Psychophysiology, 32 (1995). 274-285. Cambridge University Press. Primed in ihe DSA.Copyright <y 1995 Society for Psychophysiological Research

The N400 as a function of the level of processing

DOROTHEE J. CHWILLA, COLIN M. BROWN, AND PETER HAGOORTMax Planck Institute for Psycholinguistics. Nijmegen. The Netherlands

AbMraci

In a semantic priming paradigm, the effects of different levels of processing on the N400 were assessed by chang-ing the task demands. In the lexical decision task, subjects had to discriminate between words and nonwords. andin the physical task, subjects had to discriminate between upperca.se and lowercase letters. The proportion of relatedversus unrelated word pairs differed between conditions. A lexicality test on reaction times demonstrated that thephysical task was performed nonlexically. Moreover, a semantic priming reaction time effect was obtained onlyin the lexical decision task. The level of processing clearly affected the event-related potentials. An N4(X) primingeffect was only observed in the lexical decision task. In contrast, in the physical task a P300 cflecl was observedfor either related or unrelated targets, depending on their frequency of occurrence. Taken together, the resultsindicate that an N40() priming effect is only evoked when the task performance induces the semantic aspects ofword.s to become part of an episodic trace of the stimulus event.

I)eM;riplors: Levels of processing. Semantic priming. N400, P300

In this study, we investigated the influence of task demands onthe N400 semantic priming effect. In particular, we focused onthe impact on the N400 of different levels of processing of lexicalstimuli. In this report, we first discuss the relevant semantic prim-ing effects and the mechanisms responsible for these primingeffects then we introduce the levels of processing frameworkwithin which semantic priming effects were investigated.

One of the most consistent findings in the psycholinguisticliterature is that words are processed faster and more accuratelywhen they are preceded by a semantically related or a.s.sociatedword than by an unrelated word (e.g.. Meyer & Schvaneveldt.1971; see Neely. 1991. for a review). More recently. i( has beendemonstrated that semantic priming effects can also be recordedwith the use of the event-related brain potential (LRP) method.Kutasand Hillyard (1980. 1984) identified an ERP component,the N4D0. a negative peak with a mean latency of 400 ms anda centroparietal distribution, that is larger in amplitude forwords that are semantically incongruent with a preceding sen-tence context. Subsequent research has shown that the N4()0 istied more to semantic expectancy than to anomaly (e.g., Kutas.Lindamood. & Hillyard. 1984).

This research was supported by grant 560-256-048 from the Neth-erlands Organization for Scientific Research (NWO).

We thank Marta Kutas. two anonymous reviewers, and Tony Gail-lard for helpful comments on earlier versions of this paper. We are alsograteful to Johan Wcustink for technical assiMance and to Inge Doehringand Uli C'hwilla for graphical support.

Address reprint requests to: Dorothee .1. Chwilla. Nijmegen Institutefor Cognition and Information. P.O. Box 9104. 6500 Hh Nijmegen. TheNetherlands.

A modulation in N400 atnplitude as a function of semanticcontext is not only obtained in sentences but also when a singleword provides the context (see Kutas & Van Petten. 1988. fora review). If a target word Ls preceded by a semantically relatedprime, it elicits a stnaller N40() than when it is preceded by asemantically unrelated word (e.g.. Bentin. McCarthy. & Wood.1985; Holcomb. 1988; Holcomb & Neville, 1990; Kutas & Hill-yard, 1989). This difference in amplitude is referred to as theN4(X) priming effect. Because of the sensitivity of this ERP com-ponent to language processes, the N400 has been u.sed as a newmeasure for psycholinguistic research (Kutas & Van Petten.1988). It is by now well established that the amplitude of theN4(X) varies as a function of the degree to which the elicitingword relates to its preceding semantic context. On the basis ofthe literature, however, il is less clear which processing aspectsof the language coniprehen.sion system are reflected in the N4(K),

In functional tnodels of word recognition (e.g.. Frauenfelder& Tyler, 1987; Marslen-Wilson. 1989; Zwitserlood, 1989), threeprocesses are comtnonly distinguished: lexical access, selection,and integration. Lexical access involves the process of cotnputinga form representation of the physical signal and of tnapping thisrepresentation onto corresponding entries in the mental lexicon.This processing results in the activation of a subset of lexical ele-ments and their semantic and syntactic attributes. The processof selection refers to the process of selecting from the subset ofaccessed elements the element that best matches the input avail-able to the system. Integration refers to the process of integrat-ing a lexical element into a higher order meaning representationof the entire sentence or discourse.

Within this functional tnodel of word recognition, lexicalacce.ss is conceived of as a reflexlike automatic process. In con-

274

Levels of processing and ERPs 275

trast, lexical ituegration is thought of as a tnore controlled pro-cess that can be guided by the subject'.s awareness of theinforrnatiotial cotitetit of the discour.se and that requires moreprocessing re.sources than does lexical access. However, at the.satne time lexical integration is a tnandatory process (Fodor,1983); in normal discourses, subjects of necessity engage in theintegration of individual word meanings into message-levelrepresentations.

To assess the value of the N400 as a tneasure for psycholin-guistic research, how the N40() relates to these different lexicalprocesses must be clarified. A well-defined testing giound withinpsycholinguistics for clarifying this issue is that of semanticpriming (cf. Neely, 1991). Frotn the vast amount of reaction titnesemantic priming studies, a fairly well-developed picture hasetiierged of the mechanisms responsible for pritning effects.According to Neely and Keefe (1989), at least three mechanismsare needed to account for the full spectrum of reaction timepriming effects.

The first mechanism is automatic spreading of activation(ASA; Collins & I.oftus, 1975). ASA assumes that in .semanticmemory strong or direct litiks exist between words that areclosely related in tneaning. Presentation of a word activates thecorresponding node of this word in semantic memory, and viathe links to nearby nodes part of this activation automaticallyspreads to words that are related in meaning. As a consequence,the activated nodes representing related words need less time forsub.sequent processing. ASA has all the characteristics of anautomatic process. It is fast acting, of short duration, does notrequire attention or awarene.ss, and presuppo.ses no or only min-imal re.source capacity.

The second tnechanism is expectancy-induced priming(Becker, 1985; Posner & Snyder, 1975). This is a predictive strat-egy in which the subject uses the information provided by theprime to generate an expectancy set for related target words. Ifthe target is included in this set, it will be recognized morequickly. If it is not, the recognition for the target word will beslowed down. 1 his mechanism can be intluenced by instructionsas well as by the list structure of the material. Especially whenthe stimulus list contains a large proportion of related wordpairs, the probability that the expectancy mechanism contrib-utes to the overall priming effect is large (Keefe & Neely, 1990).

The third priming mechanism is semantic matching (De Groot,1984; Neely & Keefe, 1989). If letter strings are processed lexi-cally, subjects are assumed to mandatorily match primes and tar-gets post lexically (i.e., following lexical access and selection) forsemantic similarity. In a lexical decision task, the word/nonworddecision is inlluenced by the result of this matching process. Thedetection of a relationship between primes and targets leads toa bias to respond yes. If no such relation is detected, then thereis a bias to respond no.

In sumtnary, three mechanisms are supposed to underlysemantic pritning. The assutnption is that two of these mecha-nisms share core characteristics with mechanisms involved inordinary language processing. That is, ASA shares core char-acteristics with lexical access, and setnantic tnatching shares corecharacteristics with postlexical integration proce.sses (De Groot,1984; Henderson, 1982; Neely, 1991). Insight into the contribu-tion of these different mechanisms to the N400 setnantic prim-ing effect will therefore be of substantial value in relating theN4(X) lo the processes of access, selection, and integration. Theaim of the present study was to separate the contribution of thesepriming mechanisms to the N40() effect in the semantic prim-

ing paradigtn. In particular, we attempted to separate the effectsof ASA from the effects of semantic matching by varying thelevel of processing.

