A research topic mind map for your literature search

Sharpen the topic into a question

A topic is a container; a question is a constraint. Climate change gives you a library, while which policies reduce household emissions in large cities gives you a search you can run. Write the question as the root label in one sentence, with the population, place or period you are claiming about.

Two versions of the question usually help. Put the broad one on a child node and the narrow one on a second child, then write in the note why the narrow one won. Most dead ends in a search begin as a topic that was never sharpened, and the note is where you recover that reasoning later.

Make the boundaries explicit with child nodes for the population, the period, the geography and the kind of evidence you will accept. When a source falls outside a boundary, you can see immediately why it was excluded, which saves you from reading it twice and from arguing with yourself about whether it counts.

• The root states a question rather than a subject area
• Population, period and place are explicit child nodes
• One broad and one narrow version, so you can zoom in
• The note says why the narrow version won

Note: If the question cannot be answered with a finding, it is still a topic. Add a child node stating what evidence would change your mind, and the question usually sharpens itself.

List the search terms you will use

A search plan needs its vocabulary written down before you start, because you will not remember every synonym after the third database. Make one child node for each term group: the everyday phrase, the academic phrase, the abbreviation, the older term, and the word a neighboring discipline uses for the same idea.

Keep the strings exactly as you type them, including the quotes around a phrase and the boolean operators. A node labeled urban heat mitigation is a different search from a node labeled urban AND heat AND mitigation, so the note on the node should hold the exact string you ran.

Because the vocabulary changes as you learn the field, search the map often. Ctrl/Cmd+F while the map is active looks through labels and notes, so you can check that a term has not been used with two different spellings. A search returning both the old label and the new one is telling you the map has drifted, and the shortcut list has the key for it.

• One node per term group: plain, academic, abbreviation, older, neighboring field
• Exact search strings live in the note, quotes included
• Record which database each string was run in
• Re-run the map search after every term change

Note: No database is built into Canvasola, so the map records what you intend to run and what you have run. Running the search is still your job; the map just keeps the plan and the results in one place.

Set your inclusion rules

Inclusion rules are what keep a review honest, because they are the reason you kept some papers and dropped others. Put them on their own branch: peer reviewed or not, published within a date range, a specific method or population, and a quality threshold such as a minimum sample size.

Write each rule so you can apply it without thinking. A rule like good quality study will drift; more than 30 participants and published after 2015 will not. Put the rule in the label and any exceptions you have agreed with yourself in the note, so the exceptions are visible rather than invisible.

Add a sibling branch for exclusion rules, because the reasons you rejected a source are the ones you forget. Most of the awkwardness in a literature review comes from having quietly dropped a paper that does not fit, and a recorded exclusion rule turns that into a decision rather than a gap.

• Each rule is checkable against a specific source
• Exclusions are recorded as rules, not as private judgments
• The date range and the method threshold are written down
• Exceptions live in the note of the rule they break

Note: Rules you can apply in a minute are worth more than rules you can defend in a paragraph. If applying a rule takes judgment every time, it is too loose to be a rule.

Map what each source contributes

Add one node per source you have screened, including the ones you rejected. Label it with the citation and give it children: the question it addresses, the method, the sample, and the finding you are using or the reason you set it aside. A rejected source with a recorded reason stops you finding it again next month.

Use cross-links to connect each source to the question it bears on. A source linked to two questions is doing more work than the rest and deserves a close reading. A source with no link is either off-topic or you have not decided yet. The cross-links guide covers the mechanics.

The reading itself goes in the note, not in the label. Notes hold the quotes, the page numbers and the caveats you want to cite later, and they are searchable with Ctrl/Cmd+F while the map is active, so a phrase you half remember can be found by its words rather than by its source.

• One node per source, including the rejected ones
• Children cover question, method, sample and finding
• Cross-links show which sources serve which question
• Quotes and page numbers go in the note

Note: Node images are for figures you want to keep in view, up to 1.5 MB each, and a freehand drawing saved into a thumbnail slot works for a hand-sketched version of a study model. Neither replaces reading the paper.

