Critical Summary and Paraphrasing
The whole reading-to-writing problem is compressed into a single cell of a five-row table. Unpacked, it holds nine clauses, eight distinct instructions, and one instruction that appears to contradict another.
Nine clauses in one table cell
The supplied teaching source sets out a five-step process for developing a literature review, and gives each step a one-line task description. Four of those descriptions are a sentence or two. The fifth is not. It carries the entire craft of reading critically and writing accurately, and it is worth reading slowly.
Nine clauses, but not nine separate jobs. Two pairs overlap in ways worth noticing before you try to act on the list.
"Identify main ideas" is a heading with two branches under it: take the main idea when the author states it, and work it out when the author does not. Those are different skills with different failure modes, and treating them as one instruction is how the second gets skipped.
"Combine similar and related ideas" and "combine similar or related details into categories and provide a label or heading" are the same operation stated twice, the second time with an extra requirement attached. The requirement is the label. Combining without naming the result leaves you with a bigger pile, not a category. The five-step process this instruction sits inside is covered separately on the five-step literature review development process.
THE EIGHT INSTRUCTIONS, AND WHAT EACH ONE PRODUCES
| Instruction, as stated | What it asks of you | What you should be holding afterwards |
|---|---|---|
| Write a critical review for each article you read | One evaluative note per item, not one note per reading session | A per-item record you can retrieve without reopening the article |
| Combine similar and related ideas | Merging across items, not within them | Clusters of claims that several works are making at once |
| Combine related details into categories and label them | Naming each cluster in your own words | A labelled category that can become a paragraph or a section |
| Reflect the author's emphasis | Preserving proportion — what the work leads on, and what it treats as an aside | A summary whose weighting matches the original's weighting |
| Identify main ideas | Deciding what the work is actually claiming | A one-sentence statement of the work's central claim |
| Select the main idea when it is explicitly stated | Locating the author's own statement of it and using that | A claim you can point to in the text |
| Determine the main ideas when the author is not explicit | Inferring the claim from structure, emphasis and conclusions | An inferred claim, recorded as inferred |
| Paraphrase accurately, and exclude personal opinions | Restating without distorting, and separating judgement from preference | Text you can defend line by line against the original |
Column one reproduces the source's clauses. Columns two and three are synthesis added here; the source states the instructions without elaborating on any of them.
A critical review for each article, not for each sitting
The unit is the article. That is a stronger instruction than it looks, because the natural habit is to read three or four items in one session and write one set of impressions afterwards. What survives that is a blur, and the individual works become irrecoverable a month later.
"Critical" here carries the meaning the same source gives it elsewhere: a critical review shows how prevailing ideas fit into your own research topic, and how your topic or theory agrees or differs from them. Criticism is a relational operation. It is not fault-finding, and a note that only reports what an article says has not performed it.
Three things a critical note must do that a summary does not
Place the work against the others
State which existing position the work supports, extends or contradicts. A work that sits alone in your notes has not been reviewed critically; it has been described.
Say how well the claim is supported
Note the evidence the claim rests on and the conditions the study ran under. A finding from one organisation and a finding from a whole population are not the same kind of finding.
State the relevance to your own topic
Agrees with, differs from, or bears on your question in some specific way. Without this line the note is a general summary of somebody else's work.
One note per article also protects the later stage. Combining is only possible across records that exist separately; you cannot merge four claims you never wrote down individually. The questions worth asking of each item before you write the note are covered on evaluating an article before you cite it.
Combining, and the label that finishes the job
Combining is the operation that turns a reading log into a review. The source states it twice and the second statement is the operative one, because it adds the requirement to "provide a label or heading for ideas".
The label is the test. If four works can be gathered under a name you can write down, you have found a theme in the literature. If they cannot, you have found four works that happen to be adjacent, and forcing them into one paragraph will produce the list-of-articles paragraph the source elsewhere warns against — see what a literature review is not.
From per-article notes to a labelled category
Strip each note back to its claims
One line per claim, with the source attached. A single article usually yields two or three claims worth carrying, not ten.
Sort the claims, ignoring which article they came from
Sorting by author reproduces your reading order. Sorting by claim reveals where the field converges and where it splits.
Name each pile in your own words
The name has to be a statement about the field, not a topic word. "Handover timing" is a topic. "Handover is planned late and staffed from the delivery team" is a category.
Split any pile that resists a single name
A pile that needs a name with "and" in it is usually two categories. Splitting now is cheaper than discovering it mid-paragraph.
Record which claims disagree, and on what
Disagreement inside a category is the most useful thing in it. The source's own review sequence asks you to point out comparisons and dissonances, not just agreements.
Reflecting the author's emphasis
This is a proportionality requirement and it is easy to fail while writing nothing false. If a study spends most of its length on one relationship and one paragraph on a secondary observation, a summary that gives them a sentence each has misreported the work.
The distortion usually runs one way. You are reading for your own topic, so the passage that speaks to your question feels central even when the author treated it as an aside. That is the moment to check where the emphasis actually sits.
