Readiness checks before you consider a unit finished: (1) Given a stem that describes a sampling procedure in plain words, you can name the technique without keyword hints. (2) You can state, in one sentence each, what evidence would support and what would undermine a validity claim versus a reliability claim. (3) Given a researcher's decision, you can classify a Type I or Type II error from the decision itself, not from memorised labels. (4) You convert absolute figures to rates before comparing groups in any data interpretation item. (5) You identify the logical form of an argument before judging whether its conclusion feels plausible. (6) Your error ledger groups every miss under a named concept, and no concept has three or more recent uncorrected misses. Treat these as learning milestones for your own tracking, not as predictions of any score or outcome. For administrative details, mode and the subject-wise syllabus, rely on the NTA examination portal linked at the end of this guide.
Which paper tests what: Paper I general units and your Paper II syllabus
UGC-NET has two parts: Paper I tests general units including research aptitude, reasoning, data interpretation, ICT and communication, while Paper II tests only your chosen subject's syllabus. Outcomes depend on aggregate performance across both parts.
The NTA states that eligibility for the Junior Research Fellowship, appointment as Assistant Professor, and admission to Ph.D. depends on aggregate performance in Part I and Part II, and that the qualification categories carry different entitlements: qualifying under the Ph.D.-admission-only category does not confer JRF or Assistant Professor eligibility. Plan preparation for both parts rather than treating either as the whole test.
The connection between the parts is practical rather than structural. Paper II rewards command of your discipline's own research tradition, while Paper I's methodological concepts appear in neutral framing. Build a two-column glossary: the neutral concept on one side, its typical rendering in your subject — respondents, texts, firms, specimens, sites — on the other. Each week, write five practice stems in your subject's nouns that each use one concept, then answer them a day later. The expected observation is that recognition speeds up when the concept arrives wrapped in familiar subject language, and the same glossary supports both papers.
Sampling: naming the technique from the procedure, not a keyword
Identify the population, the frame, and how units were actually selected before naming the technique. Stems describe the selection procedure in words; the correct sampling name follows from that description, not from a single trigger word.
Probability designs give every population member a known chance of selection: simple random, systematic, stratified, cluster, and multistage. Non-probability designs do not: convenience, purposive or judgemental, quota, and snowball. The dividing question is whether selection involved randomisation over a defined frame. Stratified sampling divides the population into subgroups and then randomly selects from every subgroup; cluster sampling randomly selects whole groups and often studies all or a sample of members inside the chosen groups only.
Worked scenario: a stem describes a researcher who divides a college's students into Science, Arts and Commerce groups and then selects a random sample from each group in proportion to its size. A plausible mistake is to see the word 'groups' and choose cluster sampling. The better decision is stratified sampling, because members are drawn from every group and the randomisation happens within groups, not at the level of choosing entire groups. The distinction matters because the two designs differ in what they represent: stratification guarantees coverage of each subgroup, while clustering limits coverage to the groups actually selected, which changes what conclusions the sample can support.
| What the stem describes | Likely technique | Key check |
|---|---|---|
| Names drawn randomly from a full list of the population | Simple random sampling | Every member had an equal, known chance |
| Population split into subgroups; random selection from every subgroup | Stratified sampling | All strata represented; randomisation within each |
| Every k-th unit chosen from an ordered list | Systematic sampling | Fixed interval over a frame |
| Whole groups randomly chosen; members studied within chosen groups | Cluster sampling | Randomisation at the group level, not member level |
| Participants recruit further participants through contacts | Snowball sampling | No frame; chain referral, so non-probability |
| Researcher hand-picks information-rich cases on purpose | Purposive sampling | Selection by judgement, not chance |
Validity and reliability are two different claims — test them separately
Validity asks whether an instrument measures the intended construct; reliability asks whether it measures consistently. A stem can describe a highly consistent instrument that misses its target, so judge each property on its own distinct evidence.
The named subtypes point to different evidence. Content validity asks whether items cover the full domain, judged by expert review of coverage. Criterion validity, concurrent or predictive, asks whether scores track an external benchmark now or later. Construct validity asks whether the instrument behaves as the underlying theory of the trait predicts. Reliability subtypes are equally distinct: test-retest checks stability across occasions, split-half and internal-consistency approaches check coherence among items, and inter-rater checks agreement between scorers.
Exercise: take one instrument actually used in your field, such as an attitude scale, a rubric, or a diagnostic test. In two columns, list what evidence you would seek for each validity subtype and each reliability subtype. Expected observations: every reliability entry involves consistency across time, items, or raters, while every validity entry involves either an external criterion, expert domain judgement, or theoretical behaviour of scores. Self-check rubric: you pass when you can state, for any subtype, one sentence describing what finding would seriously undermine that claim — for example, items sampling only half the syllabus undermine content validity.
Hypotheses and error types: read what the null actually claims
The null hypothesis states no effect or no difference; the alternative states the effect the researcher expects. A Type I error rejects a true null; a Type II error fails to reject a false one. Classify from the decision taken.
