Behavioral end points for neurotoxicity risk assessment have been deve
loped and examined over the past three decades. They are now ready to
move from simple qualitative guidelines, such as exemplified by refere
nce doses, to more quantitative models, such as benchmark doses, based
on dose-response information. Risk assessors, confronted by a wider a
rray of methodologies and data than in the past, should be offered gui
dance in interpretation because now they have to deal with unaccustome
d questions and problems. These include reversibility, susceptible pop
ulations, multiple end points, and the details of dose-response and do
se-effect distributions.