Models of inflation in a gravitational background with an anisotropic space-time scaling are studied. The background is a higher-dimensional Lifshitz throat with the anisotropy scaling $z\neq 1$. After the dimensional reduction, the four-dimensional general covariance is explicitly broken to a three-dimensional spatial diffeomorphism. As a result the cosmological perturbation theory in this set up with less symmetries have to be formulated. We present the consistent cosmological perturbation theory for this set up. We find that the effective four-dimensional gravitational wave perturbations propagate with a different speed than the higher dimensional gravitational excitations. Depending on the model parameters, for an observer inside the throat, the four-dimensional gravitational wave propagation can be superluminal. We also find that the Bardeen potential and the Newtonian potential are different. This can have interesting observational consequences for lensing and CMB fluctuations. Furthermore, we show that at the linearized level the inflaton field excitations vanish.

A new model of DBI inflation is introduced where the mobile brane, the inflaton field, is moving relativistically inside a Lifshitz throat with an arbitrary anisotropic scaling exponent $z$. After dimensional reduction to four dimension the general covariance is broken explicitly both in the matter and the gravitational sectors. The general action for the metric and matter field perturbations are obtained and it is shown to be similar to the classifications made in the effective field theory of inflation literature.

We analyze the spectrum of density perturbations generated in models of the recently discovered "D-cceleration" mechanism of inflation. In this scenario, strong coupling quantum field theoretic effects sum to provide a DBI-like action for the inflaton. We show that the model has a strict lower bound on the non-Gaussianity of the CMBR power spectrum at an observable level, and is thus falsifiable. This in particular observationally distinguishes this mechanism from traditional slow roll inflation generated by weakly interacting scalar fields. The model also favors a large observable tensor component to the CMBR spectrum.