Prefrontal and posterior cortical activation during auditory working memory

I E L S E V I E R Cognitive Brain Research 4 (1996) 27-37

COGNITIVE BRAIN

RESEARCH

Research report

Prefrontal and posterior cortical activation during auditory working memory

Linda L. Chao *, Robert T. Knight Department of Neurology and Center for Neuroscience, Universi~ of California, Davis VAMC-Department of Neurology (127), 150 Muir Road, Martinez,

CA 94553, USA

Accepted 3 October 1995

Abstract

The present study investigated brain mechanisms underlying auditory memory. In a modified Steinberg memory scanning task, 11 subjects indicated whether a probe sound was part of a previously presented 4-item memory set by a button press. Behaviorally, subjects responded fastest and most accurately to probes that matched the last memory set items and slowest and least accurately to negative probes and to positive probes to the first two memory set items. Electrophysiologically, probes to the last memory set items elicited the largest amplitude and earliest latency P3 components while other probes elicited smaller amplitude, prolonged P3s as well as a negativity around 400 ms. These results suggest that subjects utilized a trace strength/self-terminating search model to perform the memory scanning task. Subjects only generated the P3 component during the matching phase of the auditory memory task while a sustained frontal negativity was elicited during both the encoding and matching phase. Taken together these findings provide evidence of differential activation of distributed neural activity during non-linguistic auditory memory.

Keywords: Evoked potential; Serial position effect; Memory scanning; P3; N4

1. Introduct ion

Sternberg [68] developed a classic paradigm to study short-term memory (STM) in which he presented subjects with a series of digits, referred to as the memory set. Two seconds after presentation of the memory set, which varied in size from 1 to 6 digits, subjects were shown a single digit and had to indicate whether or not this was a member of the memory set. Positive trials occurred when the probe matched a member of the memory set; whereas in negative trials the probe did not match memory set items.

Sternberg theorized that retrieval from STM required 4 operations. First, subjects must encode the probe. Next, subjects must compare the probe with each item in STM. Third, subjects must decide whether the probe matched any of the items in STM. Finally, subjects have to make an overt decision and act upon it. It is possible for subjects to accomplish the second and third operations in a variety of

* Corresponding author. Fax: (1) (510) 229-2315; E-mail: I [email protected]

0926-6410/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved SSD10926-6410(95)00043-7

ways. They could employ a parallel processing strategy where they simultaneously compare the probe with all items in STM. Subjects could also utilize a self-terminat- ing strategy where they stop comparing the probe with memory set items as soon as a match is encountered. Finally, subjects could engage in a serial exhaustive search where all comparisons are considered before a decision is reached. Investigators using both the classic and modified versions of the Sternberg paradigm reported that reaction time (RT) was linearly related to the number of items in the memory set regardless of whether the probe was positive or negative. Based on this evidence, Sternberg [69] postulated that memory scanning in this task was serial and exhaustive.

When recognition accuracy is plotted as a function of the serial position of an item in the list, items in the first (primacy effect) a n d / o r last (recency effect) positions are better recognized [1]. RTs can also have a similar U-shape function; however, in STM tasks involving memory scan- ning, recency effects are usually more prominent.

Many investigators have recorded event-related poten- tials (ERPs) in variations of the Sternberg paradigm to study the neural basis of STM [47,49,50,67]. When ERPs

28 L.L. Chao, R.T. Knight/Cognitive Brain Research 4 (1996) 27-37

are collected in the Sternberg task, the probe typically generates a large amplitude P3 component. Several studies have observed a linear increase in the latency of the P3 to probes as the memory set size increases [49,50,67]. Ampli- tude effects due to memory set size are more variable [49].

Patterson and colleagues [46] examined serial position effect in both auditory and visual modality for verbal and non-verbal stimuli. They found significant recency effects in RT for the auditory, but not the visual modality. Further, they reported recency effects in both RT and ERPs for verbal auditory stimuli, but not for non-verbal stimuli. The authors concluded that behavioral and electrophysiological evidence of serial position effects during memory scanning tasks was dependent on both stimulus modality (auditory vs. visual) and stimulus type (verbal vs. non-verbal).

The Steruberg task can be considered a working mem- ory task since it utilizes the type of memory that is active and relevant for only a short period of time. The process measured by the delayed-response task in non-human pri- mates has also been viewed as a working memory task because the animal must retain memory of the location of the stimulus during the period of the delay. In delayed-re- sponse tasks, correct performance is controlled by memory rather than by sensory information.

There is substantial evidence that the prefrontal region is crucial to working memory. Early monkey research revealed that bilateral frontal lesions involving the sulcus principalis resulted in marked impairments in discrimina- tion tasks involving a delayed response [12]. Single unit studies also provided strong evidence of a prefrontal con- tribution to working memory. Neurons in the principal sulcus increase their discharge rates following stimulus presentation and continue to fire tonically during the delay period until the response is executed [13].

Recognition memory for non-linguistic auditory stimuli has been dissociated into two temporally distinct processes [6]. A rapid recognition memory process, engaged by repetition of stimuli at short delays (2 s or less), is associated with enhanced P3 amplitude, shortened P3 la- tency, and improved behavioral performance while a de- layed recognition memory process is associated with gen- eration of an N4 component, reduced and delayed P3 responses, and impaired performance.

This rapid auditory memory appears similar to the recency effect observed in the Sternberg memory scanning task. However, these results contradict the findings of Patterson and colleagues [46], who did not find recency effects in performance for visual stimuli and in ERPs for non-verbal auditory stimuli. To further address this issue, we employed ERP and behavioral techniques to investigate serial position effects in a modified Sternberg memory scanning task with non-linguistic auditory stimuli. Serial position effects for both probes and memory set items were examined since Sternberg [69] suggested that such effects may arise during the encoding of memory set items, rather than in the comparison stage.

2. Materials and methods

2.1. Subjects

Auditory event-related potentials (ERPs) were obtained from 11 paid subjects (8 males, 3 females). All of the subjects were young (21.7 ___ 1.8 years), right-handed, and had no history of audiological or neurological disease. All aspects of the research were explained to the subjects who signed statements of consent approved by the Institutional Review Boards of the Martinez Veterans Administration Hospital and the University of California at Davis.

2.2. Stimuli and procedure

Subjects were binaurally presented with sets of 4 stim- uli (60 dB HL, 700 ms duration, 1200 ms ISI) to memorize in a modified Sternberg memory scanning paradigm. Fol- lowing the memorized set, a probe stimulus was presented after a 2500-ms delay. Subjects indicated whether or not the probe was a member of the memorized set by pressing a 'yes' or 'no' button (see Fig. 1 for schematic of paradigm). Instructions stressed both speed and accuracy. Subjects were not instructed on the use of any particular strategies prior to testing; however, after testing was com- plete, subjects were questioned about how they performed the task. Five of the 11 subjects reported utilizing semantic labels to help them to remember the sounds, however they reported that they could not label all of the sounds. The remaining subjects claimed that they simply tried to re- member the sounds without consciously labeling them.

