PSYC 575 Cognitive Psychology

Cognitive reserve modulates ERPs associated with verbal working memory in healthy younger and older adults

Megan E. Speer1 and Anja Soldan2

Megan E. Speer: [email protected]; Anja Soldan: [email protected] 1Department of Psychology, Rutgers University, Newark, NJ 07102

2Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, MD, 21205

Abstract

Although many epidemiological studies suggest the beneficial effects of higher cognitive reserve

(CR) in reducing age-related cognitive decline and dementia risk, the neural basis of CR is poorly

understood. To our knowledge, the current study represents the first electrophysiological

investigation of the relationship between CR and neural reserve (i.e., neural efficiency and

capacity). Specifically, we examined whether CR modulates event-related potentials (ERPs)

associated with performance on a verbal recognition memory task with three set sizes (1, 4, or 7

letters) in healthy younger and older adults. Neural data showed that as task difficulty increased,

the amplitude of the parietal P3b component during the probe phase decreased and its latency

increased. Notably, the degree of these neural changes was negatively correlated with CR in both

age groups, such that individuals with higher CR showed smaller changes in P3b amplitude and

less slowing in P3b latency (i.e., smaller changes in the speed of neural processing) with

increasing task difficulty, suggesting greater neural efficiency. These CR-related differences in

neural efficiency may underlie reserve against neuropathology and age-related burden.

Keywords

Cognitive reserve; Verbal working memory; Neural efficiency; Event-related potentials (ERPs); P3; Cognitive aging

© 2015 Elsevier Inc. All rights reserved.

Corresponding Author: Dr. Anja Soldan, Ph.D., Division of Cognitive Neuroscience, 1620 McElderry Street, Reed Hall, Baltimore, MD 21205, Telephone: 410-502-2188, FAX: 410-502-2189, [email protected]. Megan E. Speer, Department of Psychology, Rutgers University, Newark; Anja Soldan, Cognitive Neuroscience Division, Department of Neurology, Johns Hopkins University School of Medicine.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Disclosure Statement There are no actual or potential conflicts of interest.

HHS Public Access Author manuscript Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

Published in final edited form as: Neurobiol Aging. 2015 March ; 36(3): 1424–1434. doi:10.1016/j.neurobiolaging.2014.12.025.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

1. Introduction

The concept of cognitive reserve (CR) has been proposed as an explanation for why

individuals with similar levels of brain pathology or injury can differ markedly in the

clinical manifestation of that pathology, with some individuals being symptom free and

others showing cognitive and/or functional impairment. CR is a theoretical construct that

postulates that certain lifetime experiences, including education, occupational breadth and

complexity, and engagement in activities that are cognitively, socially, and physically

stimulating increase the efficiency, capacity, and flexibility of brain networks. As a result,

individuals with higher levels of CR are thought to be able to sustain greater levels of brain

pathology or damage before showing clinically significant levels of impairment (for a

review see Stern, 2009).

In support of the concept of CR, many studies have shown that higher levels of educational

and occupational attainment, as well as greater pre-morbid intelligence are associated with

better clinical outcomes across a variety of conditions, such as a reduced risk of mild

cognitive impairment or dementia (Pettigrew et al., 2013; Soldan et al., 2013; Wilson et al.,

2002), better recovery from traumatic brain injury (Fay et al., 2010; Levi et al., 2013), and

less cognitive impairment in multiple sclerosis (Sumowski et al., 2013) or Parkinson’s

disease (Perneczky et al., 2008).

Despite the strong evidence for the beneficial effects of CR, the neural mechanisms by

which it operates are poorly understood. It has been proposed that there are two different

ways in which CR is implemented in the brain: neural reserve and neural compensation

(Stern, 2009). Neural compensation refers to the reliance on alternative brain networks that

are not normally used by healthy individuals to maintain or improve cognitive performance

in the face of age or pathology-induced changes. Neural reserve, by comparison, refers to

individual differences in the efficiency and capacity of brain networks underlying task

performance in unimpaired individuals that provide reserve against the impact of brain

injury. Such individual differences in neural efficiency and capacity are thought to be

present before the onset of pathology or injury and therefore exist in both young and older

individuals.

In support of the concept of neural compensation, functional neuroimaging studies have

reported CR-related activation of brain regions among cognitively normal older adults that

are not typically activated by young subjects (Scarmeas et al., 2003; Springer et al., 2005;

Steffener et al., 2011; Stern et al., 2008). Likewise, high performing older individuals (who

presumably have higher CR) often recruit additional brain regions (usually in the

contralateral hemisphere) when engaging in strategies associated with better task

performance (reviewed in Reuter-Lorenz and Park, 2010). There is also some evidence for

greater neural efficiency among cognitively normal young and older adults, such that

subjects with high CR show less task-related activation as a function of increasing task load

than subjects with lower CR (Habeck et al., 2003; Habeck et al., 2005; Steffener et al., 2011;

for similar evidence also see Bosch et al., 2010).

Speer and Soldan Page 2

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

To our knowledge, prior investigations that have directly examined the neural basis of CR

have been conducted using H2 15O positron emission tomography (PET) or functional

magnetic resonance imaging (fMRI). Although these methods are very useful for identifying

mechanisms of neural compensation (due to their high spatial resolution), they are less

useful for addressing another potentially important aspect of CR, neural processing speed,

owing to their poor temporal resolution. That is, to the extent that CR is associated with

neural efficiency (e.g., higher CR, greater neural efficiency), one would predict not only a

relationship between CR and the magnitude of neural activation, but also between CR and

the speed of neural processing (e.g., higher CR, greater neural speed). Therefore, the

primary aim of the current study was to investigate the association between CR and neural

reserve (as indexed by neural efficiency and capacity) in young and older adults using event-

related potentials (ERPs), which have excellent temporal resolution. We did not examine

neural compensation because of the poor spatial resolution of ERPs. ERPs, which measure

synchronized post-synaptic potentials of pyramidal cortical neurons, have the additional

advantage that unlike the fMRI BOLD response and H2 15O PET, they are not influenced by

potentially confounding neurovascular-coupling mechanisms that change with aging (Ances

et al., 2008; Ances et al., 2009; Fabiani et al., 2014; Fleisher et al., 2009; Hutchison et al.,

2012).

Our investigation focused on the central-parietal P3b (or P300b) ERP component because

both its amplitude and latency are modulated by the cognitive demands of a task (e.g., Kok,

2001; Polich, 2007). In particular, P3b amplitude is often thought of reflecting cognitive

resource allocation (Donchin and Coles, 1988; Kok, 2001; Linden, 2005; Polich, 2007),

whereas P3b latency appears to be related to information processing speed (i.e., stimulus

evaluation time), independent of motor response preparation and execution processes

(McCarthy and Donchin, 1981; Walhovd and Fjell, 2003). Moreover, a number of studies

have reported associations between intelligence and P3b amplitude and/or latency (Gevins

and Smith, 2000; Houlihan et al., 1998; Jausovec and Jausovec, 2000; Liu et al., 2011;

Pelosi et al., 1992; e.g., Wronka et al., 2013) as well as between intelligence and neural

efficiency as measured by EEG alpha-band desynchronization (ERD, reviewed in Neubauer

and Fink, 2009), which is functionally related to the P3b (Peng et al., 2012), suggesting that

the P3b would also be related to CR.

The current study used a Sternberg verbal working memory paradigm to measure individual

differences in neural reserve (i.e., neural efficiency and capacity). Task load was

manipulated parametrically by varying the number of items participants had to encode and

retain in memory (e.g., 1, 4, or 7 letters). Varying task difficulty is advantageous for

studying the neural basis of CR because CR is hypothesized to modify how the brain copes

with increasing task demands, as could be the case following brain insult: tasks that are easy

when an individual is healthy may be difficult following brain damage or pathology (Stern,

2009). Previous work using the Sternberg paradigm with letters has shown that P3b

amplitude and latency during the encoding phase increase with the number of letters to be

encoded, reflecting extra stimulus processing at higher set sizes (Houlihan et al., 1998).

