This chapter describes the automatic recognition of and response to human emotion within intelligent tutors. Tutors can recognize student emotion with more than 80%accuracy compared to student self-reports, using wireless sensors that provide data about posture, movement, grip tension, facially expressed mental states and arousal. Pedagogical agents have been used that provide emotional or motivational feedback. Students using such agents increased their math value, self-concept and mastery orientation, with females reporting more confidence and less frustration. Low-achieving students-one third of whom have learning disabilities-report higher affective needs than their higher-achieving peers. After interacting with affective pedagogical agents, low-achieving students improved their affective outcomes and reported reduced frustration and anxiety.