Psychedelics (5-HT2AR agonists) show great promise in treating mental disorders but their effects on neuronal population activity are not clear. Latent dimensions identified by Principal Component Analysis (PCA) capture the dynamics of this interrelated neuronal population activity. We asked whether latent dimensions emerge or disappear as a result of psychedelic administration (acute), and across other naturally occurring brain state transitions (sleep, arousal). Contrastive PCA (cPCA) [Abid et al. (2018), Nature Comm.] was applied to spontaneous brain activity recordings in rodents before vs. after drinking a psychedelic (Neuropixels, mPFC), wakefulness vs. non-REM sleep (Ephys silicon probe, V1), and low vs. high arousal (2-photon calcium imaging, V1). We identified directions that explain considerably more variance in one brain state (target) than in the other state (background), i.e. dimensions present after drug administration but not before (emerging dimensions), or vice versa (disappearing dimensions). Contrastive directions were computed by conducting the eigenvalue decomposition on the difference between background and target covariance matrices with contrastive parameter $\alpha (\Sigma_{diff} = \Sigma_t - \alpha \times \Sigma_b)$. We analysed the variance explained by the first contrastive components in target and background data, for the entire spectrum of α. We found robust contrastive dimensions of neuronal population activity across different brain state transitions, recording methodologies, and brain areas. However, contrastive dimensions did not explain a large amount of variance in one brain state (e.g. on par with principal component 1) and little variance in the other state (e.g. on par with a random direction) and were merely up/downregulated. These shared contrastive dimensions could emerge or disappear. After psychedelic administration, more shared dimensions disappeared, which might indicate less connectedness or variability of the acute psychedelic state.