Impact of Depressive Symptoms on Memory for Emotional Words in Mild Cognitive Impairment and Late-Life Depression

Abstract

Background: Amnestic mild cognitive impairment (aMCI) and late-life depression (LLD) are associated with increased risk of Alzheimer’s disease (AD). This is also true for aMCI with concomitant depressive symptoms (aMCI/D+), but few studies have investigated this syndrome.

Objectives: We aimed to clarify the association between cognitive and depressive symptoms in individuals at risk for AD by examining episodic memory for emotional stimuli in aMCI, aMCI/D+, and LLD.

Methods: Participants were 34 patients with aMCI, 20 patients with aMCI/D+, 19 patients with LLD, and 28 healthy elderly adults. In an implicit encoding task, participants rated the emotional valence of 12 positive, 12 negative, and 12 neutral words. Immediately and 20 minutes later, participants recalled as many words as possible. They were also asked to identify previously presented words during a yes/no recognition trial.

Results: At immediate recall, aMCI participants displayed better recall of emotional words, particularly positive words. aMCI/D+ and control participants displayed better recall of positive and negative words compared to neutral words. LLD participants recalled more negative than neutral words. At delayed recall, emotional words were generally better-remembered than neutral words by all groups. At recognition, all subjects responded more liberally to emotional than to neutral words.

Conclusion:We find that the type of emotional information remembered by aMCI patients at immediate recall depends on the presence or absence of depressive symptoms. These findings contribute to identifying sources of heterogeneity in individuals at risk for AD, and suggest that the cognitive profile of aMCI/D+ is different from that of aMCI and LLD. Future studies should systematically consider the presence of depressive symptoms in elderly at-risk individuals.

Keywords

Aging Alzheimer’s disease dementia depression emotion episodic memory

INTRODUCTION

Alzheimer’s disease (AD) is a rapidly growing public health problem. To better understand its progression, most early diagnostic studies focus on individuals with amnestic mild cognitive impairment (aMCI) or late-life depression (LLD), two syndromes associated with increased dementia risk [1].

From a cognitive standpoint, aMCI is known to be quite heterogeneous. Episodic memory may be exclusively affected or be accompanied by deficits in other cognitive domains [2]. Consideration of neuropsychiatric factors, which have been reported in a large number of these patients [3], results in additional heterogeneity: it is known that the presence of neuropsychiatric symptoms increases the risk of dementia [4]. Thus, continued characterization of this particularly heterogeneous population is essential to detect dementia as early as possible [5].

Patients with LLD exhibit cognitive deficits like those seen in aMCI, including deficits in episodic memory [6]. LLD refers to a major depressive episode occurring generally after age 60. Even when depressive symptoms are subclinical, their presence has been shown to increase the risk of cognitive decline [7] and to modify the pattern of brain atrophy [8]. From a neuropsychological perspective, individuals with depressive symptoms and aMCI (aMCI/D+) present a cognitive profile that is distinct from that of aMCI or LLD. For instance, LLD [9] and aMCI/D+ participants [10 , 11, but see 12] generally perform better on tests of episodic memory than non-depressed aMCI individuals, but worse on tests of inhibition [11], set-shifting [13], and semantic memory [14, 15]. Interestingly, aMCI/D+ participants in these studies typically did not meet DSM-IV criteria for major depression. Rather, they were identified on the basis of self-rating scales or questionnaires, suggesting that depressive symptoms do not have to reach clinical severity to exert an effect on cognition. Much work remains to be done to clarify how depression contributes to cognition in patients with aMCI/D+, as well as how aMCI/D+ distinguishes itself from aMCI and LLD.

The investigation of emotional memory may be an interesting avenue to characterize the cognitive profiles of individuals at risk for developing AD. Findings from studies using emotional stimuli to test cognitive performance in subjects with aMCI and major depression suggest that these groups show different cognitive biases for certain types of emotional stimuli. So far, ten studies have investigated emotional memory in aMCI. Most have found memory enhancement for emotional relative to neutral stimuli [16 –23], and in some cases this phenomenon is particularly marked for positive stimuli [17 , 23]. Only two studies found no memory enhancement for emotional relative to neutral material [24, 25]. In healthy older adults, an emotional enhancement of memory performance has similarly been reported [26]. It may manifest as increased memory for positive compared to negative [e.g., 27, 28] or neutral [e.g., 29] material, or to an equal memory advantage for positive and negative relative to neutral information [e.g., 30].

Conversely, depressed adults typically show a robust advantage in remembering negative compared to positive and neutral material [for a review, see 31]. However, research on this topic in LLD is surprisingly scarce. The only studies, to our knowledge, that have examined emotional biases in LLD have done so using tasks measuring biases in attention [e.g., 32, 33] or emotion recognition [e.g., 34], not memory. In some of these studies, LLD patients exhibit the same negative emotional bias as do depressed younger adults [32, 33], but this is not always the case [34]. It seems possible to expect that on tests of emotional memory, LLD patients will exhibit better memory for negative compared to positive and neutral stimuli, although this question has never been directly addressed.

This study aimed to compare the cognitive profiles of aMCI, aMCI/D+, and LLD patients using an emotional episodic memory task. The specific objectives were twofold. The first objective was to compare overall memory performance between groups. It was hypothesized that, relative to healthy controls, patient groups would show decreased memory performance at immediate and delayed free recall compared to healthy controls, and that recognition performance would be more severely affected in aMCI than aMCI/D+, LLD and controls. The second objective was to investigate whether certain types of emotional words (positive or negative) are better remembered than neutral words. This question specifically addressed within-group performance (i.e., the groups were not compared to each other). We hypothesized that healthy controls and aMCI patients would recall and recognize more positive than neutral words, while aMCI/D+ and LLD participants would remember and recognize more negative than neutral words.

