A study by Jonathan Gardner and Andrew Oswald also follows individuals over time but asks a different question. It considers what happens to people's happiness when marriages end. The study looks at two ways a marriage might end: divorce or death of a spouse. The results are summarized in the figure below.

The horizontal axis shows years relative to an important event (divorce or widowhood) at time 0. The vertical axis shows life satisfaction. Life satisfaction is shown in black for those who became divorced and in gray for those who became widowed.

a) Notice the initial difference in life satisfaction between those who became widowed and those who got divorced, even before the event occurred. Does this difference make you more or less confident in Gilbert's causal interpretation? Why?

b) Now consider the widows and widowers (gray line). How does their happiness change before, during, and after the year in which their spouses passed away? Does this make you more or less confident in Gilbert's causal interpretation? What does this comparison make you think might be going on in Gilbert's original correlation?