Background Music and Memory in Mild Cognitive Impairment: The Role of Interindividual Differences

Music therapy and background music in aging and cognitive decline

Music therapy is the clinical use of music aimed at treating cognitive and behavioral disorders. It may include a variety of therapeutic interventions based on listening, singing, playing instruments, or composing music. It has been widely used in aging and cognitive decline, but evidence of its efficacy is still controversial [1, 14]. The Cochrane review by Vink et al. [14] examined ten studies of patients with dementia undergoing music therapy; their review was inconclusive on the efficacy of music therapy because of the poor methodological quality in these studies. In a subsequent Cochrane review by van der Steen et al. [1], 22 studies were included. They concluded that at least five sessions of music-based intervention may produce benefits for depression and improve emotional being for individuals with dementia. However, both Cochrane reviews arrived at the conclusion that music therapy had little or no effect on cognition in this patient population. Conversely, a review by Moreno-Morales et al. [15], which included eight studies on patients with dementia, concluded that music-based therapies may improve cognition. Specifically, they found that shorter interventions (less than 20 weeks) and those based on listening to music improved general cognition outcomes in patients with mild to moderate dementia.

In MCI patients, strong conclusions on the positive effects of music-based interventions on cognition and behavioral symptoms cannot be drawn given the low quality of the methodology employed in the studies and the small effect sizes [16, 17]. Multitask movement music therapy seems to more beneficial for patients with MCI than single-training task as it has been shown to modulate the prefrontal cortex activity [18].

Finally, considering healthy older adults, a metanalysis conducted by Xu et al. [19] concluded that music-based therapies are moderately effective in improving cognitive outcomes, decreasing depressive symptomatology, and particularly effective for reducing disruptive behaviors. More robust effects for cognition were found for older adults who have had both short- and long-term musical training [20].

These reviews and metanalyses have analyzed data from music interventions conducted across multiple sessions and by comparing test scores pre- and post-intervention. An alternative way to investigate the effect of music on cognition is to use background music. In this condition, participants are passively listening to the music while they are performing a cognitive or behavioral task. A metanalysis by Kämpfe et al. [21] in older adults found that, based on a global analysis including 97 studies, there was no effect of background music on cognition. However, in a series of sub-analyses, these authors found different effects on other outcomes outside of cognition. Specifically, they found trending positive effect of background music on emotional experiences after music exposition and larger positive effects on sports performance. Additionally, one study by Reave et al. [22] reported negative effects of background music on associative memory. Older adults judged the background music played during the task as distracting and had poorer performed in recognizing previously learned face-name associations when exposed to music compared to silence.

Ferreri et al. [6] investigated the effect of background music on verbal learning (word lists) in older adults, finding it to be beneficial for the encoding. When older adults were exposed to background music, they committed less false alarms and showed less activation of the dorsolateral prefrontal cortex than in the silent condition. The authors interpreted the benefits of background music in memory as a product of creating a musical learning context with efficient strategies for binding items to one another in healthy older individuals. This interpretation fits with the associative deficit hypothesis, which proposes that older adults’ deficiencies in episodic memory are related to their inability to form and retrieve links among single bits of information (for a similar argument in patients, see [23]).

Zooming in to consider those older adults experiencing cognitive decline, the evidence of background music’s effects on cognition is limited [10]. Two studies have reported positive effects of background music on autobiographical memory [11, 12] and an another reported benefits in lexical retrieval [13] for patients with AD, but there is scarce evidence on episodic memory. In MCI patients, in addition to the therapeutic effects of music, research has focused mainly on musical memory (e.g., [24]) but not on the benefits of background music for episodic memory. Thus, our research focused on this cognitive domain since it is one of most frequently affected in patients with MCI.

How background music may improve memory

For the present study, we based our two main hypotheses on two theories that have been proposed to explain how background music influences memory performance: the ‘arousal-mood hypothesis’ and the ‘encoding specificity principle’ [4].

According to the ‘encoding specificity principle’, music could act as a contextual factor that cues the associated memory. This may function in two similar yet distinct ways. First, playing the same music during learning and testing phases improves memory because similar contextual cues assist memory retrieval (‘context-dependent memory’) [25–27]. Alternatively, memory may be significantly better when the same mood or arousal is experienced by participants at both encoding and retrieval (‘state-dependent memory’) and music can act as the contextual cue for these states [26, 28]. Supporting the overarching encoding specificity principle, it has been shown that participants have higher performance in memory recall if they are exposed to it during both encoding and recall that during encoding only [4].