Spot the overlap in the literature

Overlap is the signal you are looking for. When several sources attach to the same question node, you have a cluster of findings, and a cluster from one group of authors is a finding of its own: the field agrees, or one group is driving it. Both are worth a sentence in your write-up.

Count the children under each question node and look for the thin ones. A question with one source is not settled, it is barely discussed, and that is often the most useful gap on the map. A question with no source at all is a hole you can name in a review.

Layers help once the literature gets large. Put one layer for sources you have read, one for those still on the list, and one for your own synthesis, then use the layer tabs to scope the map to the set you are working on. The stack holds up to 8 layers.

• Children per question node show where the field agrees
• A question with one source is thin, not settled
• Layers separate read, unread and your own synthesis
• The 3D stack shows the sets as separate planes

Note: Deleting a layer keeps the nodes on it, so a layer is a way of scoping the view rather than a way of deleting work. Nothing on the map is lost when you tidy the stack.

Decide what to read in full

The map should end with a reading list you could hand to someone else. Look at each source node and ask whether the abstract was enough. Where it was not, the node's note should say what you expect to find, so the reading has a purpose and you know what would make the source worth less than expected.

Export the plan so it can live elsewhere. OPML 2.0 export carries the tree into other outlining tools, and the plain-text outline export gives you a list you can paste into a reference manager. Markdown export is the option for a document you will write against later.

Keep the map as the living plan rather than a snapshot. Each new source arrives as a node with the same four children, which keeps the shape comparable. The thesis guide covers how the same root carries a longer project.

• Every read-in-full decision is recorded on the node
• Each node states what the full text is meant to settle
• OPML 2.0 for other tools, outline text for a reference manager
• The map is updated as sources arrive, not rebuilt

Note: A search plan cannot tell you what the literature says, only what you looked at and what you decided. That is exactly why it is worth keeping separate from the summary you will write.

A checklist you can reuse

• The root states the question the search must answer
• Population, period and place are written down as child nodes
• Search terms are recorded as exact strings in the notes
• Inclusion and exclusion rules can each be applied in under a minute
• Every screened source has a node, including the rejected ones
• Sources are cross-linked to the questions they bear on
• Layers separate sources read, sources pending and your synthesis
• Each node states what reading it in full is meant to settle

Common mistakes to avoid

• Searching before sharpening. A search run on a topic returns a pile of papers and no way to know when to stop. Sharpen the question first, then run the search.
• Storing the summary on the map. A map full of summaries stops being a search plan. Keep findings in the notes attached to the source node, not in a parallel branch.
• No exclusion rules. Rejected sources disappear into memory, and next month you meet the same paper again. Write the rule that rejected them.
• One node per author. Authors publish many papers, and a node per author flattens what each paper contributes. Make the node per source.
• Reading a thin branch as settled. One supporting paper does not settle a question. Check the child count before you describe a branch as a consensus.

Frequently asked questions

How many search terms do I need?

Five groups is usually enough: the everyday phrase, the academic phrase, an abbreviation, an older term, and the word a neighboring discipline uses. Each group is run separately in each database, because a term that returns pages in one index often returns nothing in another.

Can I import the sources I already have?

You can import Markdown and OPML 2.0 files into a map, and export the map as OPML 2.0 or as a plain-text outline. Most reference managers export a bibliography rather than a tree, so the usual route is to import an outline you have shaped by hand and then add source nodes to it.

Should every source get its own node?

Every source you screened, including the ones you set aside. A rejected source with a recorded reason takes seconds to add and saves a search later. Sources you never screened do not need a node until you have looked at them.

Is the map a substitute for a proper literature review?

No. The map is a plan, a record of what you screened, and a record of why you kept or dropped things. The synthesis still has to be written, and the map cannot tell you whether a paper's method is sound. That judgment stays with you.

This guide is part of the mind map use cases collection. For the wider product tour, read the complete Canvasola guide, and keep the keyboard shortcut cheat sheet open while you work.