Choosing a dissertation topic you can still stand in month six

Supervisors can usually tell in the first meeting whether a dissertation will be painful, and it is almost never about ability. It is about the question. A topic that is too broad, too dependent on access you do not have, or too dull to sustain interest will cost you months that better planning would have saved.

Test one: can it be answered?

A question like What is the impact of social media on society cannot be answered by anyone, let alone in twelve thousand words. Narrow until the question names a specific population, a specific outcome and a specific context. The narrowing feels like a loss of ambition. It is actually the only route to a substantial piece of work.

Test two: can you get the data?

Plenty of excellent questions die on access. If your design depends on interviewing senior executives, on a dataset behind a paywall your library does not subscribe to, or on ethics approval that takes longer than your timeline, the question is not viable regardless of its merit. Confirm access before committing, not after.

Test three: is there a literature to argue with?

A genuinely untouched topic sounds appealing and is usually a warning. If nobody has written about it, you have no framework, no methods to borrow and no debate to enter. The workable position is a question adjacent to an active literature: enough published work to build on, and a gap within it you can defend.

Test four: will you still care in month six?

This sounds soft and it is the most practical of the four. A dissertation is a long piece of work carried largely on your own motivation. A topic chosen because it sounded impressive, rather than because you want the answer, becomes very hard to pick up again in the difficult middle stretch.

• Write the question in one sentence and show it to someone outside your field
• Check three recent papers exist that you could genuinely build on
• Name your data source and confirm you can reach it
• Sketch the chapter structure before committing, to test the question carries one