Those are all important questions in regards to the flight apparatus that they use in this experiment. The technique has been around for a number of years now and there have actually been tests done to establish the role of "attention" and other factors in the flys' behavior. A Neurosciences Institute (NSI) researcher, Bruno van Swinderen, published an excellent paper in Science recently using this flight simulator to investigate some of these interesting aspects of fly learning:

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd= Retrieve&db=pubmed&dopt=Abstract&list_uids=1736367 5

They can gauge the fly's learning based on a large number of criteria (crosses, colors, shapes, etc) and this learning won't occur with flies that are exposed to the "default" setup of simply a rotating drum with the glue and wire. Also, the limitations of the fly's viewing field are known to a fairly precise degree. The temperature is controlled and a valve releases air directly onto the fly from a single direction. All of these setups have been tested for just about any variation you could possibly imagine...otherwise the data would not be accepted as valid information.