MATERIALS AND METHODS

Participants

Seventy-three patients with aMCI or LLD were referred by physicians in Quebec City. aMCI participants were referred to our study by their family physician because of concerns primarily relating to episodic memory, whereas LLD participants were referred because of concerns primarily related to mood. aMCI participants met Petersen’s [2] criteria for single- (n = 9) or multiple-domain aMCI (n = 38), which include 1) a subjective memory complaint, corroborated by an informant; 2) an objective memory impairment, based on a cutoff score of≥1.5 SD below age-adjusted norms on at least one standardized memory test in the neuropsychological battery (see Materials, below); 3) preserved general cognitive functioning; 4) largely intact functional activities; 5) not demented. LLD participants (n = 19) met DSM-IV criteria for major depressive disorder [35], which include 1) depressed mood and/or a loss of interest/pleasure in daily activities for at least two weeks; 2) impaired social and/or occupational functioning as a result of depressive symptoms; 3) at least five of nine additional symptoms relating to appetite, sleep, etc. These criteria were confirmed by consensus of a four-member panel of researchers and experienced clinicians.

Next, aMCI participants were split into “depressed” (aMCI/D+) and “non-depressed” groups on the basis of the 30-item Geriatric Depression Scale (GDS [36]). The validity of the GDS in cognitively impaired persons has been quite extensively studied, and it is generally agreed that the standard GDS cut-off of 11 [36] is inappropriate for use in patients with cognitive impairment for various reasons (lack of insight, reliance on self-report which may be inaccurate, memory impairment which may cause subjects to not remember how they were feeling in the previous week, etc.). Thus, as recommended by Debruyne and colleagues [37], we used a cut-off of 8 to classify aMCI/D+ participants in our sample (n = 20), as this cut-off was found to have the best sensitivity (95%) and specificity (67%) for detecting depression in these individuals [37]. aMCI participants who obtained a GDS score of≤7 (n = 34) were classified as non-depressed. While some previous research has examined the relationship between depression and cognition using a continuous measure of GDS (e.g., [10]), the decision to use a splitting method to classify aMCI and aMCI/D+ here was based on prior research proposing that aMCI and aMCI/D+ are two distinct groups, phenotypically and perhaps also etiologically [8 , 15]. LLD participants had to obtain a minimum standard GDS cut-off score of≥11 [36] in order to be included in the study. Of note, this study required that aMCI/D+ patients not meet DSM-IV criteria for major depression. Finally, 28 healthy elderly adults (CONT) were recruited through community advertisements. The final sample consisted of 34 aMCI, 20 aMCI/D+, 19 LLD and 28 CONT.

All participants were 55 years or older and reported French as their usual language. No participant self-reported a history of neurological disease, traumatic brain injury, stroke, psychiatric illness (other than depression), substance abuse, untreated metabolic conditions or uncorrected visual/auditory impairments, intracranial surgery, nor general anesthesia or oncologic treatment in the past six months. After a complete description of the study, participants provided written informed consent. The research protocol was approved by the Research Ethics Board of the Institut universitaire en santé mentale de Qué bec.

Materials

A comprehensive clinical and neuropsychological battery was first administered in order to confirm inclusion/exclusion criteria and to characterize all participants. General cognitive functioning was assessed using the Montreal Cognitive Assessment (MoCA) [38], and the Dementia Rating Scale (DRS-2) [39]. Depressive symptoms and vascular risk were assessed using the GDS [36] and Hachinski’s scale [40], respectively. The Test de rappel libre/rappel indicé á 16 items (RL/RI-16) [41] and the immediate recall (3 min) condition of the Rey-Osterrieth Complex Figure Test (ROCFT) [42] were used to assess verbal and visual episodic memory, respectively. The ROCFT also served to test visuo-constructive abilities. Visuo-perception was assessed using the size-match task from the Birmingham Object Recognition Battery (BORB) [43] and processing speed was assessed with the Coding subtest from the WAIS-III [44]. The 15-item Boston Naming Test (BNT) [45] as well as phonemic (T-N-P) and semantic (animals) fluency tests [46] were used to assess language, while the Pyramids and Palm Trees Test (PPTT) [47] served to assess semantic memory. Lastly, the Stroop from the D-KEFS battery [48] measured executive functioning. Results from these tests are presented in Table 1.

The experimental task was a computerized (E-Prime 2.0; Psychology Software Tools, Pittsburgh, PA) episodic memory task comprised of 72 French words, divided into three lists of 24 positive, 24 negative, and 24 neutral words (word lists and their English translation are found in the Supplemental materials section). Due to the lack of published norms for affective French words, the word lists were generated by members of our laboratory drawing upon examples of emotional words among those listed in Affective Norms for English Words [49]. The lists were piloted with elderly volunteers from our participant database to ensure the task’s validity and feasibility, as well as to ensure that positive, negative and neutral items were indeed perceived as such by participants. Each of the three emotional lists was then divided into two pseudo-random lists of 12 words each; one was presented during the study phase and the other served as a distracter list during the recognition phase.

Procedure

Experimenters were blind to participants’ diagnosis. During the study phase, words were presented individually in the center of the screen in complete random order (i.e., non-blocked design). Subjects were asked to read each word aloud and to assess the degree to which each word was negative or positive, on a scale from 1 (very negative) to 9 (very positive). This ensured that the words conveyed the intended valence, which is important because subjective interpretation of emotional material can influence memory performance for these items [50]. Secondly, our procedure allowed for implicit encoding of words, which is associated with better memory performance in adults with cognitive impairment [51].