Tlie levels of processing framework as described by Craik andLockhart (1972) have often been used to investigate episodicmemory. This literature has consistently demonstrated a betterrecognition and recall for words that have to be processed onthe basis of semantic features (a deep level of processing) thanfor words that have to be processed on the basis of visual or pho-nemic features (a shallow level of processing; see Jacoby & Dal-las, 1981). The basic idea is that shallow processing of the wordstimuli discourages analysis of their semantic aspects. As aresult, the semantic context might not be incorporated into theformation of the episodic trace of the stimulus event. If seman-tic priming effects are neverthele.ss obtained, they probablyresult from automatic activation spreading within semanticmemory. Consequently, semantic priming effects obtained in ashallow task most likely arise from ASA. In contra.st. deeperprocessing presumably involves the proce.ssing of the semanticaspects of the presented words. In this case, the semantic con-text does exert its inlluence on the formation of the episodictrace. Because both expectancy-induced priming and semanticmatching presuppose the formation of an episodic trace for aparticular wordlike stimulus event, priming effects obtained ina situation that is compatible with deeper levels of processingmight, therefore, also reflect the effects of these additional prim-ing tnechanisms.

In contrast to the popularity of the levels of processingframework within memory research, there are only a few psy-cholinguistic studies that have assessed the effects of the levelof processing on semantic pritning. In these studies, the effectof the level of processing was assessed by manipulating thenature of the information (e.g., orthographic, semantic)required to accurately make a lexical decision on target letterStrings. This manipulation was done in two ways: (a) by vary-ing the kind of nonwords that were presented as targets (DeGroot, 1987; Schulman & Davison, 1977) or (b) by varying thekind of discrimination required on the prime word (Smith, The-odor, & Franklin, 1983) or the target word (De Groot, 1987).Taken together, the results of these studies reveal that the seman-tic priming effect is much larger for tasks that impose a deeperlevel of processing than for tasks for which a more shallow levelsuffices. Although the size of the reaction-time priming effectsdepends on the depth of processing, significant priming effectshave also been reported with shallow processing tasks. Smithet al. (1983) obtained a priming effect when a phonemic analy-sis was performed on the prime. Mitchell et al. (1991) found areaction-time pritning effect in a task in which subjects had todiscriminate words on the basis of physical features. These find-ings demonstrate that the u.se of shallow processing tasks in itselfdoes not necessarily exclude the processing of a word's seman-tic aspects.

The levels of processing approach have also been used todetermine the proces.sing nature of tbe N400. The results of theseERP studies were not conclusive. Some studies demonstratedN400 priming effects with shallow processing tasks (Besson,Fi.schler, Boaz, & Raney, 1992; Kutas & Hillyard, 1989). Otherstudies, however, did not observe N400 priming effects undershallow processing conditions (Bcntin, Kutas, & Hillyard, 1993;Deacon, Breton, Ritter, & Vaughan, 1991). The fact that in somestudies N400 priming effects have been obtained in shallow pro-cessing tasks has been taken as evidence that the N400 is sensi-

276 D.J. Chwilla, C^.M. Brown, and P. Hagoorl

tive to automatic lexical semantic proces.ses (Kutas & Van Petlen,1988). However, this conclusion critically depends on the claimthat the N4(X) priming effects obtained in these shallow process-ing tasks cantiot be attributed to priming mechatiisms other thanASA. To examine this issue these studies will be described inmore detail.

Kutas and Hillyard (1989) obtained an N4(X) primitig effectin a letter search task. Iti this task, the ftrst word was followedafter a stimulus onset asynchrony (SOA) of 700 ms by the sec-ond word, which in turn was followed after I,2(X) ms by a sin-gle letter. Subjects had to indicate whether the letter had beenpresent in either or both of the words. In response to the sec-ond word, an N400 priming effect was observed. AlthoitghKutas and Hillyard (1989) correctly assumed that performingthis task does nol require semantic processing, subjects may stillhave matched the words for semantic similarity. The delaybetween the presentation of the words and the moment subjectswere asked to re.spond might have given subjects ample oppor-tunity to in fact perform some kind of semantic matching, inwhich case the observed N400 priming effect does not necessar-ily stem from ASA. A similar argument holds for the study ofBesson et al. (1992), in which subjects had to indicate whetherthe first and the last letter of two words that were presented withan SOA of 30() ms were the same or different, ln this task, asmall (< I ii\) but reliable N4(K) priming effect was observed.The authors attributed this small N4(X) priming effect to auto-matic lexical semantic processing. However, becau.se the subjectswere also required to perform memory tests on the stimuli, pay-ing attention to word meanings was probably reinforced by thedesign.

The contrasting results of various studies testing the effectsof depth of processing on the N4(X) and the N4<X) priming effectsobtained with shallow levels of processing in some studies mightreflect the fact (hat (he shallow task did not in ail cases effec-tively prevent the occurrence of semantic matching. Alterna-tively, these shallow N4(X) priming effecis might result fromASA. To decide between these two possibilities, more objectivecriteria are needed to ascertain whether the performance of aparticular ta.sk involves lexical processing. One objective methodis computation of the le.xicality effect, that is, the difference inoverall reaction times between all word and all nonword targets.In general, words arc responded to faster than are nonwords.In models of word recognition, lexlcalily effects are atlributedto operations at a lexical level of prcKessing (cf. De Ciroot, 1987).

The aim of the present study was to separate the effects ofA.SA from the effects of semantic matching on (he N4()() prim-ing effect by varying the level of processing. It was asstimed thatN4(K) priming effects for tasks that are compatible with deeperlevels of processing are mediated by ASA and by semanticmatching. In contrast, N4(X) priming effects for shallow process-ing tasks were assumed lo exclusively reflect ASA. To controlfor task-relaled levels of processing, we established whether ornot the task performance depended on lexical characteristics ofthe presented letter strings. If the task were performed nonlex-ically, that would suggest thai indeed the lask discouragessemantic analysis. If N4(K) semantic priming effects wereobtained in the absence of a reaction lime lexicality effect, thenthis N400 effect would most likely be due to automatic spreadof activation in semantic memory.

In the current study, the effects of the level of processing onthe reaction time and the N400 measure wa.s assessed by com-paring the processing of the same words and nonwords in a lex-

ical decision (ask and a physical task. The choice of the taskswas based on previous studies that demonstrated differences inERPs between a lexical decision and a physical task. In thesestudies, a clear reduction in (he size of repetition pritning (Kugg,Furda, & Lorist, 1988) and setnantic pritning effects (Mitchellet al,, 1991) was observed for the physical task, which requiredsubjects to respond on the basis of certain physical aspecis oithe letter strings. Therefore, it was assutned that the level of pro-cessing for the two tasks is different. A visual wotd priming par-adigm was used, in which a prime word was followed aflcr7(K) ms by a target word. The level of processing was manipu-lated by varying the target discritrtination between tasks. In thelexical decision (ask, the target discrittiina(ion was based on aword/nonword discrimination. In the physical task, the targetdiscrimination was based on purely physical features (lettercase), which does not require lexical processing. The u.se of thelexical decision task does not necessarily guarantee access loword meaning, because this task could be performed solely on(he basis of the word form. However, there i.s ample evidencefrom the semantic priming literature thai normally this taskyields semantic pritning effects. Therefore, it is reasotiable toassume (hat (he lexical decision (ask is compatible with .seman-tic analysis, whereas Ihe physical task discourages semantic anal-ysis, because paying attention to the relationship between thewords does not help to perform (he letter ca.se discrimination.

To test whether the physical task is indeed performed noti-lexically, the lexicalily effect was a.ssessed. If lexical processingwere involved in (he (ask perfortnance, the proces.sing o'i wordswottid be faster than the processing of nonwords.' If, however,a lask were performed nonlexically, the reaction titnes to wordtargets would not be differeni from the reaclion limes lo non-word targets. In the absence of a reaction time difference be-tween words and nonwords, we can address the question ofwhether an N4(K) priming effect occurs when the task perfor-mance does nol involve lexical processing. The absence of alexicalily reaction litne effect does not necessarily itnply thatsemantic aspects of the stimuli go unnoticed at all levels of Iheprocessing system. Whether or not these aspecis ate processedto some degree tnusi be inferred frotn the EiRf waveform. If theERP data allow us to infer that semaniic proce.ssing has goneon at sotne level in the langttage system, then we can use Ihe pres-ence or absence of N4()0 effects to make claims about the pritn-ing mechanisms (hat the N4(K) is pjirlictilarly sensitive lo.Assuming Ihat we can etisure thai the physical task is perfortnednonlexically, the following predictions cati be made. If the N4(K)priming effect is mediated by semaniic malching and expecuincybul not by ASA, we expect an N4(K) pritnitig effect in Ihe lexi-cal decision task but no effect in the physical task. In contrast,if Ihe N400 effeci also reflects ASA, an N40() priming effectshould additionally be obtained iti Ihe physical task. An N4(K)priming effect in the physical task wottid itnply thai N4(X) prim-ing effecis occur relatively automatically, ihal is, ihal they areindependent of the task demands.

