Using spaced retrieval and Montessori-based activities in improving eating ability for residents with dementia


  • This article was published online on 6 January 2010. An error was subsequently identified in the Key points section. This notice is included in the online and print versions to indicate that both have been corrected [26 January 2010].



To construct a training protocol for spaced retrieval (SR) and to investigate the effectiveness of SR and Montessori-based activities in decreasing eating difficulty in older residents with dementia.


A single evaluator, blind, and randomized control trial was used. Eighty-five residents with dementia were chosen from three special care units for residents with dementia in long-term care facilities in Taiwan. To avoid any confounding of subjects, the three institutions were randomized into three groups: spaced retrieval, Montessori-based activities, and a control group. The invention consisted of three 30–40 min sessions per week, for 8 weeks.


After receiving the intervention, the Edinburgh Feeding Evaluation in Dementia (EdFED) scores and assisted feeding scores for the SR and Montessori-based activity groups were significantly lower than that of the control group. However, the frequencies of physical assistance and verbal assistance for the Montessori-based activity group after intervention were significantly higher than that of the control group, which suggests that residents who received Montessori-based activity need more physical and verbal assistance during mealtimes. In terms of the effects of nutritional status after intervention, Mini-Nutritional Assessment (MNA) in the SR group was significantly higher than that of the control group.


This study confirms the efficacy of SR and Montessori-based activities for eating difficulty and eating ability. A longitudinal study to follow the long-term effects of SR and Montessori-based activities on eating ability and nutritional status is recommended. Copyright © 2010 John Wiley & Sons, Ltd.


Eating difficulty in dementia may involve impaired food transfer from the meal tray to the mouth, impaired transfer of food from the mouth into the stomach, or both. Once self-feeding is impaired, the patient becomes at risk for malnutrition and its sequelae. When self-feeding ceases altogether, the patient becomes dependent on the motivation, commitment, and skill of caregivers for adequate nutrition and hydration (Priefer and Robbins, 1997; Castellanos et al., 2003). Maintaining the eating ability of patients with dementia is a major concern of health professionals, but only a few published training programs for nurses and nursing assistants focusing on feeding institutionalized residents with dementia exist (Roberts and Durnbaugh, 2002; Chang and Lin, 2005). Moreover, the myth that patients with dementia cannot learn limits the training of patients with dementia to deal with their eating problems (Brush and Camp, 1998a).

Spaced retrieval (SR) training is regarded as one method for patients with dementia that can enhance learning and retention of information by recalling that information over increasingly longer periods of time (Bourgeois et al., 2003; Loewenstein et al., 2004). SR involves the use of well-learned processes and the unconscious, relatively effortless acquisition, and retrieval of new information (Bourgeois et al., 2003). Several studies have reported the efficacy of SR in improving eating difficulties among patients with dementia (Camp and Stevens, 1996; Brush and Camp, 1998a; Brush and Camp, 1998b). However, the few studies about SR as a treatment for eating difficulty of patients with dementia were all limited in that they did not use an adequate sample size, had no control group, no assessor blinding, no statistical comparison between pre- and post intervention, no standardized protocol, and/or no objective outcome measure to reflect improvement in eating difficulty, such as self-eating ability and nutritional status (Camp and Stevens, 1996; Brush and Camp, 1998a; Brush and Camp, 1998b; Bourgeois et al., 2003; Hopper et al., 2005).

The Montessori methods are regarded as being a successful approach for special care dementia units in clinical practice (Buckbee, 1999). Montessori methods are regarded as capable of stopping or reducing residents' problem behaviors when residents participated in Montessori-based programming (Orsulic-Jeras et al., 2000a; Camp, 2001). However, problem behaviors often occur when residents are carrying out daily activities such as eating, toileting, or bathing (Ryden et al., 1991) and these daily activities were not observed during previous studies (Orsulic-Jeras et al., 2000a; Orsulic-Jeras et al., 2000b; Judge et al., 2000; Camp, 2001). Additionally, like the studies on SR, previous studies regarding Montessori-based activities had similar limitations in their research design (Camp et al., 1997; Orsulic-Jeras et al., 2000a; Orsulic-Jeras et al., 2000b; Camp, 2001; Camp and Skrajner, 2004; Mahendra et al., 2006; Jarrott et al., 2008).

Therefore, the aim of this study was to investigate the effectiveness of training of SR and Montessori-based activities in decreasing feeding difficulty and nutritional status for residents with dementia.


