As someone who is published, I can tell you that it depends entirely on the field. One possibility is obtaining data from other people and analyzing it in new ways. There are many free public sources of data, and a lot of researchers will share old data sitting on their hard drives if they think it could result in publications. Off the top of my head, genomics, bioinformatics, microscopy, medical imaging/radiology, and biometrics are all fields where there is a glut of data and people would gladly welcome new, more powerful analysis tools and procedures.