The items presented were drawn from a repertoire of 256 possible digitized sounds. The sounds consisted of non-speech human sounds (e.g. sneeze or cough), animal vocalizations (e.g. a dog barking), musical instruments (e.g. piano or guitar), and noises that occur in the environ- ment (i.e. dishwasher noises or garage door opening). The sounds had previously been rated as either verbally encod- able or non-verbally encodable by 10 subjects who did not

Memory Set Probe

I

-~700 m s ~-

Fig. 1. Schematic diagram of the memory scanning paradigm showing timing of stimulus presentation. This example shows a positive probe to the second memory set item.

L.L. Chao, R.T. Knight/Cognitive Brain Research 4 (1996) 27-37 29

participate in the current experiment. The raters were asked to listen carefully to each sound and to write down what they thought the sound was. The stimuli were pre- sented at the same speed as in the paradigm and subjects had to rate the stimuli in this period. If the sound could not be named within the designated period of time, it was considered to be verbally unencodable. Using this proce- dure, 45% of the sounds were rated as verbally encodable.

Twenty percent of the probes were negative (not a member of the memorized set) and 80% of the probes were positive. The items contained in each trial were presented in a pseudorandom fashion with the following restrictions: (1) the same probe stimulus could not occur on two consecutive trials; (2) no more than 3 consecutive positive or negative probes occurred in sequence; and (3) the proportion of positive probes relative to the position of the matching item (i.e. first, second, etc.) was adjusted to be equal. There were a total of 200 trials.

2.3. Recording

Electroencephalogram (EEG) was recorded from Ag/AgC1 electrodes placed at Fpl, Fpz, Fp2, F3, Fz, F4, C3, Cz, C4, P3, Pz, P4, T3, T4, T5, T6, O1, Oz, and 02 according to the international 10-20 system. The elec- trodes were referred to a balanced non-cephalic, sterno- vertebral reference electrode. Vertical and horizontal EOG recorded from electrodes placed below eye and on the outer canthus were used to monitor eye movements and blinks. Continuous EEG was amplified (50 K) and filtered (0.1-100 Hz) via a Grass Neurodata acquisition system. EEG was continuously digitized at 128 Hz/channel and stored on magnetic tape for analysis. The recording epoch, including a 200-ms prestimulus baseline, was 1999 ms.

2.4. Data analysis

Data averaging was performed off-line after sorting by response type (correct or incorrect) and probe position (positive probes to the first, second, third, or last memory set item and negative probes). ERPs to the memory set items corresponding to positive probes were also sepa- rately averaged. Epochs containing eye movement or elec- tromyographic artifact over 100 /zV were automatically rejected from the averaged data. The N1 (75-125 ms), P2 (175-225 ms), N2 (150-250 ms), N4 (350-450 ms), P3 (350-750 ms) and SFN (500-600 ms) were identified from individual subjects and grand averaged waveforms. Mean amplitude measures were referred to a 200-ms pres- timulus baseline while latency measures were referred to stimulus onset.

Analysis of variance (ANOVA) were made by response and probe position. Scalp potentials were normalized for comparisons of scalp distributions across conditions and repeated-measures ANOVAs were carried out with the Greenhouse-Geisser correction. In these cases, original

degrees of freedom, the Greenhouse-Geisser coefficient and corrected probability levels are reported. The New- man-Keuls procedure was used to perform post-hoc com- parisons of the means when appropriate.

3. Results

3.1. Response times and errors

Mean reaction times (RTs) and error rate are summa- rized in Fig. 2. Analysis of variance (ANOVAs) examining RT and error rate as a function of probe position yielded significant effects of RT (F3,30 = 12.83, P < 0.0001) and error rate (F3.30 = 22.41, P < 0.0001). Subjects responded fastest to probes to the last memory set item and second fastest to probes to the third memory set item (see Fig. 2). RTs to different types of probes were further examined by means of Newman-Keuls comparisons. The comparisons revealed that RTs to negative probes and positive probes to the first two memory set items were significantly (P < 0.05) slower than RTs to probes to the third and last memory set items. Subjects made the most errors to probes to the first two memory set items. Newman-Keuls con- trasts showed that there was no significant difference between error rates to negative probes or positive probes to the last two memory set items.

3.2. Event-related potentials to memory set items and to probes

Grand averaged ERPs elicited by memory set items matching correctly detected probe items and correctly de- tected probes are shown in Figs. 3 and 4. All memory set items elicited a sustained frontal negativity (SFN) and a central N4 component. Correctly detected probe items elicited a late posterior maximal P3 in addition to a SFN.

16130

1400 ' 20

~ 12®. 1.~ ~ E ~

e¢ 1(30o. 10 v

8 0 0 - 5

o I I [ I o

I s t 2 n d 3 r d las t n e g a t i v e

P r o b e P o s i t i o n

Fig. 2. Summary of behavioral data showing reaction time (solid circles) and percent error (open squares) as a function probe position. Note the significant recency effect for both RT and error rate (P = 0.00131).

30 L.L. Chao, R.T. Knight/Cognitive Brain Research 4 (1996) 27-37

Fpl ~ Fpz . ~ Fp2 , ~ . J AN,,...:L~ . . . --1 a/~". . . :" \ " X . . . . . ,_,1 ~.~.'" "k~,~ . . . . .

""~ SFN " ' ~ ~ - ~ q,

F7 ,~. Fz F 4 ~ , . k ' ~ k L F8

"" ~'N4 ~ ' -"re . . . . . .

v ~ ,.._..I . . , ; : . . . . . . . . . .

Oz ~ ~/ . . . . .

: : I : , : , , . . . . I . . . . . is memoryset • ~ 1000 ms + ~ -__" L-L-. negative pro~s ms S - - probes to memory set items 1-3

- - probes to last memory set items

Fig. 3. Grand-averaged ERPs to memory set items (thick dotted line), correctly detected negative probes (thick dashed line), positive probes matching memory set items 1-3 (thick solid line), and positive probes matching last memory set items (thin solid line). Both probes and memory set items elicited a SFN that was a maximal at frontal sites. Positive probes to memory set items 1-3, negative probes, and all memory set items elicited a centrally distributed N4 component as well as a posterior maximal P3. Positive probes to last memory set items only elicited a P3 component.

Subjects also generated a central N4 component to all correctly detected correctly detected negative probes and to positive probes to the first 3 memory set items.

YP~ZL ~ . . . . . .

