Semantic Processing of Spoken Words in Alzheimer's Disease: An

Semantic Processing of Spoken Words in Alzheimer’s Disease: An Electrophysiological Study

Antti Revonsuo, Raija Portin, Kirsi Juottonen, and Juha O. Rinne University of Turku, Finland Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021

Abstract ■ Patients suffering from Alzheimer’s disease (AD) have severe studies is whether word-elicited ERPs other than N400 remain difªculties in tasks requiring the use of semantic knowledge. normal in AD. In the present study our aim was to ªnd out The semantic deªcits associated with AD have been extensively whether the ERP waveforms N1, P2, N400, and Late Positive studied by using behavioral methods. Many of these studies Component (LPC) to semantically congruous and incongruous indicate that AD patients have a general deªcit in voluntary spoken words are abnormal in AD and whether such abnor- access to semantic representations but that the structure of the malities speciªcally reºect deªciencies in semantic activation representations themselves might be preserved. However, sev- in AD. Auditory ERPs from 20 scalp sites to semantically con- eral studies also provide evidence that to some extent semantic gruous and incongruous ªnal words in spoken sentences were representations in AD may in fact be degraded. Recently, a few recorded from 17 healthy elderly adults and 9 AD patients. The studies have utilized event-related brain potentials (ERPs) that early ERP waveforms N1 and P2 were relatively normal for the are sensitive to semantic factors in order to investigate the AD patients, but the N400 and LPC effects (amplitude differ- electrophysiological correlates of the semantic impairment in ence between congruous and incongruous conditions) were AD. Interest has focused on the N400 component, which is signiªcantly reduced. We interpret the present results as show- known to reºect the on-line semantic processing of linguistic ing that semantic-conceptual activation and other high-level and pictorial stimuli. The results from studies of N400 changes integration processes are defective in AD. However, a word in AD remain somewhat controversial: Some studies report congruity effect earlier than N400 (phonological mismatch normal or enlarged N400 components in AD, whereas others negativity), reºecting lexical selection processes, is at least to report diminished ones. One issue not reported in previous some extent preserved in AD. ■

a variety of behavioral tasks such as naming, generating INTRODUCTION deªnitions, real/unreal object decision, and forced- One of the principal cognitive deªcits associated with choice questions about the appearance and functional AD is the impaired performance of AD patients in a attributes of objects and animals. AD patients were se- variety of semantic tasks. The patients have difªculties in verely impaired on both verbal and visual tasks and word ªnding (poverty of content words in spontaneous disproportionately impaired on items concerned with speech, decreased ability to generate instances of a given animate objects. The authors interpret their ªndings as category, and difªculties in object naming and naming to indicative of disruption at the structural level of visual deªnition), deªcient knowledge of concept meaning representation where the visual features and spatial re- (loss or disrupted organization of attribute knowledge), lations of objects are represented and at the phonologi- and anomalies in the effects of semantic context (e.g., cal output level where object names are stored for abnormal semantic priming in lexical decision)(Chert- speech production. However, because the patients per- kow & Bub, 1990; Chertkow, Bub, & Seidenberg, 1989; formed relatively well in some tasks not requiring effort- Chertkow et al., 1994; for a review, see Nebes, 1989). It ful processing, the authors conclude that semantic is unclear whether these impairments are due to a genu- networks are partly preserved in AD and that the pa- ine disruption of semantic memory or whether they tients can access semantic knowledge to some degree in merely indicate a difªculty in effortful, voluntary access an indirect way. Laatu, Portin, Revonsuo, Tuisku, and to semantic representations (Bayles, Tomoeda, Kaszniak, Rinne (1997) investigated whether AD patients are able & Trosset, 1991; Chan et al., 1993; Hodges, Salmon, & to get voluntary access to semantic representations if the Butters, 1992; Nebes, 1994). Recent studies show that AD retrieval demands of a task are gradually eased without patients probably have a combined access and storage essentially changing the nature of the information to be deªcit. Daum, Riesch, Sartori, and Birbaumer (1996) used retrieved. In some tasks the deªcits persisted even in

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 passive recognition and forced-choice tasks. The authors’ congruous or incongruous with the phrase. Visual in- conclusion was that AD patients have a generalized ac- spection of the ERPs to target words showed that the cess deªcit but some indications of a storage deªcit as N400 effect (amplitude difference between ERPs to con- well. gruous and incongruous words) was smaller for the AD These observations concerning the existence of se- patients than for age-matched controls. The N400 was mantic access and storage deªcits in AD are exclusively delayed and smaller for the elderly controls compared based on behavioral measures of semantic function, with the young. However, no statistical analyses of the which do not allow observation of cognitive processes effects were presented. In the study of Schwartz, Kutas, while they occur, but only as they are inferred from Butters, Paulsen, and Salmon (1996), the experimenter

