Differential Age Effects of Transcranial Direct Current Stimulation on Associative Memory

Abstract

Objectives

Older adults experience associative memory deficits relative to younger adults (Old & Naveh-Benjamin, 2008). The aim of this study was to test the effect of transcranial direct current stimulation (tDCS) on face-name associative memory in older and younger adults.

Method

Experimenters applied active (1.5 mA) or sham (0.1 mA) stimulation with the anode placed over the left dorsolateral prefrontal cortex (dlPFC) during a face-name encoding task, and measured both cued recall and recognition performance. Participants completed memory tests immediately after stimulation and after a 24-h delay to examine both immediate and delayed stimulation effects on memory.

Results

Results showed improved face-name associative memory performance for both recall and recognition measures, but only for younger adults, whereas there was no difference between active and sham stimulation for older adults. For younger adults, stimulation-induced memory improvements persisted after a 24-h delay, suggesting delayed effects of tDCS after a consolidation period.

Discussion

Although effective in younger adults, these results suggest that older adults may be resistant to this intervention, at least under the stimulation parameters used in the current study. This finding is inconsistent with a commonly seen trend, where tDCS effects on cognition are larger in older than younger adults.

Aging, Associative Memory, Non-Invasive Brain Stimulation, Transcranial Direct Current Stimulation, Recall, Recognition

Forgetting the name of an acquaintance or forgetting where you parked your car are frustrating daily events that reflect a failure of associative memory. For many, these events become more common with advancing age (Reese, Cherry, & Norris, 1999). Indeed, relative to younger adults, older adults suffer from deficits in associative memory, which is memory for associations between simultaneously presented information (Old & Naveh-Benjamin, 2008), such as a face and a name (Naveh-Benjamin, Guez, Kilb, & Reedy, 2004). According to Hebbian principles, learning such associations occurs when ensembles of neurons work together, or wire together, through long-term potentiation (LTP; Hebb, 1949). Importantly, work suggests that transcranial direct current stimulation (tDCS) induces LTP-like changes in stimulated cortex (Stagg & Nitsche, 2011). Thus, tDCS may be uniquely suited to improve associative memory. In addition, past work has indicated that tDCS can improve episodic memory in older adults (see Coffman, Clark, & Parasuraman, 2014, for a review).

tDCS works by sending a slight electrical current through the scalp which alters neural activity in regions located under the electrodes (Nitsche & Paulus, 2000). tDCS effects operate via two mechanisms (see Stagg & Nitsche, 2011): first, changes in neuronal membrane polarization, or online effects, increase or decrease the likelihood of an action potential and in turn affect cognition. Second, increasing neuronal firing rates leads to LTP-like synaptic changes, which lead to after-effects, or offline effects, of stimulation that outlast tDCS application on the order of hours and days (for a review, see Passow, Thurm, & Li, 2017). Offline effects are particularly relevant to tests of memory, as these effects may result in memory improvements that persist after a period of consolidation (Au, Karsten, Buschkuehl, & Jaeggi, 2017). Indeed, recent meta-analytic evidence suggests that offline effects of tDCS are stronger than online effects when used in older adults (Hsu, Ku, Zanto, & Gazzaley, 2015). For example, tDCS applied during an object-location associative memory task led to improved performance in older adults when tested after a delay (Flöel et al., 2012), yet did not result in immediate performance gains. Although Flöel et al. (2012) hints at the effectiveness of tDCS for improving associative memory in older adults, the researchers used a probabilistic learning paradigm (i.e., object-location pairs were shown multiple times; sometimes the pairing was “correct” but sometimes the pairing was “incorrect”), and thus it is unknown whether tDCS would increase associative memory in older adults in a more traditional associative memory procedure.

Given our primary focus on improving associative memory through tDCS, we chose to apply stimulation over the left dorsolateral prefrontal cortex (dlPFC) because this region plays an essential role in associative memory (Murray & Ranganath, 2007), and further because altered dlPFC activity in older adults relative to the young is correlated with decreased associative memory performance (Dennis et al., 2008). In the current work, we investigate whether application of tDCS is effective in improving face-name associative memory in younger and older adults when stimulation is applied during encoding. To test both online and offline effects, we assessed memory immediately after encoding and after a 24-h delay. We shortened this delay from 1 week as done in past research (Flöel et al., 2012) to minimize floor effects with our associative memory task. We expected that active relative to sham tDCS would improve associative-memory for both retrieval sessions (immediate Day 1—and after a delay of 24 h—Day 2).

