Copyright 1992 by The Geronlutogical Society of America
Journal of Gerontology: PSYCHOLOGICAL SCIENCES 1992, Vol. 47, No. 6, P395-P4O5
Event-Related Potential Correlates of Repetition Priming and Stimulus Classification in Young, Middle-aged, and Older Adults Maria Hamberger and David Friedman New York State Psychiatric Institute.
F
ORMAL investigations of memory have typically involved "explicit" or direct measures of memory (e.g., free recall, recognition) in which subjects are fully cognizant of their participation in a memory task, and successful performance requires deliberate recollection or recognition of a previous event. In contrast, performance on "implicit" or indirect memory tests does not require deliberate recollection. Tasks are designed to divert attention away from the memory component of the task and, ideally, to prevent subjects from becoming aware that they are undergoing memory evaluation. Although the particular task may vary (e.g., lexical decision, word-stem completion), the common element across tasks is that items are repeated at some point during the course of testing. It is taken as evidence that retention has occurred when performance enhancement (e.g., faster response time and/or increased accuracy) is associated with item repetition, a phenomenon labeled "repetition priming." Implicit Memory and Aging Although there is substantial evidence that performance on explicit memory tasks declines with increasing age (Burke & Light, 1981; Madden, 1986), investigations of implicit memory and aging have generated equivocal findings. For example, on word-stem completion (Light & Singh, 1987), lexical decision (Moscovitch, 1982), picture naming (Mitchell, 1989; Mitchell, Brown, & Murphy, 1990), and biased homophone spelling (Rose, Yesavage, Hill, & Bower, 1986), response facilitation for previously presented items is of comparable magnitude in young and older adults. In contrast, age-related decline has been suggested by results from studies utilizing prolonged study-test intervals (Rabbitt, 1982), constrained study time (Howard,
Fry, & Brune, 1991), the priming of newly learned associates (Howard, McAndrews, & Lasaga, 1986; Howard et al., 1991), and larger subject samples (Chiarello & Hoyer, 1988; Hultch, Masson, & Small, 1991). To account for these discrepancies, some investigators (Chiarello & Hoyer, 1988; Graf, 1990) have suggested that older adults appear relatively impaired on implicit memory tasks because younger adults use intentional memory strategies in performing these tasks. Alternatively, since Dunn and Kirsner (1989) have suggested that different implicit memory tasks tap different processess, the possibility exists that such processes may be differentially affected by age. More specifically, Howard (1988) has asserted that implicit "item" memory requires memory for an item already represented in long-term memory, whereas implicit "associative" memory involves memory for a new association between previously unrelated items, which requires both greater attention and active processing in order to establish a novel connection between items. According to Howard et al. (1991), elderly adults are relatively impaired on implicit memory tasks that require this more elaborate processing. Although there are numerous reports in the cognitive (Jacoby & Dallas, 1981), aging (Graf, 1990), and neuoropsychological literatures (Schachter & Graf, 1986) demonstrating differential performance of individuals on implicit vs explicit memory tasks, it is currently unknown whether these dissociations are attributable to unique anatomical brain systems and/ or functionally distinct cognitive systems. Although much research has been directed toward the resolution of these questions, this experiment used only implicit memory tasks and was not designed therefore to address this issue. The purpose of this investigation was to examine implicit memory performance across the age span as assessed by item repetition P395
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
Electrophysiological and behavioral measures were concurrently recorded from young, middle-aged, and elderly subjects during two classification tasks that required different levels of processing. In both tasks, a portion of stimuli repeated after lags of 2, 8, and 32 intervening items following their first presentation, although these repetitions were incidental to the primary task. All three age groups showed similar patterns of behavioral and event-related potential (ERP) responses associated with stimulus classification and item repetition. ERPs to repeated words were characterized by greater positive amplitude relative to words at first presentation. Similarly, ERPs elicited by items in a primed category were more positive compared with unprimed items. Reaction times were faster to both repeated and primed items. Behavioral and ERP effects were more marked when subjects were required to discriminate items based on meaning rather than on orthography. This study supports previous behavioral evidence suggesting that processes associated with repetition priming and stimulus classification are preserved with age, and demonstrates that ERPs are responsive to repetition throughout the life cycle. Implications for the cognitive processes underlying repetition effects are also discussed.
P396
HAMBERGER AND FRIEDMAN
during two types of categorization tasks. The notion that implicit item memory does not change with increasing age is based on behavioral data. In addition to behavioral measures, we examined electrophysiological responses in three age groups during implicit item memory tasks.
METHODS
Subjects. — Three groups of right-handed, female, paid volunteers participated in this study. Subjects were recruited via posted notices at the medical center, and additional elderly volunteers were recruited from senior centers in the vicinity of the medical center. Demographic information for each age group is provided in Table 1. As can be seen, all subjects were in the normal range for both IQ (score of 90 or higher on the Quick Test; Ammons & Ammons, 1962), and a modified version of the Mini-Mental State Exam (a score of 50 or higher; Mayeux, Stern, Rosen, &Leventhal, 1981). The groups did not differ on the Mini-Mental State Exam [F(2/48) = 2.58, p> .05], the Quick Test [F(2/48) = 3.15, p > .05], or the laterality quotient [F(2/48) = 2.54, p > .05]. However, the youngest group had more years of education [F(2/48) = 5.34, p < .008] and higher socioeconomic status levels [F(2/48) = 3.36, p < .04] than the two older age groups. Elderly subjects were free of dementia, depression, and limitation in the activities of daily living, according to criteria of the Short CARE (Gurland, Golden, Teresi, & Challop, 1984), a structured interview standard-
Table 1. Demographic Characteristics of the Three Samples ERPs, Repetition Priming, and Aging To date, investigators of age effects during repetition priming have only used behavioral measures. Given the reliability of behavioral and ERP repetition effects in young adults, together with the somewhat variable repetition priming effects in the elderly, we sought to examine repetition priming using ERPs in conjunction with behavioral measures in an elderly population. Because a variety of agerelated neuronal changes (Terry, DeTeresa, & Hansen, 1987) may underlie differential performance on implicit memory tasks, we recorded a physiological measure that is intimately linked with the brain's processing of cognitive information. Thus, processing differences that may have been compensated for or attentuated by the time a behavioral response was executed might be revealed during on-line recording of brain activity. For instance, as noted above, because ERP repetition effects appear to be more sensitive to depth of processing manipulations compared with behav-
Age
Education
IQ
LQ» SES" MMSC DEPd DEMe
Young Adults (N == 18) Mean SD
24.94 113.50 2.96 12.50
17.44 2.15
93.22 47.94 55.56 15.73 15.32 1.65
NA
NA
NA
NA
1.56 1.33
.17 .38
Middle-aged Adults (W = 15) Mean SD
48.86 110.40 9.35 4.01
14.87 3.22
83.33 63.13 54.40 25.66 14.10 1.76
Older Adults (N =: 18) Mean SD
70.11 119.61 3.89 10.11
14.39 3.48
69.94 59.56 54.17 44.03 22.42 2.33
NA, not applicable; Education, number of years of education. "Laterality quotient (100 = completely right-handed), assessed by the Edinburgh Inventory (Oldfield, 1971). b Socioeconomic status (higher score = lower SES). 'Modified Mini-Mental State Exam (highest score is 57). d CARE depression score (cutoff is 6). e CARE dementia score (cutoff is 7).