The ‘arousal-mood hypothesis’ predicts that memory performance is modulated by background music by its ability to alter the listener’s mood and arousal states. Previous research has found that the emotional valence and the pleasantness of music are two factors that increase arousal and mood, and in turn are beneficial for improving cognitive performance when they are manipulated [29]. For instance, Thompson et al. [30] found that participants showed better performance in a spatial memory task when they were exposed to background music compared to when exposed to silence. However, all background music was not equal. Participants’ performance improved when the background music was considered pleasant (a Mozart sonata), but not if it was sad (an Albinoni adagio). Additionally, other evidence suggests that the combination of both mood and arousal is essential for improving memory performance. Greene et al. [31] manipulated these two dimensions orthogonally and found that the benefits of background music on memory were salient in positive mood-high arousal and negative mood-low arousal conditions, relative to positive-low arousal and negative-high arousal ones. Finally, Nguyen and Grahn [4] found that participants performed better on memory tasks when listening to low arousal music irrespective of valence, but only when music was presented at both encoding and recognition phases.

The ‘arousal-mood hypothesis’ is based on evidence that cognitive performance is influenced by mood and arousal in a U-shaped fashion, with better performance associated with intermediate levels of arousal and poorer performance with either low or very high levels of arousal (also known as ‘Yerkes-Dodson law’, [29, 32]). At neurophysiological level, the changes in arousal and mood induced by music are associated with specific patterns of activity over the left and right frontal areas [33]. Specifically, participants showed greater electroencephalographic activity on the left frontal areas for musical excerpts with positive valence and greater activity on the right frontal areas for those with negative valence.

Perk et al. [10] proposed that the positive effects of music on cognition in patients may be driven by the change of the basal physiological arousal and/or the reactivation of the dopaminergic system induced by the exposure to music. Thus, the combination of cognitive and neurophysiological changes induced by music might explain difference in memory as related to levels of arousal. Fernandez et al. [34] found that background music modulated the performance of the participants while they performed an attentional task. Music with positive valence and high arousal was associated with faster reaction times and greater activation of the attentional neural network (fronto-parietal areas). Conversely, music with negative valence modulated the activation of the occipital areas and negatively impacted the performance in the attentional task.

The present study

Based on the above theories of ‘context-dependent memory’ and ‘arousal-mood hypothesis’ [4], we designed three experiments to test the effect of background music on memory performance in older adults with MCI.

First, in Experiment 1 and 2, we investigated the effect of ‘context-dependent memory’, as one of two definitions of the ‘encoding specificity principle’. In Experiment 1, background music was played during encoding only, whereas in Experiment 2, background music was included during both encoding and retrieval phases. Within each experiment, we compared memory performance following exposure to music versus a silence condition. Because Experiment 2 included the same music cues during encoding and retrieval phases, we expected better memory performance following the background music condition compared to the silence condition. These expected results and hypothesis were based on the findings reported by Nguyen and Grahn [4] showing better memory recall when exposed to music during both encoding and retrieval phases. As these authors observed better memory performance with low arousal music, we used this type of background music in Experiment 1 and 2.

Second, in Experiment 3, we tested the ‘arousal-mood hypothesis’, which predicts that memory performance is modulated by background music altering the listener’s mood and arousal states. Previous studies in healthy individuals have found mixed results on whether arousal induced by background music facilitates memory consolidation, and what level of arousal is ideal [4, 31, 35]. We explored this by exposing individuals to background music that induced low or high arousal, based on the methodology employed by Nguyen and Grahn [4]. Given the results of this study, we expected better memory performance with exposure to low arousal music compared to high arousal music.

Finally, we tested the role of interindividual differences within music-induced benefits on memory, considering a) interindividual preferences toward music with the Barcelona Music Reward Questionnaire (BMRQ) [36], b) subjective feelings towards music excerpts used in the experiments, and c) individual mood changes over the course of music exposure. All these variables may potentially influence the magnitude of music’s benefits on memory processes. Indeed, musical preferences and mood changes are the two main variables that modulate the benefits of music for behavioral disorder interventions in patients with AD [37] and for cognitive benefits in healthy individuals [29].

Participants

Three separate groups of MCI patients were recruited to participate in these three experiments (Exp. 1: n = 20; Exp. 2: n = 20; Exp. 3: n = 25). Each patient participated only in one of the three experiments. Experiments were conducted at three different time points in sequential order. Sample size was determined based on the results of study by Ferreri et al. [6] which showed an effect size of background music on memory of about 10%. A power analysis run referencing Ferreri et al. [6] suggested an adequate sample size of 18 participants, with a power of 80% and an alpha of 0.05.

