Effects of musical mnemonics on working memory performance in cognitively unimpaired older adults and persons with amnestic mild cognitive impairment

Abstract

Episodic memory (EM) and working memory (WM) are negatively affected by healthy ageing, and additional memory impairment typically occurs in clinical ageing-related conditions such as amnestic mild cognitive impairment (aMCI). Recent studies on musical mnemonics in Alzheimer's dementia (AD) showed promising results on EM performance. However, the effects of musical mnemonics on WM performance have not yet been studied in (a)MCI or AD. Particularly in (a)MCI the use of musical mnemonics may benefit the optimisation of (working) memory performance. Therefore, in the present study, we examined the effects of musical presentation of digits consisting of pre-recorded rhythms, sung unfamiliar pitch sequences, and their combinations, as compared to spoken presentation. Furthermore, musical expertise was assessed with two perceptual tests and the Self-Report Inventory of the Goldsmiths Musical Sophistication Index. Thirty-two persons with aMCI and 32 cognitively unimpaired older adults (OA) participated in this study. Confirming and extending previous findings in research on ageing, our results show a facilitating effect of rhythm in both cognitively unimpaired OA and persons with aMCI (p = .001, η_p² = .158). Furthermore, pitch (p = .048, η_p² = .062) and melody (p = .012, η_p² = .098) negatively affected performance in both groups. Musical expertise increased this beneficial effect of musical mnemonics (p = .021, η_p² = .090). Implications for the future design of music-based memorisation strategies in (a)MCI are discussed.

INTRODUCTION

One characteristic of normal cognitive ageing is memory decline. With advancing age episodic memory (EM) declines, and working memory (WM) capacity and efficiency also deteriorates (Nyberg et al., 2012). By definition, the memory performance of older adults who show ‘normal’ cognitive ageing is in the unimpaired range when compared to age- and education-adjusted normative data, as this ageing-related decline is not the result of neurodegenerative disease. In contrast, cognitive performance in persons with mild cognitive impairment (MCI) is by definition impaired, that is, exceptionally low compared to age- and education-adjusted normative data (cf. Guilmette et al., 2020). Cognitive impairments may be present in one or more cognitive domains. Persons for whom cognitive impairments are in the memory domain are classified as having the amnestic subtype of mild cognitive impairment (aMCI; Petersen et al., 2014). This memory deficit may or may not be accompanied by impairments in other, non-memory domains, and individuals with aMCI are most at risk to develop Alzheimer's dementia (AD) within the next few years (Albert et al., 2011). While most studies to date have focused on EM function in aMCI patients, there is increasing evidence that WM performance, especially in WM tasks with high-executive demands, is also compromised in aMCI patients (e.g., Gagnon & Belleville, 2011; Kirova et al., 2015; van Geldorp et al., 2015).

To date there is no pharmacological treatment available for MCI, while there is some evidence for possible beneficial effects of non-pharmacological interventions (such as cognitive training or physical exercise) (Petersen et al., 2014). One way to improve memory performance and reduce limitations in everyday life is by the use of mnemonics, that is, applying memory strategies in order to ameliorate memory performance, which may promote independent living in cognitively impaired persons (Ross et al., 2022). In their systematic review, Ross et al. (2022) reported several compensation strategies of which the use depended on the cognitive status of the participants, ranging from internal strategies (e.g., verbalization, visualizing, repetition), external strategies (e.g., reminder systems like notes, calendars, lists) to behavioural strategies (e.g., accepting support and reducing expectations). Older adults with cognitive impairment reported the use of these behavioural strategies most frequently. Incorporating music into such non-pharmacological interventions may be promising.

Based on the results of a systematic review in cognitively unimpaired individuals and in individuals with memory impairment, it can be concluded that musical mnemonics may help to learn and remember verbal information (Derks-Dijkman, Schaefer, & Kessels, 2023). Musical mnemonics use a musical presentation of the to-be-remembered information (as also described in Ferreri & Verga, 2016), also referred to as ‘music as a structural prompt’ (Madsen et al., 1975). Melodic songs to teach, for example, motor and academic skills have been applied in educational settings and for therapeutic purposes, with ‘the ABC-song’ as a well-known example (Wolfe & Hom, 1993). To date, however, studies on musical mnemonics have mainly focused on WM performance in young adults (e.g., Silverman, 2007, 2010, 2012), with a growing body of literature on ageing adults and patient populations with EM impairments, such as AD patients (e.g., Ratovohery et al., 2019; for an overview, see Derks-Dijkman, Schaefer, & Kessels, 2023). However, to our knowledge, none of the previous studies have focused on the use of musical mnemonics to improve WM performance in (a)MCI or AD.