tDCS is effective in increasing memory performance in non-associative memory tasks (i.e., memory for single items) as measured by both recognition (e.g., Manenti, Brambilla, Petesi, Ferrari, & Cotelli, 2013) and recall (e.g., Flöel et al., 2012; Ross, McCoy, Coslett, Olson, & Wolk, 2011) in older adults, but rarely have both types of memory tests (recall, recognition) been used in the same study (Leach, McCurdy, Trumbo, Matzen, and Leshikar, 2016; Matzen, Trumbo, Leach, & Leshikar, 2015). Completing recall and recognition tasks may rely on different memory processes. Specifically, although recall tasks require participants to recollect information, participants may instead rely on feelings of familiarity (in addition to recollection) when completing recognition tasks (Yonelinas, 2002). Thus, it is important to assess memory via both measures to fully explore the effects of tDCS on memory. It is possible that the two types of memory measures differ in their sensitivity to the effects of stimulation. Thus, in the current work we use cued recall (seeing a face and typing in the associated name) and paired-associate recognition tasks (judging a face-name pair as intact or rearranged from other pairs) to serve as memory outcome measures.

Many tDCS studies that have included younger and older adult groups have found that tDCS effects on behavior are stronger in older adults than in the young (Perceval, Flöel, & Meinzer, 2016). This pattern of results is consistent across both motor (e.g., Zimerman et al., 2013) and cognitive domains such as fluency (Meinzer, Lindenberg, Antonenko, Flaisch, & Flöel, 2013) and name recall (Ross et al., 2011). One reason for this may be due to relatively high performance in younger adults, reducing room for improvement (i.e., ceiling effects). Another reason may stem from altered neural patterns in older adults that are more likely to be improved with tDCS (Hsu et al., 2015). Specifically, deficits in older relative to younger adults may stem from inefficient neural processing, which tDCS may be able to counteract (Meinzer et al., 2013). Although older adults tend to exhibit larger behavioral effects from tDCS across many cognitive domains (such as non-associative memory; Hsu et al., 2015), other lines of work suggest that this pattern may not extend to associative memory. Although past research indicates robust tDCS effects on associative memory in younger adults, including for object-location pairs (England, Fyock, Gillis, & Hampstead, 2015), text-font color pairs (Gray, Brookshire, Casasanto, & Gallo, 2015), and face-name pairs (Matzen et al., 2015; Pisoni, Vernice, Iasevoli, Cattaneo, & Papagno, 2015), work with older adults has been mixed. The only previous study to find associative memory improvements in older adults failed to find these effects immediately after the stimulation session (i.e., Flöel et al., 2012). Other work has shown null effects of stimulation (Cotelli et al., 2014), or that tDCS reduces associative memory performance (Leach et al., 2016). Given that older adults show pronounced deficits on associative memory compared to younger adults (Old & Naveh-Benjamin, 2008), it may be that improvements in associative memory are particularly difficult to bring about in older adults.

The main focus of this investigation was whether active relative to sham tDCS applied to the left dlPFC during encoding improves face-name associative memory in older and younger adults as measured by recall and recognition both immediately after stimulation as well as after a 24-h delay. We made two predictions: First, we hypothesized that tDCS will benefit associative memory performance, as measured by both recall and recognition. With respect to age, however, we saw different potential memory effects. Based on some previous work (e.g., Flöel et al., 2012), it may be that associative memory improves in older adults. However, given that older adults are especially vulnerable to associative memory deficits (Old & Naveh-Benjamin, 2008), it is also possible that older adults might not benefit from stimulation in line with prior mixed results (Leach et al., 2016; Cotelli et al., 2014). In contrast, because most prior work indicates that tDCS improves associative memory in younger adults, (i.e., England et al., 2015; Gray et al., 2015; Matzen et al., 2015; Pisoni et al., 2015), we expected to see associative memory improvements in the younger adults as measured by both recall and recognition. Second, we expected to find memory improvements both immediately following stimulation, and after a period of consolidation (i.e., delay), showcasing both online and offline effects of stimulation (Stagg & Nitsche, 2011) on memory. Further, the design of our study also allowed us to investigate differential age effects of tDCS depending on testing session (immediately vs. after a delay) and task type (recall vs. recognition), but we did not make strong predictions that either of these variables would interact with age.

Method

Participants

In total, 48 older adults (age 60–79; M = 65.63, SD = 4.90; 19 female) and 48 younger adults (age 18–35; M = 22.38, SD = 4.70; 29 female) participated in this study. We excluded participants with pacemakers, metal implants, abrasions to the scalp, skull fractures, brain injury, prior brain surgery, pregnancy, or a personal or familial history of epilepsy, as these are contraindications for tDCS. We also excluded participants who were unable to speak or understand English, or who were left-handed (given that hemispheric language dominance may differ between left- and right-handed individuals). All participants were recruited from the University of Illinois at Chicago or surrounding Chicago community and were paid for participation.

