Formaldehyde (CH2O) is an important intermediate in spray combustion of diesel-like fuels. Experimentally, its spatial distribution is often assessed by means of planar Laser-Induced Fluorescence (PLIF), using a frequency-tripled Nd:YAG laser for excitation and spectrally-filtered fluorescence detection. Especially in soot-laden spray flames, however, considerable spectral interference cannot be avoided, which renders interpretation of the fluorescence distribution challenging. We introduce a more selective strategy, involving dye laser excitation and background subtraction by means of off-resonant excitation in combination with a bootstrap evaluation of sets of individual events. The latter extracts persistent features out of non-simultaneously recorded snapshots of stochastic events, thereby mitigating possible interpretation issues related to the turbulent nature of the process under study. Following this approach, recurrent patterns can be distinguished from random high-intensity events (allegedly due to turbulent fluctuations). We demonstrate our method on the Engine Combustion Network (ECN) "Spray A"-benchmark, with 15 vol-% and 21 vol-% ambient oxygen; we also present spatially well-delimited CH2O fluorescence distributions, and compare these to simultaneously recorded OH* chemiluminescence intensities. (C) 2017 The Combustion Institute. Published by Elsevier Inc. All rights reserved.