In order to determine how bacteria pass through the teat canal into the teat sinus, milk samples were taken by syringe, through the teat wall, from 12 quarters of 7 cows, before and after each milking for up to 16 days. Three of the teat canals were naturally infected before the start of the experiment and the remainder were artificially infected with Staphylococcus aureus by the Hadley–Wisconsin swab technique. In 3 of the quarters the inoculum was introduced into the teat sinus through the teat canal by the inoculation technique although the swabs were inserted only 3–5 mm into the canal. The other 9 of the 12 quarters were milked a total of 149 times during the experiment but only once was a colony-forming unit of the bacterium present in a teat canal infection isolated from the milk taken from the teat sinus after milking. One quarter developed an intramammary infection, the bacterium first being detected in the teat sinus prior to the third milking after inoculation of the teat canal. The invasion of the bacteria into the teat sinus therefore occurred between milkings. Staph. aureus persisted in teat canals between 5 and 32 or more milkings after inoculation. The experiment showed that bacteria in the teat canal are seldom refluxed into the residual milk in the teat sinus during correctly conducted milking with an efficient machine.