Materials

The faces used for this study come from a database of high-quality color photographs of people taken from the neck up in front of a grey background (Ebner, Riediger, & Lindenberger, 2010). The face set featured equal numbers of younger (aged 18–30), middle aged (aged 39–55), and older adults (aged 69–80) of both genders. Names were assigned to each face using the Social Security Administration’s lists of the most common names for the decade in which the pictured person was born, as we have done before (Matzen et al., 2015).

Participants completed a battery of neuropsychology tests to compare our younger and older adult sample (Park et al., 2002) which included tests of fluid and crystallized intelligence (See Table 1). t-tests indicated that our older adult sample fit the typical profile of diminished fluid intelligence compared to younger adults, including perceptual speed (measured with the digit comparison, t(94) = 6.49, p < .05, and digit symbol, t(94) = 5.04, p < .05, tasks) although we did not see age differences for working memory span, t(94) = 0.83, n.s., or verbal fluency, t(94) = 0.62, n.s. Additionally, older adults had higher vocabulary compared to the young, t(94) = 5.26, p < .05, which is consistent with prior findings of preserved or enhanced crystallized intelligence in older adults (Park et al., 2002). Our older adult sample completed the Mini Mental State Exam to help ensure that our sample was cognitively intact (MMSE; Folstein, Folstein, & McHugh, 1975). No older adult participant scored below 26. On all neuropsychological tests, we found no differences between active and sham conditions for either age group (all p’s > .08).

Table 1.

Means and Standard Deviations for Neuropsychological Measures Split by Age and Stimulation

Age	Condition	Digit Comp.	Digit Symbol	Digit Span	Verbal Fluency	Vocabulary
Younger	Active	85.33 (14.82)	47.96 (6.93)	18.00 (3.79)	101.54 (24.42)	30.42 (5.09)
	Sham	78.63 (12.11)	45.13 (5.98)	17.50 (2.64)	93.04 (19.18)	29.21 (4.41)
Older	Active	67.50 (12.91)	35.79 (8.53)	19.04 (3.88)	96.29 (27.91)	34.00 (4.27)
	Sham	62.71 (9.80)	36.25 (6.65)	17.25 (3.96)	94.63 (22.60)	34.83 (3.67)

Note. There was no significant difference between active and sham conditions on any of these measures within either age group (all p’s > .08).

Procedure Overview

The experiment was conducted on two consecutive days allowing us to examine immediate (on Day 1) and delayed effects (Day 2) of tDCS on memory. On Day 1, participants were first prepared for stimulation and then trained on the encoding, recall, and recognition tasks. Stimulation was administered 4 min prior to the start of the encoding task to allow participants to habituate to stimulation-induced sensations, and ceased four minutes after the participant completed the encoding session. In total, stimulation lasted exactly 25 min. Participants then immediately completed the recall and recognition tasks (not stimulated). After a 24-h delay, participants returned to the laboratory to complete a second version of the recall and recognition tasks (Day 2). Participants were not stimulated on Day 2.

tDCS

For all participants, stimulation was administered with ActivaTek ActivaDose II Controllers using square sponge electrodes (3.3 cm per side), similar to previous work (Clark et al., 2012). The anodal electrode was positioned on the scalp over left dlPFC (F3 in accord with the International 10–20 system) whereas the cathodal electrode was placed on the contralateral upper arm. In a double-blind scheme, the experimenter provided either active (1.5 mA; 24 younger adults, 24 older adults) or sham (0.1 mA; 24 younger adults, 24 older adults) stimulation while participants completed the encoding task.

Day 1 Procedure

Participants gave their informed consent and then were prepped for stimulation. After electrodes were positioned but before stimulation was applied, participants were trained on the encoding, cued recall, and paired-associate recognition tasks. Stimuli used in training were not shown again during the actual memory procedures. Next, participants completed baseline measures of mood. We collected mood ratings to measure whether stimulation altered affect (Barrett, Della-Maggiore, Chouinard, & Paus, 2004). Immediately after participants completed the mood questionnaire, stimulation started. Two minutes after stimulation began the participants completed a sensation questionnaire (Time 0). This was the first of five times participants completed this questionnaire. On this questionnaire, participants rated the levels of itching, burning, tingling, and mental fatigue on a 1 (very mild amount) to 10 (extremely high amount that is incredibly uncomfortable) scale. If any participant gave a rating 7 or higher, the experiment was to be discontinued; however, no participant did so. This same sensation questionnaire was provided at four more timepoints (Times 1–4) at evenly spaced intervals (approximately every 2.5 min) during 1-min breaks in between encoding blocks.