In previous research in cognitively unimpaired young and older adults (Derks-Dijkman, Schaefer, Stegeman, et al., 2023), a modified version of the forward digit span task was used, which was partially based on a previous method described by Silverman (2007). This experimental task was presented in four task conditions: (1) digit sequences sung to a simple unfamiliar isochronous five-tone pitch sequence (‘pitch’), (2) sung to a five-tone unfamiliar pitch sequence with a rhythmic pattern with varying durations added (‘melody’), (3) spoken to an unfamiliar rhythmic pattern with varying durations (‘rhythm’), or (4) presented in the standard way (i.e., spoken at a 1 digit-per-second pace). A facilitating effect of rhythm in both young and older adults was found, and performance was negatively affected in the conditions that included pitch (i.e., the ‘pitch’ and ‘melody’ conditions) in older adults only (Derks-Dijkman, Schaefer, Stegeman, et al., 2023).

The current study aimed to examine the effects of musical versions of the digit span paradigm on WM performance both in cognitively unimpaired older adults (OA) and persons with aMCI. Specifically, we were interested in which specific aspects of musical presentation were possible key components (i.e., rhythm, pitch or melody). Furthermore, the participant's musical expertise, specifically engagement and listening experience, was taken into account in relation to the possible beneficial effects of the musical mnemonic using the Self-Report Inventory and two subtasks of the Goldsmith Musical Sophistication Index (Gold-MSI, Müllensiefen et al., 2014).

In line with previous findings of Ratovohery et al. (2019) in AD patients, we hypothesized that OA will outperform persons with aMCI on the digit span task in general, regardless of musical presentation. Second, based on previous findings in cognitively unimpaired young and older adults (Derks-Dijkman, Schaefer, Stegeman, et al., 2023) and the findings of Silverman (2007, 2010) in students, both OA and persons with aMCI are hypothesized to show the best performance in the ‘rhythm’ condition and worst in the ‘pitch’ and ‘melodic’ conditions. Furthermore, we hypothesized that persons with aMCI might show more benefit from the expected (beneficial) effects of rhythm as compared to OA, the latter having a better preserved WM performance than aMCI patients (Derks-Dijkman, Schaefer, Stegeman, et al., 2023), for whom rhythmic presentation may diminish the importance of relying on more limited executive resources due to age-related disorders including (a)MCI and AD (e.g., Gagnon & Belleville, 2011; Kirova et al., 2015) possibly providing additional structure through temporal chunking mechanisms (Purnell-Webb & Speelman, 2008; Silverman, 2012), thus reducing stimulus complexity (that is, the combined complexity of the musical and verbal stimulus together, Derks-Dijkman, Schaefer, & Kessels, 2023). Also, based on empirical evidence in OA (Derks-Dijkman, Schaefer, Stegeman, et al., 2023), we expected that in persons with aMCI – who are expected to have a worse WM performance than OA – faster WM overload would occur in the complex conditions (i.e., the ‘pitch’ and ‘melody conditions’). Possibly explained through only varying pitch (no rhythm—that is—no temporal chunking component) in the ‘pitch’ condition and the addition of a pitch component to a rhythmic pattern (i.e., in our ‘melody’ condition) leading to extra information to be processed, with as a result a higher stimulus complexity possibly causing working memory overload (for a model on musical stimulus complexity, see the systematic review of Derks-Dijkman, Schaefer, & Kessels, 2023).

Finally, we expected no moderating effect of musical expertise in OA, based on previous findings in cognitively unimpaired young and older adults (Derks-Dijkman, Schaefer, Stegeman, et al., 2023). Derks-Dijkman, Schaefer, Stegeman, et al. (2023) hypothesized that the absence of a moderating effect of musical expertise in OA may have been related to the relatively high-functioning older adults that participated in their study. As the level of education in the present sample is highly similar to that of Derks-Dijkman, Schaefer, Stegeman, et al. (2023), we expect similar results here. However, based on a small body of previous literature on musical mnemonics in AD patients (Derks-Dijkman, Schaefer, & Kessels, 2023), we hypothesized that musical expertise may moderate the beneficial effect of musical presentation in persons with aMCI (cf. Baird et al., 2017). That is, musical expertise could make it easier to effectively compensate for their WM impairment through the use of music, possibly by their engrained previous exposure to music offering additional structure, instead of being distracted by the music (causing WM overload).