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
Event-Related Potentials (ERPs) and Repetition Priming Scalp-recorded ERPs provide continuous, noninvasive monitoring of brain activity while subjects are actively engaged in a cognitive task. Unlike traditional behavioral measures, ERP recording provides a temporal readout of the brain's activity during information processing that is not necessarily contingent upon overt responding or conscious awareness. Hence, in addition to behavioral measures, the concurrent recording of ERP activity provides valuable complementary information regarding the nature, sequence, and timing of perceptual and cognitive events. Recent work has demonstrated that, in addition to behavioral measures, ERP indices are also sensitive to item repetition during implicit memory tasks. For example, in a paradigm in which subjects responded to occasional nonwords, ERPs to repeated words were characterized by greater positive amplitude from approximately 300-600 msec poststimulus relative to their first presentation counterparts (Rugg, 1985). Because superficial discrimination tasks (e.g., upper case vs lower case letters) showed small, nonsignificant ERP repetition effects compared with those recorded during lexical or semantic discrimination tasks (e.g., animal words vs nonanimal words), Rugg, Furda, and Lorist (1988) proposed that (unlike behavioral repetition effects) the magnitude of the ERP repetition effect is influenced by the "level of processing" induced by task demands (Craik & Lockhart, 1972). Rugg et al. (1988) attributed the ERP repetition effect (i.e., greater positivity to the repeated item) to an attenuation in a negativity elicited by the item's second presentation, and suggested that this negativity reflects "the consequences of reaccessing memory" and the integration of an item with its context. When an item is repeated, these processes occur to a lesser extent, which coincides with reduced negativity in the ERP elicited by the repeated item.
ioral measures (Rugg et al., 1988), ERPs in an elderly sample might be more responsive to variables such as the time course of repetition that do not ordinarily affect behavioral indices of repetition priming in either young (Dannenbring & Briand, 1982) or older adults (Moscovitch, 1982). The principal aim of this investigation was to determine whether the ERP data would mirror the findings reported in the behavioral aging literature on implicit item memory. We manipulated depth of processing during two stimulus classification tasks and the time course of item repetition by task demands and interitem lag, respectively. Similar to Rugg et al. (1988), although begun without knowledge of their study, the current orthographic task required a relatively superficial discrimination (uppercase vs lowercase words), whereas the semantic task required a "deeper" level of processing (animal vs nonanimal words).
ERPs, PRIMING, AND AGING
ized on large samples of community-dwelling elderly subjects. All subjects reported themselves to be in good physical health and did not report problems with their memory.
Electroencephalogram (EEG) recording procedures. — EEG was recorded using an electrocap from midline elec-
trodes at Fz, Cz, Pz, with lateral chains at F3, C3, P3, T3, and 01 on the left, and homologous locations on the scalp overlying the right hemisphere. The vertical electrooculogram (EOG) was recorded from an electrode located on the supraorbital ridge of the right eye, and the horizontal EOG was recorded from electrodes placed at the outer canthi of both eyes (bipolar recording). Except for the horizontal EOG electrodes, all leads were referred to nosetip. Due to a computer error, the data at the 01 electrode site were not usable for 20 of the 51 subjects. Therefore, the 01 data were excluded from all analyses. EEG and EOG were recorded with a bandpass of .01 to 30 Hz (5.3 sec time constant), with a Grass Neurodata system. Data acquisition and stimulus presentation were controlled by a PDP 11/34 computer, which digitized the EEG and EOG at 10-msec intervals for a 300 msec pre- and a 1700 msec poststimulus period separately for each word presentation, and stored the digitized records, choice responses, and RTs on 9-track digital tape. EEG was corrected for intrusion of EOG artifact using linear regression procedures based on adaptations of the methods described by Gratton, Coles, and Donchin (1983) and Verleger, Gasser, and Mocks (1982). On trials during which blink artifact was detected by computer algorithm, blink and vertical EOG artifacts were corrected first. In a second pass, horizontal EOG artifact was corrected following removal of vertical artifact from the horizontal EOG channel. In our implementation, the vertical and horizontal transmission coefficients were computed separately for each trial. Prior to computing averaged voltage measures of ERP activity, the individual averages were digitally filtered off-line at 15 Hz using the low-pass filter described by Ruchkin and Glaser (1978). ANOVAs. — ANOVAs, with Greenhouse-Geisser correction (Jennings and Wood, 1976), where appropriate, of ERP and behavioral data were performed using the BMDP4V (Dixon, 1987) program. Corrected degrees of freedom were truncated to the nearest integer. Newman-Keuls post-hoc tests were used to make pairwise comparisons when three or more conditions were contrasted. When significant interactions occurred, simple effects procedures (Winer, 1971), with reduced degrees of freedom where appropriate, were applied to explore the interactions further. RESULTS
The variables of interest were Age Group (young, middle, old), Task (orthographic vs semantic), Stimulus Type (animal vs nonanimal words in semantic blocks; uppercase vs lowercase words in orthographic blocks), Repetition (new vs old), Interitem Lag (2, 8, 32), and Electrode Site. Due to the large number of analyses, a criterion was adopted in which only F-ratios reaching alpha levels of p < .01 (after correction of degrees of freedom) were considered significant. ERP Waveforms Grand-mean ERP waveforms for Repetition are shown in Figure 1, and the Repetition and Stimulus Type ERPs of a representative subject in each age group from the midline electrodes are shown in Figure 2. As can be seen in Figure 1, the ERPs were characterized by an early negativity at about 200 msec, followed by a second negative deflection peaking
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