Encoding

Participants viewed 60 face-name pairs during the encoding session. On each encoding trial, participants saw a face and a name presented for 5000 ms (see Figure 1). Then, participants were given 3000 ms to subjectively decide whether each name “fit” the face it was paired with (1 = fits, 2 = does not fit), which is a procedure used previously (Sperling et al., 2001). This orienting task was used to ensure that participants actively attended to both the name and face. In addition, we wanted to control encoding strategies because past work suggests that older adults use sub-optimal strategies relative to the young if not given specific encoding instructions (Naveh-Benjamin, Brav, & Levy, 2007). Trials were presented in five blocks with 18 trials each, and participants completed sensation questionnaires in between blocks. Half (30) of the face-name pairs were presented only once during the encoding task, whereas the other half (30) were presented twice (we manipulated the number of stimulated encoding sessions for half of the trials to test a prevailing assumption in the tDCS literature that greater duration of stimulation leads to larger behavioral effects (e.g., Au et al., 2017). If this assumption is true, then it should be the case tDCS effects should be stronger for trials that are presented twice during a stimulated encoding session, compared to trials only stimulated once. No work to date has tested this assumption). Trials were presented in a pseudo-random order, and at least four trials separated the first from the second presentation of a pair that was presented twice.

Figure 1.

Three example trials of the encoding, cued recall, and paired-associate tasks.

Cued recall test

Immediately after encoding, participants started the recall task, where they viewed each of the faces seen during encoding. For each trial, participants were asked to type in the name that was associated with the face at encoding. Participants were told to type “no” if they did not remember the name to discourage guessing. After the response was entered, the next trial began. This continued until all 60 faces had been shown. Trials were presented in a pseudo-random order (different from the order used in the encoding task). Participants always completed the recall test before the recognition test. We did this because prior work shows that this order (recall before recognition) does not strongly influence performance on a subsequent recognition task (Darley & Murdock, 1971) compared to the reverse order (recognition before recall), and because prior tDCS work has used this same order of memory tests (e.g., Meinzer et al., 2014).

Recognition test

Immediately following the recall test, participants took the recognition test. During recognition, participants viewed each of the 60 faces shown during encoding. Each face was paired with a name. For half the trials, the correct name associated with the face was shown (intact trial). For the remaining half, a name associated with a different face during encoding (rearranged trial) was shown. For each trial, participants indicated whether the face-name pair was intact (by pressing the one key with their right index finger) or rearranged (by pressing the two key with their right middle finger). Once participants made this response, the next trial began. Trials in this task were presented in a pseudo-random order, with no more than four intact or rearranged trials presented in a row.

Post-test questionnaires

Immediately after the memory tests, the electrodes were removed and participants completed a set of questionnaires. First, participants rated their mood, which served as the posttest mood measure. Participants then took a blinding questionnaire by reporting whether they believed they received active or sham stimulation, or whether they “don’t know”, which is a procedure we have used before (Matzen et al, 2015). For these measures, 10 participants in the active condition (20.8% of active participants) guessed correctly, whereas 7 participants (14.6%) incorrectly guessed they were in the sham condition. In the sham condition, 6 participants (12.5% of sham participants) guessed correctly, whereas 7 participants (14.6%) incorrectly guessed they were in the active condition. The 66 remaining participants (68.8% of all participants; 35 active, 31 sham) reported not knowing which condition they were in. A Chi-Square Test of Independence indicated that these responses did not differ due to stimulation condition (active, sham), χ² (2, N = 96) = .85, n.s., thus we judged our blinding procedure to be effective. To end the procedure on Day 1, participants completed the neuropsychology tasks.

Day 2 Procedure

Exactly 24 h after the start of Day 1 procedures, participants returned to the laboratory for follow-up testing. During this session, participants completed the cued recall and recognition tests (not stimulated). Trials in both cued recall and recognition tests were presented in different pseudo-random orders from the Day 1 test. For the recognition task, trials that had been presented as an intact pair on Day 1 were presented as rearranged pair on Day 2 (and vice-versa).

