Skip to content

Research Process, Hypotheses and Scientific Reasoning

CSIR NET Part A research aptitude: types of research, the scientific method, hypotheses, variables, study designs and valid scientific reasoning.

  • 20 min read
  • Updated
On this page · 14 sections
  1. Prerequisites
  2. The Hook: Pasteur’s Swan-Neck Flask
  3. What Research Is
  4. The Research Process
  5. Hypotheses
  6. Variables
  7. Scientific Reasoning
  8. Memory Tricks
  9. Don’t Confuse
  10. Common Mistakes
  11. Practice
  12. Quick Revision Table
  13. Key Takeaways
  14. What’s Next

Prerequisites

The Hook: Pasteur’s Swan-Neck Flask

What Research Is

Research is a systematic, documented and reproducible search for new knowledge or for a better answer to a defined question. Three words in that definition do the work:

  • Systematic: it follows a plan decided before the data come in.
  • Documented: someone else can see what was done, with what material, and how it was analysed.
  • Reproducible: another competent worker following the same method should get a compatible result.

A one-off observation (“this plate looks contaminated”) is not research. Recording contamination rates across 200 plates under two different hood-cleaning routines, with the analysis planned in advance, is.

Types of research

Part A questions name a study and ask what kind it is. The same study can be classified along several independent axes, so read the question for the axis it wants.

AxisTypesOne-line test
PurposeBasic (fundamental, pure) vs appliedIs the aim understanding, or solving a practical problem?
Action researchPractitioners study and change their own practice, in cycles
DepthExploratory → descriptive → explanatory (analytical, causal)Mapping an unknown area, describing what is, or explaining why?
DataQuantitative vs qualitative (and mixed methods)Numbers and statistics, or words, themes and meanings?
ControlExperimental vs observational (non-experimental)Did the researcher assign the treatment?
TimeCross-sectional vs longitudinalOne time point, or the same units followed over time?
SettingLaboratory vs fieldControlled conditions, or the natural setting?
LogicDeductive (theory → hypothesis → test) vs inductive (observations → pattern → theory)Which way does the reasoning run?

Experimental, quasi-experimental and observational studies

DesignWho decides the exposure?Random assignment?Example
True experiment (e.g. randomised controlled trial)ResearcherYesSeedlings randomly assigned to 0, 50 or 100 mM NaCl
Quasi-experimentResearcher, or an existing groupingNoTwo existing classes taught by two methods
Cohort studyNature or the subjects themselvesNoFollow smokers and non-smokers forward in time to see who develops a disease
Case–control studyNatureNoStart from people with and without a disease and look back at past exposure
Cross-sectional surveyNatureNoMeasure exposure and outcome together at one time
Case report / case seriesNatureNoA description of one or a few unusual cases

Only random assignment by the researcher breaks the link between the treatment and everything else that differs between groups. That is why an experiment can support a causal claim that an observational study, on its own, cannot.

The Research Process

Textbooks list the steps in slightly different orders and with different names, but the logic is always the same. A version you can reproduce in an exam:

graph TD
    A["1. Identify a research problem
broad area to a specific question"] --> B["2. Review the literature
what is known, what is not"] B --> C["3. State objectives and hypotheses
testable, specific"] C --> D["4. Design the study
variables, controls, sample, methods"] D --> E["5. Collect data
as planned, recorded in full"] E --> F["6. Analyse data
statistics planned in advance"] F --> G["7. Interpret and conclude
support or reject the hypothesis"] G --> H["8. Report and share
paper, thesis, data deposition"] H -.->|"new questions"| A

Points that questions test:

  • The literature review comes before the hypothesis, not after the results. It tells you whether the question has already been answered and which methods work.
  • The design, including the statistical test, is fixed before the data are collected. Choosing the test after seeing the data is a classic source of false positives (see Experimental Design, Sampling and Data Analysis).
  • The process is a cycle. A rejected hypothesis or an unexpected result is the start of the next question, not a failure to hide.

