). The influence of

unequal number of trials on ERP averages. Developmental Neuropsychology, 26, 753–774.

Wagner, S. H., & Sakovits, L. J. (1986). A process analysis of infant visual and cross-modal rec-

ognition memory: Implications for an amodal code. In L. Lipsitt & C. Rovee-Collier (Eds.),

Advances in infancy research (Vol. 4, pp. 195–217). Norwood, NJ: Ablex.

Wan, H., Aggleton, J. P., & Brown, M. W. (1999). Different contributions of the hippocampus

and perirhinal cortex to recognition memory. Journal of Neuroscience, 19, 1142–1148.

Webb, S. J., Long, J. D., & Nelson, C. A. (2005). A longitudinal investigation of visual event-

related potentials in the first year of life. Developmental Science, 8, 605–616.

Wiebe, S. A., Cheatham, C. L., Lukowski, A. F., Haight, J. C., Muehleck, A. J., & Bauer, P. J.

(2006). Infants’ ERP responses to novel and familiar stimuli change over time: Implications

for novelty detection and memory. Infancy, 9, 21–44.

Wiggs, C. L., & Martin, A. (1998). Properties and mechanisms of perceptual priming. Current

Opinion in Neurobiology, 8, 227–233.

Xiang, J.-Z., & Brown, M. W. (1998). Differential neuronal encoding of novelty, familiarity and

recency in regions of the anterior temporal lobe. Neuropharmacology, 37, 657–676.

Zhu, X. O., Brown, M. W., McCabe, B. J., & Aggleton, J. P. (1995). Effects of the novelty or

familiarity of visual stimuli on the expression of the intermediate early gene c-fos in the rat

brain. Neuroscience, 69, 821–829.

NEURAL CORRELATES OF ENCODING IN INFANTS 299