Statistical analysis of life satisfaction data standardly assumes 1) linear scale use 2) interpersonal comparability 3) intertemporal comparability. To interrogate the credibility of these assumptions, we need to understand the reporting function: the process by which individuals make life evaluations and then map them to a category on the response scale. We develop a formal model of the reporting function that informs questions in long cognitive interviews of a diverse sample of 100 residents of the United Kingdom. Our results bear out previous efforts to validate life satisfaction scales. However, we also find widespread, severe, systematic, and non-random violations of all three standard assumptions. In particular, many respondents do not use the top of the scale, major shocks appear to alter the meaning of the points on the scale, older people are more likely to interpret life satisfaction questions as referring to how life went rather than how it is going, and respondents rarely describe their own scale use as linear.