MATERIALS AND METHODS

Participants

Based on previous research using similar tasks (Silverman, 2010, 2012), and our main aim to compare within-subject condition effects as well as the aMCI patient versus older control performance, we aimed for 30 participants in each group. This sample size enables us to detect small to medium effect sizes (f² = .12) for the between factors (1−β = .86), small effect sizes (f² = .02) for the within factors (1−β = .92), and small effect sizes (f² = .01) for the within-between interactions to (1−β = .81) (G*Power 3.1.9.7, α = .05, r = .7, Faul et al., 2009). OA were participants who visited the Department of Medical Psychology of the Ziekenhuisgroep Twente (a general hospital located in Almelo and Hengelo, the Netherlands) in the context of the memory clinic, who had no cognitive impairment, or were cognitively unimpaired volunteers. Thirty-two cognitively unimpaired older adults (OA; 11 men, 21 women; age: M = 72.2, SD = 6.8, range = 65–91) were included. The data of 27 OA participants of these was re-used from a previous study (Derks-Dijkman, Schaefer, Stegeman, et al., 2023). OA were included when they had sufficient vision and hearing to perform the neuropsychological tests and were able to understand, read and communicate in Dutch. Exclusion criteria were a diagnosis of mild cognitive impairment (MCI), dementia, a stroke, a psychiatric disease or excessive drug or alcohol use.

We also recruited 32 persons with amnestic mild cognitive impairment (aMCI; 19 men, 13 women; age: M = 75.0, SD = 6.4, range = 60–87), who were all outpatients of the memory clinic of the Ziekenhuisgroep Twente general hospital in Almelo and Hengelo, the Netherlands. Clinical diagnoses were made in a multidisciplinary setting by a team of geriatricians, neurologists and neuropsychologists, using the results of extensive neuropsychological testing (with all cognitive domains covered), laboratory investigations, neurological examination and neuroimaging data (MRI), the clinical interview with patient and caregiver (assessment of daily functioning) and medical and psychiatric history taking as well as other important (e.g., biographical) information. Fifteen patients met the criteria for single-domain aMCI, 17 were classified as having multiple-domain aMCI (cf. Petersen et al., 2014) and the Clinical Dementia Rating (CDR) was .5 for all patients.

We used the Dutch educational system which is based on education levels, comparable to the classification of education levels of UNESCO (2011), using a 7-point scale (Duits & Kessels, 2014), as years of education are not informative in our educational system. We have listed both education level and estimated years of education for comparison with the Anglo-Saxon educational system for descriptive purposes (Hochstenbach et al., 1998).

Materials

Neuropsychological tests and questionnaires

General cognitive functioning was assessed with the Dutch version of the Montreal Cognitive Assessment (MoCA 7.1; Nasreddine et al., 2005). The Dutch version of the National Adult Reading Test (NART, Schmand et al., 1991) was administered to estimate premorbid verbal intelligence level in all participants. We administered the WAIS-IV Digit Span subtest (Wechsler, 2008) as a descriptive measure of WM functioning. In addition, the Geriatric Depression Scale (GDS-30, Yesavage et al., 1983) was used as an index of depressive symptoms in both groups. Descriptive statistics of the OA and aMCI groups are displayed in Table 1.

TABLE 1. Descriptive statistics of the participants.

	Older adults (N = 32), M (SD)	aMCI (N = 32), M (SD)	p-value
Age	72.22 (6.78; 65–91)	74.97 (6.36; 60–87)	.099
Sex (men:women)	11:21	19:13	.045
Educational level	5^a # (1.04)	4# (1.48)	.133
Years of education	11.23^a (2.83)	11.75 (9.30)	.765
NART-IQ	107.09 (16.37; 73–140)	100.93^b (19.62; 67–144)	.183
MoCA	26.63 (2.21)	21.17^c (2.66)	<.001
GDS-30	3.93^d (3.97)	6.50^b (4.23)	.018
WAIS-IV digit span
Forward	8.31 (2.07)	7.91 (1.65)	.389
Backward	7.91 (2.05)	6.81 (1.53)	.019
Sorting	7.66 (1.81)	5.81 (2.60)	.002
Gold-MSI
Beat Perception	10.84 (2.92)	10.53 (3.14)	.682
Melody Memory	7.22 (2.37)	7.91 (1.97)	.212
Self-Report Inventory
General Sophistication	54.03 (17.10)	53.16^e (14.32)	.828
Active Engagement	24.97 (8.47)	21.39 (8.12)	.092
Perceptual Abilities	35.69 (9.08)	37.90 (7.80)	.303
Musical Training	16.59 (8.91)	15.13 (8.36)	.504
Emotions	23.78 (6.67)	23.32 (5.30)	.764
Singing Abilities	23.63 (7.11)	24.55 (5.52)	.568

• Note: Mean scores and differences between older adults and persons with aMCI. Standard deviations are shown between parentheses. Between-group differences were computed with independent-samples t-tests. A Chi-square test was used to test differences in sex distribution. Differences in the distribution of education level were tested using a Mann–Whitney U test. #, Median.
• Abbreviations: GDS-30, Geriatric Depression Scale (30-item version); MoCA, Montreal Cognitive Assessment; NART-IQ, NART IQ-estimation; WAIS-IV, Wechsler Adult Intelligence Scale-IV.
• ^a Data of one OA were missing.
• ^b Data of two aMCI participants were missing.
• ^c Data of three aMCI participants were missing.
• ^d Data of two OA were missing.
• ^e Data of one aMCI participant for the Self-Report Inventory (General Sophistication variable and all subscales) were missing.