Participants were recruited from the Neuropsychology Unit of the Hospital de la Santa Creu i Sant Pau in Barcelona. All were diagnosed with MCI according to the recommendations from the National Institute on Aging Alzheimer’s Association [38], meeting the following criteria: a) subjective or informant-based cognitive decline, b) objective cognitive decline of episodic memory as measured by a neuropsychological assessment, c) the cognitive decline is not interfering with individual independence, d) CDR score = 0 suggesting absence of dementia, and e) onset after the age of 65 (usually in the late 70 s or thereafter). Exclusion criteria consisted of the following: a) the presence of psychiatric and neurological disorders other than MCI, b) clinically known hearing or vision impairment, c) a history of alcohol abuse, and d) documented history of cerebrovascular risk that may be indicative of a mixed variant of cognitive impairment. Additionally, we excluded participants who were musicians or had professional training in music.

Material and procedure

The experimental session of the three experiments included: a) a neuropsychological assessment; b) questionnaires on mood, music use and music preferences; and c) two face recognition memory tasks.

a. Neuropsychological assessment. The neuropsychological assessment included the following tests: Mini-Mental State Examination [39], Rey’s Auditory Verbal Learning Test [40], Forward and backward digit spans [41], Block Design Test [42], Poppelreuter test [43], Boston Naming Test [44, 45], semantic and phonemic fluency tasks [46]. See Table 1 for the neuropsychological test scores.

b. Questionnaires for mood and music use and preferences. Two self-report questionnaires were used across all studies, the Scale for Mood Assessment (EVEA) and the BMRQ [36] (http://brainvitge.org/z_oldsite/bmrq.php). The EVEA consists of a self-rating of emotional state and feelings (16 items, 10-point scale) and was administered before and after the memory task [47]. The items yield ratings for four emotional states: anxiety, hostility, sadness-depression and happiness.

The BMRQ was administered between the encoding and recognition phases of the memory task to assess musical preferences. It consists of 20 items and the answers range from 1 to 5 (from completely disagree to completely agree). The items and scores are grouped into six factors: music seeking, emotion evocation, mood regulation, sensorimotor, social, music reward.

c. Face memory task

Visual stimuli. A total of 100 colored pictures of unfamiliar faces, half males and half females, all of middle-age and older people, were downloaded from electronic datasets ‘FACES’ (https://faces.mpdl.mpg.de/imeji/) [48] and processed in Adobe Photoshop. The pictures were edited in order to obtain the same resolution and dimensions for all the pictures, and to add a uniform gray background. Ten different lists of 24 stimuli each were created in order to counterbalance the presentation order of trials across participants.

Auditory stimuli. The selection of the musical pieces was based on the ratings provided by a group of volunteers different from those participants that took part in the experiments. They were asked to rate two aspects of musical excerpts on a 10-point scale: a) arousal (from ‘relaxing’ to ‘exciting’), and b) mood (from ‘sad’ to ‘happy’). Participants also reported whether the music excerpt was known or not and, if it was familiar, whether they were able to recall the title of the piece. We asked participants to rate 20 excerpts taken from popular and traditional music of the Spanish culture (mainly from the 1950 s to 1970 s). We also included one classical excerpt from ‘Adagio’ in D minor by Johann Sebastian Bach in the rating process, as this music genre was used in previous studies with background music [6, 49, 50].

Based on participants’ ratings, we selected two musical excerpts for our studies: a) ‘Adagio’ in D minor by Johann Sebastian Bach (https://www.youtube.com/watch?v=2x-OHljZzHQ) (medium-low arousal: M = 3.4; medium-low mood: M = 4.6), and b) instrumental version of ‘Un rayo de sol’ by Los Diablos (https://www.youtube.com/watch?v=IownknoPwDw) (high arousal: M = 8.2; high mood: M = 8.5). Although the original version of the ‘Un rayo de sol’ includes lyrics, we used an instrumental version of it to make it comparable to the ‘Adagio,’ which does not contain lyrics. The use of music with or without lyrics may have differential effects on cognition [51].

The selected samples were edited using Audacity software, matched on volume, and transformed into MP3 files [50]. The use of these musical excerpts in experimental conditions is explained below.

