There is growing evidence that the amygdala serves as the base for dealing with complex human social communication and emotion. Although amygdalar networks plays a central role in these functions, causality connectivity during the human lifespan between amygdalar subregions and their corresponding perception network (PerN), affiliation network (AffN) and aversion network (AveN) remain largely unclear. Granger causal analysis (GCA), an approach to assess directed functional interactions from time series data, was utilized to investigated effective connectivity between amygdalar subregions and their related networks as a function of age to reveal the maturation and degradation of neural circuits during development and ageing in the present study. For each human resting functional magnetic resonance imaging (fMRI) dataset, the amygdala was divided into three subareas, namely ventrolateral amygdala (VLA), medial amygdala (MedA) and dorsal amygdala (DorA), by using resting-state functional connectivity, from which the corresponding networks (PerN, AffN and AveN) were extracted. Subsequently, the GC interaction of the three amygdalar subregions and their associated networks during life were explored with a generalised linear model (GLM). We found that three causality flows significantly varied with age: the GC of VLA → PerN showed an inverted U-shaped trend with ageing; the GC of MedA→ AffN had a U-shaped trend with ageing; and the GC of DorA→ AveN decreased with ageing. Moreover, during ageing, the above GCs were significantly correlated with Social Responsiveness Scale (SRS) and State-Trait Anxiety Inventory (STAI) scores. In short, PerN, AffN and AveN associated with the amygdalar subregions separately presented different causality connectivity changes with ageing. These findings provide a strong constituent framework for normal and neurological diseases associated with social disorders to analyse the neural basis of social behaviour during life.