In this paper, an assessment of the criteria for the prediction of the flow excursion instabilities in vertical narrow rectangular channels is presented. The experimental database consists of 166 flow excursion points at low pressure with upward flow, and with uniform or non-uniform heat flux profiles. The test sections have gap sizes between 1.27 and 3.6 mm, hydraulic diameters between 2.37 and 6.58 mm, and length to heated diameter ratio between 70.9 and 196.8. A wide range of parameters is covered: the mass flux is between 740 and 20,325 kg/m(2)s; the outlet pressure is between 0.12 and 1.73 MPa; the heat flux is between 0.4 and 14.9 MW/m(2); the Peclet number is between 15,889 and 358,460; the outlet sub-cooling is between 4.8 and 35.1 degrees C. None of the tests reaches saturation at the exit of the test section. Several criteria for identifying the Onset of Flow Instability (OFI), were tested. Such criteria can rely on correlations for the Onset of Nucleate Boiling (ONB), the Net Vapor Generation (NVG), the onset of Fully Developed Boiling (FDB), or can relate global parameters of the system. All these models have good performances on average, with both uniform and non-uniform axial heat fluxes. The ONB-based relationships are largely conservative as expected since the ONB always precedes the OFI. The NVG criteria can provide relatively good results, but a crucial issue is related to the value of Peclet number at which the transition between the thermally and the hydro-dynamically driven bubble detachment takes place in narrow rectangular channels. In view of this, the standard Saha-Zuber correlation cannot predict OFI for Peclet numbers lower than 70,000; while the Saha-Zuber KIT correlation, whose transition Peclet number is smaller, identifies the OFI for all the experiments. The approach that makes use of a FDB correlation can capture the OFI in most of the cases, although its performance also depends on the type of correlation that is applied for the single-phase heat transfer. The Flow Instability Ratios (FIRS) like the ones developed by Whittle-Forgan or Stelling et al., are of particular interest because they only require global system parameters, and because they are shown to be a valid option for determining the flow excursion in the experiments included in this study. For instance, the Stelling FIR with the Saha-Zuber KIT correlation estimates OFI in all the tests. Finally, best-fitting procedures of the available data were also introduced in order to optimize such FIRS. (C) 2016 Elsevier Ltd. All rights reserved.