SFN ~

Fz

N4

_.--.~.II.I. ~f~,~x,_~ ,, . . . . . , ~ ~ ,

~ " ~ ' , - ' x " " " , - ~

0z P3

, s , - ° .

, , I . . . . . . . . . I . . . . S 1000 ms

S p V

+

. . . . . Memory set items - - - Negative probes

Probes to memory set items 1-3 - - Probes to last memory set items

Fig. 4. Grand-averaged ERPs to correctly detected probes and memory set items at central electrode sites.

3.2.1. N1, P2, N2 components The N1 and P2 were quantified as mean amplitude and

peak latency from 75-125 and 175-225 ms, respectively. All stimuli elicited N1-P2 components (see Figs. 3 and 4). ANOVAs that examined the amplitude and latency the components as a factors of stimulus type (memory set items vs. probes) and electrode site yielded no significant differences between amplitude or latency of the N1 and P2 elicited by memory set items compared to those elicited by probe items.

The N2 component was quantified as mean amplitude from 150-250 ms and subjected to an ANOVA taking stimulus type and electrode as factors. All stimuli gener- ated an N2 component; however N2s elicited by probes were larger in amplitude than components elicited by memory set items (for stimulus type effect: FL20 = 16.87, P = 0.0005). Because the number of incorrect trials for positive probes to the third and last memory set were too small for further analysis, only positive probes to the first two memory set items and negative probes were analyzed for response (correct vs. incorrect) effects. The amplitude of the N2 generated to correctly detected probes were larger than those generated to incorrectly detected probes (for response effect, F1.20 = 5.71, P = 0.04).

3.2.2. P3 component The P3 component was quantified as mean amplitude

from 350 to 750 ms and subjected to an ANOVA taking amplitude and electrode sites as factors. All probe items elicited P3 components that were maximal over posterior

L.L Chao, R.T. Knight / Cognitive Brain Research 4 (1996) 27-37

Table 1 Mean amplitude and peak latency ( _+ S.D.) of the P3 component as a function of probe position at Pz and Oz

31

Probe position First Second Third Last Negative

Mean amplitude Pz 5.42-Z-_ 2.73 3.40 + 3.39 6.40 + 3.69 7.58 + 2.98 7.18 + 3.55 Oz 6.11+ 3.12 4.30+ 2.50 7.77+ 3.85 9.04+ 4.38 6.48+ 3.83

Peak latency Pz 695.64 + 37.98 698.45 + 38.04 598.36 ± 26.79 502.45 + 36.77 787.36 ± 27.50 Oz 687.73 + 33.86 677.09 + 40.48 588.36 + 26.39 488.18 + 28.33 779.55 + 25.97

electrode sites (for scalp distribution: F20,200 = 109.64, • =0 .24 , P = 0.0001; see Figs. 3 and 4). There were significant effects of probe position on both the amplitude and latency of the P3 component (see Fig. 5 and Table 1). The P3 was largest in amplitude (at Pz: F4,40 = 6.81, P = 0.005) and shortest in latency (at Pz: F4,40 = 291.09, P = 0.0001) for probes to last memory set items. Because of the significant probe position effect, we further exam- ined the amplitude and latency of the P3 separately for each probe type by means of Newman-Keu l s compar- isons. These comparisons revealed that the amplitude and latency of P3 elicited by last position probes were signifi- cantly different ( P < 0.05) from the amplitude and latency of P3 elicited by other probes. An ANOVA taking stimu- lus type and electrode site as factors revealed that subjects did not generate reliable P3 components to memory set items (for stimulus type effect: Fl,10 = 12.22, P = 0.006). When we examined P3 amplitude separately for probes and memory set items, Newman-Keu l s contrasts showed that the amplitude of the P3 elicited by probes differed significantly ( P < 0.05) from the amplitude of the P3 elicited by memory set items.

3.2.3. Sustained frontal negativity Both memory set items and probes generated SFNs

onsetting at 375-400 ms and peaked at around 500-600 ms. The SFN elicited by memory set items was maximal in amplitude over frontal electrode sites for memory set items

l0

9

8

7

6

5

4

o P3 latency

I 1 st 2rid

I I I 0 3 rd last nega t ive

P robe Pos i t ion

800

6OO

5OO

400

3oo ~

200

loo

Fig. 5. P3 amplitude and latency shown as a function of probe position. Note that P3s elicited by positive probes matching last memory set items are shorter in latency and larger in amplitude than P3s elicited by other probes.

(scalp distribution: F20,200 = 11.21, • = 0 . 1 4 , P =0 .0001) while the SFN elicited by probes were maximal over fronto-polar sites (for scalp distribution, F20.200 = 107.8, e = 0.25, P =0 .0001; see Figs. 3 and 4). An ANOVA examining the amplitude of the SFN for probes and mem- ory set items at each electrode site yielded a significant interaction between stimulus type and electrode (F20.200 = 19.60, • = 0.14, P = 0.0001). This interaction was due to the difference in scalp distribution of SFN elicited by memory set items and probes. Probes to last memory set items elicited reduced amplitude SFNs compared to other probes. Mean amplitude of the SFN at Fz were - 4 . 6 6 ___ 4.52, - 5 . 7 9 + 4.19, - 4 . 8 6 + 4.17, - 1.74 + 2.76, and - 4 . 1 7 + 3.81 /xV for positive probes to memory set items 1 -4 and for negative probes, respectively. Newman-Keu l s contrasts showed that while there were no difference be- tween the amplitude of the SFN elicited by positive probes to the first, second, and third memory set i tem and by negative probes, SFNs elicited by probes to the last mem- ory set item were significantly smaller in amplitude ( P < 0.05). In addition, the SFN was larger over right frontal electrode sites for all probes (at frontal electrode sites: Fpl , Fpz, Fp2, F3, Fz, F4, F7, F8, F7,70 = 2.89, • = 0.44, P = 0.05; see Figs. 3 and 6).