Immediately after the study phase, participants were asked to count backwards from 100 for one minute. They were then asked to orally recall as many words as possible. Answers were recorded verbatim by the examiner. After a 20-minute delay, participants were asked again to recall as many words as possible. Subjects were then asked to count by 2 starting with 20 for one minute, then shown 36 targets and 36 distracters individually in random order and asked to decide (yes/no) whether each word had been presented previously.

In order to establish whether there were differences in the lists’ emotional arousal, a second, separate pool of 64 subjects (21 aMCI, 14 aMCI/D+, 7 LLD, and 22 CONT) were additionally asked to rate the words on a scale from 1 (not at all arousing) to 9 (very arousing). These subjects did not participate in any other portion of this study.

Statistical analyses

Sociodemographic and clinical variables were compared between groups using one-way ANOVAs with Bonferroni corrections. A chi-square test was performed on participants’ sex.

To verify that positive, negative and neutral items were perceived as such, mean ratings assigned to stimuli by 93 participants (data were lost for eight subjects) were compared between groups using three one-way ANOVAs with Bonferroni correction. This procedure was repeated using the arousal data obtained from the separate pool of 64 subjects.

Inspection of free recall and recognition data revealed that the dependent variables were not normally distributed (Shapiro-Wilks test p≤0.001 in all cases); thus, nonparametric analyses were performed. To verify the first main hypothesis regarding overall memory performance, the total number of words recalled at immediate recall was compared between groups using Kruskal-Wallis analysis. When this test reached significance, Wilcoxon tests were performed to compare performance pairwise between groups, using a corrected alpha level of 0.008 (Bonferroni procedure = 0.05/6). These analyses were repeated with the total number of words recalled at delayed free recall. For recognition scores, sensitivity (A′) and response bias (B′) measures were computed according to methods described by Donaldson [52] using data from 99 participants (data were lost for two participants). Values of A′ fluctuate between 0 and 1, with higher scores reflecting better sensitivity. Values of B′ fluctuate between –1 and 1, with negative values reflecting liberal responding (tendency to respond “yes”) and positive values reflecting conservative responding (tendency to respond “no”). A′ and B′ were compared between groups using Kruskal-Wallis tests followed by Wilcoxon tests.

As for our second hypothesis regarding the type of words remembered by each group, Friedman tests were used to compare word recall, sensitivity (A′) and response bias (B′) scores of positive, negative and neutral words in each group. When these tests reached significance, Wilcoxon signed-rank tests were performed to compare performance pairwise between conditions, with a corrected alpha level of 0.017 (Bonferroni procedure = 0.05/3).

In order to verify comparability of our results with those of previous research that reported parametric analyses in most cases, the above analyses were repeated using parametric analyses of variance and paired t–tests. To ensure their robustness, ANOVAs were age-, sex-, and education-corrected.

RESULTS

Participant characteristics

As shown in Table 1, the groups did not differ statistically in terms of age or education. The aMCI group included a greater men-to-women ratio than the other groups. Scores on clinical and neuropsychological tests are summarized in Table 1.

Valence and arousal ratings are summarized in Table 2. Participants’ ratings of the word lists were consistent with the intended valence of the stimuli. There were no group differences in subjective ratings between valence categories, indicating that all groups perceived the intended emotional valence of the stimuli to a similar extent (i.e., all groups gave similarly positive ratings to positive words, etc.). Similarly, there were no group differences in arousal ratings.

As determined using the OMNILEX Database (http://www.omnilex.uottawa.ca), there was no difference in subjective frequency between positive (M = 4.9, SD = 1.0), negative (M = 4.8, SD = 0.7) and neutral targets (M = 4.8, SD = 1.7), F (2, 35) = 0.044, p = 0.957. The same was true for frequency of positive (M = 4.2, SD = 2.0), negative (M = 4.3, SD = 1.5) and neutral distracters (M = 3.7, SD = 1.9), F (2, 35) = 0.316, p = 0.731. Imageability was not taken into account during the conception of the task, and post-hoc inspection of the data using one-way ANOVA revealed that neutral target words had higher imageability ratings (M = 4.9, SD = 1.7) than positive (M = 3.5, SD = 0.6) and negative words (M = 3.6, SD = 1.0), F (2, 35) = 5.030, p = 0.012. Neutral distracters (M = 3.9, SD = 2.0) were not more imageable than positive (M = 2.8, SD = 1.5) or negative distracters (M = 3.1, SD = 1.2), F (2, 35) = 1.430, p = 0.254.

Overall memory performance

At immediate recall, there were significant between-group differences in performance (χ² = 42.2, p < 0.001). Specifically, Wilcoxon signed-rank tests showed that aMCI (M = 3.3, SD = 3.0, p < 0.001) and aMCI/D+ (M = 4.3, SD = 3.2, p < 0.001) subjects recalled fewer words than CONT subjects (M = 9.7, SD = 3.1), and aMCI subjects recalled fewer words relative to LLD patients (M = 6.9, SD = 4.2, p = 0.002). There were no differences between CONT and LLD (p = 0.022), between aMCI and aMCI/D+ (p = 0.223) or between aMCI/D+ and LLD (p = 0.047). At delayed recall, there were significant between-group differences as well (χ² = 44.9, p < 0.001). aMCI (M = 2.7, SD = 2.7, p < 0.001), aMCI/D+ (M = 3.7, SD = 2.9, p < 0.001), and LLD patients (M = 5.8, SD = 3.4, p = 0.004) were impaired relative to CONT (M = 8.7, SD = 2.5), and aMCI patients were impaired relative to LLD patients (p = 0.001). There were no differences between aMCI and aMCI/D+ (p = 0.190) or between aMCI/D+ and LLD (p = 0.045). At recognition, the groups differed in their sensitivity in detecting targets from distracters, (χ² = 31.4, p < 0.001), with aMCI (M = 0.5, SD = 0.1, p < 0.001) and aMCI/D+ (M = 0.6, SD = 0.1, p < 0.001) patients displaying poorer performance than CONT (M = 0.7, SD = 0.1), and aMCI patients displaying poorer performance than LLD patients (M = 0.6, SD = 0.1, p < 0.001). There were no group differences in response bias scores (χ² = 4.2, p = 0.245).

