A Euro 2 SI (Spark Ignition) Mondeo was investigated for a fully warmed-up vehicle on a simple urban driving loop. Emissions were monitored using an on-board Horiba OBS (On-Board emission measurement System) 1300. 10 laps of a 0.6 km loop were driven by each driver and this involved 4 junctions per lap. Statistical analysis of 20 drivers was made over 27 repeat junction events for each driver. The statistical analysis of the data showed that for all drivers the CO₂, speed and throttle position were more typical Gaussian in their distribution. NOx and CO on the other hand were lognormal in their distribution. Acceleration, positive and negative throttle jerks (rate of change of throttle angle) were borderline Gaussian. HC (Hydrocarbon) emissions were not Gaussian and there was some evidence for a gamma distribution and for a lognormal distribution. Comparison of mean HC emissions between the drivers was therefore not reliable.The variation in mean speed between the drivers was small from 24-32 km/hr, but there was a much greater variation in mean acceleration rates from 0.4 - 1.2 m/s2. There was also a wide variation in average throttle position from 2.5 - 10.3% and an even greater variation in throttle positive jerk, from 5-22%/s. These different driving modes for the same road event resulted in major differences in emissions. CO₂ varied from 235 - 339 g/km for the 20 drivers with CO varying from 0.27 to 10.7 g/km, HC from 0.06 to 0.2 g/km and NOx from 0.27 to 0.8 g/km. There were even larger variations between drivers if the worst case emissions were compared for all the drivers. These results show that driver behavior has a major impact on real-world emissions and indicate that the potential of using electronic control throttle (ETC) to control acceleration and deceleration events could have a major impact on reducing real-world emissions.