Context. Linear polarisation in gamma-ray burst prompt emission is an important diagnostic with the potential to significantly constrain models. The spectrometer aboard INTEGRAL, SPI, has the capability to detect the signature of polarised emission from a bright γ-ray source. GRB 041219a is the most intense burst localised by INTEGRAL with a fluence of 5.7 ×10 −4 erg cm −2 over the energy range 20 keV-8 MeV and is an ideal candidate for such a study. Aims. Polarisation can be measured using multiple events scattered into adjacent detectors because the Compton scatter angle depends on the polarisation of the incoming photon. A search for linear polarisation in the most intense pulse of duration 66 seconds and in the brightest 12 seconds of GRB 041219a was performed in the 100-350 keV, 100-500 keV and 100 keV-1 MeV energy ranges. It was possible to divide the events into six directions in the energy ranges of 100-350 keV and 100-500 keV using the kinematics of the Compton scatter interactions. Methods. The multiple event data from the spectrometer was analysed and compared with the predicted instrument response obtained from Monte-Carlo simulations using the GEANT 4 INTEGRAL mass model. The χ 2 distribution between the real and simulated data as a function of the percentage polarisation and polarisation angle was calculated for all three energy ranges. The degree and angle of polarisation were obtained from the best-fit value of χ 2 . Results. A weak signal consistent with polarisation was found throughout the analyses. The degree of linear polarisation in the brightest pulse of duration 66 s was found to be 63 +31 −30 % at an angle of 70 +14 −11 degrees in the 100-350 keV energy range. The degree of polarisation was also constrained in the brightest 12 s of the GRB and a polarisation fraction of 96 +39 −40 % at an angle of 60 +12 −14 degrees was determined over the same energy range. However, despite extensive analysis and simulations, a systematic effect that could mimic the weak polarisation signal could not be definitively excluded. Conclusions. Our results over several energy ranges and time intervals are consistent with a polarisation signal of about 60% but at a low level of significance (∼2σ). The polarisation results are compared with predictions from the synchrotron and Compton drag processes. The spectrum of this GRB can also be well fit by a combined black body and power law model which could arise from a combination of the Compton and synchrotron processes, with different degrees of polarisation. We therefore conclude that the procedure described here demonstrates the effectiveness of using SPI as a polarimeter, and is a viable method of measuring polarisation levels in intense gamma-ray bursts.