In addition lo the lask manipulation, we also varied the pro-portion of related word pairs in Ihe slimulu.'t li.sl. The propor-

' Although it is an extremely robust nnding thut words are processedfaster than nonworils, there are a lew exceptions to this rule. I or exatti-plc, words take longer to process wheti very low-lreqiiency words areused or il siibjetis are asked lo classify ilcnis ,Kcordinj< lo wlicllier (heyhave tnore than one meaning (see 1 orstcr & Ilednall, I97f)).

Levels of processing and ERPs 111

tion of related word pairs determines the probability thatsubjects use the pritne to generate an expectancy about the tar-get (Keefe & Neely, 1990). The generation of this expectancyrequires access to the semantic aspects of the prime. Contrast-ing the eftects tor a list with a small nutiiber of related wordpairs with those for a list with a large nutnbcr of related wordpairs under both task conditions further allows us to determinewhether or not semantic factors leak through under the shallowprocessing task conditions.

Method

SubjectsThirty-six right-handed subjects. 20 wotnen and 16 men (meanage = 24 yeats) participated in the experitncnt. Hand dotninancewas asses.sed with an abridged Dutch version of the EdinburghInventory (Oldfield. 1971). Seven subjects reported left-handedness in their itiitnediate fatnily. All were native speakersof Dutch and had nortnal or corrected-to-normal vision. Theywere paid DFl. lO/hr.

Apparatus and StimuliSubjects were seated in a cotnfortable reclining chair in a dimlyilluminated, sound attenuating, and electrically shielded cham-ber. The response device containing two push buttons was fixedon a stiiall table in front of the subjects. The stimuli consistedof 960 visually presented pairs of letter strings (pritne and tar-get cotiibinations). Half of the target stitnuli were real Dutchwords and the other half were nonwords. The nonwords wereconstructed in accordance with the phonotactic constraints ofDutch and were derived from real words by substituting one ortwo letters. The word targets were preceded by a related or anunrelated prime (e.g.. hiack-white vs. soap-bird). A pair wasconsidered related if the target appeared as the highest wordassociation to the pritne in Dutch word-association nortns (DeCiroot, 1980; De Ciroot & de Bil, 1987; Lauteslager, Schaap, &Schievels. 1986). Associative strength was determined by the per-cetUage of occurrence of the target as an associate of the primeatnotig 100 university students. The tiiean percentage of asso-ciation between the pritne and the target for the related pairs was47.41 "/o (SD = I .S.̂ OVo). A pair was con.sidered unrelated if thetarget neither occurred as a word association of the pritne inthese nortns nor had any other obvious sctnantic relation to thepritne. Strings of three to eight letters were presented as targets.Across conditions, targets were balanced on mean word fre-quency (Uit den Boogaart. 1975; corpus size 720,000), word class(nouns, adjectives, verbs), and the number of letters and sylla-bles. There was a total of 160 related and 160 unrelated wordpairs.

A pilot study was conducted to select a subset of target wordsfrom the related and unrelated word pairs that are tnatched onthe basis of reaction times. In this pilot study, word and non-word targets were presented in isolation. Subjects (n = 20) per-fortiied a speeded lexical decison task: They had to indicate asfast as possible whether the target word was a real word or not.On the basis of these reaction titnes, 40 related and 40 unrelatedword pairs were selected. These pairs are referred to as the crit-ical pairs. (The appendix gives all critical pairs and the meanreaction time and language frcqtiency of the targets.) The meanreaction titne for both the critical related and unrelated targetswas 506 tns {SD = 27.6 and 28.0 ms for related and unrelatedtargets, respectively).

Two lists of 480 prime-target pairs were constructed. One listhad a high proportion of related word pairs (.80). The other listhad a low proportion of related word pairs (.20).* The 40 crit-ical related and 40 critical unrelated word pairs were the samein both lists. In the high-proportion list, the critical word pairswere supplemented by 120 related word pairs and in the low-proportion list by 120 unrelated word pairs. Two hundred prime-target pairs with a nonword target were added to each list. Addi-tionally, in each lisi 80 prime-target pairs were included withthe Dutch word blank as a prime; half of these were followedby a real word and the other half were followed by a nonword.These pairs were used as fillers in this experiment.

In the lexical decision task primes and targets were presentedin uppercase letters. In the physical task, primes were in upper-case letters, whereas half of the word and nonword targets werepresented in lowerca.se and the other half in uppercase letters.All critical targets were presented in uppercase letters. The stim-uli were presented at tnoderate contrast at the center of a PCmonitor (window of 8 x 2 cm. white on black). The stimuli sub-tended a visual angle of approximately i° x 0.8°.

The electroencephalogram (EEG) was recorded with Ag/AgCI (d = 9 tnm) electrodes tnonopolarly from three midlinepositions (Fz, Cz, Pz) and two pairs of lateral electrodes. Sym-metrical anterior temporal electrodes were placed halfwaybetween F7 and T3. and F8 and T4 sites, respectively. Symmet-rical posterior temporal electrodes were placed lateral (by iO'/tof the interaural distance) and 12.5% posterior to the vertex.The left mastoid served as reference. Electrode impedance wasless than 3 kO. The electrooculogram (EOG) was recorded bipo-larly using four Ag/AgCl (</ = 6 mm) electrodes. The verticalEOG was recorded by placing an electrode above and below- theright eye. The horizontal EOG was recorded via a right-to-leftcanthal montage. EEG and EOG signals were amplified byNihon Kohden amplifiers (type AB-601G; time constant = 8 s.low-pass filter = -.3 dB cutoff at 30 Hz). A calibration pulse(peak-peak amplitude: 2 mV. duration 1 s) was recorded beforeeach session. All physiological signals were digitized on line witha satnpling frequency of 200 Hz using a 12 bit A/D converter.Control of the presentation of stimuli and recording of per-formance data were accomplished by a Miro GD laboratorycomputer.

ProcedureSubjects were randomly assigned to task (lexical decision orphysical) and list (high proportion or low proportion). Nine sub-jects were tested in each of the four task/list combinations. Sub-jects were told that pairs of letter strings would be presented inrapid succession. The first word (the prime) was followed afteran SOA of 100 tns by a second word (the target). The stimulusduration was 2(X) ms. and the intertrial interval between the off-set of the .second word and the onset of the first word was .̂ .2 s.Subjects wore asked to pay attention to the second word, the tar-get stimulus. The kind of target discrimination varied as a func-tion of the task.

In the lexical decision task, subjects had to decide whetherthe target stimulus was a Dutch word or not. If the target wasa word, they had to press the response button on the right side;

'The relatcdncss proportion in psycholinguistic studies does noticier to the overall probability but is defined as the ratio of related andunrelated word pairs within a list (cf. Neely. t99I).

278 D.J. Chwitta, CM. Brown, and P. Hagoort

if not, they had to press the button on the left. They were askedto respond as fast as possible but to remain accurate.

In the physical task, subjects were asked to decide whetherthe target stimulus was written in upper- or lowercase. If the tar-get appeared in uppercase they had to press the button on theright side, if the target appeared in Iowerca.se they had to pressthe button on the left side. Subjects were asked to respond asquickly as possible.

In both tasks, subjects gave right-hand (right button press)and left-hand (left button press) responses. To facilitate fastresponses and avoid motor artifacts, subjects were asked to keeptheir index fingers on the response keys. Subjects were notinformed about differences in relatedne.ss between the differenttypes of word pairs. The stimuli were presented in two blocksof 240 trials each (mean duration = 16 min). There was an inter-val of 5 min between blocks. A short training session in whichthe proportion of related word pairs was 0.50 preceded theexperimental session. Subjects were trained to speed up reactiontime (<1 s) and to control their eye movements. They weretrained to make eye movements approximately 1 s after the but-ton press and to fixate on the center of the screen in anticipa-tion of the prime-target pairs.

