Can a research question change during a PhD?

The question you registered with was written by someone who had read a fraction of what you have now read. Expecting it to survive intact would be strange, and supervisors know this. What causes trouble is not the change itself but the way it happens.

There are two kinds. Deliberate revision follows a specific discovery: the data you needed does not exist, the effect you were testing was reported as a null in three unpublished studies, the concept turned out to be two concepts. That is research working correctly and it is straightforward to defend.

The methods literature lists the specific faults that force a revision, and they are mundane ones. A question gets rewritten when it is too broad, when there are not enough subjects, when answering it needs skills or equipment you do not have, when it costs too much, or when it stops looking interesting or relevant. The matching fixes are equally concrete: cut the number of variables, change your inclusion and exclusion criteria, add a data source or extend the study period, learn the skill or find a collaborator, and reduce the number of measurements or follow-ups (Ratan et al., 2019, p. 15). If one of those sentences describes your situation, you have your justification already written.

Ratan and colleagues also treat question formulation as a sequence rather than a single decision, which is why revision is built into it. You start from a broad subject, read the prior studies, identify the gap, move from a general question to specific ones, narrow the scope, evaluate what you have, and only then write the question down (Ratan et al., 2019, p. 18). They score the result against ten criteria they abbreviate as FINERMAPS, meaning feasible, interesting, novel, ethical, relevant, manageable, appropriate, potentially valuable, publishable, and systematic (Ratan et al., 2019, p. 15). A question that fails one of those tests is supposed to change.

Drift is different. Drift happens when the question is never revised on paper but the work slowly stops answering it. The introduction still asks about one thing while chapters four and five are about another, and nobody notices until the final read-through, when the thesis reads as three loosely related studies.

The protection against drift is cheap. Keep the current question in one file with a date. Every time your understanding shifts, reread it and either confirm it or rewrite it and note why. Ten minutes a month, and it also means you always have a defensible answer when someone asks what your thesis is about.

How do I justify the change to my committee?

Present it as a finding. Say what you learned that made the original question unanswerable or uninteresting, name the evidence, then state the new question and what it now makes possible. A change explained this way demonstrates judgment. An unexplained one looks like loss of control.

Bring the evidence with you. If a dataset turned out to be unusable, show the access refusal or the missingness. If the effect had already been established, bring the paper. Committees are reassured by a candidate who can point to the specific thing that forced the decision.

Say clearly what stays and what goes. A committee's real worry is timeline, so pair the change with a revised plan showing which chapters survive, which need rewriting, and what the new completion date is.

What happens to the work built on the old question?

Most of it survives in a different role. Literature becomes context, pilot data becomes justification for the redesign, and an abandoned analysis often becomes a methods appendix. Genuinely dead work is worth naming in one sentence rather than hiding, since examiners can usually see the seam anyway.

Wait before deleting anything. Work that looks useless during a redirection frequently becomes relevant again once the new question settles, and the cost of keeping files is nothing next to the cost of redoing an analysis.

A shifting question is often a sign the project is working. Marino and colleagues point out that a project going well will open new research questions and future directions, some of them beyond the scope of the PhD, and that a research project is never really finished (Marino et al., 2014, pp. 1-2). Note the difference in emphasis between the two sources. Ratan and colleagues are writing about getting the question right before the work starts, while Marino and colleagues are writing about what to do when the work keeps generating better questions than the one you registered. Both positions point at the same action, which is deciding on purpose what belongs in the thesis and what belongs in the future work section.

That decision has a deadline. Marino and colleagues advise settling on the set of agreed experiments that will provide enough evidence for the degree, then starting to write instead of returning to the laboratory for more data. They suggest fixing that plan around one year before submission and reviewing it every few weeks, with slack built in for technical problems (Marino et al., 2014, p. 1). Applied to a question that has changed, this means the new question needs a boundary and a date attached to it, not just a supervisor's agreement that it is the better question.

There is also a viva benefit. Being able to explain a path not taken, with reasons, is one of the clearest demonstrations of independent research judgment available to you, and examiners ask about it more often than candidates expect.