overt behavior. However, ERPs provide us with a fresh uttered the name of a category, which was followed after Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 methodological approach for the measurement of the about 1 sec by a visually presented target word that was on-line semantic processing of linguistic stimuli in brain- the written name of an object. Reaction times and ERPs injured patients. Especially the N400 component of the were recorded. AD patients showed large behavioral ERPs is modulated by semantic factors (Kutas & Hillyard, priming effects (faster responses to related than to un- 1980; for a review, see Hagoort & Kutas, 1994). The N400 related target words). The amplitude of the N400 effect is a negative-going wave between 300 and 600 msec (difference between ERPs to related and unrelated target poststimulus, and its scalp distribution is broad. N400 words) was, however, much smaller in the AD group and amplitude to a stimulus word is modulated by the se- peaked later than in the young and elderly control mantic expectancy (“cloze probability”) and congruity of groups. All groups showed similar patterns of ERP re- the word with the preceding semantic context. The sponses to prime words at different levels of category words most expected elicit a minimal N400, whereas the (superordinate, basic, or subordinate). According to the ones semantically incongruous and least expected elicit authors, this result indicates that the structure of seman- the largest N400 amplitudes. A standard interpretation of tic representations in AD patients is preserved at least to the N400 effect has been that it is sensitive to attention- some extent. demanding semantic integration processes, not automat- The N400 is elicited by semantically incongruous and ic aspects of lexical access (Brown & Hagoort, 1993). unexpected ªnal words in written (Kutas & Hillyard, Recent studies show that N400-like waveforms are elic- 1980) and spoken (McCallum, Farmer, & Pocock, 1984) ited from pictorial as well as linguistic stimuli (Ganis, sentences. In a recent study involving AD patients (Ham- Kutas, & Sereno, 1996; Holcomb & McPherson, 1994; berger, Friedman, Ritter, & Rosen, 1995), the ªnal words Nigam, Hoffman, & Simons, 1992; Pratarelli, 1994). These of the sentence belonged to one of four stimulus types, ªndings suggest that the N400 might largely reºect some which varied as a function of semantic relatedness to a kind of semantic search in a common modality-inde- highly expected word. The sentences and ªnal words pendent conceptual-semantic network. The issue be- were presented visually to six AD patients and to old and tween these two interpretations of N400 remains young control subjects. The N400 amplitude varied as a unresolved (Osterhout & Holcomb, 1995); assigning cog- function of semantic relatedness in young controls and nitive functions to ERPs is, in general, theoretically prob- AD patients. The elderly control subjects showed no lematic (Rugg & Coles, 1995). such orderly variation but instead showed an equally ERPs can be used to investigate the semantic process- large N400 to all kinds of unexpected ªnal words. In a ing of words in patients with linguistic or semantic simultaneous behavioral task the AD patients had difªcul- deªcits (Hagoort, Brown, & Swaab, 1996; Revonsuo & ties in distinguishing between two of the four stimulus Laine, 1996). ERPs can provide us with information about types: the best completion and its semantic associations. language processing in real time and with great temporal Thus the patients’ ERPs actually showed better discrimi- resolution, and this information can then be used to- nation of different stimulus types than their behavioral gether with results from behavioral studies to construct, responses did. The authors interpret this as suggestive of test, and constrain models of normal and deªcient lan- a semantic access deªcit that occurs in a time window guage processing (Hagoort & Kutas, 1994). Furthermore, after the N400 but before the behavioral response. ERPs can provide us with indications about the neural Ford et al. (1996) recorded ERPs to ªnal words in mechanisms of language processing, and they can be spoken sentences. Surprisingly, the N400 amplitude was successfully measured even in the absence of any addi- larger in AD patients than in elderly control subjects, tional task beyond natural listening or reading. Hagoort both for unprimed and primed words. The N400 effect and Kutas were the ªrst to present results concerning (difference between unprimed and primed conditions) N400 in AD patients. They refer to an unpublished pre- was signiªcant for both groups, but it appears to be liminary study in which AD patients, age-matched con- smaller for the AD group (Ford et al., Figure 1). The trols, and young controls were tested in an N400 authors suggest that the sentence stem did not ade- paradigm. The subjects were given a short spoken phrase quately prime the ªnal word of the sentence in AD (e.g., “A type of animal”) after which the target word was patients and thus reºects a probable impairment in se- presented visually. The target was either semantically mantic knowledge. Their interpretation, however, is in

Revonsuo et al. 409

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 contradiction with behavioral results (Nebes, Boller, & ably part of a generalized anomaly affecting ERPs across Holland, 1986; Nebes & Brady, 1991) that show that a the board. We chose to use spoken words in sentence highly constrained sentence stem produces normal context as stimuli because this is a very natural way of priming effects in AD patients when the task is to decide presenting linguistic stimuli. No concurrent overt re- whether the target word is an appropriate ending for the sponses were required from the subjects while the ERPs sentence. Ford et al. admit that the paradoxical result of were recorded in order to not contaminate the N400 by larger N400 amplitudes in AD patients than controls may an overlapping P300. However, the subjects’ recognition be due to the baseline against which the N400 was of the semantic incongruity of the stimulus sentences measured: It was signiªcantly more negative for the con- was tested separately. In interpreting our results we took

trols than the patients. into account the recent ªnding that semantically incon- Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 In sum, we now have reports with inconclusive and gruous terminal words in spoken sentences elicit an even somewhat conºicting results on the N400 in AD. early negativity called phonological mismatch negativity Hagoort and Kutas (1994) and Schwartz et al. (1996) (PMN), if the semantically anomalous word has a totally report reduced N400 amplitudes for AD patients, unexpected initial phoneme (Connolly & Phillips, 1994). whereas Ford et al. (1996) and Hamberger et al. (1995) PMN and N400 occur partially in the same time window, report normal or greater than normal N400 amplitudes but they probably reºect different stages in word recog- for AD patients. All studies report that despite the abnor- nition, and it is possible that they are preserved or mal N400 amplitudes, the amplitude still is sensitive to impaired independently of each other. In order to get a semantic modulation in AD patients. detailed picture of such dissociations, we analyzed data On the basis of these studies it remains somewhat in 100-msec time windows at 200 to 1500 msec from unclear what actually happens to N400 in AD. A part of stimulus onset. We expected to see relatively normal the differing results can probably be attributed to differ- ERPs to spoken words in AD patients with regard to ences between patient groups, tasks, stimulus presenta- components not typically modulated by semantic factors tion, and modality in the different studies. There are also (N1, P2) and abnormal ERPs in components typically some methodological difªculties with some of the stud- sensitive to semantic modulation (N400, LPC). ies, because ERP components other than the N400 to word stimuli are not usually analyzed. Therefore it remains unclear whether or not some of the earlier ERP RESULTS components such as N1 or P2 were abnormal for some Behavioral Tasks of the AD groups studied. If that is the case, it will be difªcult to draw any deªnite conclusions that speciªcally The AD patients were severely impaired in the WAIS concern the N400. Furthermore, in most of the pre- Similarities task. Mean raw score for the controls was 19 viously reported studies the subjects were required to points (SD = 3.2, range 13 to 24), and for AD patients it make an overt response to the target word. This is was 8 points (SD 4.8, range 2 to 14, t = 6.96, df = 23, known to elicit an overlapping P300 component. It in- p < 0.0001). AD patients’ explicit understanding of se- teracts with the N400 because both components occur mantic incongruity was also impaired compared with in partially overlapping time windows. A consistent the controls. Mean error score for the controls was 0.4 ªnding concerning ERPs in AD is that P300 latency is (SD 0.8, range 0 to 3), and for the AD group it was 7.2 − signiªcantly delayed compared to age-matched controls (SD 6.4, range 0 to 20, t = 3.18, df = 8.1, p < 0.05). (e.g., Polich, Ladish, & Bloom, 1990). Consequently, if However, variability was maximally large in the patient P300 and N400 are elicited by the same task, they inter- group: One patient made no errors at all and was thus act differently in AD patients and age-matched controls, perfectly able to recognize which of the 40 test sen- possibly making the amplitude and/or latency of the tences were incongruous and which were not; another N400 effect appear different for the AD patients than for patient scored at chance level in this task because he the controls, depending on the exact pattern of the obviously was unable to keep the beginning of the sen- differential overlap with P300. tence in mind until the ªnal word was presented. Five The aim of the present study is to analyze in detail of nine AD patients made more errors in this task than ERP deºections N1, P2, N400, and LPC to semantically any of the control subjects did, and the other four pa- congruous and incongruous spoken words. We aim to tients were within the range of the control subjects’ ªnd out whether the semantic deªcits of AD patients are error rate. reºected as selective anomalies of the semantic modulation of speciªc ERP components while other features of Statistical Analyses ERPs to spoken words remain relatively normal. If, however, such ERP components that presumably reºect com- Four different ERP components were analyzed and com- pletely nonsemantic functions (N1, P2) prove to be pared: N1 (negative peak amplitude and latency 50 to clearly abnormal as well, anomalies in semantically 150 msec), P2 (positive peak amplitude and latency 150 modulated ERPs represent no special case but are prob- to 300 msec), N400 (mean amplitude 300 to 800 msec),