Emotional memory biases

Results from the different conditions of the experimental task are summarized in Table 3. At immediate recall, there were differences in the type of words recalled by the CONT (χ² = 18.9, p < 0.001), aMCI (χ² = 12.7, p = 0.002), aMCI/D+ (χ² = 9.9, p = 0.007) and LLD (χ² = 8.9, p = 0.012) groups. Wilcoxon signed-rank tests subsequently revealed that CONT participants recalled more positive (p < 0.001) and negative (p < 0.001) than neutral words. aMCI participants recalled more positive (p = 0.001) and negative (p = 0.014) than neutral words, and more positive than negative words (p = 0.014). aMCI/D+ patients recalled more positive (p = 0.004) and negative (p = 0.009) than neutral words. The LLD group recalled more negative than neutral words (p = 0.015).

At delayed recall, there were differences in the type of words recalled by the CONT (χ² = 27.6, p < 0.001), aMCI (χ² = 22.3, p < 0.001), aMCI/D+ (χ² = 9.2, p = 0.010) and LLD groups (χ² = 16.9, p < 0.001). Wilcoxon signed-rank tests showed that CONT participants recalled more positive (p < 0.001) and negative (p < 0.001) than neutral words. aMCI patients similarly recalled more positive (p < 0.001) and negative (p = 0.001) than neutral words. aMCI/D+ patients also recalled more positive (p = 0.005) and negative (p = 0.011) than neutral words. Similarly, LLD patients recalled more positive (p = 0.001) and negative (p = 0.006) than neutral words.

In terms of sensitivity, Friedman tests were non-significant for all groups (all p > 0.05). Response bias scores are illustrated in Fig. 1. Friedman tests were highly significant for all groups (all p < 0.001). Wilcoxon signed-rank tests showed that CONT participants’ responses to positive (p < 0.001) and negative (p < 0.001) stimuli were more liberal than responses to neutral stimuli. aMCI patients were more liberal in their responses to positive (p < 0.001) and negative (p < 0.001) words than neutral words, and more liberal in their responses to positive than negative words (p = 0.006). aMCI/D+ patients were more liberal in their responses to positive (p < 0.001) and negative (p < 0.001) than neutral words. LLD patients also responded more liberally to positive (p < 0.001) and negative (p < 0.001) than neutral words.

All results remained unchanged when analyses were repeated using age-, sex-, and education-corrected parametric ANCOVAs.

Antidepressant medication use

Because the use of antidepressant medication is known to influence cognition [e.g., 55], information regarding the duration of antidepressant use was collected from the medical files of LLD patients for use in post-hoc analyses, to determine whether this variable was associated with experimental task performance. Data were missing from the files of four LLD patients, therefore exploratory correlational analyses were conducted on data from the remaining 15 individuals. One moderate positive correlation emerged between duration of medication use and the number of negative words retrieved at delayed recall (Pearson r = 0.66; p = 0.014). No other variables were significantly correlated with duration of antidepressant use (all p > 0.05).

DISCUSSION

The aim of this study was to compare memory for emotional words between aMCI, aMCI/D+, and LLD patients. Two main findings emerged. First, as expected, overall memory performance in patients was impaired relative to CONT participants. Second, the central finding of the present study was the manifestation of specific biases in participants’ recall and recognition of emotional stimuli.

Overall memory performance

Our finding of impaired free recall in all three patient groups is consistent with results from previous studies in aMCI and aMCI/D+ patients [10, 11] as well as in LLD [6]. At recognition, aMCI subjects showed poorest sensitivity in detecting targets, while performance of aMCI/D+ was intermediate between that of aMCI and LLD subjects. As expected, memory performance of LLD subjects was comparable to that of controls on recognition trials. Memory performance in individuals with aMCI/D+ [11] and LLD [56, 57] has previously been shown to improve with cueing compared to free recall. A study by Dierckx et al. [9] found that cued recall reliably distinguished MCI from LLD participants, as only the former were impaired on this task. Similarly, Hudon and colleagues [11] reported that immediate free recall was impaired in aMCI and aMCI/D+ participants, but improved with cueing only in the aMCI/D+ group. This pattern of results suggests that depressive symptoms are associated with inefficient self-retrieval [58]. One might speculate that abnormal prefrontal functioning, which is known to be associated with depressive symptoms [59], may account for these results. Neuroimaging techniques would allow verification of this possibility. Alternatively, depressive symptoms may appear as a reaction to cognitive change and may thus suggest preserved insight; conversely, the absence of this depressive reaction may indicate a more severe degree of pathological change in the brain which may be associated with worse episodic memory performance.