Stimuli and procedures. — Two orthographic (uppercase vs lowercase words), two semantic (animal vs nonanimal words), hereafter referred to as elaboration blocks, and four continuous recognition memory blocks were presented during one morning session that lasted approximately 1.5 hours. Only data from the elaboration blocks will be reported here. The continuous recognition memory data have been preliminarily reported (Friedman & Hamberger, 1989), with a full report currently in preparation. The continuous recognition procedure was administered two blocks at a time, whereas the semantic and orthographic blocks were administered back to back, one block at a time. The three kinds of blocks were counterbalanced across subjects, such that the orderings were identical for subject 1 in all three groups, for subject 2 in all three groups, etc., with the constraint that no more than two continuous recognition and two elaboration blocks (one semantic, one orthographic) could appear in succession. Each block consisted of 108 stimuli. Within each block, one third of the items were new and did not repeat, and one third were new and subsequently repeated. Thus, one third of all items were old. The new items that repeated were represented after equiprobable lags of 2, 8, and 32 intervening items following their initial presentation. In comparisons of new vs old, only those new items that repeated comprised the new data set. Comparisons of new items that did not repeat (i.e., control items) with those that did showed no significant differences by analysis of variance (ANOVA) for either reaction time or ERP measures. Words were randomly selected by computer from 925 nouns originally normed by Paivio, Yuille, and Madigan (1968). For the semantic blocks, an additional 100 animal words were taken from Battig and Montague (1969). The probability of each Stimulus Type (animal vs nonanimal words in semantic blocks, uppercase vs lowercase words in orthographic blocks) was equated at 50/50 within each block. Word characteristics known to influence memory (e.g., frequency, imagery, and concreteness) were balanced across lags. Words were presented (300 msec duration, 2 sec interstimulus interval) on a video monitor (HP model 2458) subtending a vertical visual angle of .98°, and horizontal angles ranging from 1° 58' to 6° 34'. Subjects were instructed to make speeded, choice reaction time (RT) responses on each trial to animal vs nonanimal words in the semantic blocks, and uppercase vs lowercase words in orthographic blocks by pressing either a right- or left-hand held button. The hands assigned to the two responses were counterbalanced across subjects. Only RTs between 200 and 2200 msec poststimulus were accepted as correct responses. Responses occurring outside this time interval were not included in any of the behavioral or ERP analyses. Subjects were instructed to refrain from facial, eye, and bodily movements during the tasks, and were not informed about the repetition of items within a block.
P397
HAMBERGER AND FRIEDMAN
P398
SEMANTIC YOUNG
MIDDLE
OLD
Cz
X ORTHOGRAPHIC MIDDLE
OLD
Fz
Cz
10
Pz *
400 I 1200 800 1600
400 I 1200 I 800 1600
msec
400 I 1200 I 800 1600
- New Old
Figure 1. Grand mean repetition ERPs recorded at the midline electrode sites averaged across subjects within each age group. (Top) ERPs elicited by new (i.e., first presentation) and old (i.e., second presentation) items in the semantic task. (Bottom) ERPs elicited by new (i.e., first presentation) and old (i.e., second presentation) items in the orthographic task. Arrows and hatchured marks indicate stimulus onset with time lines every 400 msec. Vertical portion of inverted T-bar marks mean reaction time; horizontal portion indicates mean within-subject standard deviation.
between about 350 and 400 msec at Cz. This deflection (hereafter referred to as N400) was followed by a large amplitude positive potential with a parietal focus, peaking at approximately 600 msec (P600; i.e., the classical P300 or P3b component). The ERP data from the lateral sites as a function of both Repetition and Stimulus Type were highly similar to these shown for the midline electrodes. Because analyses of these data did not reveal findings that differed from those performed on the midline ERP indices, only the midline analyses will be reported in detail. ERP deflections were measured as averaged voltages
falling within a specified time window with respect to the voltage of the 300 msec prestimulus baseline. (Latency windows for the amplitude measures of N400 and P600 were defined separately for each age group and task. In the Orthographic task, N400 criteria were 250-325 msec for the young, 250-350 msec for the middle-aged, and 250-400 msec for the elderly. P600 criteria were 325-575 msec for the young, 350-700 msec for the middle-aged, and 400-700 msec for the elderly. In the Semantic task, the N400 criteria were 300-450 msec for the young, 250-450 msec for the middle-aged, and 250-500 msec for the elderly. P600 crite-
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
Pz
ERPs, PRIMING, AND AGING
MIDDLE (46)
YOUNG (24)
P399
OLD (74) New Old
REPETITION
Fz Cz Pz Non Animal Animal
STIMULUS TYPE
Fz
Pz
I
•
I
•
I
•
l
'
I
400 800 1200 1600
msec
'
i
'
i
•
i
•
i
400 800 1200 1600
msec
T
I '
TT
|
400 800 1200 1600
msec
Figure 2. ERPs recorded at the midline electrode sites for a representative subject from each age group. Numbers in parentheses indicate the age of each subject. (Top) Repetition ERPs elicited by new (i.e., first presentation) and old (i.e., second presentation) items in the semantic task. {Bottom) Stimulus-type ERPs elicited by animal and nonanimal words in the semantic task. Arrows and hatchured marks indicate stimulus onset with time lines every 400 msec. Vertical portion of inverted T-bar marks mean reaction time; horizontal portion indicates mean within-subject standard deviation.
ria were 450-700 msec for the young, 450-750 msec for the middle-aged, and 500-850 msec for the elderly.) Visual inspection of the complete set of waveforms revealed two small negative deflections within the N400 time window. However, because these two deflections did not show differential effects as a function of the experimental variables, they were combined and treated as one overall "N400." P600 latency was measured as the most positive time point falling between 250 and 950 msec. Because differences between conditions appeared subsequent to the N200, only analyses of the N400 and P600 deflections will be detailed. The mean number of trials comprising each average did not differ systematically as a function of Age, Task, Repeti-
tion, or Stimulus Type. Across age groups, the mean number of trials (SD and range) comprising the Repetition averages was 69.3 (5.3, 35-72) in the Orthographic task, and 67.1 (5.9, 34-72) in the Semantic task. The mean number of trials (SD and range) comprising the Stimulus Type averages was 69.2 (5.5, 35-72) in the Orthographic task, and 67.1 (5.9, 35-72) in the Semantic task.