Experimental digit span task

The experimental digit span task was partly based on a previous method used by Silverman (2007). Here we integrated it with the standard procedure which is also applied in the Digit Span subtests of the WAIS-IV (that is, sequences increasing in length, with two different sequences per length presented, Wechsler, 2008). The task consisted of 32 sequences of mono-syllabic digits (1, 2, 3, 4, 5, 6, 8, 10), excluding the multi-syllabic digits numbers 7 and 9 in Dutch (i.e., ‘ze-ven’ and ‘ne-gen’), as those would have had melodic consequences (i.e., for one number, two tones would have been needed). We assigned the digits pseudo-randomly to the melodies. In each sequence, each digit occurred only once. The digit sequences increased in length (5, 6, 7, 8 digits), starting with two 5-digit sequences, followed by two sequences consisting of 6, 7 and 8 digits respectively. For each length of digit-sequence (i.e., 5, 6, 7, 8 digits), the digits were presented in four conditions: spoken (A), sung to a simple unfamiliar isochronous five-tone pitch sequence (“pitch”, B), spoken to an unfamiliar rhythmic pattern with varying durations (“rhythm”, C), sung to an unfamiliar five-tone pitch sequence with a rhythmic pattern with varying durations added (“melody”, D) (For an example of the scoring form, see Table A1 in the Appendix 1).

The pitch sequences were composed using the pitches C, D, E, G and A (C major key, started on a C, moved upward and returned back to a C). We restricted pitch intervals where possible to a major third or less. In the spoken and pitch conditions only quarter notes (quavers) were used, while quarter notes, eighth and half notes were used in the rhythm and melody conditions (for musical notation of the different conditions, see Figure 1).

FIGURE 1

Music notation of the four task conditions (8- digit sequences), showing two digit sequences per condition. (A) Spoken condition, (B) Pitch condition, (C) Rhythm condition, (D) Melody condition.

Musical expertise

We assessed musical expertise with a research version (Dutch translation used by Derks-Dijkman, Schaefer, Stegeman, et al., 2023) of the Self-Report Inventory and two perceptual tests (Beat Perception and Melody Memory) of the Goldsmiths Musical Sophistication Index v1.0 (Gold-MSI, Müllensiefen et al., 2014). The questionnaire consisted of 31 statements on musical engagement and behaviour (agreement with each statement scored on a Likert-scale from one [totally disagree] to seven [totally agree]) and some additional questions (e.g., formal music training, number of hours listening to music per day, number of musical instruments played) and comprised a general index; General Sophistication and the subscales Musical Training, Perceptual Skills, Active Involvement, Emotions and Singing Skills, with good psychometric properties for the English (Cronbach's alpha between .79 and .93 for all five factors and the General Sophistication factor, Müllensiefen et al., 2014) and German version of the questionnaire (Cronbach's alpha between .72 and .91 for all five factors and the general factor; Schaal et al., 2014).

Procedure

All tests were administered in a fixed order, but the four conditions of the musical digit span task were administered in a counterbalanced way (i.e., in one participant the administration order of the conditions might be pitch, rhythm, spoken, melody, while the order in another participant for example might be spoken, pitch, melody, rhythm); resulting in a total of 24 possible administration orders (see Table A2). Assignment to one of the 24 possible orders was random in order to control for order and carry-over (e.g., practice or fatigue) effects. As we intended to include 30 participants per group, then each possible order would have occurred at least once in both an OA and aMCI participant. The stimuli consisted of pre-recorded mp4 sound files played on a computer using a female soprano voice, presented through a previously published procedure (Derks-Dijkman, Schaefer, Stegeman, et al., 2023), in which over-ear headphones (Philips SHP6000) were used initially, but a switch was made to speakers (Philips SPA 2200/00) for practical reasons. Beforehand it was checked with the participant whether the participant could hear it well enough and if necessary the volume was adjusted. At the start of the test, a spoken example of a digit sequence was provided as an example. After each presentation of a digit sequence, the participants had to recall the digits exactly in the same order, having the free choice to speak or sing. Participants received as much time as needed to recall the sequences. The experimenter wrote down the response, the modality of the answers (spoken or sung) and the administration order. No feedback was provided. No stopping rule was applied (unless the participant could not complete the task). One point was awarded for each correctly recalled sequence (i.e., all digits in the exact order as presented), resulting in a maximum score of eight points for each condition. All data were stored and analysed in an anonymised way. The study took place in a quiet room in different settings (i.e., Medical Psychology department of Ziekenhuisgroep Twente hospital, or home of the participant), assuring there were no other people, or other possible distractions present. For more details and verbatim instructions see Derks-Dijkman, Schaefer, Stegeman, et al. (2023).