Moreover, longer P3b latencies during encoding, particularly at higher set sizes, have been

associated with higher intelligence scores and better task performance (Houlihan et al.,

1998). This suggests that greater load-related increases in P3b latency and amplitude at

Speer and Soldan Page 3

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

encoding may be beneficial to task performance and be associated with CR. By comparison,

during the probe phase P3b amplitude decreases with increasing memory load, whereas

latency increases, reflecting more resource-demanding memory search processes for the

higher set sizes (Houlihan et al., 1998; Morgan et al., 2008; Pinal et al., 2014).

Similar to the definitions provided by Stern (2009), we defined greater neural efficiency as

smaller changes in P3b amplitude or latency with increasing memory load given the same or

better behavioral performance. Greater neural capacity was defined as a larger change in

P3b amplitude or latency with increasing memory load, coupled with better performance.

The following hypotheses were tested. First, higher CR is associated with better task

performance, particularly at the higher set sizes. Second, P3b amplitude and latency during

the encoding phase increase with set size and the amount of this load-related increase is

positively correlated with CR. If the first and second hypotheses were confirmed, this would

provide evidence for greater neural capacity among individuals with higher CR. Third, P3b

amplitude during the probe phase decreases with memory load, whereas P3b latency during

the probe phase increases with memory load and the amount of these load-related changes is

negatively correlated with CR (i.e., smaller change with higher CR, in line with the results

of an fMRI study using the same task, Habeck et al., 2005). If the first and the third

hypotheses were both confirmed, this would provide evidence for greater neural efficiency

among high CR individuals. In addition, we calculated neural inefficiency indexes to more

directly relate load-related changes in P3b amplitude to behavioral performance measures

and tested whether these indexes correlate with CR. Lastly, we tested the hypothesis that the

relationships between CR, behavior, and load-related changes in P3b amplitude and latency

are the same across age groups, which would support the view that neural reserve operates

similarly in young and older adults.

2. Methods

2.1 Participants

Twenty-five healthy young adults and 21 non-demented healthy older adults participated in

this study. Two older participants performed at chance-level in the high load condition and

were excluded from analyses because their poor performance may have reflected

disengagement from the task, which cannot be interpreted meaningfully with respect to CR.

Thus, analysis was performed on 25 young adults and 19 older adults. See Table 1 for

participant characteristics. All were right-handed with normal or corrected-to-normal vision

(visual acuity cut-off of 20/50 as assessed by a Snellen chart), and none reported any history

of neurological or psychiatric diseases that would affect the central nervous system. The

older participants were recruited from the community surrounding North Dartmouth, MA,

and received monetary compensation at a rate of $15/ per hour. All older adults were

screened for dementia via the dementia rating scale (DRS; all scored at or above 137;

Mattis, 1988). Young adults were students at the University of Massachusetts, Dartmouth,

and received partial course credit for their participation. Written informed consent was

obtained prior to participation.

Speer and Soldan Page 4

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

2.2 Cognitive Reserve Composite Score

We created a CR composite score based on measures thought to reflect CR: the National

Adult Reading Task (NART; Nelson, 1982), the vocabulary subtest of the Wechsler Adult

Intelligence Scale-Revised (WAIS-R; Wechsler, 1981), and years of education. Measures of

intelligence, particularly verbal intelligence, are commonly included when attempting to

model CR (Alexander et al., 1997). Supporting this approach, a prospective study by

Richards and Sacker (2003) found that intelligence at age 53 was uniquely influenced by

adult occupation, educational attainment, and childhood cognition. Accordingly, basing a

measure of CR on variables such as verbal intelligence means that CR is not inflexible, but

rather can change over the course of one’s lifetime. Educational attainment, on the other

hand, is the most commonly used proxy for CR, and is often combined with other predictors

into a composite variable as in the present study. Because education is a strong determinant

of future employment and income level (Beckles et al., 2011), educational attainment

directly correlates with other CR proxies such as socioeconomic status and occupational

attainment.

All three proxy measures of CR were correlated with one another for the older adults (all p <

0.05). For the young adults, who had not yet completed their education, the CR measure was

based on the NART and WAIS-R vocabulary scores only, which were also correlated (p <

0.05). To calculate the composite CR score, these individual measures were transformed to

z-scores and then averaged. Use of CR composites such as these has been shown to have

construct validity (Siedlecki et al., 2009).

2.3 Stimuli and Procedures

All participants performed a delayed item recognition task measuring verbal working

memory (Sternberg, 1966) on a computer in a sound attenuated booth. Each trial began with

the presentation of a variable inter-stimulus interval lasting 2.25 s to 2.75 s. A memory set of

one, four, or seven uppercase letters was presented for 2.5 s. Regardless of set size, the

stimulus geometry was a two-row array with one row containing three stimuli and the other

four stimuli. For set size one and four, each absent letter in the array was replaced with an

asterisk to keep the geometric array consistent for each set size. This way, differences in

neural responses to different set sizes at encoding cannot be attributed to differences in the

amount of visual information presented, but must reflect the active processing of this

information. The memory set was followed by a 5 s delay that served as a retention period in

which only a blank screen was shown. Then, a probe stimulus was presented for 2 s that

consisted of a single lowercase letter. Participants were instructed to press a key indicating

whether or not the probe was part of the memory set. Memorizing uppercase letters and then

responding to a lowercase letter forced participants to base their decision on the

phonological or pre-lexical representation of the letters rather than simply on the shape of

the letters. The total length of one trial was 12 s (see Figure 1).

A total of 240 trials were divided into eight blocks, each containing 10 trials of each set size

(1, 4, 7) for a total of 80 trials at each set size. Within each block, there were 5 targets (i.e.,

the probe letter was in the memory set) and 5 non-targets (i.e., the probe was not in the

memory set) at each set size. Each block contained a different list of stimuli. The order of

Speer and Soldan Page 5

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

block presentation was counterbalanced across participants and each block’s contents were

randomized. Participants were given a 1-minute break between each block in order to help

prevent fatigue. The first two blocks served as training and were excluded from analysis to

minimize the effects of task-related skill learning on performance and neural responses. The

last six blocks (i.e. 60 trials of each set size) were included in data analysis. Participants

were instructed to answer as quickly and accurately as possible. They were not given

feedback about their performance.

2.4 Behavioral Data Analysis

The behavioral accuracy data was analyzed in terms of the signal-detection theory measure

of sensitivity, dL, a measure of accuracy without response bias, (Snodgrass and Corwin,

1988). The measure of dL can be thought of as standardized hits minus standardized false

alarms. It is based on logistic distributions and is functionally equivalent to d’ (d-prime),

which is based on normal distributions. Both dL and mean reaction time (RT) were assessed

for effects of set size, age group, and group by set size interaction via repeated measures

analysis of variance (ANOVA). Incorrect trials were excluded from the RT analysis. For

each subject, linear regression analysis was used to calculate the slope of RT across set size,

the RT intercept (i.e., imputed value at set size = 0), the slope of dL across set sizes, and the

dL intercept. The slope of performance variables measures how well one adapts to the

changes in task difficulty, whereas the intercept of performance variables measures the

baseline performance level when there is no memory load (i.e., at set size = 0). These

measures were used to calculate the neural inefficiency scores (see below).

2.5 EEG Recording and Analysis

Brain electrical activity was recorded from 32 scalp sites (ActiveTwo electrodes, Biosemi)

using an elastic cap with mounted active electrodes positioned according to the International

10/20 System. The electrode offset was kept below 40 mV. Electrodes were initially

referenced to the common mode sense (CMS) electrode and then converted to an average

reference offline. EEG was amplified and continuously sampled at 512 Hz with a bandpass

filter of 0.05–100 Hz. The electrooculogram (EOG) was recorded by means of electrodes

placed approximately 1 cm below each eye as well as lateral to the outer canthi of each eye.

