Association between balance ability and memory function in community-dwelling older adults Authors: Masataka Ohki, Kyoko Minehisa, Keiko Sakai, Maki Takei, Masakazu Imaoka, Masatoshi Takeda, Misa Nakamura Published: 2026 ABSTRACT Exercise is a commonly recommended non-pharmacological intervention for the prevention of dementia. However, selection of the most appropriate and effective exercise requires clarification of the relationship between types of motor and cognitive functions. We aimed to determine motor functions that might be related to cognitive function. The study participants were 49 community-dwelling older adults with normal cognitive function (74.33 ± 6.16 years old). Cognitive function was evaluated using the Addenbrooke cognitive examination-III (ACE-III) and motor function by hand grip strength, 30-second chair stand test, walking speed, and unipedal stance time with eyes open (UPST). Physical activity was measured by wireless activity meter. Univariate regression analysis showed that age, skeletal muscle mass index, UPST, step count, and Exercise (METs・h) were significantly associated with ACE-III total scores. In multiple linear regression analysis, ACE-III total scores and UPST were significantly associated in Model 1 (β = 0. 337, P = .025), Model 2 (β = 0.349, P = .016), Model 3 (β = 0.337, P = .023), and Model 4 (β = 0.379, P = .007). The ACE-III memory domain and UPST had a significant association in Model 1 (β = 0.317, P = .045), Model 2 (β = 0.350, P = .022), and Model 4 (β = 0.301, P = .037). Better balance ability was associated with better cognitive function, particularly memory performance, in our cohort of cognitively normal older adults. Balance assessment may serve as a simple adjunctive indicator of cognitive health. 1. Introduction Motor functions, such as walking speed and cognitive function, are known to decline with age.[1] Cognitive impairments are associated with falls and impairment of activities of daily living, and they are thought to greatly affect activities of daily living.[2,3] Furthermore, combined cognitive impairments in hospitalized patients are associated with especially poor functional outcomes.[4] Prevention and remediation of cognitive impairments are thus thought to be extremely important. Physical inactivity is a risk factor for the onset of dementia, so increasing physical activity may reduce the risk and prevalence of dementia.[5,6] Exercise is a commonly recommended non-pharmacologic intervention for cognitive impairments. It may contribute to the prevention of dementia by increasing cognitive reserve and reducing neuroinflammation and brain damage.[7] Exercise regimens for patients with mild cognitive impairment and various types of dementia have been shown to have the potential to improve cognitive function. In patients with dementia, the effect of exercise is comparatively small,[8] but it may be more effective as an intervention in earlier conditions, such as in mild cognitive impairment[9–13] and in individuals with no objective cognitive impairment.[14,15] Types of exercise include aerobic training,[13] resistance training,[11] or multicomponent training, which is a combination of both.[9] Mindfulness-based stress reduction[15] and dual tasks[14] are also occasionally added. However, the mechanism underlying the improvement in cognitive function in relation to these exercises is unclear. The effects of exercise on the subdomains of cognitive function have been reported in participants with subjective but not objective cognitive impairment. These effects include improved delayed memory,[16,17] improved verbal fluency,[18] and improved prospective memory[19] through aerobic training, resistance training, and multicomponent training. To elucidate more effective exercise interventions aimed at preventing cognitive impairment, we suggest it is necessary to clarify the association between each motor domain and cognitive functions. In this study, we aim to determine the motor functions that are related to cognitive functions in community-dwelling people without cognitive impairment and to clarify how cognitive domains relate to particular motor functions. 2. Methods 2.1. Study design and participants We employed a cross-sectional approach to examine the relationship between motor and cognitive function in community-dwelling older adults. Study subjects, all Japanese, were participants in a measurement session co-sponsored by Osaka Kawasaki Rehabilitation University and Kaizuka city in 2022, and participants in a health class, and were willing to be included in the study. The exclusion criteria were those with a cardiac pacemaker, those with Addenbrooke cognitive examination-III (ACE-III) total scores ≤ 88 (see below), and those with missing data. In total, 49 participants were analyzed. The calculated sample size required for a 4-group comparison in a cross-sectional study was 48 (effect size, 0.25; α error, 0.05; power, 0.8), so the actual sample size of 49 was considered to be appropriate. The study was approved by the Osaka Kawasaki Rehabilitation University Ethics Committee (Reference No. OKRU-RA0038) and was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent prior to the start of the study. 