A Bayesian approach to particle identification is reviewed and its implementation in the ALICE PID analysis framework is presented, including the procedures to treat priors across different ALICE sub-detectors. We evaluate the performance of the Bayesian PID approach, compared with standard cuts operating on the single detector signals, for charged pions, kaons and protons in the central barrel of ALICE using the specific energy loss dE/dx in the TPC gas and information from time-of-flight measurements provided by the TOF detector. PID efficiencies and contaminations are extracted and compared with Monte Carlo simulations using the $V^0$ decay channels $K^{0}_{s})\rightarrow \pi^{+}\pi^{-}, \phi \rightarrow K^{+}K^{-}$, and $\Lamnda \rightarrow p \pi^{-}$. We then compare the results from standard methods (fit to PID signals to extract yields or nσ cuts) with our results using Bayesian PID for single-particle spectra of pions, kaons and protons, and in the analysis of $D^{0}\rightarrow K^{-}\pi^{+}$.. In both the full analyses considered in the paper, no significant deviation is found with respect to results using “standard” PID methods used by ALICE so far. Because of this, we conclude that the Bayesian framework is robust enough to be used for analyses in ALICE. The case of the $\Lambda_{c}^{+}$ baryon is finally presented as an example where the Bayesian approach is able to combine information from the different detectors quickly and effectively, while a simple nσ approach is unable to find a positive signal.