Approval for the study was obtained from the Institutional Review Board (IRB) of the Taipei Veterans General Hospital. Written consent was obtained from participants if they were able to understand the implications of the study. Failing this, their family or guardian provided proxy consent for participating in the study.


During the study period, a total of 15 dementia special care units existed in Taiwan. Subjects were chosen from dementia special care units at three long-term care facilities in Metropolitan Taipei. Subjects chosen met the following criteria: (1) Being diagnosed with dementia on their chart; (2) Scored ≥ 2 on the EdFED; (3) Able to stay in the institutions during the entire study period; and (4) their MMES results ranged from 10–23.

Eighty-five subjects participated this study, with 82 subjects finishing the 8-week study. The three dropouts were due to hospitalization and/or an unwillingness to continue, which yielded a 96.5% completion rate.


A single evaluator, blind, randomized control trial was used. To avoid residents confounding, the three institutes were randomly assigned to the SR, Montessori-based activity, and control groups. The interventions for the SR, Montessori-based activity, and control group were done in 35–40 min sessions, three times per week, for 8 weeks. The principal researcher explained the research, purpose, and procedures to the managers of the nursing home and/or the subjects' relatives and obtained informed consent.

Baseline data on eating difficulty, nutritional status, and eating amount were collected by the research assistants before the interventions began. Each resident was observed for three meals to calculate their length of eating time and eating amount as well as assistance levels given during meals. Once the baseline data collection during mealtimes was completed, two researchers led SR and Montessori-based activities prior to the care activity for a period of 8 weeks. Following the 8 weeks intervention, the same type of data collected for the baseline data were collected during mealtimes for post-intervention behaviors. No data were collected during the 8 weeks intervention period itself. Routine activity was a control intervention so that the effects could be differentiated from the effects obtained from the other two interventions. The data collectors did not know which group the subjects belonged to. To control internal validity of the experimental design, this study arranged for the same research members to perform the SR and Montessori-based group interventions. In terms of the control group, the leader of routine activities was based on staff's working schedule.

Preparation of trainers

Prior to the study, the training manual of SR constructed by the Myers Research Institute was reviewed. Two research assistants, who were in doctoral and master programs and had been conducting research on residents with dementia, completed basic training in SR and Montessori-based activities. To control the quality of SR and Montessori-based activity performance, the following controls were implemented to ensure internal validity.

  • (1)The training manual for SR was constructed by the research team who then trained the research assistants. The content of the training manual consisted of the categories of memory, errorless learning, definition of SR, implementation of SR, a videotape, and demonstration and return demonstration.
  • (2)Demonstration and return demonstration of SR and Montessori-based activities
  • (3)After finishing the two-day trainer course, the two research assistants were requested to lead SR or Montessori-based activities at an institution for 8 days. The principal investigator evaluated the two research assistants' skills in SR and Montessori- based activity during this period.


After the protocol for intervention was initially constructed, it was sent to five multidisciplinary experts including a member of a gerontological nursing faculty, a psychogeriatrician, neurologist, occupational therapist, and speech pathologist, whose additional suggestions were used to determine the final protocol of intervention. Their input resulted in some changes, such as using real fruits instead of pictures of fruits during intervention sessions to enhance motivation, and buying desserts, which the residents had no opportunity to eat after admission to the facility, as a reward for participating at each intervention session.

The SR group received 8-weeks of training. The training consisted of two dimensions: eating procedure and eating behavior. SR used immediate, 1, 2, 4, 8, 16, and 32 min time interval trials to train subjects. The Montessori-based activity program for persons with dementia was developed by Camp (2001), but only hand-eye coordination, scooping, pouring, and squeezing were employed. This study added the matching and differentiating of edible and not-edible items to the program. The control group received and participated in the daily routine normally followed by their institution's schedule.


The Chinese versions of the MMSE to assess cognitive status and the Barthel index to assess activities of daily living (ADL) were used. Both of these have been used previously by many institutions as screening tools to determine if residents meet set standards of admission into long-term care facilities.

Edinburgh feeding evaluation in dementia (EdFED)

The EdFED was developed to assess feeding difficulty in older people with dementia, and consists of items that measure passivity and obstinacy, feeding difficulty, and nursing intervention (Watson, 1994). Lin and Chang (2003) translated the EdFED into Chinese. In the Chinese version, items were rated from 0–2, which were recoded from the 1–3 (never to occurs often) that occurred in the original English version. A higher score represents more eating difficulty. A modest equivalence, internal consistency and construct validity were found (Lin and Chang, 2003; Lin et al., 2008).