ERP and Behavioral Analyses The main effects of Task and Age apply to both Stimulus Type and Repetition analyses and, therefore, will only be reported for the Repetition analyses.
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
A. I A/y'""'
HAMBERGER AND FRIEDMAN
P400
Behavioral data. — RTs for Repetition within each task and each age group are shown in Table 2. Although RTs were faster in the Orthographic than the Semantic task [F(l/48) = 368.85, p < .0001] and faster to old than new items [f (1/48) = 57.47, p < .0001], these main effects of Task and Repetition were modulated by a significant interaction of these two variables [F(l/48) = 27.21, p < .0001]. Simple effects procedures indicated that, although repetition priming (RT faster to old compared with new items) occurred in both the Semantic and Orthographic tasks, the effect of Repetition was greater in the Semantic condition [Semantic F(l/48) = 51.52, p < .0001; Orthographic F(\l 48) = 9.5,/? < .0003]. There were no main effects of either Lag or Age. The interactions of Repetition and Lag, Repetition and Age, and Repetition, Lag, and Age were not significant, indicating that the magnitude of the Repetition effect was equivalent at each lag and did not differ as a function of Age. ERPs Amplitude. — As can be seen in Figure 1 (see individual subject waveforms in Figure 2) and Table 2, the Repetition effect (old items more positive relative to new items) was most prominent in the Semantic task. Both N400 [F(l/48) = 13.03, p < .0007] and P600 [F(l/48) = 16.80, p < .0002]
Table 2. Behavioral and ERF' Data as a Function of Repetition RT and P600 Latency at Pz Middle
Young Orthographic New Old Grand means
New Old Grand means
Grand Means
RT
P600
RT
P600
RT
P600
RT
P600
566(112) 554(107) 560(109)
481 (87) 467 (77) 474(81)
541 (100) 533(89) 537 (87)
509 (54) 499 (57) 503 (55)
575(100) 573 (92) 574 (75)
553(55) 530 (56) 542 (53)
562(94) 554 (90) 558 (92)
514(73) 498 (67) 507 (70)
RT
P600
RT
P600
RT
P600
RT
P600
709(143) 680(141) 695(123)
575 (48) 553 (47) 564 (48)
687(126) 658(123) 672(102)
624(119) 607(105) 616(110)
704(128) 674(120) 689 (89)
644 (83) 659 (98) 652(90)
701(105) 671 (104) 686(105)
614(90) 606(96) 610(92)
Middle
Young Semantic
Older
Older
Grand Means
Amplitudes of N400 at Cz and P600 at Pz Middle
Young Orthographic New Old Grand means
N400 .05 .34 .19
N400
P600
5.83 6.65 6.24
-2.35 -1.60 -1.98
5.28 6.44 5.86
New Old Grand means RT, reaction time.
N400 .59 .60 .60
Middle
Young Semantic
Older
P600
Grand Means P600
N400
P600
7.11 7.27 7.19
-.46 -.13 -.30
6.12 6.81 6.46
P600
N400
P600
5.14 7.24 6.19
-1.22 .32 -.45
5.15 7.08 6.11
Older
N400
P600
N400
P600
-1.40 1.17 -.11
5.95 8.27 7.11
-2.41 -1.80 -2.11
3.11 5.48 4.83
N400 -.03 1.24 .60
Grand Means
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
showed main effects of Repetition, but only P600 was larger in the Orthographic task [F(l/35) = 7.19, p < .01]. For N400 the interaction of Task with Repetition was significant [F(l/48) = 6.23,/? < .01]. Although for P600 this interaction was marginally significant [F(l/48) = 3.78, p < .06], visual inspection of the waveforms in Figure 1 appears to indicate a much larger Repetition effect in the Semantic compared with the Orthographic task. Accordingly, simple effects procedures were performed for the P600 data. Unlike the RT data in which Repetition was significant in both tasks, simple effects procedures indicated that the effect of Repetition was significant only in the Semantic task [N400: F(l/48) = 24.69, p< .0001; P600: F( 1/48) = 20.32, p< .0001]. For P600, the interaction of Repetition and Electrode Site [F(2/7) = 9.29, p < .0008] was also significant. Simple effects procedures indicated that the effect of Repetition was significant at both Cz [F( 1/48) = 18.89,/? < .0001] andPz[F(l/48) = 17.63,/? < .0001], but not at Fz. Similar to the behavioral data, the absence of significant interactions of Repetition and Lag, Repetition and Age, and of Repetition, Lag, and Age indicated that the magnitude of the Repetition effect was the same at each Lag and did not differ as a function of Age. To illustrate the consistency with which individual subjects conformed to the pattern shown in the grand means of Figure 1 and Table 2, Table 3 presents the number of subjects in each age group who showed faster RTs, smaller N400s, and larger P600s to the repeated items in the Semantic task. As can be seen, the majority of subjects in each group exhibited the effect of Repetition on RT, as well as on N400 and P600 amplitudes.
Repetition. — Because there were no Stimulus Type by Repetition interactions for either the behavioral or ERP data, Repetition averages elicited by new and old items were collapsed across Stimulus Types.
P401
ERPs, PRIMING, AND AGING
Although there were no main effects of Age on the amplitudes of N400 or P600, Age by Electrode location interactions were significant for both deflections [P600: F(3/71) = 3.53,/?
