GAM · Research Strategies — Strategy & Tricks

📐 The rules of the game

Every Research Strategies item is a short passage about a study — a survey, an experiment, a batch of interviews — carrying three or four questions. The passage defines the terms it tests. Nobody wants your textbook recited; they want you to read a design carefully and refuse to claim more than it supports. That is the whole game.

The vocabulary, as the exam uses it. The tell column is how you spot each concept inside a passage in five seconds.

The rule that decides more items than any other: correlation is not causation. If nobody manipulated anything, “associated with” is the honest verdict, and any option saying “causes”, “improves” or “leads to” is reaching.

👀 Five ways the same passage gets asked

Passages repeat; stems rotate. Name the variant from the stem alone and you know which box of your map to read from.

Underneath sit three kinds of thinking: definition-application (rows 1 and 4 — reuse the passage's own words), which-design-fits (row 3), what-the-text-implies (rows 2 and 5 — how far the evidence reaches). The last pair carries most medium and hard items, and is where a confident wrong answer feels best. Give it ten extra seconds.

🧠 Map the study: five boxes, forty seconds

Stem and options first, then five boxes. About forty seconds, and every passage fills them — even the ones that pretend to be about theory.

1. Read the stem, skim the options, then the passage. Ten seconds. You learn what is at stake — a variable, a flaw, a design, a conclusion — and the passage becomes a search, not a story.
2. Who? Sample size and source. “200 March borrowers at one library” — note the narrowness; it decides what you may generalise.
3. What was manipulated? Whatever the researcher set (“assigned to 20 or 40 minutes”) is the IV. If nothing was set, the study is observational and causal options are already dead.
4. What was measured? The outcome — recall, marks, completion. That is the DV; conclusions must be about this measure, not a cousin of it.
5. How were groups formed? Random assignment, self-selection, or pre-existing groups (small vs large firms). Only the first supports “causes”.
6. What was compared? Group vs group, before vs after, subgroup vs subgroup. A conclusion that compares something the study never compared is wrong.

Now hold each option against the map. The right one describes it. Wrong ones add a box the passage never filled — a cause with no manipulation, a population with no sampling, a measure nobody took, or a fact that is true in the world but absent from the text.

✂️ Shortcuts that are actually safe

• No manipulation, no causation. Box 2 empty? Cross out every option with “causes”, “raises”, “leads to” or “proves”. That usually kills two before you have thought at all.
• The option that repeats the passage's definition wins. If the passage defines an inference as a conclusion drawn from evidence, the right option paraphrases that sentence. Item writers test reading, not improvisation.
• Extreme words are red flags. “Proves”, “always”, “all”, “never”, “permanently”, “must”. Social science rarely earns them. Read the option before you kill it — but expect to kill it.
• Fewest added assumptions. Between two survivors, prefer the one that needs nothing the passage didn't say: “72 of 100 respondents” beats “most students”.
• “Not supported” beats “disproved”. When a prediction fails, the correct option says it was not supported under these conditions. Options that bury the theory forever are wrong.

⚡ Speed: read the stem first, then hunt

Ninety seconds a question sounds tight until you notice passages are shared: a passage plus three questions is allowed four and a half minutes. Take under three and bank the rest.

When the stem says “dependent variable”, your eyes go straight to the measurement sentence and skip the recruitment paragraph entirely.

🪤 Traps the item writers love

• Reversed variables. An option names the outcome as the IV: “the number of words recalled was manipulated”. Set things come first in time, measured things last. If you can't say what was set, you haven't mapped.
• Sampling swapped for assignment. “Randomly selected” says nothing about which group they joined; a random sample that then self-selects into treatment still cannot support a causal claim.
• Generalising beyond the sample. Fifty nurses at one hospital do not become “health workers nationwide”. Match the population in the option to box 1, word for word.
• Reliability dressed as validity. A scale that gives the same wrong answer twice is reliable and invalid. A 0.9 test–retest correlation says nothing about measuring the right thing.
• The outside-knowledge trap. “Larger samples reduce error” is true, and usually irrelevant. The only question is what this passage supports; a textbook truth with no hook in the text is the most tempting distractor in the family.
• Subgroup used to erase the whole. A link that vanishes in one subgroup does not mean “no effect”; it means the relationship may depend on something. Say “depends”, never “nothing”.
• Stuck at 60 seconds? Stop weighing, start deleting: kill causal verbs the design can't support, then options naming people the study never sampled. Take the survivor with the softest verb — guesses are free.

