The solar reflectance bands (SRB) of the Advanced Very High Resolution Radiometers (AVHRR) flown onboard NOAA satellites are often referred to as non-calibrated in-flight. In contrast, the Earth emission bands (EEB) are calibrated using two reference points, deep space and the internal calibration target. In the SRBs, measurements of space count (SC) are also available, however, historically they are not used to specify the calibration offset ("zero count", ZC), which does not even appear in the calibration equation. A regression calibration formulation is used instead, equivalent to setting the ZC to a constant, whose value is specified from pre-launch measurements. Our analyses supported by a review of the instrument design and a wealth of historical SC information show that the SC varies in-flight and it differs from its pre-launch value. We therefore suggest that (1) the AVHRR calibration equation in the SRBs be re-formulated to explicitly use the ZC, consistently with the EEBs, and (2) the value of ZC be specified from the onboard measurements of SC. This study emphasizes the importance of clear discrimination between the SC (which is a measured quantity and therefore takes on a range of values, characterized by the empirical probability density function, PDF), from the ZC (which is a parameter in the calibration equation, i.e. a number whose value needs to be estimated from the measured SC as a mean, median or other statistic of the measured PDF). The ZC-formulation of the calibration equation is physically solid, and it minimizes human-induced calibration errors resulting from the use of a regression formulation with an un-constrained intercept. Specifying the calibration offset improves radiances, most notably at the low end of radiometric scale, and subsequently provides for more accurate vicarious determinations of the calibration slope (inverse gain). These calibration improvements are important for the products derived from the AVHRR low-radiances, such as aerosol over ocean, and particularly critical when generating their long-term climate data records.