Background: The sharing of excitation energy between the fission fragments is one of the key issues in studying nuclear fission. One way to address this is by studying prompt-fission neutron multiplicities as a function of other fission observables such as the mass, $\overline{\ensuremath{\nu}}\left(A\right)$. These are vital benchmark data for both fission and nuclear deexcitation models, putting constrains on the fragment excitation energy and hence on the competing prompt neutron/$\ensuremath{\gamma}$-ray emission. Despite numerous detailed studies, recent measurements done at JRC-Geel with the SCINTIA array in the epithermal region show surprisingly strong discrepancies to earlier thermal fission data and the Wahl systematics.Purpose: The purpose was to perform measurements of the prompt-fission neutron multiplicity, as a function of fragment mass and total kinetic energy (TKE), in $^{235}\mathrm{U}\left({\text{n}}_{\text{th}},\text{f}\right)$ and $^{252}\mathrm{Cf}\left(\text{sf}\right)$, to verify and extend the SCINTIA results. Another goal was to validate the analysis methods, and prepare for planned investigations at excitation energies up to 5.5 MeV.Methods: The experiments were conducted at the former 7 MV Van de Graaff facility in JRC-Geel, using a Twin Frisch-Grid Ionization Chamber and two liquid scintillation detectors. A neutron beam with an average energy of 0.5 MeV was produced via the $^{7}\mathrm{Li}$(p,n) reaction. The neutrons were thermalized by a 12 cm thick block of paraffin. Digital data acquisition systems were utilized. Comprehensive simulations were performed to verify the methodology and to investigate the role of the mass and energy resolution on measured $\overline{\ensuremath{\nu}}\left(A\right)$ and $\overline{\ensuremath{\nu}}\left(\text{TKE}\right)$ values. The simulation results also revealed that the $\ensuremath{\partial}\overline{\ensuremath{\nu}}(A)/\ensuremath{\partial}A$ and $\ensuremath{\partial}\overline{\text{TKE}}/\ensuremath{\partial}\overline{\ensuremath{\nu}}$ are affected by the mass and energy resolution. However, the effect is small for the estimated resolutions of this work. Detailed Fluka simulations were performed to calculate the fraction of thermal neutron-induced fission, which was estimated to be about $98%$.Results: The experimental results on $\overline{\ensuremath{\nu}}\left(A\right)$ are in good agreement with earlier data for $^{252}\mathrm{Cf}\left(\text{sf}\right)$. For $^{235}\mathrm{U}\left({\text{n}}_{\text{th}},\text{f}\right)$, the $\overline{\ensuremath{\nu}}\left(A\right)$ data is very similar to the data obtained with SCINTIA, and therefore we verify these disclosed discrepancies to earlier thermal data and to the Wahl evaluation. The experimental results on $\overline{\ensuremath{\nu}}\left(\text{TKE}\right)$ are also in agreement with the data at epithermal energies. For $^{252}\mathrm{Cf}\left(\text{sf}\right)$ a slope value of $\ensuremath{\partial}\overline{\text{TKE}}/\ensuremath{\partial}\overline{\ensuremath{\nu}}=\left(\ensuremath{-}12.9\ifmmode\pm\else\textpm\fi{}0.2\right)\phantom{\rule{0.28em}{0ex}}\mathrm{MeV}/\mathrm{n}$ was obtained. For $^{235}\mathrm{U}\left({\text{n}}_{\text{th}},\text{f}\right)$ the value is $\left(\ensuremath{-}12.0\ifmmode\pm\else\textpm\fi{}0.1\right)\phantom{\rule{0.28em}{0ex}}\mathrm{MeV}/\mathrm{n}$. Finally, the neutron spectrum in the center-of-mass system was derived and plotted as a function of fragment mass.Conclusions: This work clearly proves the lack of accurate correlation between fission fragment and neutron data even in the best-studied reactions. The new results highlight the need of a new evaluation of the prompt-fission multiplicity for $^{235}\mathrm{U}\left({\text{n}}_{\text{th}},\text{f}\right)$.