Data AnalysisEEG and EOG records were examined for artifacts and forexcessive EOG amplitude during the epoch from 150 ms preced-ing the prime to I s after the onset of the target. Only trials inwhich the EOG amplitude did not exceed 100 //V and in whichno other artifacts were present were used for averaging. ERF'swere averaged time locked to the target, relative to a 100-ms pre-target baseline.

The window for scoring ERP components was based uponvisual analysis and depended upon the time interval in whichmaximal differences between conditions were obtained. The fol-lowing measures were obtained for each derivation: N400 (aver-age amplitude in the 300-400-ms epoch after the target stimultis)and P300 (average amplitude in the 400-500-ms epoch after thetarget stimulus).

Unless explicitly stated otherwise, all analyses were carriedout on the 40 critical items per condition only. Analyses of theERP data involved analyses of variance (ANOVAs) with task(lexical decision, physical task) and proportion (high, low) a.sbetween-subject variables and relatedncss (related, unrelated)and electrode (Fz, Cz, Pz, anterior temporal bilateral sites, pos-terior temp<3ral bilateral sites) a.s within-subject variables. Whereinteractions with the electrode variable are reported, ANOVAswere performed after performing a Z-score normalization pro-cedure to equalize the mean amplitudes across experimentalcondition.^. This normalization procedure is equivalent to thenormalization procedure suggested by McCarthy and Wood(1985).'

At the reaction time level, the lexicality effect was assessedto reveal whether the physical task was indeed performed non-lexically. For each subject, the mean reaction time was calcu-lated for correct responses to all word target.s, related andunrelated combined (cf. De Groot, 1987). Also, each subject'smean reaction time for all correct nonword target respon.se.s wa.scalculated.

'The Z-«core normalization procedure wa» <le.<K;ribed by Rdiler,Heil, and Glowalla (1993).

To analyze reaction time priming effects, separate ANOVAswere performed for both tasks, with proportion (high, low) andrelatedness (related, unrelated). To control for an increase inType I error in within-subjects tests, the degrees of freedom forthe E tests were adjusted using the procedure described byGreenhouse and Geisser (1959; cf. Winer, 1971). The adjusteddegrees of freedom and p values are presented. The significanceof contra.sts was assessed by post hoc Newman-Kculs tests. Alleffects mentioned are significant at the .05 level or beyond.

Results

Reaction Time DataThe lexicality test revealed that the physical task was performednonlexically, that is, on the basis of nonlexical form informa-tion only. In Ihe physical task, no difference in mean reactiontime was found between the 240 word targets (419 ms) and the240 nonword targets (416 ms). For the lexical decision task, alexicality effect was found: reaction times were shorter for wordtargets (525 ms) than for nonword targets (592 ms) (F[ 1,161 =99.38, p < .0001).

The reaction time results for the critical word pairs are sum-marized in Table 1. A priming effect was obtained in the lexi-cal decision task (/-"11,I6| = 72.05, p < .0001) but not in thephysical task (F[ 1,161 = 0.30, n.s.). In the lexical decision task,a decrease in reaction time was found for related word pairs(519 ms) as compared with unrelated word pairs (559 ms). Forthe lexical decision task, a Proportion x Relatedness interaction(Fll,16| = 20.16. p< .001) revealed that the priming effect wasmore pronounced for the high-proportion condition (59 ms)than for the low proportion condition (21 ms). Post hoc testsdemonstrated that the priming effect in the lexical decision taskwas significant for the high-proportion (p< .01) as well as forthe low-proportion ( p < .01) condition. In the physical task, nosignificant differences in reaction times between related andunrelated word pairs were found.

ERPxGrand averages for the lexical and the physical tasks and forhigh and low proportions are presented in Figure I and Fig-ure 2, respectively, in which related and unrelated word targetsand nonword targets are superimposed.

Inspection of the waveforms suggests that in the lexical deci-sion task an N400 priming effect clearly developed, which wasabsent in the physical task. In both tasks, a P3(X) was evoked,reflecting that a decision was required on the target. The N400in the lexical decision task was affected by relatedness; largeramplitudes were observed for unrelated than for related targets.

Table 1. Reaction Time (ms) for High- and Low-ProportionConditions

Tank

Lexical decisionHighLow

PhysicalHighLow

Related

510527

44H4()4

Unrelated

569548

46040K

Dtfference

3921

124

Levels of processing and ERPs

LEXICAL DECISION

279

LOW

KlRiirc I, Grand ERP averages of the lexical decision task (n = 9 sub-jects) for the left and right anterior (AL, AR), the three midline (Fz, Cz,I'z), and posterior (Pl., PR) electrodes, superimposed for related (thinline), unrelated (broket) line), and notiword (thick line) targets separately(or high and low proportion.

The N400 and Ihe P3(K) were widely distributed. The N400 inIhc lexical decision task was largest at the vertex, whereas theP300 in the physical task showed maxitiial amplitudes at Pz.

The main results of the overall analyses are sutnmarized inTable 2." The main effect of task indicated that overall ampli-tudes were tnore positive in the physical than in the lexical deci-sion task, in particular within the N4(X) window. The interactionbetween task and relatedtiess revealed that the N40() in the lex-ical decision task was larger for unrelated than for related wordpairs, whereas no such difference was found in the physical task.A main effect of proportion within the N400 window reflectedIhat overall atnpliludes were less positive in the high-proportioncondition (2.4 ;iV) than in the low-proportion condition (5.1 fiV).A Task X Proportion x Relatedness interaction was present.This interaction revealed that for the two tasks a different inter-play between proportion and relatedness was observed.

In addition to the overall ANOVA, .semantic primitig effeciswere analyzed for each task separately. These ANOVAs involvedproportion, relatedness, and electrode.

Lexical decision task. The ANOVA yielded a significantlylarger N400 amplitude for unrelated than for related word pairs

^Additional ANOVAs with broader latency regions, in which theN40() was tneasured as the meati amplitude within the .•(00-500-ms epochand the P.KXI was measured as the mean amplitude of the 400-600-msepoch, confirtned all effects reported in Table 2,

Table 2, F Values for ANOVAs for theN400 and the P300

Effect

TaskProportionRelatednessElectrodeTask X RelatednessTask X Proportion x Relatednessrask X ElectrodeRelatedness x Electrode

,32,32,32

1,321,321,32,32

1,32

N400

35.31"*5.97*

22.I3'**15.77***20.52" '

5.81*10.43**9.78**»

P300

8.62"ns

I2.47»»28.26'"22.21***7.26*4.77*5.13*

Note: Interactions with p values > .05 are omitted."Degrees of freedom were adjusted according to Greenhouse andGeisser (1959).•p< 05; "p< .01; ••*/?< .001.

(F[I,161 = 28,50, p < .001), An interaction of Electrode xRelatedness (F[ 1,161 = 10.51, p < .01) indicated that the N400primitig effect was most pronounced at the midline sites, in par-ticular at Cz and Pz. The N4(X) priming effect was smaller atright and left posterior sites and further decreased at lateral ante-rior sites (Figure 3), The N4(X) tended to b)e larger (about 0,5 ;iV)above right posterior than above left posterior sites. In parallel

Figure 2, Grand ERP averages of the physical task (/j = 9 subjects) forthe left and right anterior (AL, AR), the three midline (Fz, Cz, Pz), andposterior (PL, PR) electrodes, superimposed for related (thin line), unre-lated (broken line), and nonword (thick line) targets separately for highand low proportion.

280 D..I. Chwilla. CM. Brown, and P. Hagoort

to the reaction time results. Figures I and 3 suggest that the dif-ference in N400 amplitude between related and unrelated wordpairs was larger for the high-proportion (4.4 jiV) than for thelow-proportion (2.5 fi\) condition. This increase in primingeffect was caused by a larger N400 amplitude for unrelated wordpairs than for related word pairs in the high- than in the low-proportion condition. Although in the overall ANOVA the Pro-portion X Relatedness interaction was not significant {p< .20),a significant interaction was obtained at the left posterior site(p < .05). In a recent extension of this study with a larger num-ber of subjects (n = 15). we observed a similar increase in N4(K)priming effect for the high-proportion condition, which was sta-tistically reliable (Brown. Hagoort. & Chwilla. 1994).