WHERE AUTHORS SIGNAL EMPHASIS
| Signal | What to look at | What it tells you |
|---|---|---|
| The abstract | Which finding is named first and which is named at all | The author's own ranking of their results |
| Proportion of length | How many pages each theme or result receives | Where the argument's weight was placed |
| Repetition across structural points | A purpose or claim restated in abstract, introduction and conclusion | The spine of the paper; in some works it does the job research questions would otherwise do |
| Position in the conclusion | What is carried forward and what is dropped | What the author thinks the work established |
| Hedging language | "May", "can be considered", "suggests", "was not examined" | The author's own confidence level, which your paraphrase must preserve |
General reading guidance, consistent with the source's instruction to reflect the author's emphasis. The source names the requirement without naming these signals.
Main ideas, stated and unstated
The source splits this task cleanly. When the author states the main idea, select it. When the author does not, determine it. The second is the harder job and the source offers no method for it.
When the claim is explicit
- Use the author's own statement rather than your reconstruction of it
- Record where it appears, so you can return to it
- Check that the abstract, the introduction and the conclusion state the same claim
- Where they differ, note the difference — it is a finding about the work
- The risk here is low: you are quoting a position, not building one
When the claim is not explicit
- Work from the conclusions backwards — what did the author think was established?
- Use proportion: the theme that consumed the most space is usually the claim
- Look at what the design was built to detect; a method reveals an intention
- Write the inferred claim as a sentence and mark it as your inference
- The risk here is high: an inferred claim can be attributed to an author who never made it
Marking the inference is the safeguard, and it costs one word. "The paper argues" and "the paper appears to treat X as its central concern" are different claims about the work, and only one of them is checkable against the text.
Works with no stated research question make this common rather than exceptional. Where a study expresses its purpose only in prose, no finding can be cleanly said to answer a question, and the main idea has to be reconstructed by the reader. That structural pattern is examined on where research questions and hypotheses live.
Paraphrasing accurately
Accuracy has two components that are usually treated as one. The paraphrase must not reproduce the original's wording, and it must not change the original's claim. Most attention goes to the first. The second is where the damage is done.
FIVE WAYS AN ACCURATE-LOOKING PARAPHRASE GOES WRONG
| Failure | What it looks like | Why it matters |
|---|---|---|
| Too close to the original | Synonyms substituted into the author's sentence structure | It is the author's sentence with different words in it; quote it instead, or restate the claim from your own note |
| Scope inflation | A finding from one organisation restated as a finding about the sector | Changes what the evidence supports. The commonest and least visible of the five |
| Hedge stripping | "Can be considered to exist" becomes "exists" | Transfers the author's uncertainty into your certainty, and your reader has no way to detect it |
| Direction added | An observed association restated as one thing causing another | Attributes a causal claim to an author who reported an association |
| Threshold laundering | A value the authors chose and described as having no scientific basis, restated as a standard | Converts a study-specific selection into a benchmark that does not exist |
The first failure is about wording; the other four are about meaning. The source's instruction is "paraphrase accurately" and does not distinguish between them.
The last two rows are not hypothetical. Among the published works read structurally for this library, one paper states plainly that there is no systematic procedure for setting the critical thresholds it uses, selects values by scaling them from an earlier study, and then lists that decision in its limitations. A paraphrase that reports those values without the caveat would manufacture a standard the paper explicitly denies having.
Another states that a relationship "can be considered" to exist while conceding in the same paragraph that it was not quantified. Both sentences belong in a faithful paraphrase. Only the first usually survives.
The contradiction: exclude personal opinions, write a critical review
These two instructions sit six clauses apart in the same cell. Taken literally they cannot both be followed, because a critical review is an evaluative act and evaluation is a judgement made by a person.
The resolution is not to soften either instruction. It is to notice that they are about different things. "Exclude personal opinions" rules out preference. "Write a critical review" requires judgement. A judgement is a claim about the work that is answerable to criteria and to evidence in the text; a preference is not answerable to anything.
Opinion — excluded
- "This is a weak paper" with nothing attached
- "I found this unconvincing"
- "The author clearly does not understand delivery practice"
- Approval or disapproval of the author's position rather than their argument
- Anything that would change if you were in a different mood
Judgement — required
- "The claim is drawn from two projects in one organisation and is stated as a general finding"
- "The instrument is described but not reproduced, so the result cannot be replicated"
- "The stated aim and the analysis address different constructs"
- "This contradicts the position taken by the works in the previous category, on this specific point"
- Anything a second reader could check against the article and agree or disagree with
The practical rule that falls out of this: every evaluative sentence you write about a work should be followed, or preceded, by the thing in the work that warrants it. If nothing in the article can be pointed at, the sentence is an opinion and the source's instruction removes it.
A worked pass over one theme
Four articles have been read and noted individually. Reduced to claims, and with the article identities set aside, they look like this.
- Note 1 — handover planning begins after commissioning starts; treated as a scheduling observation, stated in one paragraph.
- Note 2 — operations staff join the project late; the study's central finding, occupying most of its length.
- Note 3 — handover documentation is produced to satisfy contract closure rather than operational need; a strong claim, evidenced from a single site.
