In neuroscience, time-frequency analysis is widely used to investigate brain rhythms in brain recordings. In event-related protocols, it is applied to quantify how the brain responds to a stimulation repeated over many trials. We here focused on two common measures: the power of the transform for each single trial averaged across trials, avgPOW; and the power of the transform of the average evoked potential, POWavg. We investigated the influence of additive noise on these two measures. We quantified the expected effect using theoretical calculations, simulated data and experimental brain recordings. We also considered the case of color noise. We extracted the main factors influencing the effect of noise on POWavg and avgPOW, such as the noise variance, the number of trials, the sampling rate, the type of noise, the type of time-frequency transform and the frequency of interest. When dealing with time-frequency analysis, the impact of noise on the neuroscientist's work can drastically vary depending on these factors. The present results should help researchers improve their understanding and interpretation of time-frequency diagrams, as well as optimize their experimental designs and analyses based on their neuroscientific question.