2.2. Measurements of body composition Body composition parameters (body mass index and skeletal muscle mass index [SMI]) were measured using a bioelectrical impedance analysis device (InBody 270, InBody) at 20 and 100 kHz.[20] Participants stood barefoot on the Inbody device, with their feet not touching each other. Their elbows were extended, their shoulders slightly abducted, and they held handle electrodes in each hand. Skeletal muscle mass was assessed using SMI, which was calculated as muscle mass (kg) divided by the square of height (m) (kg/m2). 2.3. Serum biochemical measurements All blood collections were taken at least 2 hours after the participants had eaten, between 09:00 AM and 3:00 PM. Apolipoprotein E (ApoE) genotyping was performed by Japan Clinical Laboratories, Inc., to identify the genetic risk distribution for dementia in the study population. 2.4. Measurements of cognitive function ACE-III, first validated by Hsieh et al,[21] is used to evaluate cognitive function. We used the validated Japanese version of the questionnaire.[22] ACE-III consists of attention, memory, fluency, language, and visuospatial subdomains. The highest total score is 100 points, and the highest scores for each domain are 18, 26, 14, 26, and 16 points for attention, memory, fluency, language, and visuospatial domains, respectively. Normal cognitive function is indicated by a total score of ≥ 89, with a score of 76 to 88 indicating a suspicion of mild cognitive impairment, and a score of ≤ 75 indicating a suspicion of dementia.[22] 2.5. Measurements of motor function Hand grip strength was evaluated as an indicator of whole-body muscle strength.[23] The measurement was performed in a standing position, with the shoulder joint slightly abducted and the elbow joint at 0° so that the upper limb did not come into contact with the trunk. The left and right sides were each measured twice, and the maximum value was used. A digital hand dynamometer (Grip-D, Takei Equipment Industry) was used for the measurement. To evaluate lower extremity performance, we used the 30-second chair stand test (CS-30) and maximum walking speed. CS-30 is a simple tool for evaluating lower-limb muscle strength.[24] Participants were instructed to repeat as many standing and seated repetitions as possible during a 30-second period from a seated position in a chair. Considering the effects of fatigue, the test was performed only once. Gait speed is a simple measure of motor function and is associated with survival and fall risk in older adults.[25,26] As in a previous study,[27] a walking path was set up with a measurement section of 2.4 m and an aiding section of 1 m before and after. Walking time at maximum speed was measured twice and then averaged to calculate walking speed. Unipedal stance time with eyes open (UPST) was used to measure balance ability. The UPST is a balance ability test that has been validated for reliability, regardless of age or sex.[28] We selected UPST because it is a well-validated and reliable measure of balance ability that can be safely administered to older adults without increasing the risk of falls. The participants were instructed to place their hands on their hips, and after a signal, slightly flex 1 side of their hip and knee joints, lift their foot off the floor, and maintain that position for as long as possible. The maximum time was set at 60 seconds. Measurements were taken twice on each side, and the maximum value was used. 2.6. Measurements of physical activity A wireless activity meter (AM510N; ACOS Co., Ltd.) was used to measure physical activity. Wireless activity meters were distributed 3 months prior to the measurement date, and participants were requested to wear them for 3 months, except when bathing. Five items were collected via the activity meter: number of steps, number of active steps (number of steps during activities of 4 or more metabolic equivalent of task [METs] units), active activity duration (duration of 4 or more activities), METs activity duration, and Ex (METs・h). 2.7. Questionnaires A paper-based self-administered questionnaire was used to collect information on educational background (years of formal education), smoking history, and alcohol consumption in midlife. 2.8. Statistical analysis In the linear regression analysis, ACE-III was the dependent variable, and the independent variables were sex, age, educational background, body mass index, SMI, grip strength, CS-30, walking speed, and UPST. In the multiple linear regression analysis, the total ACE-III scores or each subitem was the dependent variable. The independent variables were sex, age, and UPST for Model 1; sex, educational background, and UPST for Model 2; age, educational background, and UPST for Model 3; and sex, SMI, and UPST for Model 4. Owing to the relatively small sample size, we constructed these 4 parsimonious models instead of a single fully adjusted model to minimize overfitting while examining the robustness of the association between UPST and cognitive function across clinically relevant confounders. Statistical analysis was performed using SPSS (version 28.0 for Windows; SPSS, Inc.) with a significance level of 5%. 