Inflationary cosmology provides a causal mechanism for the generation of super-Hubble cosmological fluctuations. There have been many alternative proposals suggested to accomplish this feat, however these all seem to share the need to violate at least the Null Energy Condition. I will attempt to make this statement more precise, and focusing on the case of string motivated models that contain a gravi-scalar in their spectrum (such as the string theoretic dilaton) we will find a "no-go" theorem. This provides an important challenge for models such as String Gas Cosmology and the Pre-Big Bang, if such models are to become more predictive.