Split-beam correlation is a technique for passive bearing estimation of underwater sound sources. Split-beam correlation processing can be an attractive alternative in surveillance by means of towed arrays or when using several linear arrays installed at the sea bottom. The main advantage compared to classical beamforming, evident in high Signal to Noise Ratio (SNR) situations, is improved bearing resolution. This work demonstrates the development of a computer code for split-beam processing. A future scope with the implementation of this method is to study its applicability in shallow waters. The horizontal arrays to be utilized should contain many wavelengths of the source signal components and still be populated by sensors separated by a moderate fraction of a half-wavelength. A rule of thumb is derived here for accessing the multipath interference that may occur in such environments. Field measurements carried out in coastal shallow waters are analysed. Coherence measurements indicate that for frequencies below 250 Hz the experimental site most often supports coherence lengths of the order of 200 m. The expected trend of coherence improvement with decreasing frequency is not evident. For the higher frequencies this behavior could be explained by changes in the speed of sound and sea surface roughness, affecting the multipath propagation within the shallow water channel in a different way during each measurement. With a few exceptions, coherence did improve with decreased hydrophone separation distances. Split-beam processing is carried out using signals from an approximately horizontal linear array, with the same aperture as the experimentally deduced coherence lengths. The array is sparsely populated with hydrophones. The sensor spacing is up to fifteen times larger than what should be appropriate for the source signal bandwidth utilized. Grating lobes are introduced as a consequence, and the array gain becomes severely degraded. The coastal environment hosting the array, causes source signal reflection from shore, further increasing the destructive impact of the grating lobe presence. The beam power versus bearing plots did indicate the position of the sound source but they did also show a number of false bearing indications most likely due to the grating lobe issue. No statement about the ultimate performance of the split-beam correlation technique can therefore be made on the basis of the processed data.