Eye movements were modeled and compensated for using 2–4 ocular source components

(BESA; MEGIS Software GmbH, Grafelfing, Germany). Incorrect trials, as well as trials

with muscle or skin artifacts were excluded from analysis. For each participant, baseline-

corrected, artifact-free trials, time locked to the onset of a stimulus (memory set or probe),

were averaged separately for each set size (1, 4, or 7) from 100 ms before stimulus onset

until 1,000 ms thereafter. A low-pass filter of 40 Hz was applied after averaging.

For the encoding phase, the mean number of trials for set size 1 was 43.0 (SD = 7.5, Min =

25, Max = 57), for set size 4 it was 43.7 trials (SD = 8.0, Min = 25, Max = 59), and for set

size 7 it was 37.1 trials (SD = 7.1, Min = 22, Max = 57). For the probe phase, the mean

number of trials for set size 1 was 48.0 (SD = 7.4, Min = 31, Max = 59), for set size 4 it was

47.3 trials (SD = 7.0, Min = 29, Max = 59), and for set size 7 it was 40.9 trials (SD = 7.0,

Min = 24, Max = 57). Overall, the number of trials was lower for the older than the young

Speer and Soldan Page 6

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

adults [F(1, 42) = 4.58, p = .04], due to the lower number of correct trials among the older

adults (see Results).

Mean amplitude of the P3b ERP component was calculated in 100 ms bins for the interval

300 – 800 ms post stimulus onset. P3b peak latency was measured using the half area

latency, which corresponds to the time point that divides the area under the P3b waveform

(from 300 to 800 ms) into two equal regions. The half-area latency works well on large

components, like the P3b, and is less sensitive to noise than peak latency (Luck, 2005).

The overall analysis approach was to first confirm that the P3b component was indeed

modulated by task demands in both age groups, as we hypothesized. To do so, we tested for

the presence of an effect of set size on P3b amplitudes in each 100 ms time window that

encompassed the duration of the P3b (i.e., 300 – 800 ms) during the encoding phase

(memory set presentation) and the probe phase in each age group. Next, only for those time

bins that showed a significant set size effect in one or both age groups, we averaged across

electrodes (where appropriate) and time bins to calculate the mean P3b amplitude for each

set size. We then calculated the difference between mean P3b amplitude at set size 1 and 7

and used this as an index of neural processing that is sensitive to task demands. Likewise,

for P3b latency, we first tested for an effect of set size and, if present, proceeded to calculate

the difference in latency between set size 1 and 7 as an index of neural processing that is

sensitive to task demands. The next step was to examine whether the amount of demand-

related amplitude or latency modulation (i.e., difference between set size 1 and 7) correlated

with CR composite score.

We also calculated neural inefficiency scores that directly relate the degree of demand-

related P3b amplitude modulation to behavioral performance measures and tested if these

scores were associated with composite CR. Neural inefficiency was defined as the amount

of change in task-related neural processing as a function of behavioral performance. Neural

inefficiency scores (the reciprocal of efficiency) were computed for each subject by dividing

the change in task-related neural processing (i.e., slope of P3b mean amplitude with respect

to set size) by behavioral performance values (where a higher value means better

performance): dL intercept, dL slope, RT intercept−1, and RT slope−1. Neural inefficiency

was used because it may be a more stable measure than efficiency (Zarahn et al., 2007). By

this measure, less efficient individuals show greater changes in neural activation with

increasing task demands to achieve the same or lower performance than more efficient

individuals.

An alpha level of .05 was adopted for all statistical analyses. Greenhouse–Geisser

corrections were applied where appropriate to correct for violations of the sphericity

assumption. Significant main effects and interactions were followed by post hoc contrasts

and Bonferroni-Holm corrected pairwise comparisons.

Speer and Soldan Page 7

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

3. Results

3.1 Behavioral results

Sensitivity (dL) was marginally lower for the older than the younger adults [F(1, 42) = 3.08,

p = .086]. There was also a main effect of set size [F(2, 84) = 127.62, p < .001], and an age

by set size interaction [F(2, 84) = 7.24, p = .001]. As set size increased, dL decreased for

both the younger [F(2, 48) = 50.33, p < .001] and older adults [F(2, 36) = 84.42, p < .001],

but the amount of decrease in dL was greater for the older than the younger adults, see

Figure 2a. Pairwise comparisons showed a significant difference in dL between set sizes 1

and 7 and between 4 and 7 and in both age groups (all p < .0001), while the difference

between set sizes 1 and 4 was only significant in the young adults (p < .005). The slope in dL

was marginally greater in the older relative to the young adults (t(42) = 1.82, p = .076), but

there was no difference in dL intercept (t < 1).

For RT, there was a main effect of age group [F(1,42) = 44.33, p < .0001], indicating slower

RTs in the old than young adults, a main effect of set size [F(2, 84) = 291.47, p < .0001] and

an age by set size interaction [F(2, 84) = 19.27, p < .0001]. As set size increased, there was a

significant increase in RT for younger [F(2, 48) = 150.31, p < .0001] and older adults [F(2,

36) = 135.23, p < .0001], with older adults showing a greater increase than the young adults,

see Figure 2b. Pairwise comparisons showed reliable differences in RT between set sizes 1

and 4 and 4 and 7 in both age groups, all p < .0005. Both the RT-intercept (t(42) = 3.20, p

= .004) and the slope in RT with respect to set size (t(42) = 5.13, p < .0001) were greater in

the older than the younger adults.

Combining across age groups and partialing out the effects of age, the CR composite score

was positively correlated with dL at set size 4, dL at set size 7, and change in dL from set size

1 to 7 (dL slope), all r > .32, all p < .04. The size of these correlations did not differ across

age groups (as determined by the Fisher r - to – z transformation, all p > .24), except for the

correlation between composite CR and dL slope, which was significantly greater in the

young than in the older adults (z = 1.95, p = .05). There were no significant correlations

between composite CR and any of the RT measures (for age groups separately and when

combining across age groups).

3.2 Electrophysiological results

P3b amplitude and latency were assessed at posterior parietal electrodes (PZ, P3 and P4),

where the component was maximal (see Figure 3). Repeated measures ANOVA was used

with factors for age group, set size, electrode, and time (for amplitudes only, 100 ms bins

from 300 to 800 ms). Effects of time and electrode are not mentioned unless they involved

interactions with set size and/or age group. The P3b latency data from one older participant

was excluded from analysis because of excessive noise. Exclusion of this participant from

all other analyses did not alter the results.

3.3 Encoding Phase

P3b Amplitude—P3b amplitudes to encoded items showed a significant interaction

between age group and set size [F(2, 84) = 3.83, p =.026], between age group, set size, and

Speer and Soldan Page 8

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

time [F(8, 336) = 2.52, p =.011], as well as between set size, time, and electrode [F(16, 672)

= 5.92, p < .001], indicating set size differences based on age and time. Separate follow-up

ANOVAs for each age group and time bin showed no significant effects involving set size

for the older adults [all p >= .08], suggesting that set size did not reliably modulate P3b

amplitudes for this age group during encoding. For the young adults, there were significant

set size effects during the 500–800 ms time window (all p < .05), but no interaction between

set size and electrode (all p > .35). Post-hoc t-test (averaging across electrodes and time

bins) indicated that P3b amplitude increased from set size 1 to 7 in the young adults from

500–800 ms [t(1, 24) = 2.45, p = .022]. Set size 4 was intermediary and did not differ from

set size 1, but was lower than set size 7 (p = .007).

P3b Latency—There was a significant main effect of set size [F(2, 84) = 3.12, p = .05]

and a significant interaction between set size and electrode [F(4, 168) = 4.87, p = .002].