3.2.4. N4 component

The N4 component was quantified as a mean amplitude from 350 to 450 ms. All memory set items elicited an N4 component. An ANOVA examining the amplitude of the N4 elicited by memory set items over all electrode sites yielded a main effect of scalp distribution effect (F20.200 = 11.70, e = 0.16, P = 0.0001). This was due to the fact that N4s to memory set items were maximal at fronto-central sites. N4s were also observed for positive probes to mem- ory set items 1-3 and for negative probes. An ANOVA examining the amplitude of the N4 as a function of probe position (first, second, third, last, negative) and electrode site yielded a significant effect of scalp distribution (F20,200 = 98.61, • = 0.24, P = 0.0001), probe position (F4.20 = 9.73, • = 0.68, P = 0.0002) and a significant interaction between probe position and electrode (F80,800 = 2.41, • = 0.09, P = 0.0274). The scalp distribution effect reflects that the N4 component was most prominent over fronto- central sites. Because of the significant probe effect and probe by electrode interaction, we further examined the

32 LL. Chao, R.T. Knight/Cognitive Brain Research 4 (1996) 27-37

7.83

-7.15

llV

Fig. 6. Topographical map of the SFN (window 525-575 ms) generated to all correctly identified probes. Note that the SFN shows enhanced negativity over the right hemisphere (P = 0.05).

amplitude of N4 to each probe separately by means of Newman-Keuls comparisons. Theses comparisons re- vealed that the amplitude of the N4 elicited by positive probes to the first, second, and third memory set item were significantly different (P < 0.05) from the N4 elicited by probes to the last memory set item and by negative probes. The mean amplitude of the N4 was further examined at Cz alone and Newman-Keuls comparisons confirmed that positive probes to last position memory set items elicited significantly ( P < 0.05) smaller amplitude N4s while neg- ative probes elicited significantly larger amplitude N4s. The amplitude of the N4 at lateral electrodes were sub- jected to a separate ANOVA to evaluate scalp distribution, laterality, and probe position. While the ERPs were gener- ally more negative over the right than the left hemisphere, the main effect of laterality did not reach significance (Fl.10 = 2.95, P = 0.12). The effect of probe position on N4 amplitude was also not significantly asymmetric (for probe position X laterality: F4.40 = 1.42, P = 0.26).

4. Discussion

Serial position effects during non-linguistic auditory memory result in distinct behavioral and electrophysiologi- cal effects. Significant recency effects were observed for RTs, response accuracy and ERPs to probes items. Probes to the last memory set items were associated with larger amplitude, earlier latency P3s, reduced amplitude sustained

frontal negativity (SFN) and no N4 generation. Probes to memory set items 1-3 were associated with smaller ampli- tude, prolonged latency P3, larger amplitude SFNs and N4 generation. Memory set items elicited SFNs as well as an N4 component, but did not generate P3 components. These results indicate that distinct neural circuits dependent on the delay period are engaged during auditory memory.

4.1. Behavioral measures and serial position effect

In the current study, behavioral measures revealed a systematic decrease in RT as a function of probe recency (RTs to negative probes and 1st and 2nd positive probe > RTs to 3rd probe > RTs to last probe). This relationship between RT and recency is consistent with a self-terminat- ing serial search where the memory scan ceases as soon as a match is encountered. The trace strength model, in which familiarity or the strength of the memory trace influences the speed and accuracy of retrieval, has also been proposed as a method by which subjects perform memory scanning tasks [39]. The trace strength model, in conjunction with a self-terminating serial search model, appears to best ex- plain the current behavioral data. The relationship between RT and probe recency suggests that subjects scanned memory set items in reverse serial order. According to the trace strength model, the most recently presented memory set item would have the strongest memory trace, explain- ing why subjects respond fastest and most accurately to last position positive probes. Items presented at the begin-

L.L. Chat, R.T. Knight/Cognitive Brain Research 4 (1996) 27-37 33

ning of the memory set would have decreased memory traces by the time the probe is presented explaining the increase in error rates and RTs for probes to early memory set items.

4.2. ERP correlates of the serial position effect

4.2.1. P3 component The P3 ERP component is a positive potential with

peak latencies around 300-600 ms generated following detection of events that are relevant to the subject's task (i.e. the 'oddball' paradigm [66]). P3 amplitude has been shown to be sensitive to stimulus probability, subjective probability, stimulus meaning, and task relevance. Atten- tion allocation, information delivery, cognitive closure, stimulus categorization, context-updating, and template matching have been proposed to underlie P3 generation, although no consensus has been reached [9,76].

The current P3 findings support the trace strength model of memory processing which has received additional sup- port from behavioral measures in memory scanning tasks [20]. Subjects generated the largest amplitude, earliest latency P3 components to probes to last memory set items. 'Stronger' memory traces, accompanied by rapid stimulus classification times and increased accuracy have been asso- ciated with larger P3 responses [23]. Other investigators have proposed that subjects' confidence that a match has occurred is reflected in both P3 amplitude and latency [25,65]. Assuming that subjects scanned memory set items in reverse serial order and utilized a memory trace strength strategy to recognize the probes, the most recently pre- sented memory set item would have the strongest memory trace and these stimuli would be the easiest to recognize. Electrophysiologically, this is reflected by larger amplitude and earlier latency P3s. Memory trace strength for items presented earlier in the memory set and for items never presented would tend to be weaker and subjects may not be as confident in those responses. This may account for the smaller amplitude and prolonged latency of the P3s associated with positive probes in positions 1-3 and nega- tive probes.

The amplitude of P3 has also been associated with the amount of attention allocation to a task [9]. Using this schema of the P3, Patterson and colleagues [46], postulated that the more recently presented memory set items receive more attention and possibly 'deeper' processing during memory scanning. They further suggest that this deeper processing is the reason why last position probes were associated with larger amplitude P3 components.

Models that stress attention or 'depth of processing' strategies have been proposed as a possible method by which subjects recall items in a memory task [7,38]. Two kinds of rehearsal in the context of levels of processing have been distinguished [8]. The continual repetition of analyses that have already been carried out is known as type I or rote rehearsal. The principal function of this type

of rehearsal is to retain the availability of an item in memory and it typically does not lead to stronger or more permanent memories. Type II rehearsal, also known as elaborative rehearsal, refers to successively deeper process- ing of the stimulus and tends to result in more durable memories.

Since subjects were not instructed on the use of any particular strategies in the current study, they engaged in whatever processing they thought would best lead to suc- cessful remembering. Depending on their meta-memory skills, they could have, among other possibilities, semanti- cally labeled the sounds and repeated the labels to them- selves verbatim, attempt to organize the material into coherent structures, or form visual images. In a Sternberg- type task, where the presentation of the memory set item is rapid and the delay before the probe presentation is brief, shallow rote rehearsal seems more likely to occur. Even though subjects had enough time to semantically label the sounds, only 5 of the 11 subjects reported utilizing such a strategy for only some stimuli. In addition, the rapid rate of stimulus presentation probably did not allow the sub- jects to rehearse any labels in a deep or elaborate manor. It is more likely that subjects repeated the labels to them- selves in rote rehearsal manner. Last position memory set items probably receive the least amount of attention and rehearsal since the shortest amount of time elapsed be- tween presentation of the probe and the last memory set item.

