This paper investigates the subjects' ability to perceive trumpet sound changes, the sounds being either simulated by physical modeling or played by a real musician. The objective of the paper is to study to what extent simulations are able to create perceptually distinct sounds and to compare the discrimination on simulated sounds to the discrimination obtained when the instrument is played. Two experimental factors are considered to study perceptual discriminations by a panel of subjects: changes in the resonator geometry (instrument discrimination) and changes in the sound dynamics of the same instrument (dynamics discrimination). The changes in the resonator geometry are made with tiny geometrical variations of a parameterized leadpipe, whereas changes in dynamics are generated modifying control parameters for simulated sounds and playing dynamic instructions for real musicians. The same note (Bb4) is simulated using the harmonic balance technique and played by a real musician. To assess the discrimination, a "same-different" test has been carried out involving a panel of 24 listeners and pairs of sounds coming from the simulations or the recordings (for the sake of the experiment, the recorded sounds are resynthesized). For the different experimental conditions (differences between trumpets or between dynamics), the abilities of the listener to discriminate different sounds and to identify identical sounds are analyzed. Using the Receiver Operating Characteristics plots, the results show that even if the differences are more noticeable with sounds played by the musician, simulations are able to produce noticeable differences between the instruments, mainly due to playing frequency differences. They confirm the capacity of the physical model to represent realistic playing conditions for the trumpet.