Hourly infiltration rates in three adjacent unoccupied test houses have been correlated to determine the relative contribution of wind and stack effects. The data base, obtained using tracer gas techniques over 2,000 hourly readings in each house was sorted to a single 45° wind sector. Different weather sealing methods gave mean infiltration .rates of 0.19, 0.45 and 0.59 ACH (air changes per hour) for the three houses . Addition of independent wind and temperature induced pressures correlated only part of the weather induced data variability, leaving a residual RMS scatter of about 0.04 ACH.