A previous recognition memory study involving envi- ronmental auditory stimuli [6] yielded similar P3 effects as those found in the current study. The P3 was largest in amplitude and earliest in latency to the short delay repeti- tion stimuli. It was proposed that P3 generation indexed neocortical template-matching mechanisms that subjects used to identify stimuli repeated after a short interval without intervening distracting stimuli. That explanation can also account for P3 generation in the present study. The delay between the last memory set item and the probe is short (2.5 s) and free of distracting stimuli. The neocor- tical template-matching mechanism could be likened to a strategy of sensory matching of the probe to the memory trace. Stemberg [68] and Forrin and Cunningham [10] suggested that such effects could account for matching very recent items in a memory scanning task.

Although the P3 component has been proposed to be a reflection of memory encoding [9], temporal lobe struc- tures do not appear to be crucial to generation of the brunt of the scalp-recorded P3. Unilateral temporal lobectomies which included removal of anterior hippocampus, amyg- dala, and anterior temporal lobe [22], destruction of hip- pocampal CA1 cells due to hypoxia [48], and bilateral mesial temporal damage due to herpes simplex encephali- tis [43] do not significantly affect midline scalp-recorded P3b in auditory or visual tasks, although reductions at far lateral temporal sites can be observed [36].

Convergent evidence from scalp recordings in brain-le-

34 L.L. Chat, R.T. Knight/Cognitive Brain Research 4 (1996) 27-37

sioned patients and intracranial recordings supports multi- ple neocortical generators in frontal and posterior associa- tion cortex for the P3 component. Polarity inversions of potentials in the P3 time window have been recorded intracranially from the frontal and posterior cortex [63,77]. Large amplitude P3-1ike potentials have also been ob- served in the posterior neocortex of monkeys and human, particularly in the region of the inferior parietal lobe [44,63]. Reductions of the scalp-recorded P3 in patients with dorsolateral prefrontal and posterior association cor- tex lesions further supports multiple neocortical generators for this potential [26,78]. These data suggest that the template matching mechanism associated with generation of the scalp recorded posterior P3 component occurs pri- marily in the neocortex.

4.2.2. Sustained frontal negativity An SFN was generated to all memory set items and

probes. Slow wave activity has been widely reported to occur in tasks requiring sustained mental activity [56] and short-term memory tasks [31,49,50,55]. One property of these slow waves is that their amplitude increases with task demand. This effect has been observed in tasks such as mental rotation [21], mental arithmetic [56], and learn- ing verbal associations [30]. In a recent study, Ruchkin et al. [55] dissociated negative show waves associated with working memory from slow waves related to general preparatory operations.

The SFN recorded in the current study may be related to the early contingent negative variation (CNV). Numer- ous investigators have argued that the CNV is not a unitary phenomenon, but rather consists of two general categories of waves, early and late [33,52,71]. The early wave, or the 'O wave' originates at a latency of 300-350 ms and peaks at 500-600 ms after the warning signal in a typical CNV paradigm and has a frontal distribution. Tecce [71] sug- gested that the early CNV is related to 'arousal processes' while Loveless and Sanford [33] presumed it to be a component of the orienting response.

There are numerous instances in which the negative waves have been elicit in non-CNV situations [29,32,53]. During a signal detection task, Rohrbaugh et al. [53] recorded a frontal negative afterwave that was small under conditions of passive observation and enhanced when the stimuli provided information and required attention. The amplitude of the slow negative wave has also been linked with task difficulty. In a pitch discrimination task, the amplitude of the negative slow wave was larger for the more difficult discriminations [29].

In the current experiment, the SFN was larger in ampli- tude to the more difficult stimuli, negative probes and positive probes to memory set items 1-3, which were associated with higher error rates. Probes to the last mem- ory set items, associated with the shortest RTs and lowest error rates, elicited significantly smaller amplitude SFNs than other probes. The neocortical template-matching hy-

potheses fits this data well. If subjects utilized a trace strength/template-matching strategy to recognize and identify last position probes, they would have exerted less effort, and thus would be predicted to generate a reduced amplitude SFN. Memory set items also elicited SFNs. This may relate to Rohrbaugh's [53] finding that stimuli which provide information and require attention generate a frontally distributed negative component.

The frontal cortex has long been considered a possible generator of the CNV both because the CNV has a fronto- central scalp distribution and since this evoked potential occurs in situations involving putative 'frontal lobe' func- tions [34]. Lesions in the dorsolateral prefrontal cortex centered in Brodmann areas 9, 45, and 46 reduce the later phases of the CNV over the lesioned cortex and at distant scalp sites over the lesioned hemisphere [54]. These results are consistent with a late CNV generator in the dorsolat- eral prefrontal cortex which also modulates generation of the potential throughout the ipsilateral hemisphere. Given the similarities between the CNV and the SFN, it is possible that the dorsolateral prefrontal cortex may con- tribute to generation of the SFN as well. In the current study, the SFN was larger over the right frontal lobe in accord with PET reports of right prefrontal activation in tasks of sustained attention [24,45].

4.2.3. N4 component Subjects generated a negative wave around 400 ms to

memory set items, negative probes, and positive probes to the first 3 memory set items. This negativity could be related to the N400, or N4 first described by Kutas and Hillyard [27]. Theories concerning the cognitive basis of the N4 have addressed whether modulation of the compo- nent reflects retrieval of information from long-term mem- ory [28] and whether the N4 is associated with purely linguistic or more general semantic processing. Deviant endings to melodies [5] do not elicit the N4 component; however, other non-verbal stimuli such as faces [2], pic- tures [11], and environmental noises [6,75] have been reported to elicit the N4.

The N4 component has been recorded during recogni- tion memory experiments [ 11,16]. One study reported that N4 amplitude increased with increases in the number of pictures to be named [70]. The authors proposed a role for the N4 in memory search. Intracranially, an N4-1ike poten- tial sensitive to stimulus repetition has been recorded in the medial temporal lobe [64]. Based on the proposed role of the medial temporal lobe in formation and retrieval of recent memory, the authors suggested that the medial temporal lobe-N4 may be involved with memory forma- tion and retrieval processes.

Subjects generated large amplitude negativity peaking at 400 ms to negative probes. This negativity was similar to the N400 which has been reported in ERPs to mismatch- ing items [18]. It is interesting to note that in this study, the N400 was elicited in a task that could have been per-

L.L. Chao, R.T. Knight/Cognitive Brain Research 4 (1996) 27-37 35

formed without access to word meaning. Subjects did not generate an N4 component to positive probes that matched the last memory set item. This may relate to the reductions in N4 observed with repetition in lists [3,57], sentence final position [4], and text [74]. It has been proposed that reductions in N4 amplitude with stimulus repetition may be associated with reductions in the explicit memory de- mands of the task [61]. At short delays (2.5 s), the pro- posed neocortical template matching mechanism may be sufficient for subjects to recognize the repeated stimuli and explicit memory search is not needed. Positive probes to memory set items 1-3 elicited N4s. While some investiga- tors have reported that the number of intervening items does not affect the repetition effect on the amplitude of the N4 [58], the current results are similar to the results of recognition memory studies involving environmental sounds [6] and line drawings [41]. In these studies, the N4 was generated in conditions that involved a long delay interval filled with intervening items. After a delay of more than 4 s, the neocortical template (or strength of the memory trace) may not be sufficient for stimulus recogni- tion and subjects need to engage additional memory search mechanisms. A few subjects reported utilizing semantic labels to remember some of the stimuli; thus it is possible that N4 generated may be due, in part, to memory compar- isons of semantic encodings of the stimuli.

