Model predictions for the dendrite growth velocity at low undercoolings are deviating significantly from experimental data obtained in electromagnetic levitation with a capacitance proximity sensor (CPS) [K. Eckler, D.M. Herlach, Mater. Sci. Eng. A 178 (1994) 159]. In addition to that, previous data sets obtained by different techniques are not in good agreement with each other. For instance, growth velocity data for nickel melts obtained with a high-speed camera system [D.M. Matson, in: Solidification 1998, TMS, Warrendale PA, 1998, p. 233] show higher values at low undercoolings than data obtained with the CPS. Within this work new measurements of dendritic growth velocity in levitated undercooled nickel samples were performed as a function of undercooling Delta T to investigate this discrepancy. Solidification of the undercooled melt was detected at undercooling levels within the range of 30 K < Delta T < 300 K. The new data reveal high accuracy and low scattering. These data are compared with two independent growth velocity data sets and discrepancies are discussed. For verification of the new CPS data dendrite growth velocity was also measured by using a high-speed camera where the morphology of the intersection of the solidification front with the sample surface was investigated. The new experimental data are analyzed within the model of dendrite growth obtained on the basis of Brener's theory [E. Brener, J. Crystal Growth 99 (1990) 165] and the model of dendrite growth with melt convection in a solidifying levitated drop, presently being developed. Special attention is paid to the effects of convection and small amounts of impurities on the growth dynamics at small undercoolings. (c) 2006 Elsevier BN. All rights reserved.