All participants gave their written informed consent and voluntarily participated in this study. The data collection for this study has been approved by the Ethics Committee of the Faculty of Social Sciences of the Radboud University (ECG2012-1304-025) and the local ethical committee of the Ziekenhuisgroep Twente hospital (September 12, 2016, ZGT16-22).

Analyses

A (mixed-model) 2 × 4 repeated-measures analysis of variance (ANOVA) was computed using SPSS (Statistical Package for the Social Sciences, IBM) version 28.0. The between-subject factor was group (two levels: OA versus aMCI), experimental digit span task condition (four levels: spoken, rhythm, pitch, melody) was the within-subject factor, and the score on the experimental digit span task was the dependent variable, with planned follow-up comparisons. We planned independent-samples t-tests based on our specific a-priori hypotheses and we used the one-tailed p-value as we had specific hypotheses regarding the direction of these effects in the different groups. In addition, to examine the effect of musical expertise and whether administration order might be a confounder, we conducted an analysis of covariance (ANCOVA) with administration order and musical expertise as covariates. We used only the General Sophistication factor of the Gold-MSI as this scale most closely fits our definition of musical expertise, covering the general engagement of a participant with/a participant's interest in music. Effect sizes (η_p² or Cohen's d) were reported for all factors for the analyses of variance. Alpha was set at .05 throughout, as all comparisons were planned. Effect sizes were interpreted as small, medium and large based on convention (Cohen, 1988; Cohen's d: .2, .5, .8, η_p²: .01, .06, .14 respectively).

RESULTS

Demographics and mean test values for both groups are shown in Table 1. All assumptions for the analysis of variance were met. Figure 2 shows the performance in the OA and in persons with aMCI on all four conditions (See also Table A3 for means and SDs).

FIGURE 2

Mean number correctly reproduced digit sequences (± SEM) in the four task conditions for the cognitively unimpaired older adults and persons with aMCI, showing the differences between the groups and between task conditions. *p < .05, **p < .01 (one-tailed).

Overall, OA outperformed aMCI patients in their performance on the experimental digit span task (F(1, 62) = 6.77, p = .012, η_p² = .098). Furthermore, the performance of both groups differed across the four conditions (F(3, 186) = 14.80, p ≤ .001, η_p² = .193); planned simple contrasts revealed a better performance in the rhythm versus spoken condition (p = .001, η_p² = .158), rhythm versus pitch (p ≤ .001, η_p² = .330) and rhythm versus melody condition (p ≤ .001, η_p² = .360). Furthermore, a better performance was found in the spoken versus pitch (p = .048, η_p² = .062) and spoken versus melody condition (p = .012, η_p² = .098). No significant difference was found between the pitch and melody conditions (p = .309, η_p² = .017). No significant Group × Condition interaction effect was found (F(3, 186) = 1.23, p = .300, η_p² = .019). Planned independent-samples t-tests revealed a significantly worse performance for the aMCI group compared to OA in the spoken (t(62) = −1.98, p = .026, d = 1.89), rhythm (t(62) = −2.72, p = .004, d = 1.75) and melody conditions (t(62) = −2.79, p = .004, d = 1.61) but no significant difference between OA and aMCI in the pitch condition (t(62) = −1.31, p = .097, d = 1.62).

We included the General Sophistication factor of the Self-Report Inventory of the Gold-MSI and administration order as covariates in the ANCOVA. Administration Order (F(1, 57) = .58, p = .449, η_p² = .010) did not significantly affect the performance on the experimental musical digit span task. General Sophistication (F(1, 57) = 5.63, p = .021, η_p² = .090) significantly affected the performance on the experimental digit span task. Post-hoc correlation analysis showed a positive direction of the correlations between General Sophistication and task performance on musically presented materials (spoken: r = .272, p = .031, rhythm: r = .368, p = .003, pitch: r = .214, p = .093, melody: r = .144, p = .259) (See Table A4 for the correlations per group). Despite the significant difference in depressive symptoms (p = .018) between both groups, the absolute difference between the groups was small and only four aMCI participants scored in the clinically relevant range. Therefore we argue that the findings are not confounded by group differences in mood.

DISCUSSION

The present study aimed to examine the effects of musical presentations (i.e., pitch, rhythm and melody) of a forward digit span task on WM performance of cognitively unimpaired OA and participants with aMCI. In addition, possible beneficial effects of musical expertise were also assessed. Results showed that OA outperformed participants with aMCI on the experimental digit span task, regardless of musical presentation. Rhythm facilitated digit span performance in both the OA and the aMCI participants, whereas pitch and melody hampered the performance as compared to spoken presentation. Additionally, a significantly worse performance for the aMCI as compared to OA was found in the spoken, rhythm and melody conditions. Finally, musical expertise was shown to contribute to task performance when materials were musically presented.