Emotional memory biases

The central question of interest in this paper was whether depressive symptoms would influence the type of information remembered by cognitively impaired patients. Immediately after viewing a list of emotional words, controls remembered more positive and negative words compared to neutral words, largely consistent with the positivity effect reported in many studies (reviewed by [60]). Immediate recall in the aMCI group was better for positive than negative and neutral items, which was somewhat consistent with our predictions that positive items would be better-remembered than neutral items only. As outlined earlier, there has been some debate in the literature regarding how memory for positive material compares to memory for negative and neutral material: each of the ten studies that have addressed this issue in aMCI have found slightly different results. While some studies reported better memory for positive items only [17 , 23 (delayed recall)], others found enhanced memory for emotional over neutral items, regardless of whether they were positive or negative [16 , 23 (immediate recall)]. Still others found no benefit of emotional content on memory performance in either aMCI, AD, or healthy control groups [24, 25]. In some cases, comparability of results between studies is limited by the fact that a “neutral” experimental condition is compared to either a positive condition only [25] or a negative condition only [19], not both.

The concept of MCI is heterogeneous, and this could partly explain the differences between the studies. Additional explanations for these inconsistencies include variability in the stimuli used between studies (words [16 , 24]; sentences [17]; images [23, 25]; faces [19 , 22]; events [21]). The experimental procedure also likely played a role in performance: In some studies, encoding was explicit [16 , 22–24] whereas in others it was implicit [21, 25]. Processes used at encoding have been shown to influence subsequent memory performance in older adults, particularly those with memory impairments [51]. The results we report here suggest a third interpretation: the type of recall paradigm used to test episodic memory may influence the emotional enhancement effect. Papers in the extant literature have used immediate [18 , 24] and delayed [21 , 24] free recall to test the emotional enhancement effect, but the majority have employed recognition paradigms [16 , 25]. It is crucial to consider these differences for two important reasons.

First, the passage of time is known to affect emotional enhancement effects in young [61, 62] and older [63] healthy adults, which may explain why in this study and others, emotional information was better recalled as a whole in relation to neutral information after a 20-minute delay, whereas this was less apparent at immediate recall. According to authors who have studied this effect [61 –63], emotionally salient information remains especially accessible as the memory trace fades with time. Our results suggest that this emotionally-mediated memory consolidation effect over time is also present in patients with memory impairments.

Second, immediate recall, delayed recall, and recognition tasks impose different demands on retrieval processes. Delayed recall requires effortful processing involving self-initiated retrieval strategies, in contrast to immediate recall which relies more on attention and short-term/working memory capacities than delayed recall, and recognition which may rely on familiarity or gist-based strategies. It is altogether plausible to suppose that mood-congruent episodic memory biases may manifest themselves more readily when retrieval cognitive demands are lower (for instance, at immediate recall or recognition). In these cases, the lighter cognitive load allows cognitive resources to be allocated to regulate or re-interpret affective responses to emotional stimuli, which has been shown to account for emotional enhancement effects in older adults [64]. One neuroimaging study recently reported that, specifically in patients with aMCI, different brain regions are solicited in the retrieval of emotional information at immediate and delayed recall [24]. Additional well-controlled studies are needed to directly compare circumstances under which aMCI individuals may exhibit memory biases for emotional information.

Results from aMCI/D+ and LLD participants were partly consistent with our hypotheses, which were that both these groups would show enhanced free recall of negative relative to neutral words. Only LLD participants showed this pattern at delayed recall, while aMCI/D+ showed a benefit for both types of emotional information. At delayed recall, positive words were just as well-remembered as negative words by both depressed groups. These findings are surprising considering the wealth of prior evidence for a negative memory bias in depressed individuals (for a review, see [31]). However, previous studies have all been conducted with young adults; the present study is the first to investigate the emotional enhancement effect on memory performance in aMCI/D+ and LLD. It has been suggested that the negative cognitive bias reported in young adults lessens as depressed individuals get older [65], perhaps due to age-related brain changes within the limbic system, which is known to modulate memory performance [66]. It may also be possible that these participants were able to successfully engage goal-directed processes aiming to downregulate the emotional impact of the negative words, a phenomenon which has been associated with aging [60]. Additional studies are needed to replicate and extend these findings using other emotional memory tasks.

All participants displayed a liberal bias in responding to emotional stimuli at recognition, while a conservative bias was apparent for neutral words. Only aMCI participants showed a more liberal bias to positive than negative words. Strong semantic associations between some emotional words (e.g., happiness, joy) likely produced a false-memory effect by activating semantic networks, as has been shown in other studies [e.g., 67]. Neutral items, in contrast, were more semantically distinct (e.g., shoelace, counter) and non-emotional distracters may have been less likely to elicit feelings of familiarity during the recognition task.

Imageability values differed between the neutral and emotional target words, with neutral words having higher imageability ratings. These differences are unlikely to have influenced our results or conclusions, for several reasons. First, there is some evidence that imageability does not affect recall for word lists in elderly participants (e.g., [53]). Second, in studies conducted with younger adults, imageable/concrete words are typically better-remembered than abstract words [54]; thus, any emotional memory bias seen here may actually be underestimated.

Interestingly, valence ratings assigned to positive, negative, and neutral word lists at encoding were similar between groups. Considering that depressed participants typically tend to misattribute negative (sad) labels to positive (happy), neutral or ambiguous stimuli [68], the fact that the LLD group did not differ from other groups in their assigned ratings was rather surprising. However, previous reports of negative cognitive bias in depression have come largely from studies using facial or prosodic stimuli [69], which are arguably more open to subjective interpretation than are words, which often have an unambiguous definition. Further exploration of emotion identification in word lists is warranted.

