Continuous measurements of carbon monoxide (CO) and particulate matter (PM10, PM2.5 and PM1) were made inside and outside two naturally ventilated (NV) and two mechanically assisted (MV) spaces overlooking the same six-lane highway in Central London. The indoor/outdoor ((I) over bar/(O) over bar) ratio was calculated for each site for the whole monitoring period in the usual way, and then cumulatively at each hourly time-step from the start of monitoring up to the whole period in order to track the evolution of the I/O ratio with length of monitoring or averaging period. Hourly data were also sorted by wind speed and direction and I/O ratios calculated for each direction/speed set from ensemble means.I/O ratios for the whole period were generally lower in the NV spaces than the MV, however, the indoor-outdoor relationship in the NV spaces was found to vary substantially with wind direction. For a constant above-roof wind speed, I/O ratios of CO varied by 50-60% about the mean value with direction, and by 20-30% for particulate matter. Consequently, I/O ratios for NV spaces depended on the distribution of wind direction within the calculation period and, hence, on the length of monitoring period. The I/O ratio in one NV space changed by a factor of three between the early stages and final stages of monitoring, with a "stable" final value (+/-5%) achieved after 900 h of monitoring. By contrast, in the MV spaces under constant fan speed, constant values for the I/O ratio were achieved for CO within 4 It of the start of monitoring. It is argued here that field assessments of the filtration performance of naturally ventilated spaces need to consider exposure to and distribution of natural ventilation drivers within the monitoring period in order to draw meaningful comparisons for design and exposure assessment purposes. (C) 2003 Elsevier Ltd. All rights reserved.