Results

In this section, we report encoding responses, memory performance (recall and recognition), sensation ratings, and, finally, mood ratings. At encoding, participants found that 66.8% of the faces “fit” their name, 31.2% did not fit, and failed to respond to 2% of the face-name pairs. For recall, we calculated performance as the percentage of trials where participants correctly recalled the name, disregarding spelling errors. For recognition, we calculated a corrected measure of performance, A’, using the hit rates (HR), and false alarm rates (FAR; for formula, see Stanislaw & Todorov, 1999; for analyses of HR and FAR separately, see Supplementary Analyses). We also computed means for each age and stimulation group on both the sensation and mood questionnaires. Results from the cued recall task are presented in Table 2, and recognition results are shown in Table 3. To assess the extent to which stimulation influenced memory, we conducted 2 (Stimulation: active vs. sham) × 2 (Presentations: 1 vs. 2) × 2 (Session: Day 1 vs. Day 2) × 2 (Age: younger vs. older) mixed ANOVAs on both recall and corrected recognition (A’).

Table 2.

Means and Standard Deviations for Recall as a Function of Age, Day, and Stimulation Group

		One Presentation			Two Presentations
Age	Condition	Correct	Miss	Error	Correct	Miss	Error
Day 1
Younger	Active	0.17 (0.11)	0.64 (0.13)	0.19 (0.10)	0.39 (0.19)	0.38 (0.17)	0.23 (0.12)
	Sham	0.10 (0.09)	0.73 (0.13)	0.17 (0.09)	0.27 (0.12)	0.54 (0.14)	0.19 (0.10)
Older	Active	0.07 (0.09)	0.68 (0.13)	0.25 (0.10)	0.19 (0.15)	0.50 (0.18)	0.31 (0.18)
	Sham	0.09 (0.08)	0.68 (0.17)	0.23 (0.14)	0.20 (0.14)	0.52 (0.18)	0.28 (0.15)
Day 2
Younger	Active	0.23 (0.12)	0.55 (0.19)	0.22 (0.12)	0.39 (0.20)	0.39 (0.18)	0.22 (0.13)
	Sham	0.12 (0.12)	0.72 (0.18)	0.16 (0.10)	0.27 (0.13)	0.56 (0.17)	0.17 (0.11)
Older	Active	0.08 (0.10)	0.65 (0.18)	0.27 (0.16)	0.16 (0.14)	0.55 (0.18)	0.29 (0.17)
	Sham	0.09 (0.07)	0.62 (0.18)	0.29 (0.17)	0.18 (0.15)	0.54 (0.19)	0.28 (0.14)

Table 3.

Means and Standard Deviations for Recognition as a Function of Age, Day, and Stimulation Group

		One Presentation			Two Presentations
Age	Condition	Hit Rate	FA Rate	A’	Hit Rate	FA Rate	A’
Day 1
Younger	Active	0.74 (0.15)	0.32 (0.16)	0.78 (0.12)	0.87 (0.12)	0.28 (0.16)	0.88 (0.08)
	Sham	0.73 (0.14)	0.39 (0.15)	0.74 (0.14)	0.86 (0.11)	0.31 (0.15)	0.86 (0.08)
Older	Active	0.74 (0.16)	0.59 (0.23)	0.62 (0.20)	0.86 (0.10)	0.54 (0.21)	0.74 (0.13)
	Sham	0.83 (0.14)	0.61 (0.18)	0.69 (0.16)	0.86 (0.13)	0.48 (0.20)	0.78 (0.14)
Day 2
Younger	Active	0.72 (0.17)	0.32 (0.15)	0.77 (0.13)	0.83 (0.13)	0.20 (0.11)	0.88 (0.06)
	Sham	0.70 (0.16)	0.38 (0.15)	0.73 (0.13)	0.76 (0.14)	0.32 (0.18)	0.79 (0.13)
Older	Active	0.76 (0.17)	0.55 (0.21)	0.68 (0.15)	0.78 (0.16)	0.56 (0.26)	0.69 (0.15)
	Sham	0.75 (0.18)	0.52 (0.23)	0.70 (0.15)	0.79 (0.12)	0.51 (0.22)	0.72 (0.12)

Turning first to recall, we found a significant main effect of Stimulation, F(1, 92) = 4.10, p < .05, η² = .04, and a Stimulation x Age interaction, F(1,92) = 6.30, p < .05, η² = .06. Follow-up tests revealed that the Stimulation effects (active > sham) were driven by younger adults F(1,46) = 8.38, p < .05, η² = .15, (active M: .30, sham M: .13), but not older adults, F(1,46) < 1, n.s. (active M: .19, sham M: 0.14), showing that tDCS improved associative memory in the younger adults only. The main effect of stimulation was evident on both Day 1 (active M: 0.28, sham M: 0.18) and Day 2 (active M: 0.31, sham M: 0.19). The effect of Presentation was also significant, and driven by better memory for trials presented twice (M: 0.26) compared to once (M: 0.12), F(1,92) = 172.51, p < .05, η² = 0.65, but this did not interact with the effect of Stimulation, F(1,92) < 1, n.s. We also observed a Presentation × Age interaction, F(1,92) = 14.19, p < .05, η² = 0.13. Although both age groups showed greater memory for pairs presented twice than for those presented once, this effect was stronger for younger adults, F(1,46) = 158.80, p < .05, η² = 0.78 (twice M: 0.33, once M: 0.15) than for older adults, F(1,46) = 39.86, p < .05, η² = 0.46 (twice M: 0.18, once M: 0.08).

