Sandeep Gautam has a nice post on his blog about the effects of depression on hippocampal cells.
The story is that chronically high levels of cortisol, a stress hormone, kills these neurons. Depression leads to elevated average cortisol levels, as lower moods make negative emotional responses, such as anger and anxiety, more severe. Worse, the hippocampus is also responsible for down-regulating stress levels, the shrinking mass thereof creating an inertia with respect to increases in mood, both short and long term. So, climbing out of depression is harder, the more severe the depression and the longer its duration. Fortunately, hippocampal cells are born anew with increased average mood levels over a sufficient period of time, increasing the brain's ability to downregulate stress.
I interpret this as representing the physiological mechanism by which low average mood levels demand higher net levels of reinforcement over the longer term to reduce depression-related risk aversion, in the risk averse. This serves the behavioral economic purpose of requiring more evidence that an environment long seen as hostile to the goal of secure reproduction has become more hospitable. This is the equivalent of not trusting someone who is long seen as a jerk, but acts somewhat nicer on a given day. The trust in the change doesn't come overnight, and nor should a trust in what is normally an inhospitable enviornment.
Quantitative Psychological Theory and Musings