As a portion of the overall research work for NCHRP Project 1–44, an examination of temperature effects on the onboard sound intensity method of measuring tire–pavement noise at the source was conducted. The study focused on whether a single, average normalization would be beneficial when applied to a full range of pavement types. Testing was conducted over 10 months, over a temperature range of more than 60°F (16°C), and on 10 pavements representing a variety of both asphalt and portland cement concrete design categories and covering a range in noise level of about 10 dB. On the basis of a statistical approach, temperature normalization was found to reduce uncertainty of the data and is recommended to be included in the method of test. Linear air temperature normalization is recommended, since a logarithmic regression did not further improve the fit of the data. Less scatter and improvements in the data were seen with the correlation to air temperature as opposed to pavement temperature. Theoretical and empirical assessments of the air density correction found the air density correction did not reduce uncertainty in the data and was not needed.