The test the chapter has to pass
Give the draft to a colleague from another discipline and ask, once they have finished, for one sentence naming what nobody yet knows. If they can, the chapter works. If they can only tell you which authors you read, you have written an annotated bibliography in continuous prose, which is the single most common form of a returned second chapter.
Everything else follows from that test. Each source must be doing a job in the argument, and a source you cannot say the job of should come out however long it took to find. The chapter ends precisely where your research question begins, so that the question reads as the obvious next move rather than as a change of subject. A committee that reaches your question and thinks of course tends to approve the proposal that follows.
Record the search while you are searching
The search is a method, and methods are reported. Keeping the record afterward is impossible, since nobody remembers which combination of terms produced the article that turned out to matter. Open a second document on the first day and keep these entries in it.
- ◆Which databases were searched, and on what dates, since a search repeated three months later returns a different set.
- ◆The exact search strings, including the operators and any truncation, so the search can be run again by somebody else.
- ◆What you chose to include and exclude, and why: language, publication window, study design, setting, population.
- ◆How many records each stage produced and how many survived it. Committees ask for these counts and they cannot be reconstructed later.
- ◆Where each retained source came from, including the ones found by following a reference list rather than by search, which are often the best ones.
The matrix is the actual method
Build a table with one row for every source you keep and a fixed set of columns: the citation, the question that study asked, its design, who was in it and how many, the setting, how the outcome was measured, what was found, what the authors admit limits it, and one line on what it contributes to your own argument. Fill it as you read, never afterward.
The reason for the table is not tidiness. Prose is written down the columns, while a stack of article summaries can only be written across the rows, and writing across the rows is what produces the chapter that summarizes. Once thirty rows exist you can see, at a glance, that most of the field studied one population, that the two studies disagreeing with each other used different instruments, and that nobody has looked at your setting at all. Those observations are your paragraphs, and none of them is visible while reading one article at a time.
Appraise as you fill it. Note the sample and whether its size was justified, whether the design can carry the claim the abstract makes, and whether the instrument was ever validated in a population resembling yours. This is where a review earns the word critical, and it is the same reasoning you will apply to your own analysis later, tried out first on somebody else's data.
Writing by theme, in one shape
Every synthesis paragraph has the same four moves and they can be rehearsed until they are automatic. Begin by stating, in your own words, what the assembled literature collectively shows. Support it with the studies, cited in parentheses, described in a clause each rather than a sentence each. Then present the disagreement or the limit honestly, because a field where everyone agrees is a field you have read too narrowly. Close with what that leaves unsettled.
Note what is absent from that shape. No paragraph begins with an author's name, no paragraph handles one study only, and no paragraph reports a finding without evaluating it. Organize the chapter by these claims, and let the headings name the themes rather than the variables, since a heading that repeats your instrument list tells a reader nothing about the argument they are about to read.
The opening you are claiming, and keeping it current
Finish with a gap stated in one paragraph, and be exact about which kind it is. A population nobody has yet examined, a setting nobody has tried, a disagreement nobody has resolved, an instrument nobody has validated where you intend to use it: these are four different claims and they lead to four different studies. Vague gaps produce vague designs, and a committee will hear the vagueness in your methods chapter long before you do.
Then keep the thing alive. Re-run the saved searches before the proposal defense and again before the final version, because a review is a claim about the present and doctoral work takes years. Where a chapter has already come back twice, the difficulty is nearly always structural rather than industrious, and patrons at that stage bring it to a mentor who rebuilds it from the matrix outward rather than adding further sources to a chapter that cannot hold the ones it has.
Questions put to the house
How many sources should a review contain?
Enough to support the argument and no more, which is why programs rarely publish a number. What committees examine is coverage: whether the seminal work is present, whether recent work is represented, and whether the studies that disagree with your position have been included. A chapter with fewer sources, each doing visible work, reads better than a longer one padded for weight.
How recent do the sources need to be?
Recent enough that a reader believes you know the present state of the field, with older work included where it is foundational or where it is the origin of a measure you use. Some programs publish a window and some leave it to the chair. Where a window exists, apply it to the empirical studies rather than to theory, and say so in the chapter.
What is the difference between a review and a synthesis?
A review is the chapter; synthesis is the thing it must do. Summarizing reports each study in turn. Synthesizing groups studies by claim, compares their designs and findings, and reaches a judgment the individual papers do not state on their own. If your paragraphs open with authors' names rather than with claims, the chapter is still at the summarizing stage.
Should the search strategy appear in the chapter?
In most doctoral programs, yes, at least briefly: which databases, what terms, what was included and excluded, and roughly how many records survived each stage. It demonstrates that the review was systematic rather than convenient. Keep the full log separately for the appendix and for your own use, since committees frequently ask for it after the chapter has been read.