From problem to research question

A good research problem is narrowed until it can be answered with the resources at hand. A common acronym for a good objective is SMART: specific, measurable, achievable, relevant, time-bound.

Too broadResearchable
“How does salt affect plants?”“Does 100 mM NaCl reduce the root length of 7-day-old rice seedlings compared with 0 mM, under the same light and temperature?”
“Is the new drug good?”“Does drug X lower systolic blood pressure after 8 weeks more than placebo in adults with stage 1 hypertension?”

Hypotheses

A hypothesis is a tentative, testable statement about the relationship between two or more variables, made before the test. It is a proposed answer to the research question.

What makes a hypothesis good

PropertyMeaningFails if…
TestableSome observation could be made to check itIt refers to something that cannot be measured
FalsifiableSome possible result would show it to be wrongEvery possible outcome “fits” it
SpecificNames the variables and the direction of the effect“Salt affects plants in some way”
Consistent with what is knownDoes not ignore established facts without reasonIt contradicts well-tested results with no new evidence
Simple (parsimonious)Uses no more assumptions than neededSeveral extra assumptions are added to rescue it

Falsifiability is the criterion associated with Karl Popper: a scientific claim must forbid some outcomes. “This drug will either improve or not improve the condition” forbids nothing, so it cannot be tested.

Null and alternative hypotheses

Null hypothesis \( H_0 \)Alternative hypothesis \( H_1 \) or \( H_a \)
StatesNo effect, no difference, no associationThere is an effect, difference or association
ExampleMean root length at 100 mM NaCl = mean root length at 0 mMMean root length at 100 mM NaCl ≠ (or <) mean at 0 mM
Role in a testThe hypothesis the statistical test is set up to assessAccepted only if the data make \( H_0 \) untenable
  • A directional (one-tailed) alternative names the direction (“less than”); a non-directional (two-tailed) one says only “different”.
  • A test can reject \( H_0 \) or fail to reject it. Failing to reject is not proof that \( H_0 \) is true: the study may simply have been too small to detect the effect.

Hypothesis, theory, law

TermWhat it isExample
HypothesisA testable proposal for one question“Enzyme E is inhibited by Cu²⁺”
TheoryA well-tested, broad explanation that unifies many observations and makes predictionsCell theory; evolution by natural selection
LawA concise description (often mathematical) of a regularity, without by itself explaining whyMendel’s law of segregation; Boyle’s law
ModelA simplified representation used to reason or predictMichaelis–Menten model of enzyme kinetics

“It is only a theory” misuses the word. In science, a theory is the strongest kind of explanation available, not a guess.

Variables

A variable is any characteristic that can take different values. Naming the variables in a described experiment is one of the most common research-aptitude questions.

VariableDefinitionIn the NaCl–rice example
Independent (explanatory, predictor, treatment, factor)The one the researcher changes or uses to group the subjectsNaCl concentration
Dependent (response, outcome)The one measured to see the effectRoot length
Controlled (constant)Kept the same in all groups so they cannot explain a differenceLight, temperature, seed batch, age of seedlings
ConfoundingVaries along with the independent variable and also affects the dependent variable, so the two effects cannot be separatedNaCl trays kept on the warmer window shelf
ExtraneousAny variable other than the independent one that might affect the outcome; becomes a confounder if it differs between groupsSmall differences in watering
Mediating (intervening)Lies on the causal path: independent → mediator → dependentNaCl → reduced water uptake → shorter roots
ModeratingChanges the strength or direction of the effectThe effect of NaCl is larger in one rice variety than another

Operational definitions

An operational definition says exactly how a variable is measured. “Stress” is a concept; “plasma cortisol at 08:00 h, measured by ELISA” is an operational definition. Two studies that use different operational definitions of the same concept may disagree without either being wrong, and a question may hinge on spotting this.