In line with previously reported positive effects of musical mnemonics on WM performance in university students and cognitively unimpaired young and older adults (e.g., Derks-Dijkman, Schaefer, Stegeman, et al., 2023; Silverman, 2007, 2010, 2012; Silverman & Schwartzberg, 2014, 2019) our hypothesis concerning the expected beneficial effects of rhythm in both groups was confirmed. Thus, rhythmic presentation positively affects WM performance, not only in cognitively unimpaired OA but also in participants with cognitive impairment (that is, aMCI). Therefore, we replicate and extend previous results, and suggest here that musical mnemonics may also improve WM performance in persons with (mild) cognitive impairment. A possible explanation for these findings is that rhythm or time structure facilitates chunking, which enhances WM performance (e.g., Derks-Dijkman, Schaefer, & Kessels, 2023; Derks-Dijkman, Schaefer, Stegeman, et al., 2023; Silverman, 2012), an explanation that is also supported for persons with aMCI.

In line with our expectations that both groups would also perform worst in our “pitch” and “melody” (that is, pitch-varying) conditions (Derks-Dijkman, Schaefer, Stegeman, et al., 2023; Silverman, 2007, 2010) we indeed found a significantly worse performance in the pitch and melody conditions compared to the spoken conditions. We previously argued based on findings of OA compared to younger adults that adding pitch to the pitch and melody conditions - even though created in a predictable tonal context – could result in ‘extra’ information that needs to be processed, thereby increasing the complexity of the stimulus (Derks-Dijkman, Schaefer, & Kessels, 2023) resulting in WM overload, which could also apply here to not only the OA, but also the aMCI participants.

Furthermore, we hypothesized that aMCI would benefit to a larger extent from rhythmic presentation than OA, the latter having a better-preserved WM performance than aMCI patients, thus performing already at a higher level than aMCI and at the level of young adults (Derks-Dijkman, Schaefer, Stegeman, et al., 2023). Specifically, rhythmic presentation is thought to decrease the executive task demands, which would especially benefit the cognitively impaired aMCI group (e.g., Gagnon & Belleville, 2011; Kirova et al., 2015). However, both groups showed similar benefits of rhythmic presentation, suggesting that rhythm is an effective element of a musical mnemonic regardless of expected (and observed) differences in WM performance between OA and aMCI.

Also, based on empirical evidence in OA (Derks-Dijkman, Schaefer, Stegeman, et al., 2023), we hypothesized that persons with aMCI would experience a faster WM overload in the conditions with only varying pitch (but no rhythm) or a rhythmic pattern with an added pitch component (i.e., our ‘melody’ condition), as a result of the stimulus complexity. However, planned comparison revealed that both groups did not differ in their performance on the pitch condition. Also, the aMCI showed no evident additional performance hindrance in the melody condition. These results suggest that here a higher stimulus complexity results in WM overload regardless of (mild) cognitive impairment.

Our results showed that having more musical expertise enhanced the beneficial effect of musical presentation in both the OA (cf. Derks-Dijkman, Schaefer, Stegeman, et al., 2023) and in aMCI participants (cf. Baird et al., 2017). Possibly, in both OA and aMCI participants, (having more) musical expertise – here formulated as degree of engagement with/interest in music – might have made it easier to effectively recall digits through a musical mnemonic, making it a relevant factor to consider in future research. Inspection of the correlations between musical expertise and the individual conditions shows that more musical expertise is associated with a better performance after spoken and rhythmic presentation (and weakly, but non-significantly correlated with the pitch condition), but – somewhat surprisingly – the weakest correlation was found in the melody condition. This may be in line with the chunking hypothesis, in that chunking may be most feasible in both the spoken and rhythm conditions (to which musical expertise apparently also contributes), but that the melody condition results in WM overload, especially in the individuals with aMCI regardless of their musical expertise (see Table A4, showing the correlations per group).

Potential limitations to be considered are that given our modest sample size, power may be limited for the correlational analyses. Next, as our experiment was performed in a naturalistic setting, the presentation of musical stimuli was not fully controlled. However, given our within-subject design regarding the conditions of musical presentation, we argue that these limitations may have attenuated some effects but not affected the outcome of our study. Furthermore, we did not analyse the effects of the serial position of the digits, being often reported in previous research (cf. Silverman, 2007, 2010, 2012). Also, one could argue that our paradigm mainly taps the phonological loop of the working memory system, that is, the maintenance of acoustical and/or verbal information (cf. Baddeley, 2000), and as such predominantly reflects short-term memory rather than the active manipulation aspect of WM (i.e., the central executive [CE] component). However, also in Baddeley and Hitch's view (Baddeley & Hitch, 1974) longer ‘forward sequences’ may elicit strategic processing, such as chunking, for which the CE is required. Furthermore, previous research has shown that the performance on forward digit span sequences is already reduced in MCI and may therefore benefit from compensation through a musical (that is, rhythmic presentation) mnemonic (Kessels et al., 2011). Alternatively to the chunking hypothesis in accordance with Baddeley's working memory model, the model by Derks-Dijkman, Schaefer, and Kessels (2023) argues that the musical stimulus (rhythm) and the verbal stimulus (digits) in combination may have resulted in a more cohesive stimulus with a reduced stimulus complexity and thus a lower working memory load due to an enhanced accent structure fit. Finally, although we carefully translated and adapted the Gold-MSI from the English original, the psychometric properties of the Dutch version of this questionnaire need to be established in large normative samples.