Task procedure. The memory task was the same across the three experiments and was administered using DMDX software [52]. There were two phases, the encoding and recognition trials (see Fig. 1). Each participant was tested in an individual session, as opposed to group testing.

Fig. 1

Memory task and background conditions of the three experiments.

During the encoding phase, participants were presented with a series of 24 faces and were instructed to memorize them. Every trial started with a fixation point (a black cross) in the center of the screen displayed for 500 ms, followed by a picture for a maximum of 2500 ms. In order to promote task engagement, participants were instructed to indicate whether the faces were female or male by pressing one of two keys on the keyboard.

After a 10-min break, there was a recognition phase, where participants were presented with 24 faces that were previously presented and 24 faces that were new in random order. They were instructed to judge whether each face was ‘new’ (not presented during encoding) or ‘old’ (previously seen during encoding) by pressing one of two keys on the keyboard within a maximum of 2500 ms.

Experiments 1, 2, and 3. All experiments followed a within-subjects design, where participants performed the memory task twice over two distinct experimental conditions. Two sets of different pictures were used to avoid interference across conditions. The difference between experimental conditions was for the presence or absence of music (Experiments 1 and 2) or the use of two separate music excerpts inducing different levels of arousal (Experiment 3).

In Experiment 1, participants were exposed to ‘Adagio’ in D minor by Johann Sebastian Bach during the encoding phase in one condition (‘Music’) and they were not exposed to music in the other condition (‘Silence’).

In Experiment 2, participants were exposed to ‘Adagio’ in D minor by Johann Sebastian Bach during both encoding and recognition phases in one condition (‘Music’) and they were not exposed to music in the other condition (‘Silence’).

In Experiment 3, participants were exposed to ‘Adagio’ in D minor by Johann Sebastian Bach during the encoding phase of one condition (‘Low Arousal’) and they were exposed to ‘Un rayo de sol’ by Los Ladros in the other condition (‘High Arousal’). The order of presentation of the musical excerpts (i.e., Music-Silence versus Silence-Music) was balanced across participants.

The duration of the music excerpt was two minutes longer than the encoding phase of the Music conditions. The auditory presentation started after participants received task instructions, two minutes before the beginning of the experiment. This pre-task exposure was based on previous work determining that one to two minutes of music exposure is needed to elicit a change in mood and arousal [4].

The BMRQ [36] was administered between the encoding and recognition phases of the memory task. The EVEA [47] was completed before and after music exposure. For Experiments 1 and 2, EVEA was used for the Music condition and in Experiment 3 was employed in both Music conditions (Low and High Arousal). In Experiment 3, participants were also required to rate the music excerpts on three dimensions (happiness, arousal, pleasantness) on a 10-point scale (1 being low and 10 high in degree).

The procedures involving experiments were approved by the Ethics Committee of the Universitat Oberta de Catalunya for the proposal titled “The Mozart Effect on memory in patients with cognitive decline (MEM-COG)”.

Statistical analysis

Individual measures of sensitivity [d^′= Z(Hits) - Z(False Alarms)] and response bias {c = - [z(Hits) + z(False Alarms)]/2} were calculated separately for each condition and experiment. Repeated measures ANOVA was performed to compare the d’ and c scores with condition (Experiment 1 and 2: Silence versus Music; Experiment 3: Low-Arousal Music versus High-Arousal Music) as a within-subject factor. The normality assumption was checked with the Shapiro-Wilk test and the results showed the data were normally distributed.

Mood changes after music exposition were analyzed by comparing the four subscale scores of the EVEA (sadness-depression, anxiety, anger-hostility, and happiness) with a repeated measures ANOVA, where time (pre versus post) served as a within-subject factor. We have applied the False Discovery Rate (FDR) with the Benjamini and Hochberg [53] procedure for multiple comparisons when necessary. This technique has the advantage of controlling for Type I errors and is applicable for both within- and between-subjects comparisons. The significance level is reported as a corrected p value.

The effect size was calculated as partial eta squared ( ηp2 ) [SS_effect / (SS_effect + SS_error)].

To analyze individual differences in the effect of music on memory, we ran a stepwise linear regression analysis with the d’ values as a dependent variable. For Experiments 1 and 2, the predictors were the subscale scores of the BMBQ (music seeking, emotion evocation, mood regulation, sensorimotor, social, and music reward) and the mood changes (sadness-depression, anxiety, anger-hostility, and happiness). For Experiment 3, the predictors were the subscale scores of the BMBQ, mood changes, and the music ratings for pleasantness, arousal, and happiness.