Analysis of the P300 revealed a main effect of electrode(F[ 1,16] = 8.14. /7 < .05), with the largest amplitude at Pz. AsFigure I shows, the N400 priming effect extends up to at least600 ms posttarget. yielding larger P300s for related than forunrelated word pairs (F | 1,161 = 28.35, p < .001). Thereforeoverlap of N400 and P300 cannot be excluded. Because of thisoverlap, it is difficult to a.ssess the effects of the experimentalmanipulations on the P300 in the lexical decision task.

Physical task. A main effect of electrode was observed forthe physical task (F[1.I61 = 24.02, p < .001). Post hoc testsdemonstrated that P300 amplitude was larger (p < .01) at Pzthan at all other electrodes (p < .01) except Cz.

The P300 was affected by the proportion of related wordpairs (FI1.16) = 4.66. p < .05); its amplitude was larger in thelow-proportion (9.1 ^V) than in the high-proportion (5.4 fiV)condition. The ANOVA also yielded an interaction of Propor-tion X Relatedness (F[ 1.16] = 12.18, p < .01). This interactionindicated that the P300 was affected by probability (see Figure 3).In the high-proportion condition, in which related word pairswere presented with a high probability, P300 was larger in re-sponse to unrelated (6.5 fiV) than to related (4.2 jiV) word pairs.A post hoc test revealed that this difference in P30() amplitudewas significant at the 1*̂0 level. In contrast, in the low-proportioncondition. P300 was larger for the infrequent related (9.7 /iV) thanfor the frequent unrelated (8.4 ^V) word pairs. A post hoc testdemonstrated that this difference in P300 amplitude was signif-icant at Pz (p < .05) but not at other electrodes.

Lexicality TestTo investigate the effect of lexicality, ERPs were averaged sep-arately for word and nonword targets. Separate ANOVAs werecarried out for the lexical and the physical task with between-subject variable proportion (high, low) and within-subject vari-ables lexicality (word, nonword) and electrode (seven levels).

The ANOVAs of the lexical decision task yielded a maineffect of lexicality for both the 30O-40()-ms epoch (F[l,161 =71.39./7< .0001). and the 400-500-ms epoch (F|1.16| =93.58,p < .0001). In the early and the late epoch, the waveform wasmore negative going for nonwords than for words. The lexical-ity effect was widespread across the scalp, with a trend towarda larger difference between words and nonwords above the righthemisphere.

A significant effect of lexicality was also obtained in the phys-ical taisk (300-400-ms epoch: F[ 1,161 = 74.18,/?< .0001; 400-500-ms epoch: FI 1,16] = 65.04. p < .0001). In both epochs, thewaveform wa.s less positive for nonwords than for word targets.No interaction of proportion and lexicality was obtained, indi-

cating that the lexicality effect was of .similar size for the high-and the low-proportion condition,'

Topography of the Priming Effectand the Lexicatity EffectInspection of Figures I and 2 suggests that there might be topo-graphical differences between the priming effect and the lexi-cality effect. In the lexical decision task, for instance, botheffects result in a modulation of the N400. The priming effectis largest at centroparietal midline sites and much stnaller at ante-rior sites (see also Mgure 3). In contrast, the lexicality effect hasa more frontal distribution.

To test for possible differential distributions of priming atidlexicality effects, for each task separate ANOVAs were per-formed in which the scalp distributions of the pritning effect andthe lexicality eftcct were compared directly. This analysis wasperformed for the early (N400) and the late (P300) latency region(300-400 ms and 400-500 tns. respectively).

The priming effect was computed by subtracting the meanamplitudes for related targets frotn the mean amplitudes forunrelated targets. The lexicality effect was computed by sub-tracting the mean amplitudes for unrelated targets from themean amplitudes for nonword targets. I or each task, ANOVAswere perfortned on these difference scores, with proportion asthe between-subject variable and ditnension (semantic vs. lexi-cal) and electrode as within-subject variables.

For the lexical decision task, a marginally significant Dimen-sion X Electrode interaction was obtained for the early window(F|l ,16| = 3.19,/7< .10) but not for the late window. No sig-nificant Proportion x Dimension x Electrode interactions wereobtained. These results confirm that the N400 lexicality effectwas more frontally distributed than the ccntroparietally distrib-uted N400 priming effect. Also in the physical task, a tnargin-ally significant Ditnension x Electrode interaction was obtainedfor the early window (F| 1.161 =3.62,/J< .10)but tiot for thelate window.

Whether the distributional differences between priming andlexicality effects have any functional significance is, however,unclear. Because both N4(X) and P300 generator ensembles seem

' However, a possible problem Is that In the previous test unrelatedand related words were collapsed atid contrasted with nonwords and thatthe inclusion of related words confounds the priming effect with the lex-icality effect. Therefore, it tnight be argued that to get a more appropri-ate estimate of the lexicality effect, nonwords should be contrasted onlywith unrelated words. To control for Ihe possibility that the lexicalityeffect actually arises frotn a confounding with the priming dimension,additional ANOVAs were performed in which only unrelated words werecontrasted with nonwords. These analyses essentially revealed the sameresults. Significant lexicality effects for both epochs were obtained inthe lexical decision task (?(K)-4(X)-ms epoch: /-|I.I61 = 16.89./;< .(X)l;4OO-5O()-ms epoch: /- (1. lft| = 17.55. p < .(X)l) and also in the physicaltask (1(X)-400-ms epoch: /-11.I6| = 16.91./>< .(X)l; 4(K) 500-ms epoch:/-|I.I61 =2.3..19./>< .(X)l). The only difference with respect tothe orig-inal analyses was that tor the late epoch in the physical task a Propor-tion X Lexicality interaction (/• (1.16] = 7,.17. p < .02) was obtained. Thisinteraction indicated that the lexicality effect was more pronounced inthe high- thati in the low-proportion condition (see Figure 2). However,the probability for nonwords and unrelated words differed t>etween con-ditions such that in the high-proportion condition unrelated words werepresented with a low probability, tf the lexicality effect in the physicaltask is due to a modulation in P.1(K) atnplitude. then these differencesin probability might affect the amplitude of this positivity. l-igure 2indeed suggests that probability is the most important detertninant ofthe differential amplitude modulation of the positivity.

Levels of proeessing and ERPs 281

LEXICAL TASK

AL AR F J CI PI PL PR *L AR Fj C» Pj PL PR

AL AR Fi Cl PI PL PR AL AR F i C I P I PL PR

Figure 3. Difference in amplitudebetween unrelated and related targetsfor the N400 in the lexical decisiontask and for the P.'OO in the physicallask separately for the high and thelow proportion condition for the threemidline (Fz, Ci. Pz), left and rightanterior t AL. AR), and posterior (PL,PR) electrodes.

to have their effects in at lea.st partly overlapping time windows,their differential contributions to the waveforms might haveresulted in overall topographic differences between lexical andpriming effects. It, therefore, seems premature to interpret thesedifferences in a functional sense.

Discussion

The present findings replicate earlier results that showed thatN400 priming effects are evoked in a lexical decision task. Theduration of the N4(X) observed in Ihis study, its topography, andthe size of ihe pritning effect are in agreement with thosereported in the literature (sec Holcomb, 1988; Holcomb &Neville, 1990). The N4()0 is also quite similar to the cla.ssicalN4(X) observed in sentence cotiipletions (Kutas& Hillyard, 1980).