Strengths are the use of carefully designed stimuli, and carefully described clinical diagnoses of the aMCI participants. Furthermore, to our knowledge, this is the first study using this type of ‘musical’ digit span task in aMCI participants, thereby extending previous research in students and cognitively unimpaired young and older adults to a cognitively impaired population.

Future research focusing on the effects of musical mnemonics in (a)MCI and AD patients is recommended as both groups might have WM impairments (cf. Kirova et al., 2015), thereby focusing on the enhancement of WM performance, especially by means of rhythm. Open questions include the role of familiarity with the presented rhythms, as well as examining the effects of different types of rhythms (e.g., simple, more complex, related to variations in durations of tones or grouping of tones within the given pulse, cf. Derks-Dijkman, Schaefer, & Kessels, 2023). As recommended previously, extending the research from digits to other verbal stimuli, such as words, would also be relevant for future studies (cf. Derks-Dijkman, Schaefer, Stegeman, et al., 2023). Furthermore, in order to examine the effects of musical expertise, follow-up research conducting a moderation analysis using larger groups is recommended and could include behavioural measures of musical expertise as a covariate.

In conclusion, with the current study, we extend prior research findings in students (e.g., Silverman, 2007, 2010, 2012) and cognitively unimpaired young and older adults (Derks-Dijkman, Schaefer, Stegeman, et al., 2023) to a cognitively impaired population. Here we provide the first findings of the facilitating effect of rhythm (and hampering effect of pitch and melody) in a musical presentation of a WM task in cognitively unimpaired OA and aMCI participants to a similar extent. Future research is needed to further unravel whether next to rhythm – facilitating temporal chunking – tonal aspects of music could also contribute to chunking the verbal and musical stimulus together (Derks-Dijkman, Schaefer, & Kessels, 2023). As rhythm in particular seemed to result in a positive effect on WM performance in both OA and aMCI participants, we recommend using a rhythmically spoken presentation of to-be-remembered information in designing a musical mnemonic together with the individual who experiences memory problems. In daily life, this might be a potentially helpful strategy for remembering a limited amount of information for a short period of time.

AUTHOR CONTRIBUTIONS

Marije W. Derks-Dijkman: Conceptualization; formal analysis; methodology; project administration; visualization; writing – original draft; writing – review and editing. Rebecca S. Schaefer: Formal analysis; methodology; supervision; visualization; writing – original draft; writing – review and editing. Lisa Baan-Wessels: Investigation; project administration; writing – original draft; writing – review and editing. Ilse A. D. A. van Tilborg: Conceptualization; data curation; supervision; writing – original draft; writing – review and editing. Roy P. C. Kessels: Conceptualization; formal analysis; methodology; supervision; visualization; writing – original draft; writing – review and editing.

ACKNOWLEDGEMENTS

We would like to thank Maike Alberts and Maartje Stegeman, for their contribution to the data collection of this study in the context of their master thesis research.

CONFLICT OF INTEREST STATEMENT

No potential conflict of interest was reported by the author(s).

APPENDIX 1

TABLE A1. Example of scoring form of the musical digit span.