Converging evidence for the anatomic locus of neural activity underlying generation of N4 potentials has been produced by studies of surgical resection patients, intracra- nial recordings, fluoro-deoxyglucose (FDG) and positron emission tomography (PET). Scalp-recorded N4 subcom- ponents have widespread midline-central and midline- posterior distributions consistent with a deep neural gener- ator [42]. Observations that unilateral anterior temporal lobectomy [59,62] attenuates the N4 component suggests that the temporal lobe plays an important role in its generation. Intracranial ERPs recorded from medial tempo- ral structures support this conclusion. These intracranial- recorded ERPs have been reported to be sensitive to the congruity of sentence endings [35], to lexical and repetition effects with single word presentations [64], and to verbal and visuospatial recognition memory tasks [14,51]. The large amplitude of the depth N4, the steep voltage gradi- ent, and local polarity reversals within the medial temporal structures [17,35,51,64], together with correlated unit activ- ity [19] suggest that these components were locally gener- ated. In addition, there is some evidence to suggest that the parietal lobe may contribute to N4 generation. Steep volt- age gradients and local polarity reversals of the N4 have been recorded within the superior and inferior parts of the parietal lobe [14]. This is consistent with FDG/PET stud- ies of cerebral metabolism which have shown a correlation between N4 amplitude and metabolism activation in the angular gyrus [40]. Taken together, these findings suggest that medial temporal structures (amygdala, hippocampus, parahippocampal gyrus, and possibly enthorhinal cortex)

and parietal areas may be involved in the generation and modulation of the scalp-N4 [14].

Afferents from posterior association cortical areas con- verge on medial temporal lobe structures both directly and via bi-directional limbic-neocortical connections [60,73]. Based on this evidence, Smith et al. [64] proposed that the medial temporal lobe-N4 may represent excitation of hip- pocampal synapses in response to convergent activation from these projections. If the medial temporal lobe-N4 did result from convergent activation of hippocampal synapses, such activation would be expected to result in long-lasting enhancement of their efficacy [37]. Multiple feedback loops have been reported to relay information from the medial temporal lobe back to association cortex [72]. Long-term enhancement of hippocampal synapses would likely result in changes in this feedback. Neural models of recent memory [15] predict that medial temporal lobe feedback, in response to presentation of an item for memory recogni- tion, invokes episodic representation of that item, produc- ing a feeling of familiarity for it. In the current study, negative probes and positive probes to early memory set items may have to elicit the N4 component because after a long delay, the neocortical template is not sufficient for stimulus recognition and medial temporal lobe feedback loops need to be engaged.

In summary, the current experiment found behavioral and electrophysiological evidence of serial position effects in an auditory memory scanning task supporting a trace strength/self-terminating search process. The ERP results indicate that distinct neural circuits dependent on the delay period are engaged during the processing of non-linguistic auditory stimuli.

Acknowledgements

This work was supported by the Veterans Administra- tion Research Service, Javits Award NS21135, PO NS17778 from the NINDS, and predoctoral award F31 MH10958-01 from the NIMH.

References

[1] Baddeley, A.D. and Ecob, J.R., Reaction time and short-term mem- ory: implications of repetition effects for the high speed exhaustive scan hypothesis, Q. J. Exp. Psychol., 25 (1973) 229-240.

[2] Barrett, S.E., Rugg, M.D. and Perrett, D.I.. Event-related potentials and the matching of familiar and unfamiliar faces, Neuropsychol- ogy, 26 (1988) 105-117.

[3] Bentin, S. and Feldman, L.B., The contribution of morphological and semantic relatedness to repetition priming at short and long lags: evidence from Hebrew, Q. J. Exp. Psychol. [A], 42 (1990)693-711.

[4] Besson, M., Kutas, M. and Van Petten, C., An event-related poten- tial (ERP) analysis of incongruity semantic congruity and repetition effects in sentences, J. Cogn. Neurosci., 4 (1992) 132-149.

[5] Besson, M. and Macar, F., An event-related potential analysis of incongruity in music and other non-linguistic contexts, Psychophysi- ology, 24 (1987) 14-25.

36 L.L. Chao, R.T. Knight/Cognitive Brain Research 4 (1996) 27-37

[6] Chao, L.L., Nielsen-Bohlman, L. and Knight, R.T., Auditory event- related potentials dissociate early and late memory processes, Elec- troencephogr. Clin. Neurophysiol., 96 (1995) 157-168.

[7] Craik, F.I.M., The fate of primary memory items in free recall, J. Verb. Learn. Verb. Behav., 9 (1970) 143-148.

[8] Craik, F.1.M. and Lockhart, R.S., Levels of processing: A frame- work for memory research, J. Verb. Learn. Verb. Behav., 11 (1972) 671-684.

[9] Donchin, E. and Coles, M.G.H., Is the P300 component a manifesta- tion of context updating? Behav. Brain Sci., 11 (1988) 357-374.

[10] Forrin, B. and Cunningham, K., Recognition time and serial position of probe item in short-term memory, J. Exp. PsychoL, 99 (1973) 272-279.

[11] Friedman, D., Cognitive event-related potentials during continuous recognition memory for pictures, Psychophysiology, 27 (1990) 136- 148.

[12] Goldman, P.S. and Rosvold, H.E., Localization of function within the dorsolateral prefrontal cortex of the rhesus monkey, Exp. Neu- rol., 27 (1970) 291-304.

[13] Goldman-Rakic, P.S., Bates, J.F. and Chefee, M.V., The prefrontal cortex and internally generated motor acts, Curr. Opin. Neurobiol., 2 (1992) 830-835.

[14] Guillem, F., N'Kaoua, B., Rougier, A. and Claverie, B., Intracranial topography of event-related potentials (N400/P600) elicited during a continuous recognition memory task, Psychophysiology, 32 (1995) 382-392.

[15] Halgren, E., Human hippocampal and amygdala recording and stim- ulation: evidence for a neural model of recent memory. In: L.R. Squire and N. Butters (Eds.), The Neuropsychology of Memory, Guilford, New York, 1984, pp. 165-181.