Antidepressant medication use

Exploratory post-hoc analyses revealed that patients with the longest antidepressant use also showed the strongest negative bias at delayed recall. These patients may be non-responders to antidepressant medication and may present a negative memory bias that is absent in patients who successfully respond to antidepressant treatment [70]. Our data did not allow for exhaustive analyses between antidepressant use and task performance, and no information was collected regarding other types of medication that may be associated with neuropsychiatric symptoms (e.g., cholinesterase inhibitors), but future studies should consider these variable in cohorts of depressed participants.

Limitations

It is possible that duration and/or number of depressive episodes may in part explain the relationship between medication use and negative bias. Regrettably, data regarding the age of onset of the first depressive symptoms or the number of previous episodes in the LLD group were not collected; as such, we were unable to examine possible differences related to between early- and late-onset cases of LLD, duration or number of previous depressive episodes. In addition, the lower age limit for this study was set to 55 years because in many cases the onset of the prodromal stage of AD occurs around this time; however, a limitation to this approach is that it may have increased the number of false positive aMCI cases. Using different GDS cut-off scores to assess depression in aMCI/D+ and LLD groups may be viewed as a third limitation, as this procedure may have introduced imbalance between the groups. Moreover, in interpreting our findings the higher male:female ratio in the aMCI group must be considered, as gender effects on emotional responses are well-known [71]. We attempted to correct for this difference by using sex-corrected parametric analyses and found that our results were unchanged, suggesting that sex differences are not likely to account for our findings. A final limitation is the relatively small size of our sample, although the present study is the largest one conducted to date examining emotional memory performance in individuals with mild memory difficulties.

Conclusions

This study is the first to show that the type of emotional information remembered by aMCI patients depends on the presence or absence of depressive symptoms. Memory performance in non-depressed aMCI individuals appears to be influenced more strongly by positive information, particularly when retrieval cognitive demands are lower (for instance, at immediate recall or recognition). This is not true in the presence of depressive symptoms.

Globally, the results support the perspective that individuals with aMCI and aMCI/D+ show different cognitive profiles, a question which has received only little attention until now. Our data contribute to the growing body of literature aiming to clarify the nosological characteristics of aMCI by clarifying sources of heterogeneity within this syndrome (i.e., depressive symptoms) that may account for, or help to explain, certain cognitive symptoms. Finally, future studies aiming early identification of AD should systematically consider the presence of depressive symptoms in elderly at-risk individuals.

Footnotes

ACKNOWLEDGMENTS

This study was funded by a research grant (#10 59 to CH) and a doctoral scholarship (#10 81 to BLC) from the Alzheimer Society of Canada, and a Chercheur-boursier, Senior salary award from the Fonds de la recherche du Québec - Santé (CH).

Authors’ disclosures available online ().

Appendix

References

Ownby

, Crocco

, Acevedo

, John

, Loewenstein

(2006) Depression and risk for Alzheimer disease - Systematic review, meta-analysis, and metaregression analysis. Arch Gen Psychiatry 63, 530–538.

Petersen

(2004) Mild cognitive impairment as a diagnostic entity. J Intern Med 256, 183–194.

Apostolova

, Cummings

(2008) Neuropsychiatric manifestations in mild cognitive impairment: A systematic review of the literature. Dement Geriatr Cogn Disord 25, 115–126.

Rosenberg

, Mielke

, Appleby

, Oh

, Geda

, Lyketsos

(2013) The association of neuropsychiatric symptoms in MCI with incident dementia and Alzheimer disease. Am J Geriatr Psychiatry 21, 685–695.

Hughes

, Snitz

, Ganguli

(2011) Should mild cognitive impairment be subtyped? Curr Opin Psychiatry 24, 237–242.

Herrmann

, Goodwin

, Ebmeier

(2007) The cognitive neuropsychology of depression in the elderly. Psychol Med 37, 1693–1702.

Steffens

(2012) Depressive symptoms and mild cognitive impairment in the elderly: An ominous combination. Biol Psychiatry 71, 762–764.

Morin

J-F

, Mouiha

, Pietrantonio

, Duchesne

, Hudon

(2012) Structural neuroimaging of concomitant depressive symptoms in amnestic mild cognitive impairment: A pilot study. Dement Geriatr Cogn Dis Extra 2, 573–588.

Dierckx

, Engelborghs

, De Raedt

, De Deyn

, Ponjaert-Kristoffersen

(2007) Differentiation between mild cognitive impairment, Alzheimer’s disease and depression by means of cued recall. Psychol Med 37, 747–755.

10.

Bruce

, Bhalla

, Westervelt

, Davis

, Williams

, Tremont

(2008) Neuropsychological correlates of self-reported depression and self-reported cognition among patients with mild cognitive impairment. J Geriatr Psychiatry Neurol 21, 34–40.

11.

Hudon

, Belleville

, Gauthier

(2008) The association between depressive and cognitive symptoms in amnestic mild cognitive impairment. Int Psychogeriatrics 20, 710–723.

12.

Johnson

, Mauer

, Jahn

, Song

, Wyshywaniuk

, Hall

, Balldin

, O’Bryant

(2013) Cognitive differences among depressed and non-depressed MCI participants: A project FRONTIER study. Int J Geriatr Psychiatry 28, 377–382.

13.

Zahodne

, Tremont

(2013) Unique effects of apathy and depression signs on cognition and function in amnestic mild cognitive impairment. Int J Geriatr Psychiatry 28, 50–56.

14.

Brunet

, Hudon

, Macoir

, Belleville

, Rousseau

, Bouchard

, Verret

, Chertkow

, Chayer

, Kergoat

M-J

, Joubert

(2011) The relation between depressive symptoms and semantic memory in amnestic mild cognitive impairment and in late-life depression. J Int Neuropsychol Soc 17, 865–874.