Item	Sequence	Response	Score	Item score
1. P	4 – 5 – 6 – 3 – 8		0 1	0 1 2
P	1 – 3 – 6 – 8 – 4		0 1
2. M	2 – 4 – 5 – 10 – 3		0 1	0 1 2
M	5 – 3 – 2 – 1 – 10		0 1
3. R	8 – 5 – 1 – 3 – 6		0 1	0 1 2
R	3 – 4 – 10 – 2 – 8		0 1
4. S	6 – 8 – 2 – 3 – 1		0 1	0 1 2
S	10 – 2 – 5 – 4 – 8		0 1
5. P	6 – 3 – 4 – 8 – 5 – 10		0 1	0 1 2
P	1 – 4 – 2 – 5 – 3 – 8		0 1
6. M	10 – 1 – 2 – 4 – 5 – 3		0 1	0 1 2
M	8 – 10 – 2 – 6 – 5 – 4		0 1
7. R	4 – 2 – 8 – 3 – 5 – 6		0 1	0 1 2
R	6 – 1 – 4 – 2 – 10 – 8		0 1
8. S	3 – 2 – 8 – 10 – 4 – 6		0 1	0 1 2
S	2 – 1 – 8 – 6 – 4 – 5		0 1
9. P	6 – 3 – 4 – 1 – 8 – 5 – 10		0 1	0 1 2
P	1 – 4 – 2 – 5 – 6 – 3 – 8		0 1
10. M	10 – 1 – 2 – 4 – 5 – 3 – 6		0 1	0 1 2
M	8 – 10 – 2 – 3 – 6 – 5 – 4		0 1
11. R	5 – 3 – 8 – 6 – 4 – 2 – 1		0 1	0 1 2
R	2 – 6 – 4 – 10 – 5 – 3 – 8		0 1
12. S	3 – 2 – 8 – 10 – 4 – 6 – 5		0 1	0 1 2
S	10 – 2 – 1 – 8 – 6 – 4 – 5		0 1
13. P	2 – 3 – 5 – 6 – 10 – 1 – 8 – 4		0 1	0 1 2
P	4 – 2 – 10 – 1 – 3 – 5 – 6 – 8		0 1
14. M	10 – 3 – 6 – 2 – 4 – 8 – 1 – 5		0 1	0 1 2
M	8 – 1 – 2 – 3 – 10 – 5 – 4 – 6		0 1
15. R	10 – 4 – 3 – 6 – 5 – 1 – 8 – 2		0 1	0 1 2
R	1 – 2 – 5 – 8 – 10 – 6 – 3 – 4		0 1
16. S	5 – 6 – 8 – 2 – 4 – 3 – 1 – 10		0 1	0 1 2
S	1 – 3 – 5 – 8 – 6 – 2 – 10 – 4		0 1

• Note: Score form (one of 24 administration orders). The four conditions are abbreviated with S, ‘spoken’; P, ‘pitch’; R, ‘rhythm’ and M, ‘melody’.

TABLE A2. Administration orders of the experimental digit span task.

1.	2.	3.	4.	5.	6.
Spoken	Spoken	Spoken	Spoken	Spoken	Spoken
Pitch	Pitch	Rhythm	Rhythm	Melody	Melody
Melody	Rhythm	Melody	Pitch	Pitch	Rhythm
Rhythm	Melody	Pitch	Melody	Rhythm	Pitch

7.	8.	9.	10.	11.	12.
Pitch	Pitch	Pitch	Pitch	Pitch	Pitch
Melody	Melody	Spoken	Spoken	Rhythm	Rhythm
Rhythm	Spoken	Rhythm	Melody	Melody	Spoken
Spoken	Rhythm	Melody	Rhythm	Spoken	Melody

13.	14.	15.	16.	17.	18.
Rhythm	Rhythm	Rhythm	Rhythm	Rhythm	Rhythm
Melody	Melody	Spoken	Spoken	Pitch	Pitch
Pitch	Spoken	Melody	Pitch	Spoken	Melody
Spoken	Pitch	Pitch	Melody	Melody	Spoken

19.	20.	21.	22.	23.	24.
Melody	Melody	Melody	Melody	Melody	Melody
Spoken	Spoken	Pitch	Pitch	Rhythm	Rhythm
Pitch	Rhythm	Rhythm	Spoken	Pitch	Spoken
Rhythm	Pitch	Spoken	Rhythm	Spoken	Pitch

• Note: Grid showing all possible orders of administration with different combinations of the four conditions (i.e., spoken, pitch, rhythm and melody).

TABLE A3. Descriptive statistics for the results by group and condition.

	Spoken		Pitch		Rhythm		Melody		n
	M	SD	M	SD	M	SD	M	SD
OA	3.47	2.11	2.87	1.70	4.19	1.71	3.00	1.69	32
aMCI	2.53	1.65	2.34	1.54	3.00	1.78	1.88	1.54	32
Total	3.00	1.94	2.61	1.63	3.59	1.83	2.44	1.70	64

TABLE A4. Correlations between the Gold-MSI General Sophistication subscale and the performances on the four conditions of the musical digit span.

Condition	Total (N = 63)	OA (N = 32)	aMCI (N = 31)
	Gold-MSI General Sophistication	Gold-MSI General Sophistication	Gold-MSI General Sophistication
Spoken	.272*	.322	.196
Pitch	.214	.222	.199
Rhythm	.368**	.511**	.228
Melody	.144	.219	.039

• Note: Pearson correlations. *p < .05, **p < .01.

Open Research

DATA AVAILABILITY STATEMENT

The auditory stimuli (mp4 sound files) used in this experiment are available via the Donders Institute repository (https://doi.org/10.34973/dxq4-a452). Unfortunately, the data cannot be shared since the informed consent form that all participants signed upon recruitment did not contain a statement on data sharing, for which explicit consent is required under the EU General Data Protection Regulation 2016/679.