We propose that two related sources of variability in studies of caregiving health effects contribute to an inconsistent pattern of findings: the sampling strategy used and the definition of what constitutes caregiving. Samples are often recruited through self-referral and are typically comprised of caregivers experiencing considerable distress. In this study, we examine the health effects of caregiving in large population-based samples of spousal caregivers and controls using a wide array of objective and self-report physical and mental health outcome measures. By applying different definitions of caregiving, we show that the magnitude of health effects attributable to caregiving can vary substantially, with the largest negative health effects observed among caregivers who characterize themselves as being strained. From an epidemiological perspective, our data show that approximately 80% of persons living with a spouse with a disability provide care to their spouse, but only half of care providers report mental or physical strain associated with caregiving.