3. Results 3.1. Participant characteristics The analysis included 49 participants (mean age 74.33 ± 6.16 years, 37 women and 12 men) after exclusion of 2 individuals with missing data and 9 with ACE-III total scores ≤ 88. The results for age, ApoE genotype, body composition, cognitive function, motor function, and physical activity by sex are shown in Table 1. Table 1Participant characteristics.TotalMenWomenParticipants, n (%)4912 (24.4)37 (74.6)Age (yrs)74.33 ± 6.1676.67 ± 6.4673.57 ± 5.96Educational background (yrs)13.02 ± 1.2014.08 ± 2.0212.68 ± 1.89Body composition BMI (kg/m2)22.40 ± 2.8024.16 ± 2.8521.83 ± 2.57 SMI (kg/m2)6.11 ± 0.927.43 ± 0.545.68 ± 0.51ApoE phenotype, n (%) ε3/ε334 (69.4)12 (100)22 (59.5) ε4/ε38 (16.3)0 (0)8 (21.6) ε2/ε36 (12.2)0 (0)6 (16.2) ε4/ε21 (2.0)0 (0)1 (2.7)Alcohol*, n (%)23 (46)11 (91.7)12 (32.4)Smoking†, n (%)2 (4.0)1 (8.3)1 (2.7)Cognitive function ACE-III total score (points)94.90 ± 2.8094.00 ± 2.4995.19 ± 2.87 Attention domain (points)17.61 ± 0.6717.50 ± 0.6717.65 ± 0.68 Memory domain (points)25.14 ± 1.1925.25 ± 1.1425.11 ± 1.22 Fluency domain (points)11.18 ± 1.6410.58 ± 2.0211.38 ± 1.48 Language domain (points)25.37 ± 0.8825.17 ± 0.7225.43 ± 0.93 Visuospatial domain (points)15.59 ± 0.6715.50 ± 0.6715.62 ± 0.68Motor function Hand grip strength (kg)27.53 ± 6.8037.92 ± 2.3224.16 ± 3.51 CS-30 (count)19.55 ± 4.6421.08 ± 7.3119.05 ± 3.37 Walking speed (m/s)1.65 ± 0.231.75 ± 0.161.61 ± 0.245 UPST (s)42.64 ± 20.9134.64 ± 20.1745.24 ± 20.75Physical activity Step count (steps)5525.23 ± 2274.765617.00 ± 2940.425495.46 ± 2062.48 Active step count (steps)1475.05 ± 1364.631670.79 ± 1618.561411.57 ± 1290.57 Active activity time (min)10.83 ± 10.5913.03 ± 13.1110.12 ± 9.74 Activity time (min)218.31 ± 57.62197.60 ± 68.14225.03 ± 53.08 Ex (METs・h)2.44 ± 1.303.74 ± 1.942.35 ± 1.03Values are presented as means (SD) or prevalence (%).ACE-III = Addenbrooke cognitive examination-III, ApoE = apolipoprotein E, BMI = body mass index, CS-30 = 30-second chair stand test, n = number of participants, SD = standard deviation, SMI = skeletal muscle mass index, UPST = unipedal stance time with eyes open.*Asked about a history of alcohol consumption in middle age.†Asked about current smoking status. Participant characteristics. Values are presented as means (SD) or prevalence (%). ACE-III = Addenbrooke cognitive examination-III, ApoE = apolipoprotein E, BMI = body mass index, CS-30 = 30-second chair stand test, n = number of participants, SD = standard deviation, SMI = skeletal muscle mass index, UPST = unipedal stance time with eyes open. Asked about a history of alcohol consumption in middle age. Asked about current smoking status. 3.2. Coefficients for factors associated with the ACE-III total scores on univariate linear regression analysis The results of univariate regression analysis to predict the ACE-III total scores and related variables are shown in Table 2. Age (β = -0.302, P = 0.035), SMI (β = -0.284, P = 0.048), UPST (β = 0.413, P = .003), step count (β = 0.548, P < .001), and Ex (METs・h) (β = 0.442, P = .001) were significantly associated with ACE-III total scores (Table 2). Table 2Results of linear regression analysis for each factor and its association with the ACE-III total scores. β 95% CI P Sex−0.184−3.048–0.670.204Age−0.302−0.264–−0.010.035*Educational background0.198−0.126–0.682.173BMI−0.251−0.535–0.033.082SMI−0.284−1.724–−0.010.048*Hand grip strength−0.181−0.193–0.044.214CS-300.078−0.129–0.224.593Walking speed0.161−1.550–5.407.270UPST0.4130.020–0.091.003*Step count0.5480.000†–0.001< .001**Active step count0.2810.000–0.001.051Active activity time0.2750.000–0.027.056Activity time0.270−0.003–0.146.061Ex (METs・h)0.4420.386–1.520.001*β is the standardized partial regression coefficient.BMI = body mass index, CI = confidence interval, CS-30 = 30-second chair stand test, SMI = skeletal muscle mass index, UPST = unipedal stance time with eyes open.†0.000373.*P < .05.**P < .001. Results of linear regression analysis for each factor and its association with the ACE-III total scores. β is the standardized partial regression coefficient. BMI = body mass index, CI = confidence interval, CS-30 = 30-second chair stand test, SMI = skeletal muscle mass index, UPST = unipedal stance time with eyes open. 0.000373. P < .05. P < .001. 3.3. Coefficients for factors associated with the ACE-III total scores on multiple linear regression analysis The results of multiple linear regression analysis are presented in Table 3. The ACE-III total score was the dependent variable, and the independent variables were sex, age, and UPST in Model 1; sex, educational background, and UPST in Model 2; age, educational background, and UPST in Model 3; and sex, SMI, and UPST in Model 4. Multiple linear regression analysis showed that UPST was significantly associated with the ACE-III total scores in Model 1 (β = 0. 