The goal of this study was to exatnine which aspects of lexi-cal processing (acces.s, .selection, integration) are reflected by theN4(X) effect. This issue was addressed by assessing the effectsof different semantic priming rnechanisms on the N4(X) prim-ing effect. More specifically, we attempted to separate the effectsof automatic spreading of activation from the effects of seman-tic tnatching by varying the level of processing. We comparedthe KRPs in a lexical decision task with the ERPs in a physicaltask. At the reaction time level, the lexicality test indicated thatdifferent levels of processing were induced by performing thetwo tasks. There was no difference in the reaction times for wordand nonword targets in the physical task, indicating that this taskwas perfortned nonlexically. In contrast, a significant lexical-ity effect was obtained in the lexical decision task. In addition,a reaction time semantic priming effect was obtained in the lex-ical decision task but not in the physical task.

After having established that the physical task was performednonlexically, we can address the question whether an N400 prim-ing effect occurs when the task performance does not involvelexical processing. An N400 priming effect clearly developed in

the lexical decision task and was absent in the physical task (Fig-ures 1 and 2). In the lexical decision task, the N400 was modu-lated by the relatedness of prime and target, yielding much largeramplitudes for unrelated than for related word pairs. In con-tra.st, no indication for an N400 priming effect was obtained inthe physical task. Once it is made sure that for task-related per-formance the target words are processed nonlexically, an N400priming effect is no longer observed. This result suggests thatthe modulation of the N4(X) amplitude requires the explicit iden-tification of a word, including its meaning. If word meaningdoes not become part of the episodic trace of a particular word-like stimulation event, N400 pritning effects do not seem to arise.This result also lends credibility to the suggestion that previousstudies in which N400 priming effects were obtained in shallowprocessing tasks (Besson et al., 1992; Connolly, Stewart, & Phil-ips, 1990; Kutas & Hillyard, 1989) might have been unsuccess-ful in preventing explicit identification of the words and theirmeaning.

N400 Effect and Lexicat ProcessesWhat are the implications of these results with respect to the pro-cessing nature of the N4(X) effect? The absence of an N400 prim-ing effect in the physical task challenges the view that the N400effect reflects the automatic process of lexical access and its con-comitant ASA. This conclusion is further supported by the ab-sence of an N400 priming effect when the prime is masked,thereby preventing conscious identification of the stimulus(Brown & Hagoort, 1993). The presence of N400 priming effectsin the lexical decision task provide.s further evidence that theN400 effect is closely tied to lexical integration processes, medi-ated by the mechanistn of setnantic matching. Brown and Hagoort(1993) argued that lexical processing in the context of word prim-ing shares core characteristics with the process of lexical inte-gration in the ordinary process of sentence comprehension. Thatis, once lexical information has been accessed automatically on

282 D.J. Chwilla, CM. Brown, and P. Hagoorl

the basis of the speech signal or the orthographic input, the syn-tactic and semantic information associated with the activatedword form has to be integrated with the preceding context intoan overall interpretation of the whole utterance. The N4(X) effectis claimed to reflect this process of lexical integration(see Holcomb, 1993, and Rugg et al., 1988, lor similar views).

Processing in ihe Physical Ta,skSo far, the results for reaction time and N400 are compatiblewith two conclusions: (a) the absence in the physical task of alexicality effect and a priming effect is due to the fact that non-lexical task performance prevented word meaning from becom-ing part of the episodic trace, and (b) even accessing wordmeaning in semantic memory was prevented under the shallowtask conditions. This latter possibility is, however, excluded bythe P3(X) results.

In the physical task, we observed a modulation of the pari-etally distributed positivity as a function of the probability ofrelated and unrelated word pairs, with larger amplitudes for thestimulus category that occurred with the lower probability (seeFigures 2 and 3). Both the distribution of this positivity and itssensitivity to the probability of a certain stimulus event suggeststhat this positivity reflects the P300 component (see Johnson,1988, for a review). The important point here is that the P3(X)results demonstrate that the subjects a.ssigned related and unre-lated word pairs to different categories. This assignment presup-poses that subjects accessed word meanings when processing theprimes and targets. As such, it implies that even in the absenceof reaction time and N400 effects in the physical task, access-ing word meaning in semantic memory was not prevented. Themore reduced positivity to nonwords as compared with unrelatedand related target words Ttts well with the above picture, becausethe nonwords were presented with a higher probability (.50) thanwere both unrelated word pairs (overall probabilities of .08 inthe high-proportion condition and .33 in the low proportion con-dition) and related word pairs (.33 and .08 in high- and low-proportion list, respectively). Although on the basis of theseresults we cannot fully specify all the dimensions that might haveentered into the categorization of the prime-target pairs by thesubjects, it seems clear that lexical and semantic characteristicswere both involved.*

This study was not designed to investigate the relationshipbetween the P300 and lexical processes. Nevertheless, we ob-served an intriguing dissociation between the reaction time andN400 results on the one hand and the P300 results on the otherhand. What might these P300 effects reflect? A tentative expla-nation is that the presence or absence (in the case of nonwords)of activation of word meanings in semantic memory and theautomatic spread of activation between related nodes in seman-tic memory might have led to an implicitly generated categori-zation of stimulus events that became part of the episodic

*In the P300 literature, the probability effect in often related to thetask demands. In the physical task, the probability of the different cat-egories (related, unrelated, nonwords) was not task relevant. However,the P3(X) probability effect has also been linked to automatic process-ing. Several results indicate that the relationship between P3(X) and prob-ability meet.* criteria for automatic encoding (see Campbell, Courchesne,Picton, & Squires, 1979; Duncan-.Iohnson & Donehin, 1977). Moreover,P3(X) effects have been reported in neuropsychological patients that didnot show behavioral sensitivity to the task-relevant dimension (Renault,Signoret, Debruille, Breton, & Bolgert, 1989).

representation in tertns of more likely and less likely stitnulusevents. That is, the semantic aspects of the words did not thetn-selves become part of the episodic trace (as suggested by theabsence of N4()0 effects) but led to an explicitly recognized butunspecified sense of more and less frequent stitnulation events,resulting in the observed P300 effects.

CaveatsThe above claims can only be made if it can be shown that theobserved differences in ERP pattern between tasks could not beascribed to other factors.

A first, minor consideration is Ihat task difficulty was notcontrolled across tasks. This resulted in differences in overallreaction times between the lexical and the physical task. It couldbe argued that the observed differences in ERP pattern betweentasks are due to differences in reaction titne. However, twoaspects of the present data argue againsi this possibility. First,in the physical task, reaction times were the same for related andunrelated targets and for words and nonwords, whereas therewere clear differences in P300. This dissociation suggests thatdifferences in tiRPs cannot be attributed in a simple way lo dif-ferences in reaction time. Second, the fact that different ERPpatterns were obtained across tasks, especially the absence ofan N4(K) priming effeci in the physical task, supports the ideathat Ihe difference in task performance i.s not of a quantitalivebut of a qualitative nature.

A second issue concerns component overlap. Because boththe N40() and the P30() component occur within roughly thesame latency range, it cannot be excluded that both N4(X)s andP3OOs were present lo different degrees across the conditions.The crucial question, however, is whether the main effects ofthe present study can be explained in terms of component over-lap. We first discuss the possible itnplication of cotnponent over-lap for the N400 effects obtained in the lexical decision task andthen consider the implications for the P.3(X) effecis in the phys-ical task.

Could the N400 priming effect ob.served in the lexical deci-sion task be explained in terms of an overlapping P.300 cotnpo-nent? This idea is rejected for the following reasons. In Ihepresent study the probability of related and unrelated targets wasreversed across conditions (.80/.20 and .20/.80). Because of theinverse relationship between stimulus probability and P300 atnp-litude, we expect different N400 patterns acro.ss conditions if the?m) is overlapping. Because of the larger P300 for Ihc low-probability event, the amplilttde of the N400 should be reducedto unrelated targets in the high-proportion condition, becauseunrelated word pairs are only presented in 2O"/o of the cases. Ifthis were the case, the N400 for unrelated targets in the high-proportion condition would have been smaller than the N400for unrelated targets in the low-proportion condition. However,the opposite pattern was observed. The N400 for unrelated word-pairs even tended to be larger in the high- than in the low-pro-portion condition.