Journal of Gerontology: PSYCHOLOGICAL SCIENCES 1992, Vol. 47, No. 6, P395-P4O5
Event-Related Potential Correlates of Repetition Priming and Stimulus Classification in Young, Middle-aged, and Older Adults Maria Hamberger and David Friedman New York State Psychiatric Institute.
F
ORMAL investigations of memory have typically involved "explicit" or direct measures of memory (e.g., free recall, recognition) in which subjects are fully cognizant of their participation in a memory task, and successful performance requires deliberate recollection or recognition of a previous event. In contrast, performance on "implicit" or indirect memory tests does not require deliberate recollection. Tasks are designed to divert attention away from the memory component of the task and, ideally, to prevent subjects from becoming aware that they are undergoing memory evaluation. Although the particular task may vary (e.g., lexical decision, word-stem completion), the common element across tasks is that items are repeated at some point during the course of testing. It is taken as evidence that retention has occurred when performance enhancement (e.g., faster response time and/or increased accuracy) is associated with item repetition, a phenomenon labeled "repetition priming." Implicit Memory and Aging Although there is substantial evidence that performance on explicit memory tasks declines with increasing age (Burke & Light, 1981; Madden, 1986), investigations of implicit memory and aging have generated equivocal findings. For example, on word-stem completion (Light & Singh, 1987), lexical decision (Moscovitch, 1982), picture naming (Mitchell, 1989; Mitchell, Brown, & Murphy, 1990), and biased homophone spelling (Rose, Yesavage, Hill, & Bower, 1986), response facilitation for previously presented items is of comparable magnitude in young and older adults. In contrast, age-related decline has been suggested by results from studies utilizing prolonged study-test intervals (Rabbitt, 1982), constrained study time (Howard,
Fry, & Brune, 1991), the priming of newly learned associates (Howard, McAndrews, & Lasaga, 1986; Howard et al., 1991), and larger subject samples (Chiarello & Hoyer, 1988; Hultch, Masson, & Small, 1991). To account for these discrepancies, some investigators (Chiarello & Hoyer, 1988; Graf, 1990) have suggested that older adults appear relatively impaired on implicit memory tasks because younger adults use intentional memory strategies in performing these tasks. Alternatively, since Dunn and Kirsner (1989) have suggested that different implicit memory tasks tap different processess, the possibility exists that such processes may be differentially affected by age. More specifically, Howard (1988) has asserted that implicit "item" memory requires memory for an item already represented in long-term memory, whereas implicit "associative" memory involves memory for a new association between previously unrelated items, which requires both greater attention and active processing in order to establish a novel connection between items. According to Howard et al. (1991), elderly adults are relatively impaired on implicit memory tasks that require this more elaborate processing. Although there are numerous reports in the cognitive (Jacoby & Dallas, 1981), aging (Graf, 1990), and neuoropsychological literatures (Schachter & Graf, 1986) demonstrating differential performance of individuals on implicit vs explicit memory tasks, it is currently unknown whether these dissociations are attributable to unique anatomical brain systems and/ or functionally distinct cognitive systems. Although much research has been directed toward the resolution of these questions, this experiment used only implicit memory tasks and was not designed therefore to address this issue. The purpose of this investigation was to examine implicit memory performance across the age span as assessed by item repetition P395
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
Electrophysiological and behavioral measures were concurrently recorded from young, middle-aged, and elderly subjects during two classification tasks that required different levels of processing. In both tasks, a portion of stimuli repeated after lags of 2, 8, and 32 intervening items following their first presentation, although these repetitions were incidental to the primary task. All three age groups showed similar patterns of behavioral and event-related potential (ERP) responses associated with stimulus classification and item repetition. ERPs to repeated words were characterized by greater positive amplitude relative to words at first presentation. Similarly, ERPs elicited by items in a primed category were more positive compared with unprimed items. Reaction times were faster to both repeated and primed items. Behavioral and ERP effects were more marked when subjects were required to discriminate items based on meaning rather than on orthography. This study supports previous behavioral evidence suggesting that processes associated with repetition priming and stimulus classification are preserved with age, and demonstrates that ERPs are responsive to repetition throughout the life cycle. Implications for the cognitive processes underlying repetition effects are also discussed.
P396
HAMBERGER AND FRIEDMAN
during two types of categorization tasks. The notion that implicit item memory does not change with increasing age is based on behavioral data. In addition to behavioral measures, we examined electrophysiological responses in three age groups during implicit item memory tasks.
METHODS
Subjects. — Three groups of right-handed, female, paid volunteers participated in this study. Subjects were recruited via posted notices at the medical center, and additional elderly volunteers were recruited from senior centers in the vicinity of the medical center. Demographic information for each age group is provided in Table 1. As can be seen, all subjects were in the normal range for both IQ (score of 90 or higher on the Quick Test; Ammons & Ammons, 1962), and a modified version of the Mini-Mental State Exam (a score of 50 or higher; Mayeux, Stern, Rosen, &Leventhal, 1981). The groups did not differ on the Mini-Mental State Exam [F(2/48) = 2.58, p> .05], the Quick Test [F(2/48) = 3.15, p > .05], or the laterality quotient [F(2/48) = 2.54, p > .05]. However, the youngest group had more years of education [F(2/48) = 5.34, p < .008] and higher socioeconomic status levels [F(2/48) = 3.36, p < .04] than the two older age groups. Elderly subjects were free of dementia, depression, and limitation in the activities of daily living, according to criteria of the Short CARE (Gurland, Golden, Teresi, & Challop, 1984), a structured interview standard-
Table 1. Demographic Characteristics of the Three Samples ERPs, Repetition Priming, and Aging To date, investigators of age effects during repetition priming have only used behavioral measures. Given the reliability of behavioral and ERP repetition effects in young adults, together with the somewhat variable repetition priming effects in the elderly, we sought to examine repetition priming using ERPs in conjunction with behavioral measures in an elderly population. Because a variety of agerelated neuronal changes (Terry, DeTeresa, & Hansen, 1987) may underlie differential performance on implicit memory tasks, we recorded a physiological measure that is intimately linked with the brain's processing of cognitive information. Thus, processing differences that may have been compensated for or attentuated by the time a behavioral response was executed might be revealed during on-line recording of brain activity. For instance, as noted above, because ERP repetition effects appear to be more sensitive to depth of processing manipulations compared with behav-
Age
Education
IQ
LQ» SES" MMSC DEPd DEMe
Young Adults (N == 18) Mean SD
24.94 113.50 2.96 12.50
17.44 2.15
93.22 47.94 55.56 15.73 15.32 1.65
NA
NA
NA
NA
1.56 1.33
.17 .38
Middle-aged Adults (W = 15) Mean SD
48.86 110.40 9.35 4.01
14.87 3.22
83.33 63.13 54.40 25.66 14.10 1.76
Older Adults (N =: 18) Mean SD
70.11 119.61 3.89 10.11
14.39 3.48
69.94 59.56 54.17 44.03 22.42 2.33
NA, not applicable; Education, number of years of education. "Laterality quotient (100 = completely right-handed), assessed by the Edinburgh Inventory (Oldfield, 1971). b Socioeconomic status (higher score = lower SES). 'Modified Mini-Mental State Exam (highest score is 57). d CARE depression score (cutoff is 6). e CARE dementia score (cutoff is 7).
