Tsetse flies significantly impact public health and economic development in sub-Saharan African countries by transmitting the fatal disease African trypanosomiasis. Unusually, instead of laying eggs, tsetse birth a single larva that immediately burrows into the soil to pupate.  Where the female chooses to larviposit is therefore crucial for offspring survival. Previous laboratory studies suggested that a putative larval pheromone, n-pentadecane, attracts gravid female Glossina morsitans morsitans to appropriate larviposition sites. However, this attraction could not be reproduced in field experiments. Here, we resolve this disparity by designing naturalistic laboratory experiments that closely mimic the physical characteristics found in the wild. We show that gravid G. m. morsitans were neither attracted to the putative pheromone nor, interestingly, to pupae placed in the soil.  In contrast, females appear to choose larviposition sites based on environmental substrate cues. We conclude that, among the many cues that likely contribute to larviposition choice in nature, substrate features are a main determinant, while we failed to find evidence of a role of pheromones.

The dataset was collected in grouped non-forced two-choice experiments as described in 10.1098/rspb.2023.0030. Briefly, heavily pregnant female Glossina morsitans morsitans were given the choice between a) sand and leaf litter, b) putative pheromone and control (under leaf litter), and c) sand conditioned with fresh pupae and unconditioned sand (under leaf litter). Larviposition choice was recorded based on where the pupae were found: in the stimulus tray ("stim"), in the control tray ("con"), on the floor on the stimulus side ("stim.out"), or on the floor on the control side ("con.out"). Larvae could not cross the boundary between the two sides of the floor. Pupae were collected daily, and their average weight was determined at the end of each experiment.

A Bayesian model of the data was built to verify the results.

All data was analysed in R using custom-designed analysis scripts. 

The data is stored in .csv files. All data was analysed in R using custom-designed analysis scripts.

The code for the Bayesian model is provided as an R markdown file.