Multicollinearity
A linear relationship among the explanatory variables. Partial multicollinearity (a strong but not exact relation) leaves OLS unbiased but inflates the variance of the estimates, giving big standard errors and small t-statistics; the exact case is Perfect Multicollinearity.
When to use
The classic signature is joint significance on an F-test with individually insignificant t-test results, or neither variable significant when both are in the model. A rule of thumb is a correlation above about 0.7; check the correlation matrix before dropping a seemingly irrelevant variable.
See Lecture 8 and Lecture 13.