<b>Background:</b> Inference on pneumococcal transmission has mostly relied on longitudinal studies which are costly and resource intensive. Therefore, we conducted a pilot study to test the ability to infer who infected whom from cross-sectional pneumococcal sequences using phylogenetic inference. <b>Methods:</b> Five suspected transmission pairs, for which there was epidemiological evidence of who infected whom, were selected from a household study. For each pair, <i>Streptococcus pneumoniae</i> full genomes were sequenced from nasopharyngeal swabs collected on the same day. The within-host genetic diversity of the pneumococcal population was used to infer the transmission direction and then cross-validated with the direction suggested by the epidemiological records. <b>Results:</b> The pneumococcal genomes clustered into the five households from which the samples were taken. The proportion of concordantly inferred transmission direction generally increased with increasing minimum genome fragment size and single nucleotide polymorphisms. We observed a larger proportion of unique polymorphic sites in the source bacterial population compared to that of the recipient in four of the five pairs, as expected in the case of a transmission bottleneck. The only pair that did not exhibit this effect was also the pair that had consistent discordant transmission direction compared to the epidemiological records suggesting potential misdirection as a result of false-negative sampling. <b>Conclusions:</b> This pilot provided support for further studies to test if the direction of pneumococcal transmission can be reliably inferred from cross-sectional samples if sequenced with sufficient depth and fragment length.