Although the interaction between age group and set size did not reach significance [p = .10],

we analyzed the latency data separately for the two age groups because the amplitude data

had revealed an effect of set size for the young but not older individuals. Consistent with the

amplitude data, there was a main effect of set size on P3b latency for the young adults [F(2,

48) = 5.54, p = .001], but no effects involving set size for the older adults [all p > .10]. For

the young adults, a post-hoc t-test (averaging across electrodes) revealed no difference in

latency between set sizes 1 and 4 (t < 1), but significantly longer latencies at set size 7 than

at set size 1 (t(24) = 2.28, p = .03) and 4 (t(24) = 3.90, p = .0007, see Figure 4.

Correlation with Cognitive Reserve—Since there was a significant set size effect on

P3b amplitude from 500 to 800 ms in young adults, we averaged the amplitudes across this

time window and across electrodes, separately for each set size, to calculate the mean

change in P3b amplitude from set size 1 to 7 for each subject as an index of neural

processing that is sensitive to task demands (i.e., mean amplitude at set size 7 – mean

amplitude at set size 1). We then tested if this value correlated with the CR composite score.

No significant correlation with composite CR was observed. Likewise, the correlation

between the change in P3b latency from set size 1 to 7 (averaging across electrodes) and

composite CR in the young subjects was not significant. This suggests that the load-related

increase in P3b amplitude and latency in the young adults occurs independently of

composite CR.

3.4 Probe Phase

P3b Amplitude—P3b amplitudes to probes were smaller for older than younger adults

[F(1, 42) = 6.69 p = .013]. The effect of set size was significant [F(2, 84) = 9.12, p =.0004],

as were the interactions between set size and time [F(8, 336) = 4.38, p = .002], and between

set size, time, and electrode [F(16, 672) = 2.26, p = .037]. Separate follow-up ANOVAs for

each time bin showed a significant set size effect for each time window (all p < .03), but no

interactions between set size and electrode or set size and age (all p > .15). Post-hoc t-tests

(averaging across electrodes and time bins) confirmed that P3b amplitude decreased from

set size 1 to 7 in both the young [t(1, 24) = 2.40, p = .025] and older adults [t(1, 18) = 3.52, p

= .002]. Set size 4 was intermediary and not significantly different from set size 7 (p > .09

Speer and Soldan Page 9

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

for both age groups). For the older adults, the difference between set sizes 1 and 4 was

significant (p = .005).

P3b latency—P3b latencies were longer for older than younger adults [F(1, 41) = 8.22, p

= .007] and varied as a function of set size [F(2,82) = 15.38, p < .001]. In addition, there was

a set size by electrode interaction [F(4, 160) = 2.92, p = .034], reflecting a smaller effect of

set size at electrode P4 than at electrodes PZ and P3. Post-hoc t-tests, collapsing across

electrodes, showed that latency was significantly shorter at set size 1 than at set size 7 in

both younger [t(24) = 4.43, p < .001] and older adults [t(18) = 2.99, p = .009]. Set size 4 was

intermediary and differed from set size 1 (p <= .056 in both age groups), but not from set

size 7 (see Figure 5).

Correlation with Cognitive Reserve—Because there was a significant set size effect on

P3b amplitude in each time bin from 300 to 800 ms in both age groups, we averaged the

amplitudes across this time window and across electrodes, separately for each set size, to

calculate the mean change in P3b amplitude from set size 1 to 7 for each subject (mean

change in P3b amplitude = amplitude at set size 7 – amplitude at set size 1). We then tested

if this value correlated with the CR composite score. Combining across age groups to

increase power, but partialing out the effect of age to account for the age-related decrease in

P3b amplitude, there was a negative correlation, r(41) = −.40 p = .009, such that individuals

with higher composite CR showed less change in P3b amplitude with increasing task

demands. See Figure 5 for a scatterplot of this correlation. This correlation was significant

for the group of older adults (r(17) = −0.70, p = .001), but did not reach significance in the

young adults r(23) = −0.25, p = .23. However, the size of the correlations was not

significantly different across age groups (z = 1.07, p = .28). Additionally, the correlation

between composite CR and change in P3b amplitude from set size 4 to 7 was significant for

the young adults (r(23) = −0.47, p = .02), indicating that when task demands are sufficiently

high, there is a similar relationship between demand-related change in P3b amplitude and

composite CR in both young and older adults.

For P3b latency, we also averaged across electrodes and found that the CR composite score

negatively correlated with the amount of change in P3b latency change from set size 1 to 7,

combining across both age groups, partialing out the effects of age, r(40) = −.55, p < .001

(see Figure 5). Thus, higher composite CR was associated with a smaller change in the

speed of neural processing as task demands increased. This correlation was significant for

both age groups, r(23) = −0.54, p = .006 for young adults and r(17) = −0.49, p = .046 for

older adults.

3.5 Relationship between neural inefficiency and cognitive reserve

There was a negative correlation between the CR composite score and neural inefficiency

scores for load-related changes in P3b amplitude with respect to both accuracy (dL slope:

partial r (41) = −.33, p = .035) and reaction time (RT slope−1: partial r(41) = −.47, p = .002;

RT intercept−1: partial r(41) = −.42, p = .005), partialing out the effects of age. This

suggests that independent of age, higher levels of composite CR are associated with less

neural inefficiency for P3b amplitude change with respect to dL slope, RT slope, and RT

Speer and Soldan Page 10

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

intercept (see Figure 6). Separate analyses for each age group confirmed that these

correlations were significant in both the young and older adults (all p <= .05 one-tailed,

except for RT intercept−1 for young adults, where p = .15) and the size of the correlations

did not differ across age groups.

3.6. Does cognitive reserve modify age-related changes in P3b amplitude and latency?

We also tested if composite CR reduces the impact of aging on P3b amplitude (reduced with

aging) and latency (increased with aging). ANOVAs with group (young, old) and composite

CR (high, low) were performed for mean P3b amplitude and latency (averaging across set

sizes) and for P3b amplitude and latency at set size 7. There were no significant interactions

between age group and composite CR level, all p >= .10, indicating a similar age effect on

P3b amplitude and latency across CR levels.

4. Discussion

We examined the relationship between individual differences in CR in young and older

adults and individual differences in neural efficiency and capacity that have been

hypothesized to underlie the beneficial effects of higher CR. We investigated this

relationship by looking at the association between CR composite score and load-related

changes in P3b amplitude and latency during the performance of a verbal working memory

task that increased in difficulty from a set size of one to seven letters. As hypothesized, both

young and older individuals with higher composite CR were more accurate in the task at

higher set sizes than those with lower composite CR. Contrary to our predictions, there was

no association between CR composite score and load-related changes in P3b amplitude or

latency during the encoding phase of the task. For the probe (or retrieval) phase, we found

that independent of subjects’ age, higher levels of composite CR were associated with

smaller changes in P3b amplitude and latency with increasing working memory load, as we

had hypothesized. These findings support the view that young and older cognitively normal

adults with higher CR have neural networks that operate more efficiently when task

demands increase, consistent with prior fMRI studies (Bartres-Faz et al., 2009; Bosch et al.,

2010; Habeck et al., 2003; Habeck et al., 2005; Steffener et al., 2011; Stern et al., 2008). Our

results extend prior findings by showing that greater neural efficiency among high CR adults

not only reflects the amount of neural activation (as indexed by ERP component amplitude

or fMRI BOLD signal), but also the speed of neural processing, as indexed by ERP

component latency.