TTie second question is whether Ihe absence of an N400 prim-ing effect in the physical task may be due to overlapping com-ponents. Here, a distinction should be made between twodifferent types of overlap. The first concerns overlap of theN40() component with the P300 component, whereas the secondconcerns an overlap of N400 priming effects with P300 proba-bility effects. From the present data, it cannot be excluded thatan N400 component was evoked that overlapped the P300 in thephysical task. This possibility is suggested by the difference in

Levels of processing and ERPs 283

scalp distribution of the P300 probability effect across propor-tion conditiotis (see Figure 3). However, the distributional differ-ences might at least in part be a result of some baseline problemsin the high-proportion condition, with the larger baseline dif-ferences between the related and unrelated conditions at thefrontal leads. These problems might have contributed to the dis-tributional differences that were obtained. However, the distri-butional differences were not statistically significant. In thephysical task, neither the Proportion x Electrode nor the PritneType X Proportion x Electrode interaction approached signif-icance. We are, therefore, very reluctant to assign strong inter-pretational relevance to this apparent distributional difference.

Regarding the overlap of N400 and P30() effects, the crucialquestion is whether the absence of an N400 pritning effect in thephysical task could be explained by an overlap of the N400 pritn-ing effect with the P300 probability effect. The presence of anN4(K) pritnitig effect in these data is very unlikely. Based on thecurrent literature on N4()0 pritning effects, an underlying N400priming effect could result from the following two scenarios.

Scenario 1 is that both in the high-proportion and in the low-proportion conditions a (small) N400 priming effect had beenelicited in the physical task by the .semantically related or unre-lated itetns. In Scenario I, we assume that this effect is of simi-lar magnitude in both proportion conditions. This scenario leadsto the prediction that the positive shift in the high-proportioncondition is smaller than the positive shift in the low-proportioncondition. The rea.son for this prediction is that in the formercase the positive shift (as a function of probability) is for the.same itetns as is the negative shift (as a function of relatedness).In the low-proportion condition, the positive shift is for theinfrequent, related words and the larger negativity is for the fre-quent, untelated targets. The resulting tiet effect should there-fore be stnaller in the case where negative and positive effectsare for the satne items (high proportion), as compared with thesituation where the two effects arc for different itetns (low pro-portioti), thereby even iticrcasing the net effect. However, nosuch difference in the size of the effects was observed. If thereis a difference at all, it is in the opposite direction (overall pos-itive shift of 2.3 nV in the high-proportion condition and of1.3 nV in the low-proportion condition).

Scenario 2 is that just as for the lexical decision task, Ihe hid-den N4(X) pritning effect in the high-proportion condition islarger than in the low-proportion condition. In that case, thecancellation effect tor the high-proportion condition would havebeeti even greater.

If there had been an overlap between N400 pritning effectsand P3(X) probability effects in the physical task, we expectedit to have the predicted consequences oti the basis of our knowl-edge about factors responsible for N400 priming effects. The.se

consequences were not observed. Therefore, acknowledging theprincipled impossibility of excluding component overlap withsurface recordings, the likelihood that overlap is responsible forthe pattern of results observed is sufficiently low to warrant ourconclusions.

A possible critique of the interpretation of the P300 effectsin the physical task is that the absence of an N400 priming effectmight arise from a baseline problem. This problem indeed mightplay a role in the high-proportion condition, where the wave-fortns for related and unrelated pairs separate very early in time(Figure 2). This difference in the baseline at least enhances theP300 probability effect. The reasoning here is that if the wave-fortns were forced to align, then an N400 pritning effect mightemerge and the P3(X) effect might disappear. It is indeed diffi-cult to interpret the early onset of the effeet. and the possibil-ity cannot be discarded that the P300 probability effect in thiscondition might result from noise or baselining problems.

To check whether baseline problems might underly theobserved effects, additional analyses were performed, in whicha later baseline of 30-130 ms posttargel was used, thereby align-ing the waveforms to counter their seemingly early separation.The.se ANOVAs revealed a significant interaction of Propor-tion X Relatedness for the P300 (/^11,16] = 4.81. p < .05)(latency epoch: 400-500 ms) but not for the N400 (latency epoch:300-400 tns). Moreover, with the use of this baseline no maineffect of relatedness was obtained for the early and the latelatency epoch. The additional ANOVAs. therefore, stronglysuggest that the P300 probability effect was not evoked by dif-ferences in ba.seline but is due to modulation of the P3(X)component.

Conclusions

Under shallow processing conditions, the amplitude of the P300is sensitive to category probability, which in the present studywas related to word and nonword stimuli. The dissociation be-tween tio reaction titne differences for words and nonwords inthe physical task, with P300 differences for these same stimuli,provides evidence that despite the shallow task requirements lex-ical information can still impact the ongoing analysis process.Under these conditions, however, the N4(X) is not sensitive tolexical information.

The amplitude of the N4(X) is modulated when semanticaspects of word stimuli enter into the episodic trace of wordlikestimulus events. However, automatic spreading of activationbetween related nodes in semantic metnory does not appear toaffect the N400. These results support the idea that within thecontext of normal language processing the N400 effect primar-ily reflects lexical integration processes.

REFERENCES

Beeker. C. A. (1985). What do we really know about semamie contexteffects during readitig? In D. Besner, T O. Waller. & E. M. Mac-Kinnon (Eds.). Reailinn research: Advances in theory and practice(Vol. 5. pp. 125-169). Toronto: Academic.

Bentin. S.. Kutas, M.. & Hillyard. S. A. (1993). Electrophysiologicalevidence for task effect.̂ on setiiantie priming in auditory word pro-cessing. I'sychophy.siology. JO, 161-169.

Bentin. S.. McCarthy. C... & Wood. C. C. (1985). Event-related poten-tials associated with semantic priming. Ktectrocncephalography andClinical Neurophysiology, 60, 343-355.

Besson. M.. Fisehler. 1.. Boaz. T . & Raney. G. (1992). Effects of auto-tnatic associative activation on explicit and implicit memory tests.Journal of Experimental Psycholofiy: Learning. .Memory, and Cog-nition. IS. 89-105.

Brown. C. M..& Hagoort. P. (1993). The processing nature of the N400:Evidence frotti masked priming. Journal of Cognitiw Neuroscience,5. ?4-44.

Brown. C. M.. Hagoort. P.. & Chwilla. D. .1. (1994). .4n event-relatedbrain potential analysis of visual word priming effects. Manuscriptsubmitted for publication.

284 D.J. Chwitta, CM. Brown, and P. Hagoort

Campbell, K. B., Courchesne, E.. Picton, T. W., & Squires, K. C. (1979).Evoked potential correlates of human information processing. Bio-logical Psychology. 8. 45-68.

Collins, A. M., & Loftus. E. F. (1975). A sprcading-activation theoryof semantic processing. Psvchological Review. 82, 407-428.

Connolly. J. F., Stewart, S. H.. & Phillips, N. A. (1990). The effectsof processing requirements on neurophysiological responses to spo-ken sentences. Brain and Language. 39. .102-318.

Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A frame-work for memory research. Journal of Verbal Learning and VerbalBehavior, II, 671-684.

Deacon, D., Breton. F.. Ritter. W , & Vaughan, H. G., Jr. (1991). Therelation.ship between the N2 and N4fl(): Scalp distribution, stimulusprobability, and task relevance. Psychophysiolony. 2H, 185-200.

De Groot, A. M. B. (1980). Mondelinge woordassociatienormen. Lisse,The Netherlands: Swets & Zcitlinger.

De Groot. A. M. B. (1984). Primed lexical decision: Combined effectsof the proportion of related prime-target pairs and the stimulus-onsetasynchrony of prime and target. The Quarterly Journal of Experi-mental Psychology. 36A. 253-280.

De Groot. A. M. B. (1987). The priming of word associations: A levels-of-processing approach. The Quarterly Journal of Experimental Psy-chology, 39A, 721-756.

t)e GrcK)t. A. M. B., & de Bil. .1. M. (1987). Nederlandse woordassocia-tienormen met reacttetijden. Lisse, The Netherlands: Swets &Zeitlinger.

Duncan-Johnson, C. C , & Donchin, E. (1977). On quantifying surprise:The variation in event-related potentials with subjective probabil-ity. Psychophystology. 14. 456-467.

Fodor, J. A. (1983). The modularity of mind. Cambridge, MA: MITPress.