410 Journal of Cognitive Neuroscience Volume 10, Number 3

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 and LPC (mean amplitude 800 to 1500 msec). All of these Type effects (Congruous vs. Incongruous). Finally, corre- are readily visible in the averaged ERP waveforms of the lations between N400 peak amplitude and behavioral control group (Figure 1) and at least N1 and P2 are tasks were calculated. obvious also in the AD patients’ ERPs (Figure 2). Re- peated-measures analyses of variances (ANOVAs) were ERPs to Congruous Words used to test the main effects of Group (2 levels) and Electrode (20 levels) and their interaction on the ampli- N1 tudes and latencies of the ERP components. The Huynh- N1 peaked around 100 msec after stimulus onset. In N1 Feldt correction was applied to all repeated measures amplitude, there was a signiªcant main effect of Elec- with greater than 1 degree of freedom. Additional mean trode, F(19, 475) = 4.58, p < 0.01. Maximum amplitudes Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 amplitude analyses in 100-msec time windows from 200 in both groups were observed around the vertex. There to 1500 msec were carried out in order to get a detailed was no signiªcant Group effect, although the N1 peak picture of the time course and group differences in Word amplitude tended to be stronger for the control group.

Figure 1A. ERPs to Congruous (broken line) and Incongruous (con- Figure 2A. ERPs to Congruous (broken line) and Incongruous (continuous line) ªnal words of a sentence in the control group, 20 tinuous line) ªnal words of a sentence in the AD group, 20 scalp scalp electrodes. electrodes.

Figure 2B. ERP waveforms N1, P2, N400, and LPC as measured at Figure 1B. ERP waveforms N1, P2, N400, and LPC as measured at vertex (electrode Cz) in the AD group. ERPs to Congruous (broken vertex (electrode Cz) in the control group. ERPs to Congruous line) and Incongruous (continuous line) ªnal words of a sentence. (broken line) and Incongruous (continuous line) ªnal words of a sentence.

Revonsuo et al. 411

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 In N1 latency, there were signiªcant main effects of Effects of Semantic Incongruity on ERPs Electrode, F(19, 475) = 59286, p < 0.001, and Group F, Three different ERP components were analyzed and (1, 25) = 5.64, p < 0.05, and a signiªcant Electrode × compared: P2, N400, and LPC. These were selected be- Group interaction, F(19, 475) = 179.95, p < 0.001. The cause there is evidence that semantically incongruous N1 latency for the AD patients was signiªcantly shorter spoken words modulate ERPs in these time windows in than for the Controls (70 msec for AD and 90 msec for healthy elderly adults (Revonsuo & Laine, 1996). Eyeball- Controls in the frontal electrodes with minimum la- ing the difference waves reveals that N400 and LPC tency). effects are larger in magnitude for the controls than for AD patients (Figure 3). Repeated-measures ANOVAs were Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 P2 used to test main effects of Group (2 levels), Word Type (2 levels), and Electrode (20 levels) and their interaction P2 peaked around 230 msec after stimulus onset. In P2 on the amplitudes and latencies of the ERP components. amplitude, there was a main effect of Electrode, F(19, Here we report the effects of Word Type and Group. The 475) = 25.36, p < 0.001, with maximum amplitudes main effects of Electrode parallel those reported above observed at frontal electrodes. No signiªcant Group ef- for Congruous Words. No signiªcant interaction effects fects were observed. In P2 latency, there was a signiªcant involving Electrode were observed, indicating that the main effect of Electrode, F(19, 475) = 27756.75, p < scalp topography of the ERPs to Incongruous words was 0.001 and signiªcant interaction of Electrode × Group, similar for both groups and both word types. F(19, 475) = 65.25, p < 0.001 but no main effect of Group. The earliest peak latencies for the control group were observed at occipital electrode sites around 205 P2 msec, but for AD patients they were observed at T5. Thus, In P2 amplitude, a main effect of Word Type was found: the distribution of peak latencies were different in the P2 peak amplitude was signiªcantly smaller to incongru- two groups. ous than to congruous words, F(1, 25) = 4.84, p < 0.05. The P2 peak latency also showed a main effect of Word N400 Type: The P2 peak was signiªcantly earlier for the Incon- gruous than for the Congruous words, F(1, 25) = 5.56, Maximum amplitude peaks were observed around 460 p < 0.05. Our interpretation of these results is that the to 530 msec after stimulus onset at centroparietal elec- N400 negativity (or the PMN, see “Discussion”) partly trode sites. We extracted the N400 average amplitude at overlaps the P2 time window. Thus, P2 is not in itself 300 to 800 msec. A signiªcant effect of Electrode, F(19, sensitive to semantic modulation, although the overlap 475) = 10.93, p < 0.001, and Electrode × Group interac- with the PMN-N400 complex shows itself by “cutting” tion, F(19, 475) = 2.33, p < 0.05, were revealed. the preceding P2 positive peak and leaving it thus sig- niªcantly shorter and weaker. This effect was similar for LPC controls and AD patients; no Group interactions were observed. The LPC has been reported to follow the N400 in auditory ERPs to spoken words (McCallum et al., 1984; Revonsuo & Laine 1996; Juottonen, Revonsuo, & Lang 1996). We extracted the average amplitude at 800 to 1500 msec. A signiªcant main effect of Electrode was observed, F(19, 475) = 8.05, p < 0.001, with maximum values observed over the occipito-parietal areas.

Interim Summary: ERPs to Congruous Words With regard to the peak amplitudes of N1 and P2, the AD patients’ ERPs were no different from those of the control subjects. However, with regard to the peak latencies, some differences emerged. The N1 peak was actually earlier for the AD group than for the controls. The topographical distribution of peak latency was different between the groups as indicated by the signiªcant interactions of Electrode × Group in the peak latencies of N1 and P2. N400 mean amplitude was topographically different between the groups, but no group differences emerged in the LPC. Figure 3. Difference waves at electrode Cz.

412 Journal of Cognitive Neuroscience Volume 10, Number 3

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 N400 In N400 average amplitude, a signiªcant main effect of Word Type was found, F(1, 25) = 8.20, p < 0.01, but no interaction of Word Type × Group. In other words, N400 average amplitude at 300 to 800 msec for the Incongru- ous words was not signiªcantly different between the groups. Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 LPC In LPC average amplitude, no signiªcant effects involving Group or Word Type were found.