4.1 Neural Efficiency

We further corroborated the association between CR and neural efficiency by testing the

relationship between CR composite score and an index of neural inefficiency, which is the

ratio of the amount of demand-related neural activity (i.e., P3b amplitude slope with

increasing task load) per unit of behavioral performance (i.e., accuracy or RT slope with

increasing task load). We found that, indeed, higher composite CR was associated with less

neural inefficiency with respect to both accuracy (dL-slope) and reaction time (RT-slope and

RT-intercept). This suggests that individuals with lower composite CR showed greater

Speer and Soldan Page 11

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

changes in activation as the task became more difficult although it benefited them less in

terms of performance, reflecting less efficient processing.

The present results are compatible with those by Habeck et al. (2005), who used a verbal

working memory task very similar to ours in an fMRI investigation of CR in young subjects.

Habeck et al. (2005) identified spatial covariance patterns of brain regions for each phase of

the task whose expression increased monotonically with increasing set size in the majority

of subjects. The degree to which subjects expressed these load-related patterns during the

delay and probe phases was inversely correlated with CR, indicating that smaller changes in

load-related activation were associated with higher CR. The load-related pattern expressed

during the encoding phase did not correlate with CR, consistent with the lack of a

relationship between CR composite score and P3b amplitude and latency observed during

the encoding phase of the present study.

Interestingly, while a few regions in the pattern identified by Habeck et al., (2005) for the

probe phase showed an increase in activation with set size (cerebellum, inferior frontal

gyrus), most regions showed a decrease with set size, including medial frontal and parietal

regions, cingulate gyrus, middle temporal gyrus, and parahippocampal gyrus. Although our

ERP results cannot be directly compared to these fMRI findings owing to the difference in

methodologies and analyses, the spatial pattern identified by Habeck et al. (2005) likely

overlaps with the network of regions that generated the P3b in the probe phase of the present

study, the amplitude of which also decreased with increasing task demands. While the

spatial localization of CR-related brain activity was not of primary interest in the present

study, our results, in combination with these fMRI results suggests that the brain regions

whose activity was modulated by composite CR in the present study consisted of a frontal-

temporal-parietal network. This interpretation is bolstered by source localization studies that

have consistently identified parietal, temporal-parietal, frontal, and cingulate cortex as

generators of the posterior P3b (Frodl et al., 2000; Li et al., 2009; Moores et al., 2003).

More interestingly, the results from fMRI studies using other behavioral paradigms have

also identified activity in frontal, temporal-parietal, and cingulate areas as being correlated

with level of CR, suggesting that activity in these brain regions may be most sensitive to

CR-related changes. As noted by Stern et al. (2008), there may be a general brain network

that underlies the beneficial effects of CR across age groups and tasks and that may be

broadly involved in executive and control processes. The present results would be consistent

with this possibility, although future studies are needed to test if the P3b – CR relationships

observed in this study generalize to other experimental tasks and stimulus types. If so, this

would have potentially important clinical consequences, as the P3b is relatively easy to

measure and could be used to track the effectiveness of interventions (both pharmacologic

and behavioral) designed to increase CR and improve neural functioning among middle-

aged and older adults. Many such cognitive training programs are currently available (e.g.,

Acevedo and Loewenstein, 2007; Olazaran et al., 2010), but consistent neural outcome

measures that generalize across a variety of tasks are lacking. However, future work is

needed to evaluate the reliability of the P3b for predicting CR level. Additionally, future

research should examine the relationship between CR and other ERP components,

particularly the P3a, which indexes frontal executive functions (Barcelo et al., 2006; Dien et

Speer and Soldan Page 12

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

al., 2004). Given that this component appears to increase with aging (Alperin et al., 2014;

Fabiani et al., 1998), it might also be associated with individual differences in CR.

Another noteworthy finding is that the correlations between composite CR and load-related

changes in P3b amplitude and latency as well as between composite CR and several

behavioral measures were observed in a sample that included both younger and older adults

and were independent of age. This supports the view that CR develops across the lifespan. It

is also consistent with the view that high levels of CR could provide functional resilience in

the face of non age-related burden, such as traumatic brain injury that might occur at any age

(Fay et al., 2010; Levi et al., 2013).

Our results also are in line with the Compensation-Related Utilization of Neural Circuits

Hypothesis (CRUNCH; Reuter-Lorenz, 2008), which describes age-related changes in

performance and brain activation in terms of neural efficiency and capacity, similarly to the

concept of neural reserve. CRUNCH postulates that older adults have less efficient neural

networks, requiring them to engage a network to a greater degree than young adults when

task demands are low and leading them to reach the capacity of the network at lower levels

of difficulty than young adults (e.g., Daffner et al., 2011; Schneider-Garces et al., 2010).

Consistent with this model, we found that older adults had smaller P3b amplitudes and

longer latencies during the probe phase, even at the lower set sizes, consistent with greater

resource utilization and less efficient processing. Furthermore, the older adults showed

greater performance decrements with increasing set size than the young adults, while the

change in P3b amplitude and latency with increasing task difficulty during the probe phase

was the same in both age groups, which also suggests less efficient processing.

It is also important to consider ERP studies that have tested the relationship between P3b

amplitude or latency and intelligence. In line with our findings, Houlihan et al. (1998)

utilized a similar Sternberg WM task and found that P3b amplitude during the probe phase

decreased with set size. Furthermore, individuals with higher cognitive ability, as measured

by the Multidimensional Aptitude Battery, had greater P3 amplitudes at the higher set sizes

(and therefore smaller decrements with increasing set size), suggesting greater neural

efficiency of processing.

Studies using other experimental paradigms are more difficult to compare with our results

because they measured neural activity at a single level of task difficulty, rather than multiple

levels. As such, they do not directly address how intelligence relates to individual

differences in the way the brain copes with increasing task demands, but are more likely to

measure the association between intelligence and task-specific neural processing. For

instance, studies using oddball paradigms or cued Go-Nogo tasks have reported that

individuals with higher cognitive ability elicited greater P3b amplitudes and shorter P3b

latencies than individuals with lower cognitive ability (De Pascalis et al., 2008; Jausovec

and Jausovec, 2000; Liu et al., 2011; O'Donnell et al., 1992; Wronka et al., 2013). Assuming

that these tasks were at least moderately difficult, these results would be consistent with

ours, in that we also found that individuals with higher composite CR had faster P3b

latencies, as well as a trend toward larger P3b amplitudes, during the probe phase when task

demands were high (i.e., set size of 7). The benefit of tasks that parametrically manipulate

Speer and Soldan Page 13

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

difficulty, as was done in our study, is that they may be better able to assess how well

individuals cope with increasing task demands irrespective of specific task features. This is

important when studying CR, because it has been hypothesized that high CR promotes

greater flexibility and adaptability to increasing task demands across a variety of tasks.

4.2 Neural Capacity

During the encoding phase of the task, only the young, but not the older adults,

demonstrated an increase in P3b amplitude and latency with increasing task difficulty. This

finding, along with better task performance among the young group, may provide some

evidence of greater neural capacity in the young compared to the older adults. Specifically,

the observed relationship suggests that young adults were able to boost activation when the

task became more challenging and engage in more extensive stimulus processing, which

may have supported their superior performance at the higher set sizes. Greater neural

capacity (here defined as higher network expression at higher levels of task demand) in

young compared to older adults has been described previously (e.g., Cappell et al., 2010;

Holtzer et al., 2009; Stern et al., 2012). The biological basis of this age-related capacity

difference is currently not known, but may be related to differences in neural connectivity

and plasticity (Burke and Barnes, 2006), white matter integrity (e.g., Bennett and Rypma,

2013), or brain oxygen metabolism (Hutchison et al., 2012) that manifest with aging.

It is important to note that this age-related capacity difference was unrelated to individual

differences in CR. Theoretical models of CR have postulated that higher CR is associated

with both greater neural efficiency and greater neural capacity, the two components of

neural reserve (Stern, 2009). While there is increasing evidence for a relationship between

CR and neural efficiency, few studies have reported associations between CR and neural

capacity (e.g., Scarmeas et al., 2003). Although null findings need to be interpreted

cautiously, this might indicate that CR is more closely linked to neural efficiency than to

neural capacity. Another possibility is that the load-related processes indexed by the P3b

during the encoding phase (such as sustained attentional allocation) are less amenable to

CR-related influences than the load-related processes indexed by the P3b during the test

phase (such as categorization, or working memory updating).

