001What the board was actually doing.
A review board's first pass is procedural rather than moral. Is every required element present. Is the risk level correctly claimed against what is being collected. Do the consent document and the protocol agree with each other on every point. Could a careful stranger, reading only your documents, describe exactly what will happen to a participant and in what order.
Most first submissions fail that pass on paperwork. The protocol mentions recording and the consent form does not. The study claims exempt status while proposing to collect identifiers. Recruitment materials are described rather than attached. Instruments are summarized rather than included. None of these are ethical failures and all of them return the submission.
This matters because the frame determines the response. Read as a judgment, a letter invites you to defend your study, and defending is what produces extra rounds. Read as a list of conditions, it invites you to satisfy them, which is what closes the file.
002The two habits that cost extra rounds.
Both are common, both feel reasonable at the time, and both are expensive in a process where a cycle is measured by a committee's meeting schedule rather than by your urgency.
- Arguing rather than addressing. A response explaining why the board's concern does not apply, without changing anything, returns with the same point restated. If you genuinely believe a point rests on a misunderstanding, make the change that removes the ambiguity and explain the reasoning alongside it.
- Fixing only the instance cited. Boards routinely flag one example of a problem occurring in several places and expect the pattern corrected. Correct the flagged line alone and the next letter flags the others, at the cost of a full cycle.
- Answering out of order. Respond point by point in the board's own numbering so a reviewer can check each item without searching. Reordering to suit your narrative makes verification harder and slows the review.
- Silent changes. Anything you altered that the letter did not ask about should be stated explicitly, because an unexplained difference between versions reads as an inconsistency and generates a new question.
003The response document, structured.
One table or numbered list, matching the letter's structure exactly. For each point: what the board asked, what you changed, and where it now appears, with a page or section reference. Then attach the revised documents with tracked changes so a reviewer can see the edit without diffing anything themselves.
That is the entire technique, and its effect is out of proportion to the effort. You are removing work from the reviewer's desk. A response requiring somebody to hunt for changes across four documents takes longer to process and invites more questions; one where every change is labeled and located is checked in minutes.
Where a point genuinely cannot be complied with, say so plainly, explain why, and propose an alternative that meets the same underlying concern. Boards are far more receptive to a substitute that addresses their worry than to an explanation of why the worry is unwarranted.
004The contradiction check nobody runs.
Before resubmitting, read every document against every other document. The protocol against the consent form. The consent form against the recruitment script. The procedures section against the data handling plan. The risk level claim against what the instruments actually collect.
Internal contradiction is the most common reason a packet returns a second time, and it arises innocently: the documents were written at different times, often in different sittings, and nobody read them side by side at the end. That final cross-check takes half an hour and routinely saves weeks.
It is also the first thing we do with any packet, whether we drafted it or the candidate did, because it produces a concrete list of disagreements faster than any other review. The stage-by-stage account of the whole process is at the IRB entry, and drafting from a blank form is covered at the protocol writing page.
005Protect the timeline while you wait.
Board cycles are the part of a doctoral timeline candidates most consistently underestimate, and the waiting is not empty time. Site permission letters, instrument finalization, committee scheduling and literature updates can all proceed while a submission sits with a board, and doing them now removes them from the critical path later.
The other thing worth protecting is alignment with your methodology chapter. Every revision the board requires changes your study slightly, and if the chapter is not updated in step, you will defend a design that differs from the one you were approved to run. That divergence is uncomfortable to discover in a defense and trivial to prevent by editing both documents in the same sitting.
Questions.
Is a revision request a bad sign?
No, it is the normal outcome of a first submission and most protocols receive one. What predicts trouble is not the first letter but the pattern afterwards: a second letter repeating points from the first usually means the response argued rather than complied, or corrected a single instance where a pattern was flagged.
How long should I take to respond?
Quickly, but not so quickly that the contradiction check gets skipped. A response sent in two days that generates another round costs far more than one sent in five that closes the file. The cross-check across all documents is the step worth the extra days, since it is what prevents the second cycle.
The letter uses language I do not understand. What do I do?
Ask the board office, which is a normal request and not a mark against you. Most boards have staff whose role includes explaining what a point is asking for. Board language is formulaic once you have read enough of it, and most points reduce to a missing attachment, a contradiction between documents, or an underspecified data handling detail.
Can I start collecting data while waiting for approval?
No. Collecting before approval is one of the few genuinely serious errors available in this process, and it can invalidate the data entirely, meaning the work has to be redone even if everything else was sound. Use the waiting time for the tasks that do not involve participants, of which there are usually several.