- Note 4 — operational readiness is not measured before handover; the author states this as an aside while arguing about something else.
Sorted by claim rather than by article, notes 1, 2 and 4 cluster: all three describe operations being brought into the project late, in different registers. Note 3 does not belong with them — it is about the purpose of the documentation, not the timing of involvement.
So the pile splits into two labelled categories. The first: operational involvement is late and its lateness is rarely measured. The second: handover artefacts are designed for contractual rather than operational purposes. Both labels are statements about the field, which is what makes them usable as paragraph spines.
Emphasis now has to be preserved inside the first category. Note 2 is a central finding; notes 1 and 4 are asides in their own works. The paragraph should say so — one work makes this its argument, two report it in passing — rather than presenting three equal pieces of support. That distinction survives into the review as a claim about how well established the theme is.
Judgement, not opinion, then attaches to note 3: its claim is strong and its evidence is a single site. That is checkable, it is not a preference, and it belongs in the paragraph. How categories like these become paragraphs is taken up on writing the literature review.
Before a note becomes a paragraph
The eight instructions as a check
- Every item you intend to cite has its own note, written while the article was in front of you
- Each note states what the work claims, what supports the claim, and how it bears on your topic
- Claims have been sorted independently of the articles they came from
- Every cluster has a label that is a statement about the field, not a topic word
- The weighting in your notes matches the weighting in the originals
- Every main idea is either quoted from the author or marked as your inference
- Hedges, scope conditions and study-specific values have survived the paraphrase intact
- Every evaluative sentence has something in the article behind it
One further boundary worth stating. A per-article critical note is close in form to an annotation, and the two are easy to conflate. An annotation is a deliverable written for a reader; a critical note is working material written for you. The deliverable form is covered on the annotated bibliography.
What to carry forward
- One critical note per article, written while the article is open. Combining is only possible across records that exist separately.
- Combining is not finished until the cluster has a label, and the label has to be a statement about the field rather than a topic word.
- Preserve proportion. A summary that promotes an aside to a finding is inaccurate even when every sentence in it is true.
- Take the main idea from the author when it is stated; mark it as your inference when it is not.
- Accuracy in paraphrasing is mostly about scope, hedges, direction and study-specific values — not about wording.
- Exclude preference, keep judgement. Every evaluative sentence needs something in the article standing behind it.
Frequently asked questions
How is a critical summary different from a summary?
A summary reports what a work says. A critical summary also places it — showing how the work's ideas fit your research topic and where your topic or theory agrees or differs from them, and weighing how well the claim is supported. The source's instruction is to write a critical review for each article, so a note that only describes the article has not met it.
How can I write a critical review and still exclude personal opinions?
By separating preference from judgement. A judgement is a claim about the work that a second reader could check against the text — the sample was drawn from one organisation, the instrument is not reproduced, the stated aim and the analysis address different constructs. A preference is not answerable to anything in the article. The source's instruction removes the second, not the first.
What do I do when an author never states their main idea?
The source tells you to determine it and offers no method. In practice, work backwards from the conclusions, use proportion of length as a guide to what the work is really about, and read the design for what it was built to detect. Whatever you arrive at, write it down as your inference rather than as the author's claim.
How many categories should my combined notes produce?
The source sets no number. The useful constraint is the label: keep splitting until every cluster can be named in a single statement about the field without an "and" in it. That usually produces fewer categories than the number of articles and more than a handful.
Is it a paraphrase problem if I keep the author's numbers?
Keeping the numbers is fine; keeping them without their conditions is not. A value drawn from two projects in one organisation, or a threshold the authors selected and described as having no systematic basis, must travel with that context. Stripping it converts a study-specific finding into a standard that nobody set.
Should my critical notes go into the review as written?
No. Notes are per-article and the review is per-theme, so the notes have to be sorted and combined before they can be written up. A review built by pasting notes in reading order is the descriptive list of articles the same source warns against.
References and source attribution
- O'Leary, Z. 2017, The Essential Guide to Doing Your Research Project, 3rd ed., Sage Publications, London.
- Bryman, A. 2016, Social Research Methods, 5th ed., Oxford University Press, Oxford.
- Naoum, S. G. 2013, Dissertation Research & Writing for Construction Students, 3rd ed., Routledge.
- Trochim, W. M. K. 2006, Research Methods Knowledge Base, http://www.socialresearchmethods.net/kb/probform.php.
- The supplied teaching source: master's-level teaching notes and accompanying slides on reviewing the literature. All statements attributed to "the source" are drawn from that material, and the five-step development process including step five is quoted from its summary table.
- Structural observations about paraphrasing risk draw on five examined and published research works in project management and related fields, read for this library and referred to as Exemplars A to E. Their content is described structurally only.
Suggested questions for Ask KEVOS
- Turn my per-article notes into labelled categories and tell me which claims do not fit.
- Check this paraphrase against the original for scope inflation and stripped hedges.
- Which sentences in my draft are opinions rather than judgements?
- Help me infer the main idea of an article that never states one.
- Does my summary reflect the author's emphasis, or the part of the article I needed?