4.3 Cognitive reserve and age-related changes in neural activity

Consistent with prior studies, we found lower P3b amplitudes and longer latencies for the

older adults compared to the young individuals (McEvoy et al., 2001; Strayer et al., 1987).

This age difference did not vary as a function of CR composite score, suggesting that CR

does not directly alter the effects of aging on the P3b component. This finding is consistent

with the theoretical model of CR (e.g., Barulli and Stern, 2013; Stern, 2009), which

proposes that CR does not directly alter age or pathology-related neural changes, but rather

serves to modify the behavioral and clinical expression of those changes. In line with this

interpretation, both high and low composite CR older adults had smaller P3b amplitudes and

longer latencies than young adults even when task demands were low, but the high

composite CR individuals tended to perform better (as measured by dL) than the low

composite CR individuals. At the same time, these findings do not preclude the possibility

Speer and Soldan Page 14

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

that CR modulates other age-related neural changes not examined in the current study, as

has been proposed by other investigators (Nyberg et al., 2012).

4.4 Conclusions

In sum, this study provided further support for the concept of neural reserve by

demonstrating that higher composite CR is associated with greater neural efficiency in terms

of less neural activity and faster processing speed with increasing task demands. While our

study examined verbal intelligence and educational background as components of CR, we

acknowledge that factors related to an enriched environment – such as occupational

complexity and participation in leisure and social activities – are also important components

of CR. Therefore, future studies are needed to address the generalizability of our findings to

other proxy measures of CR. We note, however, that education, vocabulary knowledge, and

reading ability tend to have similar clinical effects on dementia risk and cognitive aging as

these other components of CR (Richards and Sacker, 2003; Scarmeas et al., 2001),

suggesting that our findings are applicable to the broader concept of CR. Additionally, this

study underscores the utility of ERPs and other electrophysiological measures for testing

neural mechanisms of CR.

Acknowledgements

This work was supported by start-up funds and the J.P. Healey Grant awarded to Anja Soldan by the University of Massachusetts, Dartmouth. We thank Kristina Monteiro and Adam Young for research assistance.

References

Acevedo A, Loewenstein DA. Nonpharmacological cognitive interventions in aging and dementia. J Geriatr Psychiatry Neurol. 2007; 20:239–249. [PubMed: 18004010]

Alexander GE, Furey ML, Grady CL, Pietrini P, Brady DR, Mentis MJ, Schapiro MB. Association of premorbid intellectual function with cerebral metabolism in Alzheimer's disease: implications for the cognitive reserve hypothesis. Am J Psychiatry. 1997; 154:165–172. [PubMed: 9016263]

Alperin BR, Mott KK, Rentz DM, Holcomb PJ, Daffner KR. Investigating the age-related "anterior shift" in the scalp distribution of the P3b component using principal component analysis. Psychophysiology. 2014; 51:620–633. [PubMed: 24660980]

Ances BM, Leontiev O, Perthen JE, Liang C, Lansing AE, Buxton RB. Regional differences in the coupling of cerebral blood flow and oxygen metabolism changes in response to activation: implications for BOLD-fMRI. Neuroimage. 2008; 39:1510–1521. [PubMed: 18164629]

Ances BM, Liang CL, Leontiev O, Perthen JE, Fleisher AS, Lansing AE, Buxton RB. Effects of aging on cerebral blood flow, oxygen metabolism, and blood oxygenation level dependent responses to visual stimulation. Hum Brain Mapp. 2009; 30:1120–1132. [PubMed: 18465743]

Barcelo F, Escera C, Corral MJ, Periáñez JA. Task switching and novelty processing activate a common neural network for cognitive control. J Cogn Neurosci. 2006; 18:1734–1748. [PubMed: 17014377]

Bartres-Faz D, Sole-Padulles C, Junque C, Rami L, Bosch B, Bargallo N, Falcon C, Sanchez-Valle R, Molinuevo JL. Interactions of cognitive reserve with regional brain anatomy and brain function during a working memory task in healthy elders. Biol Psychol. 2009; 80:256–259. [PubMed: 19022337]

Barulli D, Stern Y. Efficiency, capacity, compensation, maintenance, plasticity: emerging concepts in cognitive reserve. Trends Cogn Sci. 2013; 17:502–509. [PubMed: 24018144]

Beckles GL, Truman BI. (CDC), C.f.D.C.a.P. Education and income - United States, 2005 and 2009. MMWR Surveill Summ. 2011; 60(Suppl):13–17. [PubMed: 21430614]

Speer and Soldan Page 15

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Bennett IJ, Rypma B. Advances in functional neuroanatomy: a review of combined DTI and fMRI studies in healthy younger and older adults. Neurosci Biobehav Rev. 2013; 37:1201–1210. [PubMed: 23628742]

Bosch B, Bartres-Faz D, Rami L, Arenaza-Urquijo EM, Fernandez-Espejo D, Junque C, Sole-Padulles C, Sanchez-Valle R, Bargallo N, Falcon C, Molinuevo JL. Cognitive reserve modulates task- induced activations and deactivations in healthy elders, amnestic mild cognitive impairment and mild Alzheimer's disease. Cortex. 2010; 46:451–461. [PubMed: 19560134]

Burke SN, Barnes CA. Neural plasticity in the ageing brain. Nat Rev Neurosci. 2006; 7:30–40. [PubMed: 16371948]

Cappell KA, Gmeindl L, Reuter-Lorenz PA. Age differences in prefontal recruitment during verbal working memory maintenance depend on memory load. Cortex. 2010; 46:462–473. [PubMed: 20097332]

Daffner KR, Chong H, Sun X, Tarbi EC, Riis JL, McGinnis SM, Holcomb PJ. Mechanisms underlying age- and performance-related differences in working memory. J Cogn Neurosci. 2011; 23:1298– 1314. [PubMed: 20617886]

De Pascalis V, Varriale V, Matteoli A. Intelligence and P3 components of the event-related potential elicited during an auditory discrimination task with masking. Intelligence. 2008; 36:35–47.

Dien J, Spencer KM, Donchin E. Parsing the late positive complex: mental chronometry and the ERP components that inhabit the neighborhood of the P300. Psychophysiology. 2004; 41:665–678. [PubMed: 15318873]

Donchin E, Coles MGH. Is the P300 component a manifestation of context updating? Behavioral Brain Science. 1988; 11:357–374.