Mini-nutritional assessment (MNA)

The MNA was used to evaluate the risk of malnutrition to facilitate nutritional intervention (Vellas et al., 1994). The internal consistency of the MNA ranged from 0.74–0.83 (Bleda et al., 2002). In terms of validity, the sensitivity of the MNA to detect malnutrition was 96%, and its specificity was 26–98% (Vellas et al.,1999; Christensson et al., 2002). The Chinese version of MNA has an established usage in clinical practice. Five multidisciplinary experts including a gerontological physician, a faculty of gerontological nursing, and three nutritionists were invited to assess whether the items on the MNA could appropriately assess nutritional status of Taiwanese older people. A 4-point Likert scale was used to score items' appropriateness. The content validity of the 5 experts yielded 3.20–4.00. The intra-class coefficient for inter-rater reliability of the Chinese version of the MNA was 0.88 in this study. In terms of construct validity, the correlation coefficients of the MNA with albumin and BMI were 0.79 and 0.63, respectively (Chan et al., 2002).

Weigh and height

Body weights and heights of subjects were taken. From these values, participants' BMI was calculated.

Stopwatch and amount consumed

A stopwatch was used to measure the meal duration for residents. In terms of the amount of each meal that was consumed, the proportion of each meal residents had for each of the three daily meals was measured. These three values were then averaged, and this average value was used as the amount of each meal consumed.

Data analysis

SPSS 15.0 for window was used to analyze obtained data. Descriptive analysis was used for describing demographic data. χ2 and ANOVA were used to test if the three groups were homogeneous. A linear mixed model was used to test the differences among the three groups for eating ability and nutritional status.


The mean age for the subjects was 81.18 ± 6.37 years with ages ranging from 66–96 years. As can be seen in Table 1, except for ADL, no significant differences in demographic characteristics were found among the three groups.

Table 1. Comparison of demographic data among the SR, Montessori and control groups
VariablesControlSRMontessoriF valuep value
(n = 24)(n = 32)(n = 29)
M ± SDM ± SDM ± SD
Age81.08 ± 6.9479.69 ± 6.1082.90 ± 5.961.977.145
Length of being diagnosed with dementia37.00 ± 32.1532.88 ± 32.0225.31 ± 19.301.179.313
Length of institutionalization28.88 ± 29.4721.97 ± 21.8225.52 ± 19.340.599.552
MMSE10.46 ± 8.0213.56 ± 5.0510.83 ± 4.862.375.099
ADL55.83 ± 25.3569.53 ± 26.4748.28 ± 31.094.577.013
VariablesControl (n, %)SR (n, %)Montessori (n, %)X2p value
 Female15 (62.5%)18 (56.3%)12 (41.4%)2.589.274
 Male9 (37.5%)14 (43.8%)17 (58.6%)  
Marital status
 Married8 (33.3%)15 (46.9%)17 (58.6%)3.416.185
 Unmarried or widowed16 (66.7%)17 (53.1%)12 (41.4%)  

To determine the intervention effects, a repeated-measures analysis of eating ability and nutritional status among groups were compared. As can be seen in Table 2, the EdFED scores and assisting feeding scores for the SR and Montessori-based activity groups after intervention were significantly lower than that of the control group, which indicates that SR and Montessori-based activity may decrease the eating difficulty of residents with dementia and decrease need for feeding assistance by caregivers. However, the frequencies of physical assistance and verbal assistance for the Montessori-based activity group after intervention were significantly higher than that of the control group, which suggests that residents who received Montessori-based activity need more physical and verbal assistance during mealtimes.

Table 2. Generalized linear mixed model of eating ability and nutritional status in the three groups [Correction made to Table 2 after initial online publication]
Dependent variablesPre-testPost-testβ0β1β2β3β4β5
M ± SDM ± SD SR vs. CGMon vs. CGTimeTime*SRTime*Mon
  • Note: Regression coefficient β0 is the baseline means for the control group (or intercept); regression coefficients β1 and β2 are the mean differences in baseline scores between the spaced retrieval (SR) group and the control group (CG) (β1) and the Montessori-based activities (Mon) group and the control group (β2); regression coefficient β3 indicates the time effect observed for interventions; regression coefficient β4 indicates the mean reduction in eating difficulty between the SR group and the control group between β0 and the value at 8 weeks; β5 indicates the mean reduction in eating difficulty between the Montessori-based activities group and the control group between β0 and the value at 8 weeks.