337, P = .025), Model 2 (β = 0.349, P = .016), Model 3 (β = 0.337, P = .023), and Model 4 (β = 0.379, P = .007). The variance inflation fractions for all variables were < 4, and multicollinearity was not observed. In summary, longer UPST was associated with higher total ACE-III scores (Table 3). Table 3Results of multiple linear regression analysis for factors associated with the ACE-III total scores.Independent variable β 95% CI P Model 1UPST0.3370.006–0.084.025*Sex−0.075−2.277–1.309.589Age−0.161−0.206–0.060.273Model 2UPST0.3490.009–0.084.016*Sex−0.170−2.971–0.777.245Educational background0.204−0.116–0.688.159Model 3UPST0.3370.007–0.084.023*Age−0.159−0.203–0.059.274Educational background0.130−0.197–0.562.338Model 4UPST0.3790.014–0.087.007*Sex0.215−1.667–4.433.366SMI−0.381−2.601–0.279.111β is the standardized partial regression coefficient.CI = confidence interval, SMI = skeletal muscle mass index, UPST = unipedal stance time with eyes open.*P < .05. Results of multiple linear regression analysis for factors associated with the ACE-III total scores. β is the standardized partial regression coefficient. CI = confidence interval, SMI = skeletal muscle mass index, UPST = unipedal stance time with eyes open. P < .05. 3.4. Association between UPST and ACE-III subdomain scores The results of multiple linear regression analysis for each of the ACE-III subdomains of the attention, memory, fluency, language, and visuospatial domains for UPST are shown in Table 4. The dependent variable was the ACE-III subdomain. The independent variables were sex, age, and UPST in Model 1; sex, educational background, and UPST in Model 2; age, educational background, and UPST in Model 3; and sex, SMI, and UPST in Model 4. UPST was significantly associated with the ACE-III memory domain in Model 1 (β = 0.317, P = .045), Model 2 (β = 0.350, P = .022), and Model 4 (β = 0.301, P = .037) (Table 4). Table 4Association between UPST and ACE-III subdomain scores. β 95% CI P Attention domain Model 10.007−0.010–0.010.963 Model 20.060−0.008–0.012.691 Model 3−0.004−0.010–0.010.979 Model 40.113−0.006–0.013.460Memory domain Model 10.3170.000–0.036.045* Model 20.3500.003–0.037.022* Model 30.3070.000†–0.035.052 Model 40.3010.001–0.033.037*Fluency domain Model 10.176−0.010–0.038.247 Model 20.198−0.007–0.038.174 Model 30.182−0.009–0.038.225 Model 40.249−0.003–0.042.092Language domain Model 10.256−0.002–0.024.109 Model 20.217−0.004–0.033.158 Model 30.276−0.002–0.024.093 Model 40.218−0.003–0.021.136Visuospatial domain Model 10.070−0.008–0.013.667 Model 20.005−0.010–0.010.976 Model 30.071−0.008–0.013.659 Model 40.039−0.009–0.011.801Multiple linear regression analysis. Model 1: UPST, sex, and age as independent variables; Model 2: UPST, sex, and educational background as independent variables; Model 3: UPST, age, and educational background as independent variables; and Model 4: UPST, sex, and SMI as independent variables.β is the standardized partial regression coefficient.CI = confidence interval, UPST = unipedal stance time with eyes open.†0.000392.*P < .05. Association between UPST and ACE-III subdomain scores. Multiple linear regression analysis. Model 1: UPST, sex, and age as independent variables; Model 2: UPST, sex, and educational background as independent variables; Model 3: UPST, age, and educational background as independent variables; and Model 4: UPST, sex, and SMI as independent variables. β is the standardized partial regression coefficient. CI = confidence interval, UPST = unipedal stance time with eyes open. 0.000392. P < .05. 4. Discussion The distribution of ApoE genotypes in the present study was generally comparable with that reported for healthy Japanese populations, although the proportion of individuals with the e2/e3 genotype was slightly higher.[29] The e2/e3 genotype has been associated with a lower risk of Alzheimer disease than the e3/e3 genotype,[30] so the participants in the present study may have had a relatively low genetic risk for dementia. Regarding motor function, the mean grip strength of the participants in this study was higher than that reported in previous studies on Japanese older adults.[31] The CS-30 of the participants in this study was better than that in a previous study for both men and women.[32] Regarding walking speed, our participants of either sex did not significantly differ from the previously reported mean maximum speed of community-dwelling older adults without cognitive impairments.[33] The UPST results in this study were better than those previously reported for both men and women.