Experiment 1: background music at encoding versus silence

The d’ values were not significantly different between the Silence and Music conditions [F(1, 19) = 0.14, p = 0.51; Music: M = 0.82, SD = 0.81; Silence: M = 0.88, SD = 0.72]. Similarly, the c values were not significantly different between Silence and Music conditions in [F(1, 19) = 0.14, p = 0.71; Music: M = 0.005, SD = 0.05; Silence: M = 0.001, SD = 0.05].

Experiment 2: background music at both encoding and recognition versus silence

The d’ values were not significantly different between the Silence and Music conditions [F(1, 19) = 0.06, p = 0.81; Music: M = 0.73, SD = 0.51; Silence: M = 0.73, SD = 0.58]. Similarly, the c values were not significantly different between Silence and Music conditions [F(1, 19) = 0.005, p = 0.94; Music: M = –0.04, SD = 0.39; Silence: M = –0.04, SD = 0.45].

Experiment 3: Background music with high and low arousal

The d’ values were not significantly different between Low Arousal and High Arousal conditions [F(1, 24) = 1.30, p = 0.27; Low Arousal: M = 0.80, SD = 0.49; High Arousal: M = 0.69, SD = 0.56] (see Fig. 2 and Table 2). Similarly, the c values were not significantly different between Low- and High-Arousal conditions [F(1, 19) = 0.04, p = 0.85; Low Arousal: M = 0.14, SD = 0.44; High Arousal: M = 0.13, SD = 0.50].

Fig. 2

Memory performance (d’ values) for the three experiments separated by conditions (Experiments 1 and 2: Music versus Silence; Experiment 3: High-Arousal versus Low-Arousal Music).

Background music and mood

In Experiment 1, there was no significant difference between pre- and post-music exposure for hostility [F(1, 19) = 0.55, p = 0.62], happiness [F(1, 19) = 0.02, p = 0.88]. The scores after music exposure was significantly lower for anxiety [F(1, 19) = 13.52, p = 0.008] and sadness-depression [F(1, 19) = 5.91, p = 0.04] as compared to pre-music assessment (see Fig. 3 and Table 3).

Fig. 3

Pre- and post-music exposure mood ratings across the three experiments.

In Experiment 2, there was no significant difference between pre- and post-music exposure for hostility [F(1, 19) = 0.59, p = 0.54], happiness [F(1, 19) = 5.02, p = 0.16], anxiety [F(1, 19) = 1.37, p = 0.52], and sadness-depression [F(1, 19) = 0.39, p = 0.54].

In Experiment 3, the High-Arousal condition revealed no significant differences between pre- and post-music exposure for hostility [F(1, 24) = 2.21, p = 0.30], happiness [F(1, 24) = 0.06, p = 0.81], anxiety [F(1, 24) = 0.29, p = 0.79], and sadness-depression scores [F(1, 24) = 4.82,p = 0.16].

For the Low-Arousal condition, there was no significant difference between pre- and post-music exposure for hostility [F(1, 24) = 0.01, p = 0.99], happiness [F(1, 24) = 0.09, p = 0.99], anxiety [F(1, 24) = 0.12, p = 0.99], and sadness-depression [F(1, 24) = 3.05, p = 0.36].

Interindividual differences

In Experiment 1, we found that no variable significantly predicted the d’ scores for Music [F(10, 9) = 0.50, p = 0.83] and Silence conditions [F(10, 9) = 0.51, p = 0.84].

Likewise, in Experiment 2, we found that no variable significantly predicted the d’ scores for Music [F(10, 9) = 0.442, p = 0.90] and Silence conditions [F(10, 9) = 0.71, p = 0.69].

However, in Experiment 3, there were significant predictors of memory performance. For the Low-Arousal condition, pleasantness was a significant predictor of d’ values [B = –0.43; p = 0.02; F(1, 23) = 6.33, p = 0.02; R = 0.47], suggesting that lower scores of pleasantness were associated with better memory performance.

In contrast, for the High-Arousal condition, the Mood Regulation score of the BMRQ was a significant predictor of d’ values [B = 0.40; p = 0.04; F(1, 23) = 4.46, p = 0.04; R = 0.41], suggesting that the use of music as a strategy for mood regulation was associated with better memory performance (see Fig. 4).

Fig. 4

Partial regression plots for high-arousal music and BMRQ scores and for low-arousal music and pleasantness ratings scores.