Fabiani M, Friedman D, Cheng JC. Individual differences in P3 scalp distribution in older adults, and their relationship to frontal lobe function. Psychophysiology. 1998; 35:698–708. [PubMed: 9844431]

Fabiani M, Gordon BA, Maclin EL, Pearson MA, Brumback-Peltz CR, Low KA, McAuley E, Sutton BP, Kramer AF, Gratton G. Neurovascular coupling in normal aging: a combined optical, ERP and fMRI study. Neuroimage. 2014; 85(Pt 1):592–607. [PubMed: 23664952]

Fay TB, Yeates KO, Taylor HG, Bangert B, Dietrich A, Nuss KE, Rusin J, Wright M. Cognitive reserve as a moderator of postconcussive symptoms in children with complicated and uncomplicated mild traumatic brain injury. J Int Neuropsychol Soc. 2010; 16:94–105. [PubMed: 19835663]

Fleisher AS, Podraza KM, Bangen KJ, Taylor C, Sherzai A, Sidhar K, Liu TT, Dale AM, Buxton RB. Cerebral perfusion and oxygenation differences in Alzheimer's disease risk. Neurobiol Aging. 2009; 30:1737–1748. [PubMed: 18325636]

Frodl T, Juckel G, Gallinat J, Bottlender R, Riedel M, Preuss U, Moller HJ, Hegerl U. Dipole localization of P300 and normal aging. Brain Topogr. 2000; 13:3–9. [PubMed: 11073089]

Gevins A, Smith ME. Neurophysiological measures of working memory and individual differences in cognitive ability and cognitive style. Cereb Cortex. 2000; 10:829–839. [PubMed: 10982744]

Habeck C, Hilton HJ, Zarahn E, Flynn J, Moeller J, Stern Y. Relation of cognitive reserve and task performance to expression of regional covariance networks in an event-related fMRI study of nonverbal memory. Neuroimage. 2003; 20:1723–1733. [PubMed: 14642482]

Habeck C, Rakitin BC, Moeller J, Scarmeas N, Zarahn E, Brown T, Stern Y. An event-related fMRI study of the neural networks underlying the encoding, maintenance, and retrieval phase in a delayed-match-to-sample task. Brain Res Cogn Brain Res. 2005; 23:207–220. [PubMed: 15820629]

Holtzer R, Rakitin BC, Steffener J, Flynn J, Kumar A, Stern Y. Age effects on load-dependent brain activations in working memory for novel material. Brain Res. 2009; 1249:148–161. [PubMed: 18983833]

Houlihan M, Stelmack R, Campbell K. Intelligence and the Effects of Perceptual Processing Demands, Task Difficulty and Processing Speed on P300, Reaction Time and Movement Time. Intelligence. 1998; 26:9–25.

Speer and Soldan Page 16

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Hutchison JL, Lu H, Rypma B. Neural mechanisms of age-related slowing: the DeltaCBF/ DeltaCMRO2 ratio mediates age-differences in BOLD signal and human performance. Cereb Cortex. 2012; 23:2337–2346. [PubMed: 22879349]

Jausovec N, Jausovec K. Correlations between ERP parameters and intelligence: a reconsideration. Biol Psychol. 2000; 55:137–154. [PubMed: 11118680]

Kok A. On the utility of P3 amplitude as a measure of processing capacity. Psychophysiology. 2001; 38:557–577. [PubMed: 11352145]

Levi Y, Rassovsky Y, Agranov E, Sela-Kaufman M, Vakil E. Cognitive reserve components as expressed in traumatic brain injury. J Int Neuropsychol Soc. 2013; 19:664–671. [PubMed: 23575273]

Li Y, Wang LQ, Hu Y. Localizing P300 generators in high-density event- related potential with fMRI. Med Sci Monit. 2009; 15:MT47–MT53. [PubMed: 19247255]

Linden DE. The p300: where in the brain is it produced and what does it tell us? Neuroscientist. 2005; 11:563–576. [PubMed: 16282597]

Liu T, Xiao T, Shi J, Zhao D. Response preparation and cognitive control of highly intelligent children: a Go-Nogo event-related potential study. Neuroscience. 2011; 180:122–128. [PubMed: 21329744]

Luck, SJ. An introduction to the event-related potential technique. First ed.. Cambridge, MA: MIT Press; 2005.

Mattis, S. Dementia Rating Scale. Odessa, FL: Psychological Assessment Resources; 1988.

McCarthy G, Donchin E. A metric for thought: a comparison of P300 latency and reaction time. Science. 1981; 211:77–80. [PubMed: 7444452]

McEvoy LK, Pellouchoud E, Smith ME, Gevins A. Neurophysiological signals of working memory in normal aging. Brain Res Cogn Brain Res. 2001; 11:363–376. [PubMed: 11339986]

Moores KA, Clark CR, Hadfield JL, Brown GC, Taylor DJ, Fitzgibbon SP, Lewis AC, Weber DL, Greenblatt R. Investigating the generators of the scalp recorded visuo-verbal P300 using cortically constrained source localization. Hum Brain Mapp. 2003; 18:53–77. [PubMed: 12454912]

Morgan HM, Klein C, Boehm SG, Shapiro KL, Linden DE. Working memory load for faces modulates P300, N170, and N250r. J Cogn Neurosci. 2008; 20:989–1002. [PubMed: 18211245]

Nelson, HE. The National Adult Reading Test (NART): Test manual. United Kingdom: NFER-Nelson, Windsor; 1982.

Neubauer AC, Fink A. Intelligence and neural efficiency. Neurosci Biobehav Rev. 2009; 33:1004– 1023. [PubMed: 19580915]

Nyberg L, Lövdén M, Riklund K, Lindenberger U, Bäckman L. Memory aging and brain maintenance. Trends Cogn Sci. 2012; 16:292–305. [PubMed: 22542563]

O’Donnell BF, Friedman S, Swearer JM, Drachman DA. Active and passive P3 latency and psychometric performance: influence of age and individual differences. Int J Psychophysiol. 1992; 12:187–195. [PubMed: 1592672]

Olazaran J, Reisberg B, Clare L, Cruz I, Pena-Casanova J, Del Ser T, Woods B, Beck C, Auer S, Lai C, Spector A, Fazio S, Bond J, Kivipelto M, Brodaty H, Rojo JM, Collins H, Teri L, Mittelman M, Orrell M, Feldman HH, Muniz R. Nonpharmacological therapies in Alzheimer's disease: a systematic review of efficacy. Dement Geriatr Cogn Disord. 2010; 30:161–178. [PubMed: 20838046]

Pelosi L, Holly M, Slade T, Hayward M, Barrett G, Blumhardt LD. Event-related potential (ERP) correlates of performance of intelligence tests. Electroencephalogr Clin Neurophysiol. 1992; 84:515–520. [PubMed: 1280197]

Peng W, Hu L, Zhang Z, Hu Y. Causality in the association between P300 and alpha event-related desynchronization. PLoS One. 2012; 7:e34163. [PubMed: 22511933]

Perneczky R, Drzezga A, Boecker H, Ceballos-Baumann AO, Granert O, Forstl H, Kurz A, Haussermann P. Activities of daily living, cerebral glucose metabolism, and cognitive reserve in Lewy body and Parkinson's disease. Dement Geriatr Cogn Disord. 2008; 26:475–481. [PubMed: 18984958]

Pettigrew C, Soldan A, Li S, Lu Y, Wang MC, Selnes O, Moghekar A, O'Brien R, Albert M. Relationship of Cognitive Reserve and APOE Status to the Emergence of Clinical Symptoms in

Speer and Soldan Page 17

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Preclinical Alzheimer’s Disease. Cognitive Neuroscience. 2013; 4(3–4):136–142. [PubMed: 24168200]

Pinal D, Zurrón M, Díaz F. Effects of load and maintenance duration on the time course of information encoding and retrieval in working memory: from perceptual analysis to post-categorization processes. Front Hum Neurosci. 2014; 8:165. [PubMed: 24744715]

Polich J. Updating P300: an integrative theory of P3a and P3b. Clin Neurophysiol. 2007; 118:2128– 2148. [PubMed: 17573239]

Reuter-Lorenz PA, Park DC. Human neuroscience and the aging mind: a new look at old problems. J Gerontol B Psychol Sci Soc Sci. 2010; 65:405–415. [PubMed: 20478901]

Reuter-Lorenz PAC, KA. Neurocognitive aging and the compensation hypothesis. Current Directions in Psychological Science. 2008; 17:177–182.

