To help understand the fundamental processes involved in direct injection, a research project was conducted using a single-cylinder, two-stroke research engine at a mid-speed, boat load operating condition. A 24 statistical factorial experimental design was applied. Of the factors tested at this operating condition, spray type was the most important factor affecting hydrocarbon emissions, followed by in-cylinder flow-related factors. Injection spray was also most important for nitrogen oxide emissions, carbon monoxide emissions, and efficiency. The dominant mechanism influencing the results was misfire, with other mechanisms present for specific responses.