Turning to recognition, ANOVA results for A’ showed a non-significant main effect of Stimulation, F(1,91) < 1, n.s.; this was qualified, however, by a significant Stimulation × Age interaction, F(1,91) = 5.20, p < .05, η² = .05. For this interaction, active stimulation improved memory for younger adults, F(1,45) = 4.31, p < .05, η² = 0.09 (active M: 0.83, sham M: 0.68), but not for older adults, F(1,46) < 1, n.s. (active M: 0.78, sham M: 0.72). Similar to recall, we also found increased performance for trials presented twice (M: 0.79) versus once (M: 0.71), F(1,91) = 64.28, p < .05, η² = 0.41, but this did not interact with the effect of Stimulation, F(1,91) < 1, n.s. The Presentation x Age interaction was marginal, F(1,91) = 3.34, p = .07, η² = 0.04, reflecting the fact that the presentation effect was greater for younger adults, F(1,45) = 56.64, p < .05, η² = 0.58 (twice M: 0.85, once M: 0.75) than for older adults, F(1,46) = 16.83, p < .05, η² = 0.27 (twice M: 0.73, once M: 0.67).

Results from the sensation questionnaire are reported in Figure 2. To test effects of Age and Stimulation on reported physical sensation from stimulation, we combined measures of itching, burning, and tingling to create a sensation composite (Cronbach’s α = 0.93; we entered every physical sensation measure (itching, burning, tingling) at every timepoint (0–4) into this reliability analysis). We entered this composite into a 5 (Timepoint: time 0–4) × 2 (Stimulation: active vs. sham) × 2 Age (younger vs. older) mixed ANOVA. There were main effects of Timepoint, F(1,92) = 49.01, p < .05, η² = 0.35, Stimulation, F(1,92) = 45.46, p < .05, η² = 0.33, and age, F(1,92) = 6.73, p < .05, η² = 0.07, indicating that sensations dissipated over time, were greater in the active condition (M: 1.92) than the sham condition (M: 1.17), and were greater in younger participants (M: 1.67) than older participants (M: 1.40). These results were qualified by a significant Timepoint × Stimulation × Age interaction, F(1,92) = 5.74, p < .05, η² = 0.06, whereby the dissipation of physical sensations over time was stronger for younger adults in the active stimulation condition. To test if any of these sensations affected memory, we ran correlations between the physical sensation composites and both memory performance measures (Correct recall rate and A’). None of these correlations were significant (all p’s > .09).

Figure 2.

Self-reported sensations for active and sham groups as well as older and younger adults. Time 0 was administered 2 min after the start of stimulation, and Times 1–4 were administered during 1-min breaks in between encoding blocks. Participants rated their sensations on a 1–10 scale. Error bars depict standard error.

Mood rating results are reported in Figure 3. We entered each mood rating into a 2 (Timepoint: before vs. after) × 2 (Stimulation: active vs. sham) × 2 Age (younger vs. older) mixed ANOVA (due to experimenter error, two older adults, one from each condition, did not fill out mood measures and were excluded from this analysis). We report ANOVA results for each mood rating in this order: nervousness, excitement, fatigue, and inability to focus. A main effect of Timepoint for nervousness, F(1,90) = 48.18, p < .05, μ² = 0.35, indicated that participants became less nervous after stimulation and the memory tasks was over (i.e., pre-test compared to post-test). Similarly, we found that participants were more excited before stimulation than after, F(1,90) = 41.39, p < .05, μ² = 0.31. Results also indicated a main effect of Age for excitement, F(1,90) = 6.05, p < .05, μ² = 0.06, showing that older adults were less excited than younger adults. We found main effects of Stimulation, showing higher ratings from participants in the sham condition than those in the active condition, for ratings of fatigue, F(1,90) = 9.05, p < .05, μ² = 0.09, and inability to focus, F(1,90) = 4.04, p < .05, μ² = 0.04. For fatigue ratings, we also saw a main effect of Age, F(1,90) = 5.92, p < .05, μ² = 0.06, indicating that younger participants were less tired than older participants. Although some of these measures differed between stimulation condition and age groups, none of the mood measurements showed correlations with our memory outcome measures (all p’s > .10).