Richards M, Sacker A. Lifetime antecedents of cognitive reserve. J Clin Exp Neuropsychol. 2003; 25:614–624. [PubMed: 12815499]

Scarmeas N, Levy G, Tang MX, Manly J, Stern Y. Influence of leisure activity on the incidence of Alzheimer's disease. Neurology. 2001; 57:2236–2242. [PubMed: 11756603]

Scarmeas N, Zarahn E, Anderson KE, Hilton J, Flynn J, Van Heertum RL, Sackeim HA, Stern Y. Cognitive reserve modulates functional brain responses during memory tasks: a PET study in healthy young and elderly subjects. Neuroimage. 2003; 19:1215–1227. [PubMed: 12880846]

Schneider-Garces NJ, Gordon BA, Brumback-Peltz CR, Shin E, Lee Y, Sutton BP, Maclin EL, Gratton G, Fabiani M. Span, CRUNCH, and beyond: working memory capacity and the aging brain. J Cogn Neurosci. 2010; 22:655–669. [PubMed: 19320550]

Siedlecki KL, Stern Y, Reuben A, Sacco RL, Elkind MS, Wright CB. Construct validity of cognitive reserve in a multiethnic cohort: The Northern Manhattan Study. J Int Neuropsychol Soc. 2009; 15:558–569. [PubMed: 19573274]

Snodgrass JG, Corwin J. Pragmatics of measuring recognition memory: applications to dementia and amnesia. J Exp Psychol Gen. 1988; 117:34–50. [PubMed: 2966230]

Soldan A, Pettigrew C, Li S, Wang MC, Moghekar A, Selnes OA, Albert M, O'Brien R. Relationship of cognitive reserve and cerebrospinal fluid biomarkers to the emergence of clinical symptoms in preclinical Alzheimer's disease. Neurobiol Aging. 2013; 34:2827–2834. [PubMed: 23916061]

Springer MV, McIntosh AR, Winocur G, Grady CL. The relation between brain activity during memory tasks and years of education in young and older adults. Neuropsychology. 2005; 19:181– 192. [PubMed: 15769202]

Steffener J, Reuben A, Rakitin BC, Stern Y. Supporting performance in the face of age-related neural changes: testing mechanistic roles of cognitive reserve. Brain Imaging Behav. 2011; 5:212–221. [PubMed: 21607547]

Stern Y. Cognitive reserve. Neuropsychologia. 2009; 47:2015–2028. [PubMed: 19467352]

Stern Y, Rakitin BC, Habeck C, Gazes Y, Steffener J, Kumar A, Reuben A. Task difficulty modulates young-old differences in network expression. Brain Res. 2012; 1435:130–145. [PubMed: 22197699]

Stern Y, Zarahn E, Habeck C, Holtzer R, Rakitin BC, Kumar A, Flynn J, Steffener J, Brown T. A common neural network for cognitive reserve in verbal and object working memory in young but not old. Cereb Cortex. 2008; 18:959–967. [PubMed: 17675368]

Sternberg S. High-speed scanning in human memory. Science. 1966; 153:652–654. [PubMed: 5939936]

Strayer DL, Wickens CD, Braune R. Adult age differences in the speed and capacity of information processing: 2. An electrophysiological approach. Psychol Aging. 1987; 2:99–110. [PubMed: 3268210]

Sumowski JF, Rocca MA, Leavitt VM, Riccitelli G, Comi G, DeLuca J, Filippi M. Brain reserve and cognitive reserve in multiple sclerosis: what you've got and how you use it. Neurology. 2013; 80:2186–2193. [PubMed: 23667062]

Walhovd KB, Fjell AM. The relationship between P3 and neuropsychological function in an adult life span sample. Biol Psychol. 2003; 62:65–87. [PubMed: 12505768]

Wechsler, D. Wechsler adult intelligence scale-revised manual. New York: The Psychological Corporation; 1981.

Speer and Soldan Page 18

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Wilson RS, Mendes De Leon CF, Barnes LL, Schneider JA, Bienias JL, Evans DA, Bennett DA. Participation in cognitively stimulating activities and risk of incident Alzheimer disease. JAMA. 2002; 287:742–748. [PubMed: 11851541]

Wronka E, Kaiser J, Coenen AM. Psychometric intelligence and P3 of the event-related potentials studied with a 3-stimulus auditory oddball task. Neurosci Lett. 2013; 535:110–115. [PubMed: 23266476]

Zarahn E, Rakitin B, Abela D, Flynn J, Stern Y. Age-related changes in brain activation during a delayed item recognition task. Neurobiol Aging. 2007; 28:784–798. [PubMed: 16621168]

Speer and Soldan Page 19

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Highlights

• Healthy young and older adults performed a verbal working memory task.

• High cognitive reserve (CR) was associated with smaller changes in P3b

amplitude and latency with increasing task difficulty, regardless of age.

• Our findings underscore the utility of ERPs for testing neural mechanisms of

CR.

Speer and Soldan Page 20

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Figure 1. A schematic diagram of the Sternberg verbal working memory task illustrating the time

course of a single trial.

Speer and Soldan Page 21

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Figure 2. Behavioral performance on the verbal working memory task. These graphs depict the

parametric modulation of (A) mean sensitivity (dL) and (B) mean reaction time as a function

of increasing set size for younger and older adults. * Significant difference between groups,

p < .05.

Speer and Soldan Page 22

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Figure 3. Stimulus-locked, grand-averaged waveforms showing the P3b component during (A) the

encoding phase and (B) the probe presentation phase. For the encoding phase, there was no

reliable effect of set size on P3b amplitude and latency in the older adults, whereas young

adults showed an increase in amplitude with increasing set size from 500 – 800 ms. For the

probe phase, as set size increased, the amplitude of the P3b component decreased

monotonically from 300–800 ms and the half-area latency increased in both young and older

adults. The voltage maps on the right show the scalp distribution of activity at the time the

P3b component peaked (averaging across set-sizes).

Speer and Soldan Page 23

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Figure 4. Encoding phase. (A) Mean P3b Amplitudes by set size from 500 – 800 ms and (B) Mean

P3b latencies by set size. P3b amplitude from 500 – 800 ms increased from set size 1 to 7 in

the younger adults, but not older adults. Likewise mean P3b latency increased from set-size

1 to 7 in the young but not the older adults. However, these changes were not associated

with CR (amplitude: p = .32; latency: p = .76). * Significant difference between groups, p < .

05.

Speer and Soldan Page 24

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Figure 5. Probe presentation phase. (A) P3b Amplitudes (from 300 – 800 ms) and latencies by set size

for the probe presentation phase. (B) Scatterplots illustrating the correlation between

composite cognitive reserve score and the change in P3b peak amplitude from 300 – 800 ms

(top), and the correlation between composite cognitive reserve and P3b peak latency

(bottom) during the probe presentation phase for younger and older adults. Correlation

coefficients are for the entire group, partialing out the effects of age. * Significant difference

between groups, p < .05.

Speer and Soldan Page 25

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Figure 6. Scatterplots illustrating the correlation between composite cognitive reserve scores and

neural inefficiency with respect to dL slope (top panel), RT slope (middle panel), and the RT

intercept (bottom panel) during the probe presentation phase for younger adults (□) and

older adults (▲), partialing out the effects of age.

Speer and Soldan Page 26

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

A uthor M

anuscript A

uthor M anuscript

Speer and Soldan Page 27

Table 1

Participant Characteristics

Young (n = 25)

Old (n = 19)

Female/Male 10/15 17/2

Age ** 20.1 (2.3) 70.2 (5.1)

Education (yr) * 14.0 (1.41) 16.2 (3.3)

DRS N/A 142.6 (2.0)

NART-IQ ** 112.4 (6.1) 120.8 (6.5)

WAIS-R vocabulary

11.7 (2.8) 12.7 (2.2)

† Note: means and standard deviations (in parentheses) for demographic variables and neuropsychological test scores. DRS, Dementia Rating Scale;

NART-IQ, North American Adult Reading Test estimated IQ; WAIS-R vocabulary, Wechsler Adult Intelligence Scale, Revised, vocabulary subtest, age-scaled score.

* Significant difference between groups, p < .05.

** Significant difference between groups, p < .0001.

Neurobiol Aging. Author manuscript; available in PMC 2016 March 01.