15.

Callahan

, Joubert

, Tremblay

, Macoir

, Belleville

, Rousseau

, Bouchard

, Verret

, Hudon

(2015) Semantic memory impairment for biological and man-made objects in individuals with amnestic mild cognitive impairment or late-life depression. J Geriatr Psychiatry Neurol 28, 108–116.

16.

Brueckner

, Moritz

(2009) Emotional valence and semantic relatedness differentially influence false recognition in mild cognitive impairment, Alzheimer’s disease, and healthy elderly. J Int Neuropsychol Soc 15, 268–276.

17.

Marreneca

, Martins

, Mendonça

, Guerreiro

(2006) Emotional memory in mild cognitive impairment. Alzheimers Dement 2, S295–S296.

18.

Nieuwenhuis-Mark

, Schalk

, de Graaf

(2009) Free recall and learning of emotional word lists in very elderly people with and without dementia. Am J Alzheimers Dis Other Dementias 24, 155–162.

19.

Schefter

, Werheid

, Almkvist

, Lönnqvist-Akenine

, Kathmann

, Winblad

(2013) Recognition memory for emotional faces in amnestic mild cognitive impairment: An event-related potential study. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn 20, 49–79.

20.

Wang

, Li

, Jiang

, Bao

, Zhang

(2013) Is emotional memory enhancement preserved in amnestic mild cognitive impairment? Evidence from separating recollection and familiarity. Neuropsychology 27, 691–701.

21.

Waring

, Seiger

, Solomon

, Budson

, Kensinger

(2014) Memory for the 2008 presidential election in healthy ageing and mild cognitive impairment. Cogn Emot 28, 1407–1421.

22.

Werheid

, Gruno

, Kathmann

, Fischer

, Almkvist

, Winblad

(2010) Biased recognition of positive faces in aging and amnestic mild cognitive impairment. Psychol Aging 25, 1–15.

23.

Gorenc-Mahmutaj

, Degen

, Wetzel

, Urbanowitsch

, Funke

, Schröder

(2015) The positivity effect on the intensity of experienced emotion and memory performance in mild cognitive impairment and dementia. Dement Geriatr Cogn Dis Extra 5, 233–243.

24.

Mistridis

, Taylor

, Kissler

, Monsch

, Kressig

, Kivisaari

(2013) Distinct neural systems underlying reduced emotional enhancement for positive and negative stimuli in early Alzheimer’s disease. Front Hum Neurosci 7, 939.

25.

Parra

, Pattan

, Wong

, Beaglehole

, Lonie

, Wan

, Honey

, Hall

, Whalley

, Lawrie

(2013) Medial temporal lobe function during emotional memory in early Alzheimer’s disease, mild cognitive impairment and healthy ageing: An fMRI study. BMC Psychiatry 13, 76.

26.

Carstensen

, Mikels

(2005) At the intersection of emotion and cognition - Aging and the positivity effect. Curr Dir Psychol Sci 14, 117–121.

27.

Mather

, Carstensen

(2005) Aging and motivated cognition: The positivity effect in attention and memory. Trends Cogn Sci 9, 496–502.

28.

Mather

, Knight

(2005) Goal-directed memory: The role of cognitive control in older adults’ emotional memory. Psychol Aging 20, 554–570.

29.

Kensinger

, Brierley

, Medford

, Growdon

, Corkin

(2002) Effects of normal aging and Alzheimer’s disease on emotional memory. Emotion 2, 118–134.

30.

Charles

, Mather

, Carstensen

(2003) Aging and emotional memory: The forgettable nature of negative images for older adults. J Exp Psychol Gen 132, 310–324.

31.

Gotlib

, Joormann

(2010) Cognition and depression: Current status and future directions. Annu Rev Clin 6, 285–312.

32.

Broomfield

, Davies

, MacMahon

, Ali

, Cross

SMB

(2007) Further evidence of attention bias for negative information in late life depression. Int J Geriatr Psychiatry 22, 175–180.

33.

Dudley

, O’Brien

, Barnett

, McGuckin

, Britton

(2002) Distinguishing depression from dementia in later life: A pilot study employing the Emotional Stroop task. Int J Geriatr Psychiatry 17, 48–53.

34.

Mah

, Pollock

(2010) Emotional processing deficits in late-life depression. Am J Geriatr Psychiatry 18, 652–656.

35.

Association

(2000) Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition: DSM-IV-TR, American Psychiatric Association, Arlington, VA.

36.

Yesavage

, Brink

, Rose

, Lum

, Huang

, Adey

, Leirer

(1983) Development and validation of a geriatric depression screening scale - A preliminary report. J Psychiatr Res 17, 37–49.

37.

Debruyne

, Van Buggenhout

, Le Bastard

, Aries

, Audenaert

, De Deyn

, Engelborghs

(2009) Is the geriatric depression scale a reliable screening tool for depressive symptoms in elderly patients with cognitive impairment? Int J Geriatr Psychiatry 24, 556–562.

38.

Nasreddine

, Phillips

, Bédirian

, Charbonneau

, Whitehead

, Collin

, Cummings

, Chertkow

(2005) The Montreal Cognitive Assessment, MoCA: A brief screening tool for mild cognitive impairment. J Am Geriatr Soc 53, 695–699.

39.

Jurica

, Leittan

, Mattis

(2001) DRS-2: Dementia Rating Scale—2 Professional Manual, Psychological Assessment Resources, Lutz, FL.

40.

Hachinski

, Iliff

, Zilhka

, DuBoulay

, McAllister

, Marshall

, Russell

, Symon

(1975) Cerebral blood flow in dementia. Arch Neurol 32, 632–637.