  • *


  • **


  • ***


 Spaced retrieval (SR)4.31 ± 2.043.37 ± 1.88      
 Montessori methods (Mon)5.03 ± 3.283.54 ± 1.77      
 Control (CG)5.13 ± 3.155.04 ± 0.19      
Fed by caregivers0.96−0.61*−0.97**
 Spaced retrieval (SR)0.53 ± 1.140.37 ± 0.85      
 Montessori-based (Mon)0.41 ± 4.060.00 ± 0.00      
 Control (CG)0.96 ± 1.270.96 ± 1.33      
Physical assistance0.58−0.190.55*−0.420.770.00
 Spaced retrieval (SR)0.75 ± 1.080.40 ± 0.97      
 Montessori-based (Mon)0.72 ± 0.961.14 ± 0.93      
 Control (CG)0.17 ± 0.380.58 ± 1.14      
Verbal assistance0.29−0.140.96**−0.210.62*0.03
 Spaced retrieval (SR)0.56 ± 0.840.17 ± 0.46      
 Montessori-based (Mon)1.07 ± 1.161.25 ± 1.14      
 Control (CG)0.08 ± 0.280.29 ± 0.86      
Self eating time8.809.58***13.04***2.44−1.79−6.21*
 Spaced retrieval (SR)19.03 ± 11.8718.12 ± 7.99      
 Montessori-based (Mon)18.08 ± 8.9321.76 ± 8.92      
 Control (CG)11.25 ± 9.598.80 ± 8.73      
 Spaced retrieval (SR)21.67 ± 2.9323.97 ± 2.14      
 Montessori-based (Mon)18.31 ± 3.8817.98 ± 3.67      
 Control (CG)20.27 ± 2.1820.29 ± 2.90      
 Spaced retrieval (SR)24.70 ± 4.3124.80 ± 4.37      
 Montessori-based (Mon)21.19 ± 3.3921.13 ± 3.24      
 Control (CG)23.10 ± 2.6723.07 ± 2.48      
 Spaced retrieval (SR)58.31 ± 9.4258.24 ± 9.46      
 Montessori-based (Mon)51.11 ± 7.8250.96 ± 7.76      
 Control (CG)54.98 ± 8.5854.89 ± 8.28      
Eating amount88.062.29−10.08*−9.10**4.025.79
 Spaced retrieval (SR)85.26 ± 11.4390.73 ± 8.78      
 Montessori-based (Mon)74.67 ± 22.8578.37 ± 10.10      
 Control (CG)78.96 ± 19.2488.06 ± 17.76      

In terms of the effects of nutritional status after intervention, MNA in the SR group was significantly higher than that of the control group, while MNA in the Montessori-based activity group was significantly lower than that of the control group.


After intervention, the SR group and Montessori group revealed the effects of decreasing eating difficulty and eating dependency. That the SR group had a significant improvement in eating ability might result from the targeted step-by-step eating memory training that the protocol of SR consists of. Each session only trains one eating motion. During our training, the first learning content, ‘able to realize meal time’, took a relatively greater deal of time to establish, because we used Mozart music to ‘announce’ it was mealtime for residents. This type of mental ‘trigger’ was not in their past experience and was an unfamiliar approach, and as such, it took three or four training sessions before residents adequately made the connection of the music to mealtime. In contrast, it was easy to complete ‘able to masticate’, because when food is put into the mouth and then swallowed, it is largely an instinctual process.

In terms of Montessori activity, when the Montessori group leader conducted activities for hand-eye coordination, scooping, pouring, squeezing, matching and differentiating food, etc., she used food familiar to the subjects as the training material, but which was rarely served by the institution. Sight, smell, touch, sound and taste all influence food choice and the memory of pleasant experiences, which could facilitate learning (Mahendra, 2001; Eberhardie, 2005), as well as easily initiate the subjects' motivation for taking the food to eat. During the study, several subjects would ask the trainer ‘what do we eat today?’ when they saw the group leader. A positive attitude toward eating might affect eating behavior during mealtime and potentially decrease eating dependency. However, the Montessori group's physical assistance, and verbal assistance were significantly higher than that of the control group. This might be explained in that eating ability decreases step-by-step from a need for verbal assistance, then a need for physical assistance, and finally the need to be fed by a caregiver. Several subjects were being fed by caregivers before intervention. After intervention, they seemingly did not need to be directly fed, but they had not improved enough to completely self-feed yet and still needed physical assistance and verbal assistance.