Word Type Effect in 100-msec Time Windows Visual inspection of the group (Figures 1 through 3) and individual (Figure 4) waveforms indicates that Word Type does not have as great an effect in the case of the AD group as it does in the controls. Additional analyses were carried out on mean amplitudes of eight critical electrodes (central C3, C z, C 4, par ietal P 3, P z, P 4, and midline Oz and Fz) in 100-msec time windows between 200 and 1500 msec from stimulus onset. The signiªcance of the effect of Word Type was ªrst separately tested for each group and each time window, and then the interaction of Word Type × Group was tested in a similar manner. This analysis was carried out in order to reveal the precise time windows in which the groups signiªcantly differed in terms of the Word Type Effect (Table 1). The development of the Word Type Effect in the N400 time window is depicted in Figure 5. Note especially the relatively small (and nonsigniªcant) difference between the groups at 200 to 300 msec, the increasing effect for controls and the decreasing effect for AD patients between 200 and 600 msec, and the maximum difference between groups at 500 to 600 msec, which was also Figure 4. Individual waveforms of ERPs to Congruous (broken line) statistically signiªcant. and Incongruous (continuous line) ªnal words of a sentence; 17 control subjects and 9 AD patients.

N400 Relation to Cognition: Correlations DISCUSSION between Behavioral Tasks and N400 Amplitude The aim of this study was to ªnd out how the semantic The N400 effect was summed from six centroparietal processing of spoken words is revealed by ERPs in AD electrodes, and the relationship of this value with results patients who typically show impaired semantic process- from behavioral tasks was analyzed by calculating Pear- ing in behavioral tasks. We expected the semantic impair- son’s Correlation Coefªcients. N400 and MMSE, r = 0.45, ment to be selectively manifested in such ERP p < 0.05; N400 and explicit detection of incongruous components that are normally modulated by semantic sentences, r = 0.49, p < 0.02; and N400 and WAIS Simi- factors. Our main ªndings are in concert with this pre- larities, r = 0.60, p < 0.005. Thus, N400 peak amplitude diction: The N1 and P2 components elicited by spoken difference was signiªcantly correlated with all behav- words are fairly normal in the AD patients, but the ioral tasks. The most signiªcant correlation was observed amplitude of N400 and LPC to semantically Incongruous with WAIS Similarities, a task of semantic-conceptual words is signiªcantly reduced. The amplitude difference reasoning; the worse the performance was in the behav- between ERPs to Congruous and Incongruous words, ioral task, the smaller (or the more reversed) the N400 measured as mean amplitude in eight electrodes and effect tended to be (Figure 6). 100-msec time windows, revealed that this Word Type

Revonsuo et al. 413

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 Table 1. Statistical signiªcance of Word Type effect effect was signiªcantly different for the groups at 500 to (difference between congruous and incongruous words) 600, 1200 to 1300, and 1400 to 1500 msec. Thus, the AD within and between the study groups. The effect was patients’ impairment in semantic processing is primarily measured as mean amplitude at eight midline and associated with reduced amplitude in narrow time win- centroparietal electrodes in 100-msec time windows. dows corresponding in latency to the N400 and LPC. Time window AD vs. Furthermore, the magnitude of the N400 effect was (msec) AD Control control signiªcantly correlated with performance in behavioral tasks requiring explicit semantic knowledge, suggesting 200–300 ns p < 0.01 ns a connection between N400 and activation of semantic-

300–400 ns p < 0.01 ns conceptual representations. Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 400–500 ns ns ns 500–600 ns p < 0.05 p < 0.05 ERPs to Semantically Congruous Words 600–700 ns ns ns We ªrst investigated the ERPs to semantically congruous words in order to ªnd out whether AD patients manifest 700–800 ns ns ns any generalized abnormalities in the ERPs to spoken 800–900 ns ns ns words. The peak amplitudes of auditory ERP compo- 900–1000 ns ns ns nents N1 and P2 were not signiªcantly different for the patients and the controls. N1 latency was signiªcantly 1000–1100 ns ns ns different: N1 peaked earlier for the AD patients than for 1100–1200 ns ns ns the controls. Previous studies on auditory ERPs support 1200–1300 ns ns p < 0.05 the view that N1 and P2 are usually similar in AD patients and controls (e.g., Filipovic, Kostic, Sternic, Marinkovic, 1300–1400 ns p < 0.05 ns & Ocic, 1990). The only difference that consistently ap- 1400–1500 ns p < 0.01 p < 0.05a peared between the groups was the topographical distribution of peak latencies. This combination of normal a Signiªcant interaction with electrode position. peak amplitude with atypical distribution of peak latency may indicate that electrophysiological conduc- tance is anomalous in the brains of AD patients (i.e. the activation is abnormally propagated and summed

Figure 5. Development of N400 effect, measured as mean amplitude at eight midline and centroparietal electrodes in 100-msec time windows.

414 Journal of Cognitive Neuroscience Volume 10, Number 3

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 AD patients. However, these ªndings are in conºict with those of Hamberger et al. (1995), who reported normal N400 effects, and Ford et al. (1996), who reported enlarged N400 amplitudes for AD patients. There is no simple explanation for the differing results in terms of stimulus modality (visual/auditory) or stimulus type (sentence/word pair), task difªculty (simultaneous choice RT/no simultaneous task), or interstimulus interval because it appears that the studies with dissimilar results

