One of the common ways of conveying expression in music is via "expressive timing" -- subtle changes in tone durations not called for in the score. This paper will present a comparative analysis of four performances by well-known pianists (Gould, Rübsam, Bacchetti, Blatter) of Bach's Invention No. 9 in f minor. The novel analytical (computer-assisted) approach resulted in an accuracy of +/- 0.01 seconds in the detection of note onsets. Goals of the analysis were to answer the following questions: How do musicians perceive the rubato in comparison to the actual rubato (as standard deviation of performed tone durations from their average tone duration)? How does the tempo correlate with the perceived rubato, when the calculated rubato (as standard deviation) is similar? Are the 16th notes on downbeats, on average, performed longer (since the main motive starts on a downbeat)? How are other parts of the measure, on average, performed?