The actual electronic response of the electronics at the detector to any given concentration of ethyl alcohol in the sample chamber does not result in a direct linear relationship between alcohol concentration in the sample chamber and the voltages produced at the detector. The relationship is actually inverse and logarithmic - see Lambert-Beer law. However done in the software, the instrument needs to read those voltage fluctuations in the detector and convert them into an INDICATION (to use the metrological term)(digital display or printout) that is reasonably reliable across the MEASURING INTERVAL (metrological term). When that relationship is set up (the CALIBRATION CURVE), it is hoped that the instrument will be reliable at 10, 20, 50, 80, 100, 120, 160 mg% and so on. There may be a problem, however, on initial calibration or OVER TIME, with the values being reliable at some points in the measuring interval. We would then say that the linearizer circuitry/software is not performing its function - across the measuring interval. There is a problem with linearity between what goes in (the actual ethyl alcohol breath) and what comes out (the indication).