do not systematically differ in these respects. However, Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 dementia severity may play a role: In the present study the AD patients scored on the average 60% of maximum points in Mini-Mental State Examination (MMSE), whereas in studies by Hamberger et al. and Ford et al. AD patients scored on the average nearly 70%. Schwartz et al. do not report MMSE scores, but on the basis of the Dementia Rating Scale their patients were in the “mild to moderate” range of dementia severity. Hagoort and Kutas do not report data on dementia severity. On the basis of the available data on dementia severity in the different studies, there seems to be a tendency that the Figure 6. Diagram showing N400 effect (summed from peak amplitude differences at six centroparietal electrodes) in relation with more severely demented the AD patients are, the more WAIS Similarities scores (Pearson’s r = 0.60, p < 0.005); 17 control likely it is that reduced N400 amplitudes are observed. subjects (squares) and 8 AD patients (triangles) (1 patient was not It would be interesting to compare the N400 results given the task). with results from analogous behavioral tasks. Actually, the results on behavioral semantic priming in AD are also through brain tissue), although the neural generators of quite controversial: In some studies, no priming effects the ERP components function within normal limits. Thus, for AD patients have been observed (Ober & Shenaut, the results of the present study demonstrate that there 1988), whereas in others normal or greater than normal is no such generalized abnormality in the neural genera- effects are reported (e.g., Chertkow et al., 1994). These tion of ERPs to spoken words in AD patients that would inconsistent ªndings in studies of behavioral semantic undermine the comparison of ERPs to Congruous and priming in AD may be due to differential interstimulus Incongruous stimulus types. intervals (ISIs): The semantic activation caused by the prime may not last as long in AD patients as it does in normal subjects (Nebes, 1989). It may be that the longer the ISI is, the less likely it becomes that an N400 effect ERPs to Semantically Incongruous Words is elicited. Hamberger et al. (1995) used a relatively short For both AD patients and controls, the amplitude modu- ISI (500 msec) compared with the other studies (1000 lation induced by the Incongruous stimulus word begins msec), which may have enhanced the likelihood of at similar latency and magnitude between 200 and 300 ªnding the normal N400 they reported. Interestingly, an msec after stimulus onset. In this time window, the auditory ERP component called mismatch negativity and negativity generated by the Incongruous word interacts reºecting sensory memory decreases faster as a function with the positivity of the P2 component so that the P2 of ISI in AD than in normal subjects (Pekkonen, Jousmäki, peak amplitude remains smaller for both study groups. Kononen, Reinikainen, & Partanen, 1994). In sharp contrast, N400 peak amplitude to the Incongru- Dementia severity and ISI are thus likely to have an ous words seems to be selectively and greatly reduced effect on N400, which ought to be taken into account in the AD patients’ ERPs, and there is a signiªcant differ- in future studies. Even so, the results reported by Ford ence in the N400 effect in a corresponding narrow time et al. (1996) are still puzzling because they reported not window at 500 to 600 msec after stimulus onset. In only enlarged N400s for AD patients but also that worse addition, there is a signiªcant late difference in the LPC performance in tests of semantic knowledge were re- time windows. lated to larger N400 absolute amplitudes. However, as These ªndings taken together indicate that the elec- the authors admit, these results have to be taken with trophysiological response to semantic modulation in the caution because the baseline against which they meas- N400 time window fails to develop normally for the AD ured N400 amplitude in the case of the AD patients was patients. The present results are fairly well in concert dissimilar to that of the elderly controls. Thus, it may be with those of Hagoort and Kutas (1994), and Schwartz that what was actually measured by Ford et al. was not et al. (1996), who found reduced N400 amplitudes for a greater N400 for AD patients at all but rather a more

Revonsuo et al. 415

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 negative baseline for the control group. This baseline The N400 Effect and Semantic Processing in AD effect was due to the fact that the sentence stem was presented as natural connected speech, but the ªnal Attempts have been made to correlate N400 with differ- target word was separated by a 1000-msec silence from ent stages of word recognition (lexical activation, the stem. In effect, the silence functioned as a cue so semantic-conceptual activation, semantic integration that the subjects knew well in advance when the critical processes) and different kinds of processing (automatic and possibly anomalous word would occur. It may be spreading activation, allocation of attentional resources), that the control subjects but not the AD patients actively but no general agreement has so far emerged. In fact, it tended to build strong expectations about the ªnal word seems likely that many different cognitive phenomena

during the 1000-msec silence, resulting in a different are at work and contribute independently to some ex- Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 baseline state for the controls. In the present study, all tent to the negativity in the N400 time window. Connolly words were separated by a similar silent interval, so and Phillips (1994) showed that there is a separate nega- there was no explicit cue that would have revealed the tive-going response in the early N400 time window target word and drawn special attention to it before it modulated by the initial phoneme of the terminal word was actually presented. Additionally, Ford et al. do not in spoken sentences. This PMN occurred between 270 consider the possible effects of PMN (Connolly & Phil- and 300 msec from last word onset if the initial pho- lips, 1994; see below) on their results, although they used neme was different from that of the word with highest stimuli that are expected to elicit this component. Thus, expectancy in the given sentence context. The authors it remains unclear to what extent the N400 in AD that propose that PMN occurs at the point of lexical selec- they observed was mingled with the PMN. According to tion when it is possible to identify that the activated the present results, PMN is more likely than N400 to be word candidates do not match the constraints provided preserved in AD. by the sentence context and that the later N400 prob- One alternative explanation for the results in the pre- ably reºects an independent postlexical integration sent study is that our AD patients simply were not paying process. attention to the stimuli—for the N400 effect is not elic- The precise nature of the postlexical processes corre- ited by unattended stimuli. It has been shown that in the lated with N400 (and LPC) is still open. According to one visual modality, words presented to unattended spatial hypothesis, N400 reºects activation of semantic-concep- locations do not elicit the N400 component (McCarthy tual representations in a common modality-independent & Nobre, 1993) and that the N400 effect is not generated semantic memory system. According to a competing to target words if the prime is masked (Brown & hypothesis, N400 reºects higher-level integrative proc- Hagoort, 1993). In the auditory modality, N400 amplitude esses that bind together different kinds of lower-level is signiªcantly more negative to unattended than at- information (e.g., syntactic, semantic, pragmatic) in lan- tended target words—the N400 elicited by all unat- guage comprehension (Osterhout & Holcomb, 1995). In tended targets is as large as the N400 elicited by the present study, an interesting relationship between unrelated attended targets. For unattended words there behavioral tasks and the N400 effect was observed: N400 is no difference between related and unrelated targets was signiªcantly correlated with MMSE scores and with (no N400 effect) (Bentin, 1994). Thus, the effect of not the explicit detection of incongruity in stimulus sen- attending to the stimuli in the present experiment would tences. Remarkably, the strongest and most signiªcant be expected to erase the N400 effect from AD patients correlation was obtained between the N400 effect and by rendering their N400 amplitude to Congruous words the WAIS Similarities task. This ªnding supports the hy- more negative than that of the controls. However, this pothesis that N400, to some extent at least, reºects the was not the case: The N400 effect was not attenuated activation of semantic-conceptual representations: The because of overly negative-going responses to Congru- Similarities task requires effortful, voluntary activation of ous words but because of an absence of negative-going semantic-conceptual knowledge. responses to Incongruous words. Moreover, there is little The Incongruous ªnal words in the present study direct evidence to support the inattention hypothesis: were such that they were expected to elicit both the First, observation of the patients’ behavioral and physi- PMN and the later N400. The congruity effects of the ological alertness during the experiment does not sup- controls and the patients are fairly close to each other port it. Second, the amplitude of the N1 to auditory in magnitude in the 200 to 300-msec time window but stimuli is known to be modulated by attention so that remarkably different in the 500 to 600-msec time win- for unattended stimuli the peak amplitude is reduced dow, where the group difference was signiªcant. The AD (Mangun & Hillyard, 1995). In the present study N1 peak patients do not show a signiªcant Word Type effect in amplitude for AD patients was not signiªcantly different any time window. However, the group differences show from that of the controls. In the light of these facts, we an interesting pattern: AD patients differ from the con- regard it as unlikely that our results would be solely due trols in the time windows in which the controls show to insufªcient attention paid to the stimuli by the AD the N400 and LPC effects but do not differ in the time patients. windows in which the controls show the PMN effect.

