The complexity of human speech is an anomaly among primates, where calls have typically been thought of as stereotyped and not variable.
Recent work has shown that there is more learning and control present in nonhuman primate vocalizations than previously believed,
but adult calls have nevertheless seemed to be stable. Pomberger et al. interrupted calling marmosets with acoustic playbacks triggered by their own calls and found that the animals could extend their “phee” calls beyond normal lengths by restarting them rapidly after perturbation,
but they could only alter their calls at periodic time points. These results suggest that the phee call is not one long stereotyped series, but rather is composed of small sequential segments, like human speech.
Consumption of fecal matter is taboo, unless you have a life-threatening disease, such as Clostridium difficile infection. In these cases, fecal microbiota transplantation (FMT) is extraordinarily effective at restoring gut function to a healthier state.
However, FMT does not work so well for other indications that are less strongly associated with the microbiota. We have a weak understanding of what factors allow microbial engraftment in the gut.
Smillie et al. took genomic data from FMT-treated C. difficile patients to build a machine-learning statistical model that tells us which bacterial strains will engraft.
The main predictive factors are abundance of a strain in the donor and the species present in the patient. Donor strains whose species are also present in the patient always engraft. If the taxa are discordant, then it appears that microbial interactions prevent engraftment.