I think it is well understood that the risk in a software engineering project increases with the size of the project, and the slope of the scale exceeds linear. As the number of features increase, not only will an estimate increase, but the margin of error in that estimate increases as well.
For this reason, Agile methodology emphasizes the need to break projects into smaller efforts that will fit into a single iteration. This offers greater project predictability (and a higher level of sanity among the dev team).
Is there a name for this principle? E.g. something like "John's law?"
Has there been any research that quantifies this principle? E.g. is the relationship between size and risk geometric or exponential?