Research proposals should have a clearly defined research question and not just be a demonstration or examination of a topic. You should find a question that follows logically an existing line of inquiry or fills an existing void. The research proposal should then lay out your approach to answering this question or filling this void. Working with a faculty member can be of great assistance for topic suggestions.
The research proposal should be of sufficient focus, meaning it should be do-able and of the proper scope to finish in a timely manner. It should also not be too short. Different opportunities have different page lengths for their applications, but you should be sure to provide enough information to maximize this space. You should also be certain that your university/ department/ faculty advisor has the resources and the expertise to carry out the research you are suggesting. This will require some investigation and perhaps even some conversations with proposed institutions (if you are not already enrolled).
Few students enter graduate school knowing exactly what they want. You should remember, applying for a fellowship does not necessarily mean a lifetime commitment. If things should change in the future, that is acceptable, but when applying for a fellowship, conveying to the granting organization that your clear on what you want is very important. Granting organizations are often funding students as a researcher, not necessarily funding their research. This is a place for you to show you have a logical mind. It also may be helpful to build on what you already know or have done. You should be systematic with a clear plan and maybe even a potential timeline. The approach should also be driven by a strong rationale demonstrating your understanding of the existing literature on the topic. Citing sources is acceptable, but remember a research proposal is not a literature review. In summary, fellowship-granting organizations are looking for a clear research question, a rationale to why this question is important, and a plan on how you are going to answer this question.
You should also keep in mind your audience. Busy people who may not be experts in your specific area of interest or inquiry will be reading your proposal. You should therefore minimize jargon and state things simply in common terms (and define terms when needed). This is why it is a good idea to build on what you already know. It is easier to minimize jargon on a topic that is familiar, than one that is not. Reviewers may also be quite familiar with your current or proposed institution as well as the research conducted there (they might even be a MIT faculty member) and therefore they will probably have a good idea if the research can be accomplished.
The research proposal should be properly formatted and consistent with accepted conventions. Avoid the wall of text. The research question should be underlined, key points italicized, and big ideas in boldface type. The use of bullets is typically acceptable. All this formatting will help reviewers who are skimming and with readers reviewing large numbers of applications, they will be skimming. Use graphics and figures appropriately, a visual element can be well worth the words they are traded for, but they should also be relevant (see here for example). In summary, make it pretty!
Lastly, as with the personal statement, the research proposal is part of an application package and therefore needs to feed off of and inform the rest of the package’s contents. Please see Writing a Successful Fellowship Proposal presentation by MIT Professor Stephen Bell for more assistance and suggestions.