Learn the paired terms as a set. A research hypothesis expresses the substantive expectation; the statistical hypothesis pair translates it into testable form. Directional hypotheses specify a direction of effect, non-directional hypotheses only a difference. These labels are independent of the error labels: a one-tailed test can still produce either error type, so never infer the error type from the hypothesis style.
Mini scenario: a researcher concludes that a new tutorial format improved results, but in truth the format made no difference. The researcher rejected a null that was actually true, which is a Type I error, a false positive. A common mistake is to reverse the labels by remembering a slogan such as 'Type I is the serious one', which is context-dependent and unreliable. The better habit is mechanical: write down the decision (reject or fail to reject), the truth (null true or false), and read the error off the resulting two-by-two cell. A stem may present the truth first and the decision second, so only the mechanical mapping gets both orders right.
Data interpretation: convert and calculate before accepting an inference
Data interpretation items supply a table, chart, or case and ask what follows. Compute percentage change, shares, and averages from the figures themselves, then test each stated inference against those calculations before accepting it.
The recurring traps have names. Percentage change differs from percentage-point change, and comparing them produces wrong rankings. Averages hide distribution, so a rising mean can coexist with falling outcomes for most units. Correlation between two series never establishes that one caused the other. Rates and absolute counts answer different questions, and weighted averages differ from simple averages whenever group sizes differ. Decide which of these the stem is actually asking about before touching the options.
Worked scenario: Institution A has 400 research students of whom 40 publish during a period; Institution B has 100 research students of whom 15 publish. A plausible mistake is to pick 'Institution A shows stronger research output' because 40 exceeds 15. The better decision is to convert to rates: 40 of 400 is 10 per cent, while 15 of 100 is 15 per cent, so B's output rate is higher even though its count is lower. This matters because the question asks what the data supports, and the absolute-to-rate conversion reverses the ranking. A further check: nothing in the table shows why the rates differ, so any option attributing a cause overreaches the evidence.
- Percentage change vs percentage-point change: compute the base before comparing
- Rate vs absolute number: divide before ranking groups of different sizes
- Weighted vs simple average: check whether group sizes are equal before averaging means
- Correlation vs causation: a co-movement supports only a co-movement claim
- Averages vs distribution: a mean can rise while most individual cases fall
- Chart reading: restate axis units, note the baseline year for any growth claim, and recompute one figure by hand to confirm the scale
Reasoning items: judge the argument's form, not its plausibility
Reasoning questions are decided by argument structure, not by whether the conclusion sounds sensible. Identify premises, conclusion, and connectives first, then check validity for deductive arguments or strength for inductive ones.
Know the named forms. In a valid conditional argument, modus ponens affirms the antecedent (if P then Q; P; therefore Q) and modus tollens denies the consequent (if P then Q; not Q; therefore not P). Two look-alikes are invalid: affirming the consequent (if P then Q; Q; therefore P) and denying the antecedent (if P then Q; not P; therefore not Q). On the inductive side, watch for hasty generalisation from a thin sample, circular reasoning where the conclusion restates a premise, and attacks on the arguer rather than the argument.
Worked scenario: a stem argues that if an instrument is reliable it yields consistent results; this instrument yields consistent results; therefore it is reliable. A plausible mistake is to accept the conclusion because reliability and consistency genuinely belong together, so the argument feels right. The better decision is to flag affirming the consequent: consistency is what reliability would bring, but consistent results can arise for other reasons, so the conclusion does not follow deductively. This matters because subject knowledge actively interferes here; the true premise pair makes the invalid form harder to see, and only form analysis protects you.
ICT and communication concepts, plus a weekly cycle that ties both parts together
ICT items test what tools do for research and teaching, and communication items test how sender, message, channel, receiver and feedback interact. Both reward application to your subject; run a weekly cycle that drills that application directly.
Anchor the named concepts. On the ICT side, distinguish synchronous from asynchronous learning, know what learning management systems, MOOC platforms and digital repositories are for, and understand tools for data handling and presentation. On the communication side, trace the classic elements of sender, message, encoding, channel, decoding, receiver and feedback, and classify barriers as physical, semantic or psychological. A semantic barrier, for instance, fits a stem showing two groups using the same term differently — a situation worth mapping explicitly onto your own discipline's terminology as practice.
Adaptable weekly sequence, adjustable to the weeks you have. Days one and two: build the two-column glossary for one unit and add two subject-specific examples per concept. Day three: attempt ten stems from your subject and map each to a named concept in writing. Day four: one data interpretation set with every conversion shown by hand. Day five: a drill mixing error-type classification with fallacy identification. Day six: a timed mixed set under exam-like conditions on a computer. Day seven: update an error ledger that records each miss under a named concept, and reattempt last week's ledgered items from memory. Repeat the cycle per unit; the ledger, not the calendar, tells you when a unit is done.
References and further reading
Use these references to explore the concepts and check the latest information from the relevant organizations.