During the study Relationships of Indoor, Outdoor, and Personal Air (RIOPA*), 48-hour integrated indoor, outdoor, and personal air samples were collected between summer 1999 and spring 2001 in three different areas of the United States: Elizabeth NJ, Houston TX, and Los Angeles County CA. Air samples suitable for analyzing particulate matter 2.5 microm or smaller in aerodynamic diameter (PM2.5) were collected in 219 homes (twice in 169 homes). Indoor and outdoor air samples suitable for gas-phase and particle-phase organic analyses were collected in 152 homes (twice in 132 homes). Samples or subsets of samples were analyzed for PM2.5 mass, organic functional groups, elements, organic carbon (OC), elemental carbon (EC), gas-phase and particle-phase polycyclic aromatic hydrocarbons (PAHs), and chlordanes. Air exchange rate (AER), temperature, and relative humidity were measured for each residence; questionnaire data and time-activity information were collected from the participants. Median indoor, outdoor, and personal PM2.5 mass concentrations were 14.4, 15.5, and 31.4 microg/m3, respectively. Personal PM2.5 concentrations were significantly higher and more variable than indoor and outdoor concentrations. Several approaches were applied to quantify indoor PM2.5 of ambient (outdoor) and nonambient (indoor) origin, some using PM2.5 mass concentrations and others using PM2.5 species concentrations. PM of outdoor origin was estimated in three ways using increasingly accurate assumptions. Comparing estimates from the three approaches enabled us to quantify several types of errors that may be introduced when central-site PM concentrations are used as surrogate estimates for PM exposure. Estimates made using individual measurements produced broader distributions and higher means than those made using a single infiltration factor for all homes and days. The best estimate (produced by the robust regression approach) of the mean contribution of outdoor PM2.5 to the indoor mass concentration was 73% and to personal exposure was 26%. Possible implications of exposure error for epidemiologic assessments of PM are discussed below. Organic particulate matter was the major constituent of PM2.5 generated indoors. After correcting for artifacts, it constituted 48%, 55%, and 61% of PM2.5 mass inside study homes in Los Angeles, Elizabeth, and Houston, respectively. At least 40% but probably closer to 75% of this organic matter, on average, was emitted or formed indoors. Functional group analysis provided some insights into the composition and properties of the indoor-generated organic PM2.5. Chlordane, a very minor but mutagenic semivolatile organic mixture previously used as a termiticide, was found to be mostly of indoor origin. High emission rates were most frequently found in homes built from 1945 to 1959. Analysis of the change in gas-particle partitioning during transport of outdoor PAHs to indoor environments illustrated that chemical thermodynamics can alter the concentration and composition of outdoor PM as it is transported indoors. (This has been previously noted for nitrate [Lunden et al 2003].) In epidemiologic studies that rely on central-site monitoring data, such transformations may result in measurement error, and this possibility warrants further investigation.