[28] In summary, compared with previous studies, the participants in this study were considered to be at low risk for developing dementia and had overall good muscle strength and balance ability. In the current study, longer UPST scores were associated with higher ACE-III scores and memory domain scores in older adults without cognitive impairment. No significant associations were found between cognitive function and grip strength, CS-30 score, or walking speed. Balance training has been reported to improve memory and visuospatial cognitive function in healthy adults and to enhance attention in community-dwelling older adults.[34,35] However, the relationship between memory function and balance ability in community-dwelling older adults has not been thoroughly examined in the literature. This study is believed to be the first to attempt to clarify the relationship between memory function and balance ability in older adults. In older adults with normal cognitive function, the association between balance ability and cognitive function may reflect subtle differences in neural function rather than marked cognitive impairment. The UPST requires sustained attention, multisensory integration, and postural control, all of which depend on cortical and subcortical networks that overlap with memory-related regions.[36,37] This may partially explain why, even among older adults with normal cognitive function, better balance ability has been associated with better memory performance. When viewed as a postural maintenance task, the UPST is thought to require the integration of the visual, somatosensory, and vestibular systems. The vestibular system contributes to balance ability as a sensory system and has been reported to correlate with the volume of memory-related regions in the hippocampus.[38] In this study, balance ability was assessed using the UPST, which reflects integrated postural control rather than the function of a single sensory system. It was not, therefore, possible to statistically verify the independent associations between each of these systems and cognitive function. The association between balance ability and cognitive function observed in this study could potentially be explained by the integration of the vestibular, visual, and somatosensory systems. However, these mechanisms were not directly tested in the present study and should be interpreted as plausible hypotheses based on prior research. Regarding the relationship between CS-30 and cognitive function, low Mini-Mental State Examination (MMSE) scores were reportedly associated with low CS-30 scores when MMSE scores were used to classify normal cognitive function, cognitive decline, and cognitive impairment.[39] However, there are no studies in the literature on the association of cognitive function in participants with normal cognitive function, such as the participants in this study. It is possible that CS-30 does not cause cognitive decline within the normal range, but this is only detected after objective cognitive decline has already occurred. Walking speed has been suggested to decrease with cognitive function,[40] but we found no significant association between maximum walking speed and the ACE-III total scores in the current study. One possible explanation is that the participants maintained relatively high cognitive function, as reflected by the high mean ACE-III scores in both men and women. Gait slowing has been suggested to become more apparent during the transition from normal cognition to mild cognitive impairment, and it may precede future cognitive decline.[41–43] Most participants in the present study exhibited cognitive performance within the normal range, so the association between walking speed and cognitive function may have been too subtle to detect. In addition, differences between studies in gait assessment protocols (comfortable vs maximal walking speed) and cognitive assessment tools (MMSE/montreal cognitive assesment vs ACE-III) may also have contributed to the inconsistent findings. Hand grip strength has been positively associated with cognitive function,[44] but our results are inconsistent with this. Hand grip strength reportedly decreases with age,[31] but we found no significant association between age and grip strength. The comparatively high amount of physical activity of our participants could be a possible reason for this. In this study, active steps and physical activity times were negatively associated with age, but there was no significant association between steps, activity time, or Ex (METs × time) and age (Supplementary File 1, Supplemental Digital Content 1). Physical activity level has been reported to be positively associated with grip strength, regardless of intensity.[45] It is possible that the participants in this study maintained a relatively high level of physical activity, so there was no decline in their grip strength. This could explain the lack of association between cognitive function and grip strength. Balance exercises, especially those targeting anticipatory control, dynamic stability, functional stability limits, reactive control, and flexibility, have been shown to reduce falls.