We did not expect the result that the self-eating time of the SR group and the Montessori group significantly increased compared with the control group. A long eating time is commonly a factor related to the perception by caregivers to assist in the feeding of residents with dementia to reduce mealtime lengths and thus save on the caregivers' workload. The reason why self-eating time in the SR group and the Montessori group is significantly longer than in the control group may be because subjects in the SR group and Montessori group may have acquired a lasting positive affect toward eating and the mealtime environment, becoming more sociable or relaxed during mealtime. The mean self-eating time at post-test in the SR group and the Montessori group were 18 min and 22 min, respectively, and the literature reveals that older people tend to eat slower than young people (Hughes and Wiles, 1996). Also, 18–22 min for having a meal in an institutional setting is certainly a socially acceptable time for the activity. These factors could suggest that SR and Montessori-based activities can be used to maintain the self-care ability of residents with dementia. In contrast, the mean self-eating time in the control group at post-test was about 9 min. Nine minutes for eating one meal would commonly be considered fast for older people, particularly for those with dementia. Eating time is related to food preference and meal amount. Moreover, the control group also gets more feeding assistance to speed them up. The above factors might have influenced the faster eating time of the control group. It needs pointing out that eating too fast might cause sequential problems, such as choking, which suggests that, for future study, the consequences of eating time consumption under 10 min among residents with dementia should be investigated.

We did not have an expectation that the MNA and eating amount in the Montessori group would be significantly lower than the control group. The reason for no significant differences on the eating amount is that we used an average consumption of three meals, and the meal trays in institutions are not standardized. The caregivers in the control group would serve a lesser amount of food to residents who tended to leave food uneaten on their tray, giving an appearance of an increasing meal consumption level by the time of the post-test observation period. The use of proportionate values rather than actual quantitative food weights could actually reflect the existence of either consuming more or consuming less relatively. As to body weight, two subjects in Montessori group lost weight because of hospitalization for increased agitation intensity or frequency of dementia, and a cholecystectomy. Unfortunately, disease, or anorexia and mobility accompanied with diseases, were items on the MNA, which might also have affected the scores in the Montessori group as several subjects had diseases present in the Montessori group during study period. For any future study, it is recommended to weigh the meals to determine exact meal consumption. In addition, a longitudinal study is recommended to follow weight change after intervention.


Eating is the last daily activity that older people will lose their ability to perform, and the first daily activity for which they can regain their capability for after training. Nurses may be the first health care professionals to realize that a patient is having difficulties eating. Ayalan et al. (2006) suggest that studies should investigate the positive and negative effects of nonpharmacological interventions. This study confirms the efficacy of SR and Montessori-based activities on eating difficulty (decreasing negative effects) and eating ability (increasing positive effects). Early intervention might give the patient with dementia a better chance of maintaining independence.

Limitations in this study are acknowledged. First, single blinding was used. Although subjects had dementia and were forgetful, subjects with mild dementia might realize what type of intervention they received. It is recommended to use a cross-over design to control this confounding factor for future studies. Secondly, despite the significant differences for eating difficulty and feeding ability, insignificance was found for body weight and BMI. This may result from the 2 months intervention period not being long enough to detect changes or relationships in these variables. It is recommended that a longitudinal study be conducted to determine the long-term effects of SR and Montessori-based activities.

Key points

  • Spaced retrieval and Montessori-based activity groups could decrease eating difficulty of residents with dementia.

  • The spaced retrieval group had a significant improvement in eating ability by singing the step-by-step procedural eating memory training. [Correction made here after initial online publication.]

  • The Montessori-based activity group significantly lowered feeding by caregivers by using preferred food, smell and taste stimulation of food and task of eating linked to one another. [Correction made here after initial online publication.]

  • The self-eating time of the Montessori group significantly increased.

Conflict of interest

Beside financial support, the sponsor National Health Research Institute (NHRI) was not involved in the research process.


This study was supported by a grant from the National Health Research Institute (NHRI-EX96-9621PI). The authors thank Po-Yen Chen, Han-Yuan Chen, Ming-Ling Wu, Lin-Hsuan Weng, I-Lan Su, and Shu-Ru Jhuang for assistance with data collection. The authors also gratefully acknowledge Mr Mike Portwood for his assistance in revising the manuscript.