The Effect of Voice Output on AAC-Supported Conversations of Persons with Alzheimer's Disease
Melanie Fried-Oken, Charity Rowland, Glory Baker, Mayling Dixon, Carolyn Mills, Darlene Schultz, Barry Oken · 2009 · ACM Transactions on Accessible Computing · doi:10.1145/1497302.1497305
Summary
This study investigated whether digitized voice output on AAC (Augmentative and Alternative Communication) devices would improve conversations for people with moderate Alzheimer's disease. The researchers hypothesized that voice output might function like partner-assisted word cueing—a therapeutic technique where conversation partners supply words to help people with dementia overcome word-finding difficulties—and thereby enhance language use. Thirty adults with moderate Alzheimer's disease (mean age 74, MMSE scores 5-18) participated in two videotaped conversations using a Flexiboard AAC device. The device displayed 16 personalized messages (1-2 word phrases) related to autobiographical topics chosen by each participant and their caregivers, such as grandchildren, travel, or favorite activities. Twelve participants were randomly assigned devices with digitized voice output that spoke the phrase when pressed; eighteen used identical devices without voice output. Trained research assistants conducted 10-minute conversations in participants' homes, with the final 5 minutes coded for four measures: total utterances, percentage of 1-word utterances, percentage of topic elaborations/initiations, and references to the AAC device.
Key findings
Contrary to the hypothesis, voice output significantly depressed conversational performance. Participants using devices with voice output produced fewer total utterances (mean 46 vs. 54), more 1-word utterances (35% vs. 30%), and a lower rate of topic elaborations and initiations (22% vs. 29%). All differences were statistically significant (p < .01). The effect on AAC device references was not significant, though participants with voice output referenced the device less frequently (mean 1 vs. 3). The researchers offer several explanations for these unexpected findings. Voice output may have created perceptual and attentional interference—the spoken cue interrupted participants' train of thought, causing them to forget what they intended to say. For some participants, the novelty of voice output triggered perseverative behavior, pressing symbols repeatedly to hear the sound rather than using them communicatively. Additionally, this cohort of older adults (born in the 1930s) had little experience with "talking technology" and may have found machine-generated speech foreign and distracting rather than helpful. The study suggests that human word cueing during natural conversation creates a familiar, supportive environment that machine-generated cues cannot replicate.
Relevance
This research provides an important cautionary finding for assistive technology design: features that work well for one population may not transfer to another. Voice output, which benefits many AAC users with developmental disabilities or motor speech disorders, actively hindered communication for older adults with dementia. The finding challenges assumptions about universal benefit of speech output and highlights the need for population-specific AAC research. For practitioners working with people with Alzheimer's disease, the study suggests that simpler AAC approaches—visual memory aids without voice output—may better support conversation. The authors note that training protocols and earlier intervention (before dementia progresses) might yield different results, and that devices with integrated speakers rather than external amplifiers might be less confusing. More broadly, the research underscores that cognitive accessibility requires understanding the specific cognitive profile of users, not just adding assistive features that work for other disability populations.
Tags: AAC · Alzheimer's disease · dementia · voice output · cognitive accessibility · aging · speech synthesis