Within the first few years after chlorpromazine began to be used to treat psychosis, it was observed that it could cause many kinds of neurologic reactions that resembled those seen in idiopathic Parkinson's disease. These reactions were termed "extrapyramidal side effects" (EPS) because of their resemblance to the signs of Parkinson's disease, which were associated with degeneration of the dopamine nerve tracks located in the extrapyramidal region of the central nervous system. Eventually this association of dopamine loss, antipsychotics, and parkinsonism became a central part of the dopamine hypothesis of schizophrenia. Unfortunately, this association was also used to support the hypothesis that EPS were absolutely necessary for antipsychotic efficacy--hence the term "neuroleptic" rather than "antipsychotic." This theory, now discredited, was used to justify the practice of inducing EPS as a means to gauge whether an antipsychotic would be effective. The demonstration that clozapine, an antipsychotic virtually devoid of EPS, has better efficacy for psychosis than any other "neuroleptic" disproved the theory that EPS were fundamentally linked to efficacy. Because the idea of a relationship between EPS and efficacy was so ingrained in clinical practice, clozapine was called "atypical." Our understanding of the relationship between EPS and antipsychotic response has come full circle. With the introduction of clozapine and other newer antipsychotics, it has become clear that EPS are harmful and serve no beneficial purpose. The availability of newer antipsychotics with a lower EPS burden means that, at least in theory, it is now possible to treat psychosis without EPS in the vast majority of patients. In practice, however, EPS remain a significant problem even in the era of atypical or second generation antipsychotics (SGAs). One limitation is that the concept of "atypicality," when used to denote antipsychotic efficacy without EPS, is a relative not an absolute concept. Because all of the post-clozapine SGAs still affect the dopamine D2 receptor, it may be more accurate to say these medications have lower EPS liabilities that the earlier "neuroleptic" antipsychotics; i.e., relatively fewer patients will get EPS at therapeutic doses of one of the newer medications and, when EPS do occur, they tend to be less severe. Nonetheless, reduced EPS are not the same as no EPS, and most of the newer antipsychotics can still cause EPS in some patients. The incidence of EPS differs among the SGAs, with risperidone associated with the most and clozapine and quetiapine with the fewest EPS. The likelihood of developing EPS with a first-line SGA depends not only on the specific agent, but also on the rapidity of dose escalation, the target dose, and the patient's intrinsic vulnerability to EPS. Even with the SGAs, clinicians should not be lulled into believing EPS cannot happen, but need to be able to recognize and manage both overt and subtle manifestations of EPS. This review discusses differences among the SGAs in EPS liability, relationships between dosing and type of EPS, and situations in which differences in EPS liability among the SGAs are clinically relevant.