416 Journal of Cognitive Neuroscience Volume 10, Number 3

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 Thus, the patterns of effects are clearly dissimilar in ªndings stems from the fact that we do not yet have a these time windows: The pattern in the PMN time win- precise neurocognitive theory of the relationship of dow indicates that the responses of the AD patients to N400 to the semantic processes and structures that are incongruous words are here much closer to normal than assumed to be impaired in AD. Thus, any abnormality in any other time window. Our interpretation is that this observed in N400 in AD can be interpreted as reºecting is largely due to a residual PMN effect, reºecting rela- some kind of semantic memory impairment previously tively well-preserved lexical access and lexical selection identiªed in behavioral studies, but it would be impor- processes in AD. tant to know exactly how the two levels of analysis are What kind of conclusions can be drawn from the related to each other. Given that the existence of a

present study concerning the organization of semantic generalized access deªcit is now well established in the Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 representations and access to them? If the N400 (and/or literature on AD, and a partial storage deªcit seems likely the LPC) reºects the automatic spreading of activation as well (e.g., Laatu et al., 1997), what should we expect in a semantic network during input processing, it ap- to see in the ERPs? Should an access deªcit1 be expected pears that in AD patients the intensity with which such to show in N400 at all? What about a storage deªcit? If semantic activation spreads is abnormally low. The lack the semantic network is partially disrupted in AD, should of intensity in automatic semantic activation in AD pa- we expect the N400 to be larger or smaller in AD than tients means that a new focus of activation in the seman- in normal subjects? In order to answer these questions, tic network fails to develop and spread normally, and it we would need a neurocognitive model of semantic fails to establish the connections to the context pre- access and storage and a related model of the neurocog- viously activated. Alternatively, the old focus of activation nitive mechanisms of N400. Unfortunately, we do not brought about by the preceding context may decay know as yet what the precise mechanisms of N400-re- abnormally fast, which makes it difªcult to integrate the lated semantic activation are, even in normal subjects. target word with the context. The ERP components N1 However, the emerging picture is that many different and P2, preceding the N400 effect, appear to have nor- stages and types of cognitive processing are probably mal intensity (and P2 also has normal latency) in the AD correlated with the ERP effects elicited by semantically group. Additionally, the AD patients did not differ from incongruous words. Some of them may be to some the controls in the time window of the PMN that pre- extent preserved in AD (the PMN effect reºecting lexical sumably reºects automatic spreading of activation at the selection); others are clearly reduced (the semantic lexical level. Therefore it seems unlikely that the input N400 and the LPC), although in very mildly demented to the postlexical processes, such as semantic-concep- AD patients, an incipient reduction might not be detect- tual activation, should be missing in AD patients. We able. interpret these ªndings to show that during the auditory processing of words, the input that automatically ac- METHOD cesses the semantic-conceptual network (or other postlexical systems generating the N400 effect) is by and Subjects large normal in AD patients (as indicated by ERPs closely Normal Controls similar to the control group, prior to the N400 time window). Somehow the semantic-conceptual system it- Seventeen (nine male, eight female) elderly (mean age self seems to be incapable of developing a strong state 67.4 years, SD = 4.0, range 62 to 75), nondemented of activation, which may be due to an abnormally fast (MMSE mean score 27.7, SD = 1.8, range = 24 to 30; decay of the activated traces or to the loss of structural Folstein, Folstein, & McHugh, 1975) subjects without a links between conceptual nodes (or both). In accord- history of neurological or psychiatric disease partici- ance with this interpretation, a recent study of the or- pated in the study. Subjects had the average of 8.5 years ganization of the semantic network in AD (Chan, Butters, of formal education (SD = 3.5, range = 4 to 17). & Salmon, 1997) indicates that there is a systematic breakdown in the structure of semantic knowledge in AD Patients the course of AD. There appears to be a general theoretical difªculty in Nine patients (three male, six female) were recruited interpreting ERP data from AD patients in terms of se- from the Department of Neurology, Turku University mantic memory or processing. Although we now have Central Hospital. The mean age of the patients was 67.1 substantial knowledge of the semantic deªcits in AD as years (SD = 8.3, range 54 to 78). Mean score in the MMSE revealed by behavioral studies, it is difªcult on the basis was 18 points (SD = 6.7, range 6 to 26). Even though a of the behavioral data to generate speciªc predictions relatively crude measure, MMSE indicates that dementia concerning potential abnormalities in the N400 compo- severity varied from mild to moderate in the patient nent in AD. Should we expect to see a normal, a dimin- group. The patients had the average of 7.8 years of formal ished, or an exaggerated N400 component? The difªculty education (SD = 1.4, range 6 to 10), which was not in predicting ERP results on the basis of behavioral signiªcantly different from controls. The diagnosis of AD

Revonsuo et al. 417

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 was based on clinical history, general medical and neu- by a male voice at a slow rate, one word at a time (for rological investigation, neuropsychological assessment, a comparable procedure, see McCallum et al., 1984). laboratory tests, and neuroradiological imaging. The pa- Silence between words was about 1 sec, and between tients fulªlled the DSM-III-R (American Psychiatric end of ªnal word and beginning of the next sentence Association, 1987) criteria for dementia and the NINCDS- about 3 sec. In addition, 20 practice sentences were ADRDA criteria for probable AD (McKhann et al. 1984). similarly generated and taped. All patients had either a brain CT or MRI, or both, with results consistent with the diagnosis. The mean disease Procedure duration was 4.1 years (SD = 1.9, range = 1.5 to 7 years).

