In the first of three experiments, she had 148 of her subjects fill out a detailed questionnaire reporting how often and how intensely they experience emotions such as joy, love, compassion and awe. She also had them agree or disagree with statements like “I often notice people who need help.” Such self-reported data ought to be notoriously unreliable, since not many of us are likely to respond honestly if our answers make us look like a louse. But personality inventories are a long-standing staple of psychological testing — especially since the scoring is designed to correct for self-flattering grade inflation.

When the numbers on these inventories were crunched, Stellar and her colleagues found no meaningful personality differences among the students that could be attributable to income except one: across the board, the lower the subjects’ family income, the higher their score on compassion.

The second study involved a smaller group of 64 subjects who watched two videos — an emotionally neutral instructional video on construction techniques, and a far more charged one that involved real families coping with a cancer-stricken child. Again, the subjects filled out emotional inventories and again they scored similarly on most mood metrics, including sadness. But the lower-income volunteers continued to come out higher on the compassion-and-empathy scale.