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
Event-Related Potentials (ERPs) and Repetition Priming Scalp-recorded ERPs provide continuous, noninvasive monitoring of brain activity while subjects are actively engaged in a cognitive task. Unlike traditional behavioral measures, ERP recording provides a temporal readout of the brain's activity during information processing that is not necessarily contingent upon overt responding or conscious awareness. Hence, in addition to behavioral measures, the concurrent recording of ERP activity provides valuable complementary information regarding the nature, sequence, and timing of perceptual and cognitive events. Recent work has demonstrated that, in addition to behavioral measures, ERP indices are also sensitive to item repetition during implicit memory tasks. For example, in a paradigm in which subjects responded to occasional nonwords, ERPs to repeated words were characterized by greater positive amplitude from approximately 300-600 msec poststimulus relative to their first presentation counterparts (Rugg, 1985). Because superficial discrimination tasks (e.g., upper case vs lower case letters) showed small, nonsignificant ERP repetition effects compared with those recorded during lexical or semantic discrimination tasks (e.g., animal words vs nonanimal words), Rugg, Furda, and Lorist (1988) proposed that (unlike behavioral repetition effects) the magnitude of the ERP repetition effect is influenced by the "level of processing" induced by task demands (Craik & Lockhart, 1972). Rugg et al. (1988) attributed the ERP repetition effect (i.e., greater positivity to the repeated item) to an attenuation in a negativity elicited by the item's second presentation, and suggested that this negativity reflects "the consequences of reaccessing memory" and the integration of an item with its context. When an item is repeated, these processes occur to a lesser extent, which coincides with reduced negativity in the ERP elicited by the repeated item.
ioral measures (Rugg et al., 1988), ERPs in an elderly sample might be more responsive to variables such as the time course of repetition that do not ordinarily affect behavioral indices of repetition priming in either young (Dannenbring & Briand, 1982) or older adults (Moscovitch, 1982). The principal aim of this investigation was to determine whether the ERP data would mirror the findings reported in the behavioral aging literature on implicit item memory. We manipulated depth of processing during two stimulus classification tasks and the time course of item repetition by task demands and interitem lag, respectively. Similar to Rugg et al. (1988), although begun without knowledge of their study, the current orthographic task required a relatively superficial discrimination (uppercase vs lowercase words), whereas the semantic task required a "deeper" level of processing (animal vs nonanimal words).
ERPs, PRIMING, AND AGING
ized on large samples of community-dwelling elderly subjects. All subjects reported themselves to be in good physical health and did not report problems with their memory.
Electroencephalogram (EEG) recording procedures. — EEG was recorded using an electrocap from midline elec-
trodes at Fz, Cz, Pz, with lateral chains at F3, C3, P3, T3, and 01 on the left, and homologous locations on the scalp overlying the right hemisphere. The vertical electrooculogram (EOG) was recorded from an electrode located on the supraorbital ridge of the right eye, and the horizontal EOG was recorded from electrodes placed at the outer canthi of both eyes (bipolar recording). Except for the horizontal EOG electrodes, all leads were referred to nosetip. Due to a computer error, the data at the 01 electrode site were not usable for 20 of the 51 subjects. Therefore, the 01 data were excluded from all analyses. EEG and EOG were recorded with a bandpass of .01 to 30 Hz (5.3 sec time constant), with a Grass Neurodata system. Data acquisition and stimulus presentation were controlled by a PDP 11/34 computer, which digitized the EEG and EOG at 10-msec intervals for a 300 msec pre- and a 1700 msec poststimulus period separately for each word presentation, and stored the digitized records, choice responses, and RTs on 9-track digital tape. EEG was corrected for intrusion of EOG artifact using linear regression procedures based on adaptations of the methods described by Gratton, Coles, and Donchin (1983) and Verleger, Gasser, and Mocks (1982). On trials during which blink artifact was detected by computer algorithm, blink and vertical EOG artifacts were corrected first. In a second pass, horizontal EOG artifact was corrected following removal of vertical artifact from the horizontal EOG channel. In our implementation, the vertical and horizontal transmission coefficients were computed separately for each trial. Prior to computing averaged voltage measures of ERP activity, the individual averages were digitally filtered off-line at 15 Hz using the low-pass filter described by Ruchkin and Glaser (1978). ANOVAs. — ANOVAs, with Greenhouse-Geisser correction (Jennings and Wood, 1976), where appropriate, of ERP and behavioral data were performed using the BMDP4V (Dixon, 1987) program. Corrected degrees of freedom were truncated to the nearest integer. Newman-Keuls post-hoc tests were used to make pairwise comparisons when three or more conditions were contrasted. When significant interactions occurred, simple effects procedures (Winer, 1971), with reduced degrees of freedom where appropriate, were applied to explore the interactions further. RESULTS
The variables of interest were Age Group (young, middle, old), Task (orthographic vs semantic), Stimulus Type (animal vs nonanimal words in semantic blocks; uppercase vs lowercase words in orthographic blocks), Repetition (new vs old), Interitem Lag (2, 8, 32), and Electrode Site. Due to the large number of analyses, a criterion was adopted in which only F-ratios reaching alpha levels of p < .01 (after correction of degrees of freedom) were considered significant. ERP Waveforms Grand-mean ERP waveforms for Repetition are shown in Figure 1, and the Repetition and Stimulus Type ERPs of a representative subject in each age group from the midline electrodes are shown in Figure 2. As can be seen in Figure 1, the ERPs were characterized by an early negativity at about 200 msec, followed by a second negative deflection peaking
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