[16] Halgren, E. and Smith, M.E., Cognitive evoked potentials as modu- latory processes in human memory formation and retrieval, Hum. Neurobiol., 6 (1987) 129-139.

[17] Halgren, E., Smith, M.E. and Stapleton, J.M., Hippocampal field- potentials evoked by repeated vs. nonrepeated words. In: G. Buzsaki and C.H. Vanderwold (Eds.), Electrical Activity of the Archicortex, Hungarian Academy of Sciences, Budapest, 1984.

[18] Harbin, T., Marsh, G. and Harvey, M., Differences in the late components of the event-related potential due to age and to semantic and non-semantic tasks, Electroencephalogr. Clin. Neurophysiol., 59 (1984) 489-496.

[19] Heit, G., Smith, M.E. and Halgren, E., Neuronal activity in the human medial temporal lobe during recognition memory, Brain, 113 (1990) 1093-1112.

[20] Hockley, W.E. and Murdock, B.B., A decision model for accuracy and response latency in recognition memory, Psychol. Rev., 94 (1987) 341-358.

[21] Johnson, R., Jr., Cox, C. and Fedio, P., Event-related potential evidence for individual differences in a mental rotation task. In: R. Johnson, Jr., J.W. Rohrbaugh and R. Parasuraman (Eds.), Current Trends in Event-Related Potential Research. Electroenceph. Clin. Neurophysiol. Suppl., Elsevier, Amsterdam, 1987, pp. 191-197.

[22] Johnson, R., Jr., Scalp-recorded P300 activity in patients following unilateral temporal lobectomy, Brain, 111 (1988) 1517-1529.

[23] Johnson, R., Jr., Pfefferbaum, A. and Kopell, B.S., P300 and long-term memory latency predicts recognition performance, Psy- chophysiology, 22 (1985) 497-507.

[24] Jonides, J., Smith, E.E., Koeppe, R.A., Awh, E., Minoshima, S. and Minturi, M.A., Spatial working memory in humans as revealed by PET, Nature, 363 (1993) 623-625.

[25] Karis, D., Fabiani, M. and Donchin, E., 'P300' and memory: individual differences in the von Restorff effect, Cognitive Psychol., 16 (1984) 177-216.

[26] Knight, R.T., Scabini, D., Woods, D.L. and Clayworth, C.C., Con- tributions of temporal-parietal junction to the human auditory P3, Brain Res., 502 (1989) 109-116.

[27] Kutas, M. and Hitlyard, S.A., Reading senseless sentences: brain potentials reflect semantic incongruity, Science, 207 (1980) 203-205.

[28] Kutas, M. and Van Petten, C., Event-related brain potential studies of language. In: P.K. AcMes, J.R. Jennings and M.G.H. Coles (Eds.), Advances in Psychophysiology, JAI Press, Greenwich, CT, 1988, pp. 139-187.

[29] Lang, P.J., Gatchel, R.J. and Simons, R.T., Electrocortical and cardiac rate correlates of psychophysical judgment, Psychophysiol- ogy, 12 (1975) 649-655.

[30] Lang, W., Lang, M., Uhl, F., Kornhuber, A., Deecke, L. and Kornhuber, H.H., Left frontal lobe in verbal associative learning: a slow potential study, Exp. Brain Res., 70 (1988) 99-108.

[31] Lang, W., Starr, A., Lang, V., Lindiger, G. and Deecke, L., Cortical DC potential shifts accompanying auditory and visual short-term memory, Electroencephalogr. Clin. Neurophysiol., 82 (1992) 285- 295.

[32] Loveless, N.E., Distribution of response to non-signal stimuli. In: W.C. McCallum and J.R. Knott (Eds.), The Responsive Brain, Wright, Bristol, 1976, pp, 26-29.

[33] Loveless, N.E. and Sanford, A.J., Effects of age on the contingent negative variation and preparatory set in a reaction-time task, J. GerontoL, 29 (1974) 52-63.

[34] Low, M.D., Cerebral slow potentials in neurological disorders. In: W.C. McCallum, R. Zappoli and F. Denoth (Eds.), Cerebral Psy- chophysiology: Studies in Event-Related Potentials, Elsevier, Ams- terdam, 1986, pp. 477-481.

[35] McCarthy, G. and Wood, C.C., Scalp distributions of event-related potentials: an ambiguity associated with analysis of variance models, Electroencephalogr. Clin. Neurophysiol., 62 (1985) 203-208.

[36] McCarthy, G., Wood, C.C., Williamson, P.D. and Spencer, D.D., Task-dependent field potentials in human hippocampal formation, J. Neurosci., 9 (1989) 4253-4268.

[37] McNaughton, B.L., Activity dependent modulation of hippocampal synaptic efficacy: some implications for memory processes. In: W. Seifert (Ed.), The Neurobiology of the Hippocampus, Academic Press, London, 1983, pp. 233-252.

[38] Morris, P.E., Is output order in free recall based on the strength of the memory trace or on the subject strategies and depth of coding employed? Br. J. Psychol., 68 (1977) 447-455.

[39] Murdock, B.B., Jr. and Okada, R., Interresponse times in single-trial free recall, J. Exp. Psychol., 86 (1970) 263-267.

[40] Nenov, V.I., Halgren, E., Smith, M.E., Bather, J.M., Ropchan, J., Bland, W.H. and Mandelkem, M., Localized brain metabolic re- sponse correlated with potentials evoked by words, Behav. Brain Res., 44 (1991) 101-104.

[41] Nielsen-Bohlman, L. and Knight, R.T., Event-related potentials dis- sociate immediate and delayed memory. In: H.J. Heinze, T.F. Munte and G.R. Mangun (Eds.), Cognitive Electrophysiology: Basic and Clinical Research, Birkhauser, Boston, 1994.

[42] Nobre, A.C. and McCarthy, G., Language-related ERPs: scalp distri- butions and modulation by word-type and semantic priming, J. Cognitive Neurosci., 6 (1994) 233-255.

[43] Onofrj, M., Fulgente, T., Nobilio, D., Malatesta, G., Bazzano, S., Colamartion, P. and Gambi, D., P3 recordings in patients with bilateral temporal lobe lesions, Neurology, 42 (1992) 1762-1767.

[44] Paller, K.A. and Kutas, M., Brain potentials during memory retrieval provide neurophysiological support for the distinction between con- scious recollection and priming, J. Cognitive Neurosci., 4 (1992) 375-391.

[45] Pardo, J.V., Fox, P.T. and Raichel, M.E., Localization of a human system for sustained attention by positron emission tomography, Nature, 349 (1991) 61-64.

[46] Patterson, J.V., Pratt, H. and Starr, A., Event-related potential correlates of the serial position effect in short-term memory, Elec- troencephalogr. Clin. Neurophysiol., 78 (1991) 424-437.

