Particulate matter emissions from internal combustion engines have become an increasingly important area of focus for development teams in recent years. This is due to greater regulatory scrutiny on vehicles globally, and especially on particulate emissions. The chemical composition and bulk physical properties of the fuel have been shown to influence the particulate number emissions characteristics. Although some predictive models have been proposed, the causality of specific properties or constituents has not been demonstrated due to the co-linearity of the variables considered in previous studies.In this work, fuels were formulated to capture the expected variation in three key properties of United States (US) market gasoline fuels. Specifically, total aromatics, volatility, and particulate matter index (PMI) were varied across market extremes within regulatory limits--while holding other properties constant. Engine dyno tests were performed with the formulated fuels on a 4-cylinder 2.3 L turbocharged direct injection spark-ignited engine, operated between 2 and 16 bar BMEP.ANOVA analysis was used to demonstrate that both low volatility and high PMI independently increased particle number (PN) emissions at high load at a statistically significant level, whereas total aromatics was only significant at low-load. However, neither low-volatility nor high PMI was universally bad for PN emissions, and other properties also played an important role. For example, at <12 bar BMEP the lowest PN formulated fuel was a low-volatility, high PMI, low aromatics fuel.The nine formulated fuels from this work along with ten fuels tested in Part 1 of this work were used to correlate PN and THC emissions with various fuel properties, within the repeatability of the test. For PN emissions, most of the fuel-to-fuel differences could be explained across the entire load range using a combination of 5 parameters: FBP, T90, A9+, IP9+, and C11. A9+ and FBP were always positively correlated with emissions, whereas IP9+ was negatively correlated, and T90 and C11 switched from negative to positive depending on load. For THC, only three parameters were required, and T70 was by far the best predictor of emissions—accounting for 50-70% of the difference between fuels.The work highlights the critical properties that determine engine out emissions characteristics of different fuels, and shows a path to formulating fuels that can simultaneously reduce PN and THC from vehicles.