Despite its success as a clinical monitoring tool, pulse oximetry may be improved with respect to the need for empirical calibration and the reports of biases in readings associated with peripheral vasoconstriction and haemoglobin concentration. To effect this improvement, this work aims to improve the understanding of the photoplethysmography signal—as used by pulse oximeters—and investigates the effect of vessel calibre and haemoglobin concentration on pulse oximetry. The digital temperature and the transmission of a wide spectrum of light through the fingers of 57 people with known haemoglobin concentrations were measured and simulations of the transmission of that spectrum of light through finger models were performed. Ratios of pulsatile attenuations of light as used in pulse oximetry were dependent upon peripheral temperature and on blood haemoglobin concentration. In addition, both the simulation and in vivo results showed that the pulsatile attenuation of light through fingers was approximately proportional to the absorption coefficients of blood, only when the absorption coefficients were small. These findings were explained in terms of discrete blood vessels acting as barriers to light transmission through tissue. Due to the influence of discrete blood vessels on light transmission, pulse oximeter outputs tend to be dependent upon haemoglobin concentration and on the calibre of pulsing blood vessels—which are affected by vasoconstriction/vasodilation. The effects of discrete blood vessels may account for part of the difference between the Beer–Lambert pulse oximetry model and empirical calibration.