None of the subjects were taking medication that had EEG Recording Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 signiªcant effects on the central nervous system. All the subjects were ªrst tested for hearing. Then the procedure was explained to them by telling them that their task was to listen carefully to what the voice com- Stimuli ing through the headphones was saying. Then they lis- The set of 200 Finnish sentences reported in Revonsuo tened to the 20 practice sentences binaurally through and Laine (1996) was used. Initially, a set of 200 seman- headphones. The volume was adjusted to a level at tically congruous Finnish sentences, each consisting of 3 which the subject indicated hearing the speech clearly to 7 words, was devised. A second set of 200 semantically and effortlessly. Then the experiment proper was started, incongruous versions of the same sentences was created and the 200 sentences were played. After every 40 sen- by replacing the last word of the congruous sentences tences, a break of a couple of minutes was provided. In with a semantically incongruous word. The ªnal set of order to make sure that the subjects were paying atten- 200 stimulus sentences was constructed by randomly tion to the sentences, the control subjects were given assigning the congruous sentences into two groups: Of four printed sentences during the break and asked to the ªrst 100 sentences, the congruous version was taken, indicate if any of them were among the ones presented and of the remaining 100, the incongruous version. Thus, in the latest block. Because of their memory deªcits, the 100 sentences ended in a semantically congruous word patients were not given this task, but they were re- (e.g., “Last night I had a strange dream”) and 100 in a minded that they should carefully listen to the stimuli. semantically incongruous word (e.g., “A thick encyclope- Furthermore, the alertness of all subjects was constantly dia contains a lot of radiation”). There was no statistical observed by video monitoring the subjects and by an difference between the two stimulus types as to sen- on-line watch of the continuous electroencephalogram tence length, ªnal word length, lemma frequency, and (EEG). grammatical role (Object 37.5%, Adverbial 35%, Subject Scalp EEG was recorded with 20 silver-silver-chloride 19%, Predicative 5%, and Predicate, Attribute, Preposition, electrodes arranged according to the 10-20 system. Hori- or Postposition 3.5%). The degree of constraint of the zontal and vertical eye movements were recorded using sentence stems was estimated by giving the stems to 31 two electroculogram (EOG) electrodes. All electrodes students as a sentence-completion task. The most ex- were referenced to the linked ears. In the recording, a pected ending to stems of the congruous sentences was, time constant of 1.0 sec was used for the scalp elec- on the average, produced by 79.5% (SD = 18.5, range = trodes, and 0.3 sec for the EOG electrodes. High-fre- 32 to 100) and to stems of the incongruous sentences quency cutoff of 70 Hz was used. Digitization rate was by 78.2% (SD = 20.5, range = 29 to 100) students. Thus, 200 Hz, and the EEG data was sampled once every 5 the sentence stems were quite considerably constrain- msec. The Neuroscan recording equipment was cali- ing, and there was no signiªcant difference between the brated for each recording separately by sending 50-µV stems of the two types of stimulus sentences. In the pulses and measuring the deviation from the expected congruous sentences, the word with highest cloze prob- ampliªed value of 1.0 V. Corrections were adjusted to ability (most expected ending) was used as the target each channel separately. The EEG data were divided into word in 193 of 200 sentences, and in the remaining 7 of epochs of 1900 msec, starting from 200 msec before 200 sentences the target word was the second, third, or ªnal word onset and ending 1700 msec after it. The fourth most expected ending for the respective stem. prestimulus baseline level used was the activity from 200 The mean length of the congruous sentences was 4.8 to 100 msec before stimulus onset. Stimulus onset was words (SD = 0.9, range 3 to 7), the duration was 11.1 deªned as the point in the stimulus channel when the sec (SD 2.5, range 7 to 17), and the mean number of waveform of the last word of each sentence was ªrst syllables in the last word was 3.0 (SD = 0.8, range 2 to discernible from background noise. The ªnal-word EEG- 5). The corresponding ªgures for the incongruous sen- epochs were averaged separately for semantically con- tences were 4.7 words (SD = 0.8, 3 to 7), 11.0 seconds gruous and incongruous ªnal words. Before averaging (SD 2.5, range 6 to 18), and 2.9 syllables (SD = 0.7, range the data, all epochs showing signiªcant eye movement 2 to 5). The sentences were arranged in a random order (±50 µV) between −200 msec to +1700 msec of stimu- and digitally recorded on DAT tape. They were spoken lus onset in the EOG electrodes were rejected.

418 Journal of Cognitive Neuroscience Volume 10, Number 3

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 Behavioral Tasks N400: Evidence from masked priming. Journal of Cogni- tive Neuroscience, 5, 34–44. Immediately after the EEG recording, the subjects’ ex- Chan, A. S., Butters, N., Paulsen, J. S., Salmon, D. P., Swenson, plicit comprehension of the stimulus material was tested. M. R., & Maloney, L. T. (1993). An assessment of the seman- First, the experimenter gave at least two sentences as tic network in patients with Alzheimer’s disease. Journal of Cognitive Neuroscience, 5, 254–261. examples and showed and told the subjects that they Chan, A. S., Butters, N., & Salmon, D. P. (1997). The deteriora- were expected to either accept or reject the sentence tion of semantic networks in patients with Alzheimer’s dis- on the basis of whether or not it made sense. Then the ease: A cross-sectional study. Neuropsychologia, 35, subjects listened to a sample of 40 stimulus sentences 241–248. once again (a 20% subset of the original stimuli), one Chertkow, H., Bub, D., Bergman, H., Bruemmer, A., Merling, A., & Rothºeisch, J. (1994). Increased semantic priming in pa- Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 sentence at a time. After each sentence, the subject was tients with dementia of Alzheimer’s type. Journal of Clini- asked to indicate whether the sentence was ordinary cal and Experimental Neuropsychology, 16, 608–622. and sensible or unusual and senseless. In addition, the Chertkow, H., & Bub, D. (1990). Semantic memory loss in de- WAIS Similarities task (Wechsler, 1955) was presented to mentia of Alzheimer’s type: What do various measures the subjects in order to have a measure of semantic measure? Brain, 113, 397–417. Chertkow, H., Bub, D., & Seidenberg, M. (1989). Priming and function independent of the stimulus sentences. semantic memory loss in Alzheimer’s disease. Brain and Language, 36, 420–446. Connolly, J. F., & Phillips, N. A. (1994). Event-related potential Acknowledgments components reºect phonological and semantic processing This study was supported by the Academy of Finland and the of the terminal word of spoken sentences. Journal of Cog- Finnish Cultural Foundation. We thank the anonymous referees nitive Neuroscience, 6, 256–266. of the Journal for their helpful comments and Dr. Matti Laine Daum, I., Riesch, G., Sar tor i, G., & Birbaumer, N. (1996). Se- and Jaani Kuusela for their help with the stimulus materials. mantic memory impairment in Alzheimer’s disease. Jour- nal of Clinical and Experimental Neuropsychology, 18, 648–665. Reprint requests should be sent to Antti Revonsuo, Center Filipovic, S., Kostic, V. S., Sternic, N., Marinkovic, G., & Ocic, for Cognitive Neuroscience, Department of Philosophy, G. (1990). Auditory event-related potentials in different University of Turku, FIN-20014 Turku, Finland, or via e-mail: types of dementia. European Journal of Neurology, 30, antti.revonsuo@utu.ª. 189–193. Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). “Mini- Mental State”: A practical method for grading the cognitive Note state of patients for the clinician. Journal of Psychiatric 1. The automatic forms of access should be distinguished from Research, 12, 189–198. voluntary, effortful access to semantic-conceptual information. Ford, J. M., Woodward, S. H., Sullivan, E. V., Isaacks, B. G., Access in the former sense refers to the fast and largely invol- Tinklenberg, J. R., Yesavage, Y. A., & Roth, W. T. (1996). N400 untary stages of lexical and semantic processing that take place evidence of abnormal responses to speech in Alzheimer’s when we recognize a word and understand what it means. By disease. Electroencephalography and Clinical Neurophysi- contrast, access in the latter sense is typically operationalized ology, 99, 235–246. in terms of tasks that measure the ability of subjects to inten- Ganis, G., Kutas, M., & Sereno M. I. (1996). The search for tionally activate certain types of semantic-conceptual informa- “common sense”: An electrophysiological study of the com- tion in semantic memory (e.g., by deªning a concept). Because prehension of words and pictures in reading. Journal of AD patients often fail in such tasks, it is possible to attribute Cognitive Neuroscience, 8, 89–106. the failure to mechanisms involved in creating voluntary access Hagoort, P., Brown C. M., & Swaab, T. Y. (1996). Lexical-seman- to the relevant systems (e.g., working memory or attentional tic event-related potential effects in patients with left control functions), not to deªcits in the systems themselves hemisphere lesions and aphasia, and patients with right that are being accessed. The term access deªcit usually refers hemisphere lesions without aphasia. Brain, 119, 627–649. to difªculties in creating voluntary access to semantic-concep- Hagoort, P., & Kutas, M. (1994). Electrophysiological insights tual information. into language deªcits. In F. Boller & J. Grafman (Eds.), Handbook of neuropsychology, vol. 9 (pp. 501–530). Am- sterdam: Elsevier. REFERENCES Hamberger, M. J., Fr iedman, D., Ritter, W., & Rosen, J. (1995). Event-related potential and behavioral correlates of seman- American Psychiatric Association (1987). Diagnostic and sta- tic processing in Alzheimer patients and normal controls. tistical manual of mental disorders (DSM-III). 3rd ed., re- Brain and Language, 48, 33–68. vised. Washington DC: American Psychiatric Assoc. Hodges, J. R., Salmon, D. P., & Butters, N. (1992). Semantic Bayles, K. A., Tomoeda, C. K., Kaszniak, A. W., & Trosset, M. W. memory impairment in Alzheimer’s disease: Failure of ac- (1991). Alzheimer’s disease effects on semantic memory: cess or degraded knowledge? Neuropsychologia, 30, 301– Loss of structure or impaired processing? Journal of Cog- 314. nitive Neuroscience, 3, 166–182. Holcomb, P. J., & McPherson, W. B. (1994). Event-related brain Bentin, S. (1994) Semantic processing of ignored stimuli: The potentials reºect semantic priming in an object-decision role of attention in memory. In C. Umilta & M. Moscovitch task. Brain and Cognition, 24, 259–276. (Eds.), Attention and performance XV, conscious and Juottonen, K., Revonsuo, A., & Lang, H. (1996). Dissimilar age nonconscious information processing (pp. 551–569). inºuences on two ERP waveforms (LPC and N400) reºect- Cambridge, MA: MIT Press. ing semantic context effect. Cognitive Brain Research, 4, Brown, C., & Hagoort, P. (1993). The processing nature of the 99–107.