Stimuli and procedures. — Two orthographic (uppercase vs lowercase words), two semantic (animal vs nonanimal words), hereafter referred to as elaboration blocks, and four continuous recognition memory blocks were presented during one morning session that lasted approximately 1.5 hours. Only data from the elaboration blocks will be reported here. The continuous recognition memory data have been preliminarily reported (Friedman & Hamberger, 1989), with a full report currently in preparation. The continuous recognition procedure was administered two blocks at a time, whereas the semantic and orthographic blocks were administered back to back, one block at a time. The three kinds of blocks were counterbalanced across subjects, such that the orderings were identical for subject 1 in all three groups, for subject 2 in all three groups, etc., with the constraint that no more than two continuous recognition and two elaboration blocks (one semantic, one orthographic) could appear in succession. Each block consisted of 108 stimuli. Within each block, one third of the items were new and did not repeat, and one third were new and subsequently repeated. Thus, one third of all items were old. The new items that repeated were represented after equiprobable lags of 2, 8, and 32 intervening items following their initial presentation. In comparisons of new vs old, only those new items that repeated comprised the new data set. Comparisons of new items that did not repeat (i.e., control items) with those that did showed no significant differences by analysis of variance (ANOVA) for either reaction time or ERP measures. Words were randomly selected by computer from 925 nouns originally normed by Paivio, Yuille, and Madigan (1968). For the semantic blocks, an additional 100 animal words were taken from Battig and Montague (1969). The probability of each Stimulus Type (animal vs nonanimal words in semantic blocks, uppercase vs lowercase words in orthographic blocks) was equated at 50/50 within each block. Word characteristics known to influence memory (e.g., frequency, imagery, and concreteness) were balanced across lags. Words were presented (300 msec duration, 2 sec interstimulus interval) on a video monitor (HP model 2458) subtending a vertical visual angle of .98°, and horizontal angles ranging from 1° 58' to 6° 34'. Subjects were instructed to make speeded, choice reaction time (RT) responses on each trial to animal vs nonanimal words in the semantic blocks, and uppercase vs lowercase words in orthographic blocks by pressing either a right- or left-hand held button. The hands assigned to the two responses were counterbalanced across subjects. Only RTs between 200 and 2200 msec poststimulus were accepted as correct responses. Responses occurring outside this time interval were not included in any of the behavioral or ERP analyses. Subjects were instructed to refrain from facial, eye, and bodily movements during the tasks, and were not informed about the repetition of items within a block.
P397
HAMBERGER AND FRIEDMAN
P398
SEMANTIC YOUNG
MIDDLE
OLD
Cz
X ORTHOGRAPHIC MIDDLE
OLD
Fz
Cz
10
Pz *
400 I 1200 800 1600
400 I 1200 I 800 1600
msec
400 I 1200 I 800 1600
- New Old
Figure 1. Grand mean repetition ERPs recorded at the midline electrode sites averaged across subjects within each age group. (Top) ERPs elicited by new (i.e., first presentation) and old (i.e., second presentation) items in the semantic task. (Bottom) ERPs elicited by new (i.e., first presentation) and old (i.e., second presentation) items in the orthographic task. Arrows and hatchured marks indicate stimulus onset with time lines every 400 msec. Vertical portion of inverted T-bar marks mean reaction time; horizontal portion indicates mean within-subject standard deviation.
between about 350 and 400 msec at Cz. This deflection (hereafter referred to as N400) was followed by a large amplitude positive potential with a parietal focus, peaking at approximately 600 msec (P600; i.e., the classical P300 or P3b component). The ERP data from the lateral sites as a function of both Repetition and Stimulus Type were highly similar to these shown for the midline electrodes. Because analyses of these data did not reveal findings that differed from those performed on the midline ERP indices, only the midline analyses will be reported in detail. ERP deflections were measured as averaged voltages
falling within a specified time window with respect to the voltage of the 300 msec prestimulus baseline. (Latency windows for the amplitude measures of N400 and P600 were defined separately for each age group and task. In the Orthographic task, N400 criteria were 250-325 msec for the young, 250-350 msec for the middle-aged, and 250-400 msec for the elderly. P600 criteria were 325-575 msec for the young, 350-700 msec for the middle-aged, and 400-700 msec for the elderly. In the Semantic task, the N400 criteria were 300-450 msec for the young, 250-450 msec for the middle-aged, and 250-500 msec for the elderly. P600 crite-
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
Pz
ERPs, PRIMING, AND AGING
MIDDLE (46)
YOUNG (24)
P399
OLD (74) New Old
REPETITION
Fz Cz Pz Non Animal Animal
STIMULUS TYPE
Fz
Pz
I
•
I
•
I
•
l
'
I
400 800 1200 1600
msec
'
i
'
i
•
i
•
i
400 800 1200 1600
msec
T
I '
TT
|
400 800 1200 1600
msec
Figure 2. ERPs recorded at the midline electrode sites for a representative subject from each age group. Numbers in parentheses indicate the age of each subject. (Top) Repetition ERPs elicited by new (i.e., first presentation) and old (i.e., second presentation) items in the semantic task. {Bottom) Stimulus-type ERPs elicited by animal and nonanimal words in the semantic task. Arrows and hatchured marks indicate stimulus onset with time lines every 400 msec. Vertical portion of inverted T-bar marks mean reaction time; horizontal portion indicates mean within-subject standard deviation.
ria were 450-700 msec for the young, 450-750 msec for the middle-aged, and 500-850 msec for the elderly.) Visual inspection of the complete set of waveforms revealed two small negative deflections within the N400 time window. However, because these two deflections did not show differential effects as a function of the experimental variables, they were combined and treated as one overall "N400." P600 latency was measured as the most positive time point falling between 250 and 950 msec. Because differences between conditions appeared subsequent to the N200, only analyses of the N400 and P600 deflections will be detailed. The mean number of trials comprising each average did not differ systematically as a function of Age, Task, Repeti-
tion, or Stimulus Type. Across age groups, the mean number of trials (SD and range) comprising the Repetition averages was 69.3 (5.3, 35-72) in the Orthographic task, and 67.1 (5.9, 34-72) in the Semantic task. The mean number of trials (SD and range) comprising the Stimulus Type averages was 69.2 (5.5, 35-72) in the Orthographic task, and 67.1 (5.9, 35-72) in the Semantic task.
