A value for a continuous variable that is unusual by virtue of lying well away from the majority of the values observed. Can be a flag that there has been a data entry error but can be real. Tend to have a lot of influence in some statistical summaries of data.
Try also #
Not formally discussed in the book.