The grid has the potential to transform collaborative scientific investigations through the use of closely coupled computational and visualization resources, which may be geographically distributed, in order to harness greater power than is available at a single site. Scientific applications to benefit from the grid include visualization, computational science, environmental modelling and medical imaging. Unfortunately, the diversity, scale and location of the required resources can present a dilemma for the scientific worker because of the complexity of the underlying technology. As the scale of the scientific problem under investigation increases so does the nature of the scientist's interaction with the supporting infrastructure. The increased distribution of people and resources within a grid-based environment can make resource sharing and collaborative interaction a critical factor to their success. Unless the technological barriers affecting user accessibility are reduced, there is a danger that the only scientists to benefit will be those with reasonably high levels of computer literacy. This paper examines a number of important human factors of user interaction with the grid and expresses this in the context of the science undertaken by RealityGrid, a project funded by the UK e-Science programme. Critical user interaction issues will also be highlighted by comparing grid computational steering with supervisory control systems for local and remote access to the scientific environment. Finally, implications for future grid developers will be discussed with a particular emphasis on how to improve the scientists' access to what will be an increasingly important resource.