Individual neurons exhibit a rich repertoire of complex dynamical patterns which vary at fast timescales [1]. Despite these fast temporal dynamics, neural circuits are able to maintain stimulus representations over the course of behaviorally relevant timescales that can subsequently be decoded by time-invariant downstream circuits to perform motor actions. Previous studies have proposed that these observations can co-exist due to neural redundancy. Specifically, these frameworks suggest that task relevant information is stably stored in a low dimensional subspace [2][3]. Moreover, time invariant decoders can decode stimulus identity even though neurons exhibit dynamic activity as previously shown in the population of prefrontal cortex neurons while animals perform working memory tasks [4]. We investigated the conditions where stimuli can be decoded by a fixed decoder during a trial when the underlying activity is varying by simulating various neural dynamics. We further probed the existence of an invariant stimulus representation in calcium imaging data collected from the gustatory cortex (GC) as mice performed a cued taste paradigm. GC has been extensively studied for its role in taste representation [5]. Due to the heterogeneous responses observed in the taste evoked mean activity of neurons [6], there has been an implicit assumption of time varying coding in GC that requires downstream networks to have access to timing information during each trial [7][8][9]. Here, we fit a time varying as well as a fixed linear decoder after reducing the dimensionality of data using PCA or SemiNMF to decode taste identity. We find that we are able to stably decode the taste identity from the neural activity with a fixed decoder over the course of several seconds and the decoding performance is comparable to that of a time varying decoder. This suggests that there exists an invariant linear representation of taste in GC.