The posterior insular cortex (pInsCx) is a multimodal hub processing sensory information and aversive states. However, little is known about how specific inhibitory interneurons in the pInsCx code for fear-learning-related stimuli. Here, we performed longitudinal in-vivo Ca2+ imaging in the pInsCx during a three-day fear learning protocol with a miniature microscope; GCaMP6m was expressed in either PV+, or SOM+, or VIP+ interneurons, or in CaMKII+ (principal) neurons. We observed that positive US responses were more common in interneurons (~ 50%) than in principal neurons (~ 13%). There were various types of plastic CS coding, with positive- and negative CS responses developing- and dropping out over the three-day behavioral protocol. Concentrating on positive CS responses, we grouped the longitudinal response classes into four: Neurons that responded to the CS across the three sessions ("naive responders"), neurons that acquired a response ("CS-learners"), neurons that lost a response ("CS-unlearners"), and non-responders. VIP+ and PV+ neurons had high percentages of naive responders (27% and 12%), whereas this percentage was low in SOM+ and CaMKII+ (3-4 %). Similarly, CS-unlearners were more common in VIP+ and PV+ (24 and 21%) than in SOM+ and CaMKII+ (3 and 9%). Lastly, CS-learners were slightly more common in CaMKII+ (15%) than in the interneurons (8-11%). Our data shows, first, that VIP+ and PV+ interneurons have a large proportion of naive CS-responders and CS-unlearners; this likely causes cortical disinhibition. Second, SOM+ neurons are more similar to excitatory neurons concerning their CS-plasticity.