[47] Pfefferbaum, A., Ford, J.M., Roth, W.T. and Kopell, B.S., Age-re-

L.L. Chao, R.T. Knight/Cognitice Brain Research 4 (1996) 27-37 37

lated changes in auditory event-related potentials, Electroen- cephalogr. Clin. Neurophysiol., 49 (1980) 266-276.

[48] Polich, J. and Squire, L.R., P300 from amnesic patients with bilat- eral hippocampal lesions, Electroencephalogr. Clin. Neurophysiol., 86 (1993) 408-417.

[49] Pratt, H., Michalewski, H.J., Patterson, J.V. and Starr, A., Brain potentials in a memory scanning task. I. Modality and task effects on potentials to the probes, Electroencephalogr. Clin. Neurophysiol., 72 (1989) 407-421.

[50] Pratt, H., Michalewski, H.J., Patterson, J.V. and Stair, A., Brain potentials in a memory scanning task. II. Effects of aging on potentials to the probe, Electroencephalogr. Clin. Neurophysiol., 72 (1989) 507-517.

[51] Puce, A., Andrews, D.G., Berkovic, S.F. and Bladin, P.F., Visual recognition memory: neurophysiological evidence for the role of temporal white matter in man, Electroencephalogr. Clin. Neuro- physiol., 114 (1991) 1647-1666.

[52] Rohrbaugh, J.W. and Gaillard, A.W.K., Sensory and motor aspects of the contingent negative variation. In: A.W.K. Gaillard and W. Ritter (Eds.), Tutorials in ERP Research: Endogenous Components, Elsevier, Amsterdam, 1983, pp. 269-310.

[53] Rohrbaugh, J.W., Syndulko, K. and Lindsley, D.B., Cortical slow waves following non-paired stimuli: effects of task factors, Elec- troencephalogr. Clin. Neurophysiol., 45 (1978) 551-567.

[54] Rosahl, S.K. and Knight, R.T., Role of prefrontal cortex in genera- tion of the contingent negative variation, Cerebral Cortex, 2 (1995) 123-134.

[55] Ruchkin, D.S., Canoune, H.L., Johnson, R., Jr. and Ritter, W., Working memory and preparation elicit different patterns of slow wave event-related brain potentials, Psychophysiology, 32 (1995) 399-410.

[56] Ruchkin, D.S., Johnson, Jr., R., Mahaffey, D. and Sutton, S., Toward a functional categorization of slow waves, Psychophysiol- ogy, 25 (1988) 339-353.

[57] Rugg, M.D., The effects of semantic priming and word repetition on event-related potentials, Psychophysiology, 22 (1985) 642-647.

[58] Rugg, M.D. and Nagy, M.E., Event-related potentials and recogni- tion memory for words, Electroencephalogr. Clin. Neurophysiol., 72 (1989) 395-406.

[59] Rugg, M.D., Roberts, R.C., Potter, D.D., Pickles, C.D. and Nagy, M.E., Event-related potentials related to recognition memory: effects of unilateral temporal lobectomy and temporal lobe epilepsy, Brain, 114 (1991) 2313-2332.

[60] Seltzer, B. and Pandya, D.N., Some cortical projection to the parahippocampal area in the rhesus monkey, Exp. Neurol., 50 (1976) 146-160.

[61] Smith, M.E. and Guster, K., Decomposition of recognition memory event-related potentials yields target, repetition, and retrieval effects, Electroencephalogr. Clin. Neurophysiol., 86 (1993)335-343.

[62] Smith, M.E. and Halgren, E., Attenuation of a sustained visual processing negativity after lesions that include the inferotemporal cortex, Electroencephalogr. Clin. Neurophysiol., 70 (1988) 366-369.

[63] Smith, M.E., Halgren, E., Sokolik, M., Baudena, P., Musolino, A., Liegeois-Chauvel, C. and Chauvel, P., The intracranial topography of the P3 event-related potential elicited during auditory oddball, Electroencephalogr Clin. Neurophysiol., 76 (1990) 235-248.

[64] Smith, M.E., Stapleton, J.M. and Halgren, E., Human medial tempo- ral lobe potentials evoked in memory and language tasks, Electroen- cephaolgr. Clin. Neurophysiol., 63 (1986) 145-159.

[65] Squires, K.C., Squires, N.K. and Hillyard, S.A., Decision-related cortical potentials during an auditory signal detection task with cued observation intervals, J. Exp. Psychol. Hum. Percept., 1 (1975) 268-279.

[66] Squires, N.K., Squires, K.C. and Hillyard, S.A., Two varieties of long-latency positive waves evoked by unpredictable auditory stim- uli in man, Electroencephalogr. Clin. Neurophysiol., 38 (1975) 387-401.

[67] Starr, A. and Barrett, G., Disordered auditory short-term memory in man and event-related potentials, Brain, 110 (1987) 935-959.

[68] Sternberg, S., High-speed scanning in human memory, Science, 153 (1966) 652-654.

[69] Sternberg, S., Memory scanning: new findings and current contro- versies, Q. J. Exp. Psychol., 27 (1975) 1-32.

[70] Stuss, D.T., Picton, T.W. and Cerri, A.M., Searching for the names of pictures: an event-related potential study, Psychophysiology, 23 (1986) 215-223.

[71] Tecce, J.J., Contingent negative variation (CNV) and psychophysio- logical processes in man, Psychol. Bull., 77 (1972) 73-108.

[72] Van Hoesen, G.W., The parahippocampal gyrus: new observations regarding its cortical connections in the monkey, Trends Neurosci., 5 (1982) 345-350.

[73] Van Hoesen, G.W., Rosene, D.L. and Mesulam, M.M., Subicular input from temporal cortex in the rhesus monkey, Science, 205 (1979) 608-610.

[74] Van Petten, C., Kutas, M., Kluender, R., Mitchiner, M. and Mclsaac, H., Fractionating the word repetition effect with event-related poten- tials, J. Cognitive Neurosci., 3 (1991) 131-150.

[75] Van Petten, C. and Rheinfelder, H., Conceptual relationships be- tween spoken words and environmental sounds: event-related brain potential measures, Neuropsychologia, 33 (1995) 485-508.

[76] Verleger, R., Event-related potentials and cognition: a critique of the context updating hypothesis and an alternative interpretation of P3, Behav. Brain Sci., 11 (1988) 343-427.

[77] Wood, C.C. and McCarthy, G., A possible frontal lobe contribution to scalp P3, Soc. Neurosci. Abstr., 11 (1985) 879.

[78] Yamaguchi, S. and Knight, R.T., Gating of somatosensory input by human prefrontal cortex, Brain Res., 52l (1990) 281-288.