Figure 3.

Self-reported mood before and after stimulation for active and sham groups of both age groups. Participants reported their mood on a 0–5 scale. Error bars depict standard error.

Discussion

In this study, we examined the effects of tDCS applied during encoding over the left dlPFC on measures of associative recall and recognition in both younger and older adults. We report two primary findings: First, tDCS improved face-name associative memory in younger adults as measured by both recognition and recall, but not in older adults, suggesting that older adults might be resistant to stimulation-induced associative memory improvement (at least with the exact stimulation parameters used in the current study). Second, memory enhancement was evident in younger adults even after a 24-h delay, suggesting that tDCS may improve associative memory through offline stimulation effects. Together, these data suggest that tDCS improves associative memory in younger adults through both online and offline stimulation effects, but that tDCS is less effective in improving associative memory in older adults.

The primary aim of this study was to investigate the extent to which stimulation affected associative memory in older adults, and results indicated that tDCS did not improve associative memory in this population. Although some work suggests that behavioral effects from tDCS are stronger in older adults than in younger adults (Hsu et al., 2015; Meinzer et al., 2013; Ross et al., 2011; Zimerman et al., 2013), other work focusing on associative memory has been more mixed. In the case of associative memory, studies have variously found that tDCS either improves (Flöel et al., 2012), does not affect (Cotelli et al., 2014), or harms (Leach et al., 2016) performance. One reason associative memory may be so resistant to tDCS improvements is that older adults experience substantial deficits to associative memory that exceed performance deficits on non-associative memory tasks (Old & Naveh-Benjamin, 2008). Although poor performance often corresponds to larger tDCS effects in older adults as compared to younger adults (due to possible ceiling effects in younger adults; Hsu et al., 2015), deficits to associative memory may be more intractable and resistant to intervention. Indeed, past attempts to improve associative memory in older adults have not succeeded in improving performance to the level of younger adults (e.g., Rose & Yesavage, 1983).

A second reason we did not see improved memory performance in older adults could be that the specific stimulation parameters we used in this study were suboptimal. Aging is associated with functional reorganization of brain activity (e.g., Cabeza, 2002, Park & Reuter-Lorenz, 2009), and it may be that the location we stimulated in this investigation (i.e., left dlPFC) is not optimal for improving associative memory in older adults. Instead, it may take a different electrode location, or even bilateral stimulation of the dlPFC, to be effective at improving associative memory in older adults. Indeed, prior work has demonstrated that it takes stimulation of different cortical regions between older and younger adults to improve memory. For example, tDCS improves name recall, but only when stimulation is applied to the right hemisphere in younger adults, and the left hemisphere in older adults (Ross, McCoy, Wolk, Coslett, & Olson, 2010; Ross et al., 2011). Further, the lone study finding associative memory improvement in older adults applied stimulation to a different region (tempoparietal; Flöel et al., 2012) than we used in this study. Altogether, it may be that tDCS can improve associative memory in older adults, but only under different stimulation parameters. Future work will be necessary to investigate optimal stimulation parameters for older adults.

A third reason why we did not see improved memory performance in older adults may be due to physiological differences between younger and older adults that contributed to the differential age effects of tDCS on memory. Relative to the young, older adults have reduced cortical thickness and brain volume, particularly in prefrontal areas (West, 1996), which may impact the effectiveness of stimulation (e.g., Opitz, Paulus, Will, Antunes, & Thielscher, 2015). Specifically, greater atrophy means the distance between cortex and the scalp (and therefore the stimulating electrode) is larger, and this larger distance between scalp and cortex in older adults may render tDCS less effective compared to the young for a set dosage. Given that most studies with tDCS have used a current between 1 and 2 mA (Prehn & Flöel, 2015), it is possible that higher current strengths are needed to show equivalent effects in older adults, at least for associative memory. Further work is needed to understand how changes in brain morphology and other physiological changes might influence stimulation effects on cognition.

