Temporal dynamics of grief reactions in a Canadian bereaved sample

Background: Grief is theorised to be characterised by wave-like reactions, yet little is known about how these reactions unfold between moments. Understanding the short-term temporal dynamics among daily grief reactions could provide insight into how post-loss distress persists over time.Objective: This study examined within-person temporal and contemporaneous grief reactions networks using intensive longitudinal data.Method: Thirty adults from a multi-ethnic sample who lost a loved one on average 19.5 months ago completed a daily sampling protocol for 2 weeks (5 prompts/day; 3-hour intervals). Multilevel vector autoregressive models estimated a lagged temporal network and a contemporaneous network (within-time partial correlations) of 12 nodes from the International Prolonged Grief Disorder Scale (IPGDS).Results: In the temporal network, difficulties accepting the loss showed the highest out-strength and predicted subsequent avoidance, difficulties engaging in social activities, and guilt. Preoccupation predicted later yearning. Emotional numbness and difficulties feeling joy predicted each other at subsequent time points, suggesting a short-term feedback process. Highest in-strength was observed for difficulties engaging in social activities, avoidance, and yearning. In the contemporaneous network, grief reactions were densely interconnected; identity disruption, yearning, and sadness were most central. The strongest within-time association linked emotional numbness with difficulties feeling joy, and social-activity difficulties co-occurred strongly with guilt and blame. Difficulty accepting emerged as a key short-term predictor of downstream avoidance and withdrawal, while a numbing–anhedonia loop may sustain affective constriction throughout the day.Conclusion: This study contributes to the understanding of the daily lives of bereaved individuals, highlighting grief as an interactive, mutually reinforcing system over time. The findings may suggest candidate mechanisms for interventions. However, they should primarily be considered hypothesis-generating since temporal network estimates can be prone to low power and instability under the present small-sample condition.