Prometaphase in two large species of diatoms is examined, using the following techniques: (a) time-lapse cinematography of chromosome movements in vivo; (b) electron microscopy of corresponding stages: (c) reconstruction of the microtubules (MTs) in the kinetochore fiber of chromosomes attached to the spindle. In vivo, the chromosomes independently commence oscillations back and forth to one pole. The kinetochore is usually at the leading edge of such chromosome movements; a variable time later both kinetochores undergo such oscillations but toward opposite poles and soon stretch poleward to establish stable bipolar attachment. Electron microscopy of early prometaphase shows that the kinetochores usually laterally associate with MTs that have one end attached to the spindle pole. At late prometaphase, most chromosomes are fully attached to the spindle, but the kinetochores on unattached chromosomes are bare of MTs. Reconstruction of the kinetochore fiber demonstrates that most of its MTs (96%) extend past the kinetochore and are thus apparently not nucleated there. At least one MT terminates at each kinetochore analyzed. Our interpretation is that the conventional view of kinetochore function cannot apply to diatoms. The kinetochore fiber in diatoms appears to be primarily composed of MTs from the poles, in contrast to the conventional view that many MTs of the kinetochore fiber are nucleated by the kinetochore. Similarly, chromosomes appear to initially orient their kinetochores to opposite poles by moving along MTs attached to the poles, instead of orientation effected by kinetochore MTs laterally associating with other MTs in the spindle. The function of the kinetochore in diatoms and other cell types is discussed.