Underfitting

• error on train data high → high Bias
• error on test data high → high Variance

Also see Overfitting.