ERP and Behavioral Analyses The main effects of Task and Age apply to both Stimulus Type and Repetition analyses and, therefore, will only be reported for the Repetition analyses.
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
A. I A/y'""'
HAMBERGER AND FRIEDMAN
P400
Behavioral data. — RTs for Repetition within each task and each age group are shown in Table 2. Although RTs were faster in the Orthographic than the Semantic task [F(l/48) = 368.85, p < .0001] and faster to old than new items [f (1/48) = 57.47, p < .0001], these main effects of Task and Repetition were modulated by a significant interaction of these two variables [F(l/48) = 27.21, p < .0001]. Simple effects procedures indicated that, although repetition priming (RT faster to old compared with new items) occurred in both the Semantic and Orthographic tasks, the effect of Repetition was greater in the Semantic condition [Semantic F(l/48) = 51.52, p < .0001; Orthographic F(\l 48) = 9.5,/? < .0003]. There were no main effects of either Lag or Age. The interactions of Repetition and Lag, Repetition and Age, and Repetition, Lag, and Age were not significant, indicating that the magnitude of the Repetition effect was equivalent at each lag and did not differ as a function of Age. ERPs Amplitude. — As can be seen in Figure 1 (see individual subject waveforms in Figure 2) and Table 2, the Repetition effect (old items more positive relative to new items) was most prominent in the Semantic task. Both N400 [F(l/48) = 13.03, p < .0007] and P600 [F(l/48) = 16.80, p < .0002]
Table 2. Behavioral and ERF' Data as a Function of Repetition RT and P600 Latency at Pz Middle
Young Orthographic New Old Grand means
New Old Grand means
Grand Means
RT
P600
RT
P600
RT
P600
RT
P600
566(112) 554(107) 560(109)
481 (87) 467 (77) 474(81)
541 (100) 533(89) 537 (87)
509 (54) 499 (57) 503 (55)
575(100) 573 (92) 574 (75)
553(55) 530 (56) 542 (53)
562(94) 554 (90) 558 (92)
514(73) 498 (67) 507 (70)
RT
P600
RT
P600
RT
P600
RT
P600
709(143) 680(141) 695(123)
575 (48) 553 (47) 564 (48)
687(126) 658(123) 672(102)
624(119) 607(105) 616(110)
704(128) 674(120) 689 (89)
644 (83) 659 (98) 652(90)
701(105) 671 (104) 686(105)
614(90) 606(96) 610(92)
Middle
Young Semantic
Older
Older
Grand Means
Amplitudes of N400 at Cz and P600 at Pz Middle
Young Orthographic New Old Grand means
N400 .05 .34 .19
N400
P600
5.83 6.65 6.24
-2.35 -1.60 -1.98
5.28 6.44 5.86
New Old Grand means RT, reaction time.
N400 .59 .60 .60
Middle
Young Semantic
Older
P600
Grand Means P600
N400
P600
7.11 7.27 7.19
-.46 -.13 -.30
6.12 6.81 6.46
P600
N400
P600
5.14 7.24 6.19
-1.22 .32 -.45
5.15 7.08 6.11
Older
N400
P600
N400
P600
-1.40 1.17 -.11
5.95 8.27 7.11
-2.41 -1.80 -2.11
3.11 5.48 4.83
N400 -.03 1.24 .60
Grand Means
Downloaded from http://geronj.oxfordjournals.org/ at Universite Laval on July 2, 2015
showed main effects of Repetition, but only P600 was larger in the Orthographic task [F(l/35) = 7.19, p < .01]. For N400 the interaction of Task with Repetition was significant [F(l/48) = 6.23,/? < .01]. Although for P600 this interaction was marginally significant [F(l/48) = 3.78, p < .06], visual inspection of the waveforms in Figure 1 appears to indicate a much larger Repetition effect in the Semantic compared with the Orthographic task. Accordingly, simple effects procedures were performed for the P600 data. Unlike the RT data in which Repetition was significant in both tasks, simple effects procedures indicated that the effect of Repetition was significant only in the Semantic task [N400: F(l/48) = 24.69, p< .0001; P600: F( 1/48) = 20.32, p< .0001]. For P600, the interaction of Repetition and Electrode Site [F(2/7) = 9.29, p < .0008] was also significant. Simple effects procedures indicated that the effect of Repetition was significant at both Cz [F( 1/48) = 18.89,/? < .0001] andPz[F(l/48) = 17.63,/? < .0001], but not at Fz. Similar to the behavioral data, the absence of significant interactions of Repetition and Lag, Repetition and Age, and of Repetition, Lag, and Age indicated that the magnitude of the Repetition effect was the same at each Lag and did not differ as a function of Age. To illustrate the consistency with which individual subjects conformed to the pattern shown in the grand means of Figure 1 and Table 2, Table 3 presents the number of subjects in each age group who showed faster RTs, smaller N400s, and larger P600s to the repeated items in the Semantic task. As can be seen, the majority of subjects in each group exhibited the effect of Repetition on RT, as well as on N400 and P600 amplitudes.
Repetition. — Because there were no Stimulus Type by Repetition interactions for either the behavioral or ERP data, Repetition averages elicited by new and old items were collapsed across Stimulus Types.
P401
ERPs, PRIMING, AND AGING
Although there were no main effects of Age on the amplitudes of N400 or P600, Age by Electrode location interactions were significant for both deflections [P600: F(3/71) = 3.53,/?