Forster. K. I.. & Bednall, E. S. (1976). Terminating and exhaustivesearch in lexical access. Memory and Cognition. 4. 53-61.

Frauenfelder. U. H.,& Tyler. L. K. (1987). The process of spoken wordrecognition: An introduction. Cognition. 25. 1-20.

Greenhouse, S. W., & Geisser, S. (1959). On methods in the analysis ofprofile data. Psychometrika. 24, 95-112.

Henderson. L. (1982). Orthography and word recognition in reading.London: Academic.

Holcomb. P. J. (1988). Automatic and atlentional processing: An event-related brain potential analysis of semantic priming. Brain and Lan-guage. 35. 66-85.

Holcomb. F. J (1993). Semantic priming and stimulus degradation:Implications for the role of the N4(X) in language proces.sing. Pty-chophysiology. 30. 47-61.

Holcomb, P. } . . & Neville, H. J. (1990). Auditory and visual semanticpriming in lexical decision: A comparison using evenl-related brainp o t e n t i a l s . Language and Cognitive Processes. 5, 2 8 1 - 3 1 2 .

Jacoby, L. L., & Dallas, M. (1981). On the relationship between auto-biographical memory and perceptual learning. Journal of Experi-mental Psychology: General. 110. 306-340.

Johnson, R., Jr. (1988). The amplitude of the P30() component of theevent-related potential: Review and synthesis. In P. K. Ackles, J. R.Jennings, & M. G. H. Coles (Eds) , Advances in psychophysiotogy(Vol. 3, pp. 69-137). Greenwich, CT: JAI.

Keefe, D. E.,& Neely, J. H. (1990). Semantic priming in the pronunci-ation lask: The role of prospective prime-generated expectancies.Memory and Cognition. 18, 289-298.

Kutas, M., & Hillyard. S. A. (1980). Reading senseless sentences: Brainpotentials reflect semantic incongruity. Science, 207. 203-205.

Kutas, M., & Hillyard, S. A. (1984). Brain potentials during readingreflect word expectancy and semantic association. Nature. 307.161-163.

Kutas, M., & Hillyard. S. A. (1989). An electrophysiological probe ofincidental semantic association. Journal of Cognitive Neuroscience.I. 38-49.

Kutas, M., Lindamood, T., & Hillyard, S. A. (1984). Word expectancyand event-related potentials during sentence processing. In S. Korn-blum& J. Requin (Eds.), Preparatory states and processes (p'p. 217-238). HiUsdale, NJ: Erlbaum.

Kutas, M., & Van Petten, C. (1988). Event-related brain potential stud-ies of language. In P. K. Ackles, J. R. Jennings. & M. G. H. Coles(Eds.), Advances in psychophysiotogy (Vol. 3, pp. 139-187). Green-wich, CT: JAI.

Lauteslager, M., Schaap, T., & Schievels, D. (1986). Schr{ftelijke woor-das.s<Kiatienormen voor 549 nederlandse zetfstandige naamwoorden.t.isse. The Netherlands: Swets & Zeitlinger.

Marslen-Wilson, W. D. (1989). Access and integration: Projecting soundonto meaning. In W. D. Marslen-Wilsen (Ed.), Lexicat repre.senta-tion and process. Catnbridge, MA: MIT Press.

McCarthy, G., & Wood, C. C. (1985). Scalp distribution of event-relatedpotentials: An ambiguity associated with atialysis of variance models.Electroencephalography and Clinical Neurophysiology. 69. 218-233.

Meyer, D. M., & Schvaneveldt, R. W. (1971). Facilitation in recognis-ing pairs of words: tividetice of a dependence between retrieval oper-ations. Journal of Experimental Psychology, 90. 227-234.

Mitchell, P. I . , Andrews, S., Fox, A. M., Catts, S. V., Ward, P. B., &McConaghy, N. (1991). Active and passive attention in schizophre-nia: An ERP study of information processing in a linguistic task.Biological Psychology. 32, 101-124.

Neely, J. H. (1991). Setnantic priming effects in visual word recogni-tion: A selective review of current findings and theories. In D. Bes-ner & G. Humphreys (Eds.), Basicproce.<!.ses in reading: Vi.suat wordrecognition (pp. 264-336). Hillsdale, NJ: Erlbaum.

Neely, J. H., & Keefe, D. E. (1989). Semantic context effects on visualword processing: A hybrid prospective-retrospective processingtheory. In G. H. Bower (Ed.), The psychology of learning andmotivation: Advances in research and theory (Vol. 24, pp. 207-248).New York: Academic.

Oldfield, R. C. (1971). The a.ssessment and analysis of handedness: TheEdinburgh Inventory. Neuropsychotogia. 9. 97-113.

Posner, M. I., & Snyder, C. R. R. (1975). Attention and cognitive con-trol. In R. L. Solso(Ed.), Information proces.sing and cognition: TheLoyola symposium (pp. 55 85). Hillsdale, NJ: Erlbaum.

Renault, B., Signoret, J., t)ebruille, B., Breton, F., & Bolgert, F. (1989).Brain potentials reveal covert facial recognition in prosopagnosia.Neuropsychologia. 29. 905-912.

Rosier, 1., Hcil, M., & Glowalla, U. (1993). Moiiiloring retrieval fromlong-term memory by slow event-related brain potentials. Psvcho-physioloRV. .W. 170-182.

Rugg, M. D., Furda, J., & Lorist, M. (1988). The effects of task on mod-ulation of event-related potentials in word repetition. Psychophys-lology, 25. 55-63.

Schulman, H. G., & tJavison, T. C. B. (1977). Control properties ofsemantic coding in a lexical decision task. Journal of Verbal Learn-ing and Verbal Behavior. 16. 91-98.

Smi th , M. C T h e o d o r , t . , & Frankl in , P. I-. (1983). The re la t ionshipbetween contextual facililalion and depth of processing. Journal ofExperimental Psychology: Learning. Memory, and Cognition. 9.697-712.

Uit den Boogaart, P. C. (1975). Woordfrequenties in geschreven engesproken Nederlands. Oosthoek, The Netherlands: Scheltema &Holkema.

Winer, B. J. (1971). Statistical principles in experimental design. NewYork: McGraw-Hill.

Zwitserlood, P. (1989). The locus of the effects of sentential-setnanticcontext in spoken-word processing. Cognition, 32. 25-64.

(RECErvED March 16, 1993; ACCEPTED June 24, 1994)

Levels of processing and ERPs

Related word pairs ms

1. ader-bloed2. bakker-brood3. been-arm4. donker-licht5. gebouw-flat6. dol-gek7, storm-wind8. links-rechts9. matroos-schip

10. chef-baas11. bal-rond12. dag-nacht13. kust-zee14. laken-bed15. lang-kort16. haai-vis17. faam-roem18. dier-beest19. grocn-gras20. mi.sdaad-nioord21. dorp-stad22. gooien-werpen23. drempel-deur24. levend-dood25. ster-hemel26. berouw-spijt27, antwoord-vraag28. kil-koud29. aardig-lief30. hond-kat31. bes-vrucht32. leeg-vol33. man-vrouw34. koorts-ziek35. raak-mis36. bij-honing37. beven-trillen38. prei-groente39. papier-pen40. venster-raam

Reaction timeMSD

Language frequencyMSD

471457539474529510496488520566498510463472500523509473482502515578493467530506524461530492514503491507532529544490

519516

505.627,6

82.1»108.S

APPENDIX

Unrelated word pairs

zweet-tekstkachel-sportstof-veiligsjaal-grondhonger-struikwraak-liptuig-wandvret en-postdoel-buikuitval-slangaccu-jeugdakker-grijsstrak-puntafwas-fietskwart-gezinpakket-dakpalcis-aapkeer-kindpuinhoop-wolfmonnik-trapnatie-lolkoek-traagperron-spuitelpee-neuspaniek-blouseliefde-bolkapper-meerui-slotwoord-zakgelei-taalschedel-taxiboer-kreetrot-vrijstel-persrots-passenput-spotnevel-lekgeluid-jurkvak-steekkilo-minuut

ms

471457543477528515495490515563499514459476500525508472482506519579490469539508526457530489

509506489512532526541487

515519

505.928.0

61.S10S.9