[46,47] Single-task and dual-task balance training has also been shown in previous studies to improve gait smoothness, walking speed, fear of falling,[48] and attentional function in community-dwelling older adults.[35] These findings, together with the present results, suggest that balance performance may be associated with multiple aspects of healthy aging, including cognitive and memory function. However, although previous intervention studies have reported cognitive benefits of balance training, the present cross-sectional study cannot determine whether improving balance directly leads to improvements in memory function. Future longitudinal and intervention studies are therefore needed to clarify the causal relationship between balance ability and cognitive function. 5. Limitation and future research This study has several limitations that warrant mention. First, we targeted participants living in a specific region, and the study was limited to those who volunteered to participate. These participants are thought to be more health-conscious, more physically active, and more motivated than the general elderly population. The association between balance function and cognitive function may not, therefore, be applicable to frail older adults, older adults living in care facilities, or groups with different cultural, social, or economic backgrounds. A second limitation is that this was a cross-sectional study, so the causal relationship between UPST and cognitive function is unclear. Future research should clarify the effects of interventions to improve balance ability on cognitive and memory function in the prevention of dementia through exercise. Third, this study only used the eyes-open single-leg stance test to assess balance ability, although it is a commonly used assessment that is safe and can be performed in both community and clinical settings. Postural control is achieved through the integration of multiple senses, including vision, somatosensation, and vestibular sensation, and the weighting of these senses may change with age. In addition to the relationship between cognitive function and vestibular sensation mentioned in this study, there may also be a relationship between vision and somatosensation. However, we did not directly assess each sense, so we were unable to verify the independent relationship between each sense and cognitive function. Future research is needed to clarify the relationship between balance ability and cognitive function under different sensory conditions, such as limited visual input or disruption to somatosensation. Furthermore, investigation into the relationship between dynamic balance and cognitive function is also necessary. Fourth, the sample size satisfied the calculated minimum requirement for the planned analyses, but it remained relatively small. The absence of significant associations between cognitive function and other motor variables, including grip strength, walking speed, and chair stand performance, should therefore be interpreted with caution. The study may have been underpowered to detect weaker but clinically meaningful associations. Future studies with larger sample sizes are warranted to confirm these findings. From a clinical perspective, the UPST is an inexpensive, rapid, and easily administered assessment that can be incorporated into routine geriatric evaluations and community health screening programs. Our findings suggest that poor balance performance may serve as an adjunctive indicator for identifying older adults who may benefit from further cognitive assessment. However, this study was cross-sectional, so the observed association should not be interpreted as evidence of a causal relationship, and the UPST should not be regarded as a diagnostic tool for cognitive impairment. Rather, it may be useful as a simple screening measure that complements established cognitive assessments. 6. Conclusions Better balance performance was shown in this study to be associated with better cognitive function, particularly memory performance, in cognitively normal older adults. Balance performance may serve as a simple adjunctive indicator of cognitive health, but longitudinal and intervention studies are required to determine whether improving balance can truly contribute to maintaining cognitive function. Acknowledgements We thank Benjamin Phillis for proofreading and editing. We also thank the many students at Osaka Kawasaki Rehabilitation University and our colleagues at Kawasaki Hospital for their cooperation in testing and measurements. Author contributions Conceptualization: Masataka Ohki, Misa Nakamura. Data curation: Masataka Ohki, Misa Nakamura. Formal analysis: Masataka Ohki. Funding acquisition: Masatoshi Takeda, Misa Nakamura. Investigation: Masataka Ohki, Kyoko Minehisa, Keiko Sakai, Maki Takei, Masakazu Imaoka. Project administration: Misa Nakamura. Software: Masataka Ohki. Supervision: Masatoshi Takeda. Writing – original draft: Masataka Ohki. Writing – review & editing: Misa Nakamura.