Research in HDFS is frequently concerned with time, so studies are always designed with an orientation to time. A cross-sectional orientation makes all measurements at one point in time, with no follow-up. This provides a cross-section, but cannot identify processes of change except by relying on retrospective reports that are subject to recall bias. A longitudinal time orientation involves more than one point of contact with the study subjects, and is prospective in nature. This is clearly superior for analyzing the processes of change. Some important longitudinal research in HDFS involves long-term cohort studies. Other longitudinal research is experimental to test the impact of an intervention, and this can be long-term, repeated measures, or time series. Longitudinal research is superior to cross-sectional, but includes specific risks: attrition, instrument decay, and the practice effect. Research on marital satisfaction over time illustrates the difference in these approaches. A time-trend orientation measures a population, but not individual subjects, repeatedly over time. This can be used when the population is the unit of analysis, and the goal is to document potential historical changes in the population.