Previous research has used both natural and simulated interactions to investigate the functions of infant-directed speech (IDS), but the effect of these approaches on the results of such studies is unclear. The aim of this study was to compare F (0) contours of natural and simulated speech directed to three distinct speech recipient groups to investigate the effects of different methodologies on research findings. F (0) contours were derived from IDS, foreign-directed speech (FDS) and speech directed to a native adult speaker (ADS). The speech was produced by ten students and ten actresses in simulated 'interactions' with an imaginary speech partner and compared to an existing dataset of natural mother-infant interactions. The F (0) contours were analysed with a qualitative and a quantitative method. In the existing natural dataset, IDS F (0) contours (mainly sinusoidal-bell shapes) had been markedly different from FDS and ADS F (0) contours (mainly level shapes). However, these results were not reproducible in the simulated interactions with either students or actresses. Here, the different shapes were more evenly distributed across IDS, FDS and ADS with no difference between these three groups. Consequently, it is suggested that simulated speech should be used with caution in IDS or any other speech research.