100 CHAPTER 3 Information-Theoretic Analysis of Neural Data does little to respond to the issues. More basic and interesting would be how to assess how well a stimulus or its features are represented by a neural response and how well a neural group extracts features from its input. The Kullback- Leibler distance is the information-theoretic suggestion, but estimating it for a population response can require much data: Producing accurate estimates even for small populations could demand more data than could be reasonably measured. The reason for the "data-hungry" nature of estimating this and other information-theoretic quantities is also the same reason for employing them: their generality. They can cope with any communication and processing scenario, but this capability comes at an experimental price of demanding data to fill in the details. Do note that we have followed in the footsteps of many other authors in ignoring the existence of feedback and its corollary adaptivity. Feedback is a notoriously difficult concept to handle with the traditional tools of information theory, but it would be foolish to discount the role of feedback connections when they are so prominent anatomically even in the most peripheral sensory pathways. Some recent work has sparked interest in feedback in the informa-