🧩 Worked examples: easy → medium → hard

Run the five-box map, then the overclaim cross-out, and watch how few options survive.

🎯 Recognition drills: which concept is this?

Cover the answer, say the concept out loud, check. A few seconds per line — the goal is reflex, not reflection.

🧭 Guided walkthrough: one passage, three questions

Q1 — “What is the dependent variable?” Stem first: DV, so read box 3. Pass rate. “Whether a student had a mentor” is the grouping variable, not the outcome. 20 seconds.

Q2 — “Strongest objection to the report's claim?” Spot-the-flaw, so read box 4: self-selected. Volunteers may be more organised or better prepared already. The self-selection option wins; “sample too small” loses (600 is plenty), “pass rates are not quantitative” loses (they are percentages). 40 seconds.

Q3 — “Which conclusion is best supported?” Box 2 is empty, so cross out “raised”, “improves”, “causes”. The survivor says mentored first-years at this university were more likely to pass — association language, population pinned to box 1. 30 seconds.

• I read the stem and options before the passage
• I filled the five boxes once and reused them
• I crossed out causal options when box 2 was empty
• I matched the option's population to box 1
• I chose the survivor with the fewest added assumptions

📈 Progressive practice: untimed, by tier

Work each tier untimed until you can say why every wrong option is wrong, not just which is right. A miss you can explain beats a lucky hit.

Easy · identify variables and safe conclusions → 6 real easy questions · free, no login

Medium · confounds and design choice → 6 real medium questions · free, no login

Hard · measures, subgroups and theory revision → 6 real hard questions · free, no login

⏱️ Speed training: the clock goes on

Targets: 50 s easy, 70 s medium, 90 s hard — passage included. Beat the target with a mistake? Slow down. Miss the target with no mistakes? You are rereading. Change one thing at a time.

1. Round 1 — at target, accuracy first. Catch yourself reading a passage cold? Restart the item.
2. Round 2 — ten seconds under target. Something has to give; make it rereading, not mapping.
3. Round 3 — mixed tiers, back at target. Log where the seconds went: options, passage, or the last two survivors. Drill the slowest.

Easy · 6 items, 50 s each → 6 real easy questions, against the clock · free, no login

Medium · 6 items, 70 s each → 6 real medium questions, against the clock · free, no login

Hard · 6 items, 90 s each → 6 real hard questions, against the clock · free, no login

🏆 Exam application: spending the real 90 minutes

About 60 GAM questions in 90 minutes, these passages mixed among the numerical families. They carry no arithmetic, so they are the module's cheapest points — if you hold the pace.

1. Stem and options first, then one pass of the passage filling the five boxes.
2. Cross out overclaimers: causal verbs with nothing manipulated, populations never sampled, “always” and “proves”.
3. Take the cheap linked question first. Identify-the-variable is usually fastest; bank those seconds for the conclusion question on the same passage.
4. Trust the passage over your notes. If it defines a term differently from your textbook, the passage wins — the answer key was written from it.
5. Answer everything. No negative marking: a blank scores exactly like a wrong answer.
6. Flag and return. Still stuck at 90 seconds? Provisional answer, flag, go. Never let one passage steal three other questions.

Exam simulation · hard, 90 s each → 6 real hard questions, against the clock · free, no login

📋 Cheat sheet

• IV = what was set. DV = what was measured. Set comes before measured.
• Random sampling → who gets in → generalisation. Random assignment → which group → causation.
• No manipulation → no causal option can be right.
• Confound = groups differ in more than the IV. Self-selection is the usual one.
• Internal validity: can it claim cause? External: does it transfer? Construct: right thing measured?
• Reliable = consistent. Valid = accurate. A scale always 2 kg off is reliable.
• Operationalisation: the concept turned into a count.
• Cross-sectional = snapshot. Longitudinal = same people over time.
• Small and narrow samples limit generalisation; big samples never fix self-selection.
• Blinding hides the condition; the placebo is the inert stand-in.
• A hypothesis must be able to fail. “Not supported” ≠ “disproved”.
• Stem and options first. Map once. Locate, don't reread.
• Kill words: proves · always · all · never · nationwide · should.
• Prefer the option that paraphrases the passage's own definition.
• Match the option's population to the passage's sample, word for word.
• Subgroup differences mean “depends on”, never “no effect”.
• Fewest added assumptions wins the tie.
• Answer everything. Blank = wrong, and guesses are free.