41.

Van der Linden

(2004) L’Évaluation des Troubles de la Mémoire: Présentation de Quatre Tests de Mémoire Épisodique (avec leur Étalonnage), Solal Éditeurs Marseille, France.

42.

Rey

(1941) Psychological examination in cases of traumatic encephalopathy. Arch Psychol (Geneve) 28, 286–340.

43.

Riddoch

, Humphreys

(1993) Birmingham Object Recognition Battery (BORB), Psychology Press.

44.

Weschler

(1997) Wechsler Adult Intelligence Scale-III, The Psychological Corporation, San Antonio, TX.

45.

Kaplan

, Goodglass

, Weintraub

(1983) The Boston Naming Test, Lea & Febiger, Philadelphia, PA.

46.

Consortium des Universités de Montréal et McGill (1996) Manuel de l’examen neuropsychologique, Étude Santé et Vieillissement au Canada, Montréal, QC.

47.

Howard

, Patterson

(1992) The pyramids and palm trees test: A test of semantic access from words and pictures, Thames Valley Test Company, Bury St. Edmunds.

48.

Delis

, Kaplan

, Kramer

(2001) Delis Kaplan Executive Function System: Technical Manual, Psychological Corporation, San Antonio, TX.

49.

Bradley

, Lang

(1999) Affective norms for English words (ANEW): Instruction manual and affective ratings, University of Florida.

50.

Moritz

, Voigt

, Arzola

, Otte

(2008) When the half-full glass is appraised as half empty and memorised as completely empty: Mood-congruent true and false recognition in depression is modulated by salience. Memory 16, 810–820.

51.

Son

, Therrien

, Whall

(2002) Implicit memory and familiarity among elders with dementia. J Nurs Scholarsh 34, 263–267.

52.

Donaldson

(1996) The role of decision processes in remembering and knowing. Mem Cognit 24, 523–533.

53.

Kennet

, McGuire

, Willis

, Schaie

(2000) Memorability functions in verbal memory: A longitudinal approach. Exp Aging Res 26, 121–137.

54.

Paivio

(1971) . Imagery and Verbal Processes, Holt, Rinehart and Winston, New York.

55.

Wright

, Roumani

, Boudreau

, Newman

, Ruby

, Studenski

, Shorr

, Bauer

, Simonsick

, Hilmer

, Hanlon

, Study HA and BC (1999) Effect of central nervous system medication use on decline in cognition in community-dwelling older adults: Findings from the Health, Aging and Body Composition Study. J Am Geriatr Soc 57, 243–250.

56.

Fossati

, Coyette

, Ergis

, Allilaire

(2002) Influence of age and executive functioning on verbal memory of inpatients with depression. J Affect Disord 68, 261–271.

57.

Fossati

, Harvey

, Le Bastard

, Ergis

, Jouvent

, Allilaire

(2004) Verbal memory performance of patients with a first depressive episode and patients with unipolar and bipolar recurrent depression. J Psychiatr Res 38, 137–144.

58.

Baudic

, Tzortzis

, Barba

, Traykov

(2004) Executive deficits in elderly patients with major unipolar depression. J Geriatr Psychiatry Neurol 17, 195–201.

59.

Rogers

, Kasai

, Koji

, Fukuda

, Iwanami

, Nakagome

, Fukuda

, Kato

(2004) Executive and prefrontal dysfunction in unipolar depression: A review of neuropsychological and imaging evidence. Neurosci Res 50, 1–11.

60.

Mather

(2004) Aging and emotional memory. In Memory and Emotion, Reisberg

, Hertel

, eds, Oxford University Press, New York.

61.

LaBar

, Phelps

(1998) Arousal-mediated memory consolidation: Role of the medial temporal lobe in humans. Psychol Sci 9, 490–493.

62.

Sharot

, Phelps

(2004) How arousal modulates memory: Disentangling the effects of attention and retention. Cogn Affect Behav Neurosci 4, 294–306.

63.

Fischer

, Nyberg

, Backman

(2010) Age-related differences in brain regions supporting successful encoding of emotional faces. Cortex 46, 490–497.

64.

Leclerc

, Kensinger

(2011) Neural processing of emotional pictures and words: A comparison of young and older adults. Dev Neuropsychol 36, 519–538.

65.

Gaddy

, Ingram

(2014) A meta-analytic review of mood-congruent implicit memory in depressed mood. Clin Psychol Rev 34, 402–416.

66.

Alexopoulos

, Hoptman

, Yuen

, Kanellopoulos

, Seirup

, Lim

, Gunning

(2013) Functional connectivity in apathy of late-life depression: A preliminary study. J Affect Disord 149, 398–405.

67.

Dehon

, Laroi

, Van der Linden

(2010) Affective valence influences participant’s susceptibility to false memories and illusory recollection. Emotion 10, 627–639.

68.

Bourke

, Douglas

, Porter

(2010) Processing of facial emotion expression in major depression: A review. Aust N Z J Psychiatry 44, 681–696.

69.

Cusi

, Nazarov

, Holshausen

, Macqueen

, McKinnon

(2012) Systematic review of the neural basis of social cognition in patients with mood disorders. J Psychiatry Neurosci 37, 154–169.

70.

Harmer

(2013) Emotional processing and antidepressant action. In Behavioral Neurobiology of Depression and Its Treatment, Cowen

, Sharp

, Lau

, eds. Springer Berlin, Heidelberg, pp. 209–222.

71.

Cahill

(2010) Sex influences on brain and emotional memory: The burden of proof has shifted. Prog Brain Res 186, 29–40.