Revonsuo et al. 419

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021 Kutas, M., & Hillyard, S.A. (1980). Reading senseless sen- Nigam, A., Hoffman, J. E., & Simons, R. F. (1992). N400 to se- tences: Brain potentials reveal semantic incongruity. Sci- mantically anomalous pictures and words. Journal of Cog- ence, 207, 203–204. nitive Neuroscience, 4, 15–22. Laatu, S., Portin, R., Revonsuo, A., Tuisku, S., & Rinne, J. Ober, B. A., & Shenaut, G. K. (1988). Lexical decision and (1997). Knowledge of concept meanings in Alzheimer’s dis- priming in Alzheimer’s disease. Neuropsychologia, 26, 273– ease. Cortex, 33, 27–45. 286. Mangun, G. R., & Hillyard, S. A. (1995). Mechanisms and mod- Osterhout, L., & Holcomb, P. J. (1995). Event-related poten- els of selective attention. In M. D. Rugg & M. G. H. Coles tials and language comprehension. In M. D. Rugg & (Eds.), Electrophysiology of mind: Event-related brain po- M. G. H. Coles (Eds.), Electrophysiology of mind: Event- tentials and cognition (pp. 40–85). Oxford: Oxford Univer- related brain potentials and cognition (pp. 171–215). Ox- sity Press. ford: Oxford University Press. McCallum, W. C., Farmer, S. F., & Pocock, P. K. (1984). The ef- Pekkonen, E., Jousmäki, V., Kononen, M., Reinikainen, K., & Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/3/408/1758334/089892998562726.pdf by guest on 18 May 2021 fects of physical and semantic incongruities on auditory Partanen, J. (1994). Auditory sensory memory impairment event-related potentials. Electroencephalography and in Alzheimer’s disease: An event-related potential study. Clinical Neurophysiology, 59, 477–488. Neuroreport, 5, 2537–2540. McCarthy, G., & Nobre, A. C. (1993). Modulation of semantic Polich, J., Ladish, C., & Bloom, F. E. (1990). P300 assessment processing by spatial selective attention. Electroencepha- of early Alzheimer’s disease. Electroencephalography and lography and Clinical Neurophysiology 88, 210–219. Clinical Neurophysiology, 77, 179–189. McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, Pratarelli, M. E. (1994). Semantic processing of pictures and D., & Stadlan, E. M. (1984). Clinical diagnosis of Alzhei- spoken words: Evidence from event-related brain poten- mer’s disease. Report of the NINCDS-ADRDA Work Group tials. Brain and Cognition, 24, 137–157. under the auspices of Department of Health and Human Revonsuo, A., & Laine, M. (1996). Semantic processing with- Services Task Force on Alzheimer’s Disease. Neurology, 34, out conscious understanding in a global aphasic: Evidence 939–944. from auditory event-related brain potentials. Cortex, 32, Nebes, R. D. (1989). Semantic memory in Alzheimer’s disease. 29–48. Psychological Bulletin, 106, 377–394. Rugg, M. D., & Coles, M. G. H. (1995). The ERP and cognitive Nebes, R. D. (1994). Contextual facilitation of lexical process- psychology: Conceptual issues. In M. D. Rugg & M. G. H. ing in Alzheimer’s disease: Intralexical priming or sentence- Coles (Eds.), Electrophysiology of mind: Event-related level priming? Journal of Clinical and Experimental brain potentials and cognition (pp. 27–39). Oxford: Ox- Neuropsychology, 16, 489–497. ford University Press. Nebes, R. D., Boller, F., & Holland, A. (1986). Use of semantic Schwar tz, T. J., Kutas, M., Butter s, N., Paulsen, J. S., & Salmon, context by patients with Alzheimer’s disease. Psychology D. P. (1996). Electrophysiological insights into the nature and Aging, 1, 261–269. of the semantic deªcit in Alzheimer’s disease. Neuropsy- Nebes, R. D., & Brady, C. B. (1991). The effect of contextual chologia, 34, 827–841. constraint on semantic judgments by Alzheimer patients. Wechsler, D. (1955). Manual for the Wechsler Adult Intelli- Cortex, 27, 237–246. gence Scale. New York: Psychological Corporation.

420 Journal of Cognitive Neuroscience Volume 10, Number 3

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998562726 by guest on 28 September 2021