In France, the issue of providing adequate care for the elderly is currently in the spotlight. There has been a concerted effort to humanize institutional care by increasing the number of facilities and staff (Bonnet, 2008). However, the training of professionals working with elderly often focuses primarily on practical tasks. Caregivers, as well as medical-psychological aides, typically find themselves tasked with documenting their performed tasks, often neglecting the examination of whether their communication influences the outcome of these tasks, such as meal assistance. Yet, especially in specialized care facilities for the elderly, caregivers are required to communicate during their care routines, and often, the effectiveness of their communication determines the success of their tasks. This aspect of caregiving is frequently more complex than it may appear, particularly in settings where neurodegenerative disorders, such as Alzheimer’s disease, are prevalent, making communication with residents challenging. Caregivers are often left to adapt to these unique circumstances with minimal training (Rousseau, 2009). Using a corpus of authentic audio and video recordings collected during mealtime interactions at a specialized care facility for dependent elderly individuals, this study conducts a qualitative analysis of caregiver discourse in response to residents who sometimes encounter difficulties in their interactions. Our analysis encompasses two aspects. Firstly, we will observe verbal exchanges within the framework of the two discourse categories delineated by Fukaya et al. (2009), namely, discourse aimed at eliciting the activities of daily living (ADL)-related behavior or physical functioning (Type I talk) and discourse aimed at facilitating psychosocial life activities (Type II talk). We explore the distribution of these discourse types and residents’ reactions to each category. Secondly, we delve into the contentious phenomenon of elderspeak, as identified by Williams (2013). Elderspeak encompasses various linguistic features, including a slower speaking rate, an exaggerated intonation, an elevated pitch and volume, repetitions, employment of a simpler vocabulary, and a reduction in grammatical complexity. Additionally, this study places a special emphasis on humor, a topic rarely explored in communication studies with the elderly population. We investigate its linguistic mechanisms, the contexts in it is employed, its effectiveness, and other related aspects. (Heinemann, 2009; Joris, 2010). We demonstrate that caregiver discourse is closely correlated with the residents, particularly their ability to engage in the interaction. In addition to challenges related to purely human factors such as refusal and aggression, there are difficulties stemming from the pathologies affecting the elderly individuals. This study aligns with the “Interactional Studies of Language and Aging” theme of the conference and could contribute to the development of caregiver training programs better suited to the realities of care provision.