In contrast to the older adults, younger adults showed robust stimulation-induced memory improvements as measured by both recognition and recall. This is consistent with our prior work, and the work of others, that shows active relative to sham tDCS improves recall in younger adults (e.g., Matzen et al., 2015; Meinzer et al., 2013; Leshikar et al., 2017; Pisoni et al., 2015; Ross et al., 2010). Our results with recall, which requires active recollection of stimuli (as opposed to recognition memory, for which participants can rely on a feeling of familiarity with the stimuli; Yonelinas, 2002) suggests that tDCS improves recollection processes in younger adults. This is consistent with previous work demonstrating that tDCS enhances recollection (Gray et al., 2015). In addition, we also found evidence that tDCS improved memory in younger adults as measured by recognition. This replicates previous work that has found recognition memory improvements from tDCS in both non-associative and associative memory tasks (Chi, Fregni, & Snyder, 2010; England et al., 2015; Gray et al, 2015; Javadi & Walsh, 2012; Pisoni et al., 2015; Trumbo et al., 2016). Improving memory is an important field of study (Leshikar & Duarte, 2012, 2014; Leshikar, Duarte, & Hertzog, 2012; Leshikar, Dulas, & Duarte, 2015; Leshikar & Gutchess, 2015; Leshikar, Park, & Gutchess, 2014; McCurdy, Leach, & Leshikar, 2017), and the results of this experiment highlight the notion that tDCS can be used to improve associative memory. This work adds to emerging evidence that tDCS is effective at improving memory in younger adults.

We see two possible reasons why stimulation applied to dlPFC improved associative memory in younger adults. The dlPFC is vital for associative encoding (i.e., the building of to-be-remembered relationships between stimuli; Blumenfeld, Parks, Yonelinas, & Ranganath, 2011; Murray & Ranganath, 2007), and stimulation of this area may have facilitated associative (or relational) processing of face-name pairs during encoding, enabling better subsequent retrieval. Facilitation of this type would represent an online effect of stimulation. Second, dlPFC is also involved in the active search through memory stores during retrieval (Blumenfeld et al., 2011). Although we did not apply stimulation during the retrieval phase, it is possible that carryover effects of the stimulation may have facilitated a controlled, strategic memory search during retrieval procedures, that in turn led to improved memory performance in the young. These carryover effects would represent offline effects of stimulation on memory. Taken together, it is possible that stimulation of dlPFC led to improved memory through both online and offline effects.

Importantly, we found improved associative memory in younger adults when tested directly after encoding (i.e., Day 1), as well as after a 24-h delay (i.e., Day 2), showcasing the offline effects of stimulation evident after a period of consolidation. Finding offline effects on memory after a delay is consistent with recent work arguing that delayed tDCS effects on memory are as strong as online effects (Au et al., 2017). Our finding of delayed effects of stimulation on associative memory is consistent with one prior investigation that showed memory improvement through offline effects of stimulation (Flöel et al., 2012). It is worth noting, however, a potential limitation to this finding. Although we found stimulation-induced memory improvements for active relative to sham stimulation in younger adults on both days (Day 1 and Day 2), it is possible that the improved memory on Day 2 simply reflected better initial memory from Day 1. We see this as unlikely, however, because the stimulation effects (active > sham) were numerically larger for recall on Day 2 than Day 1, suggesting offline stimulation processes were affecting memory.

Although our blinding procedure was effective, one potential limitation of the study stems from differences in physical sensations and mood between stimulation conditions. In the current study, we used a rigorous procedure to measure physical sensations at five different timepoints throughout stimulation, allowing us to better observe differences in sensations between active and sham stimulation. This is a level of detail few other studies have used to understand differences between active and sham stimulation (though see Matzen et al., 2015 and Leshikar et al., 2017 for a similar procedure). Despite the effective blinding in this study, stimulation-induced sensation differences were evident between groups. We note, however, that sensations and mood were not correlated with memory performance, and thus unlikely to have impacted our primary memory findings. Still, the fact that sensation and mood differed between active and sham groups, suggests a need for greater care in future work to measure and account for stimulation-induced sensation differences between active and sham groups, and how that might relate to behavior.

In summary, we examined whether tDCS could be used to improve face-name associative memory, a task in which older adults show particular deficits compared to younger adults (Naveh-Benjamin et al., 2004). We found that tDCS is effective in improving face-name associative memory in younger adults, but not older adults, when applied to the left dlPFC. This shows that although tDCS is effective in some populations, tasks of associative memory may be resistant to tDCS induced performance improvements in older adults. For younger adults, improved performance compared to sham persisted after a 24-h delay, suggesting effects via offline mechanisms. This result suggests that prior tDCS results with younger adults may not generalize to older adult populations.

Supplementary Material

Supplementary data is available at The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences online.

Funding

This work was partially supported by NIA grant P30AG022849 provided through the Midwest Roybal Center for Health Promotion and Translation.

Acknowledgements

We would like to thank Drs. Bette Bottoms, David Wirtshafter, and Karl Szpunar for their invaluable input on both the project and the write up, and Diala Abughosh and Camill Burden for their help with data collection. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International, Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.