Scientific Reasoning

Deduction, induction and abduction

ReasoningDirectionCertaintyExample
DeductiveGeneral premise → specific conclusionCertain, if the premises are true and the form is validAll mammals have hair; whales are mammals; so whales have hair
InductiveSpecific observations → general conclusionProbable only; a new case can overturn itEvery swan I have seen is white; so all swans are white
AbductiveObservation → best available explanationTentative; it generates a hypothesis to testThe broth turned cloudy only when dust could reach it; the best explanation is that dust carries microbes

Hypothesis testing combines them: abduction suggests the hypothesis, deduction derives a prediction from it (“if H is true, then the swan-neck broth stays clear”), and the experiment checks the prediction.

Valid and invalid argument forms

These two pairs appear in disguise in “which conclusion follows” questions.

FormStructureValid?
Modus ponens (affirming the antecedent)If P then Q. P. Therefore Q.Valid
Modus tollens (denying the consequent)If P then Q. Not Q. Therefore not P.Valid
Affirming the consequentIf P then Q. Q. Therefore P.Invalid
Denying the antecedentIf P then Q. Not P. Therefore not Q.Invalid

Correlation, causation and the criteria for a causal claim

An association (correlation) between X and Y can arise because:

  1. X causes Y.
  2. Y causes X (reverse causation).
  3. A third variable Z causes both (confounding).
  4. Chance, especially when many associations were examined.
  5. The way the subjects were selected (selection bias).

Randomised experiments rule out 2 and 3 by design (the researcher sets the exposure, and random assignment balances other causes between groups), and a significance test puts a number on 4. For observational data, epidemiologists use a set of considerations proposed by Austin Bradford Hill to judge whether an association is likely to be causal:

ConsiderationQuestion to ask
StrengthIs the association large?
ConsistencyIs it seen in different studies, places and populations?
SpecificityIs the exposure linked to one particular outcome?
TemporalityDoes the cause come before the effect? (the one usually treated as essential)
Biological gradientDoes more exposure give more effect (dose–response)?
PlausibilityIs there a credible mechanism?
CoherenceDoes it fit what is known about the disease or process?
ExperimentDoes removing or reducing the exposure reduce the effect?
AnalogyIs there a similar, accepted cause–effect relation?

Mnemonic for the list: see Memory Tricks below.

The burden of evidence

Not all evidence weighs the same. A commonly used ranking for questions about the effect of an intervention, from weakest to strongest:

graph BT
    A["Expert opinion, anecdote"] --> B["Case reports and case series"]
    B --> C["Cross-sectional surveys"]
    C --> D["Case-control studies"]
    D --> E["Cohort studies"]
    E --> F["Randomised controlled trials"]
    F --> G["Systematic reviews and meta-analyses
of good-quality trials"]

A well-run study lower down can still outweigh a badly run study higher up, so treat the pyramid as a default, not a law. In mechanistic laboratory biology the equivalent idea is converging lines of evidence: the same conclusion reached by genetics, biochemistry and imaging is stronger than any one of them alone.

Memory Tricks

Don’t Confuse

ThisvsThat
Basic research: aim is understandingApplied research: aim is a practical solution
Experimental: researcher assigns the treatmentObservational: researcher records exposures that occurred anyway
Cross-sectional: one time point, many subjectsLongitudinal: same subjects measured repeatedly over time
Cohort: from exposure forward to outcomeCase–control: from outcome back to exposure
Hypothesis: testable answer to one questionTheory: broad, repeatedly tested explanation
Fail to reject \( H_0 \)Accept (prove) \( H_0 \): never stated this way
Confounder: affects both the exposure and the outcomeMediator: lies on the causal path between them
Inductive: specific → general, probableDeductive: general → specific, certain if valid

Common Mistakes

Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper.

Level 1: Foundation

  1. In “the effect of light intensity on the rate of photosynthesis”, name the dependent variable.
  2. A researcher starts with 100 patients who have a disease and 100 who do not, and looks back at their smoking history. What type of study is this?
  3. Is “If P then Q; Q; therefore P” a valid argument form?
Answers
  1. Rate of photosynthesis. It is what is measured; light intensity is what is varied.
  2. Case–control study. It starts from the outcome and looks back at the exposure (retrospective).
  3. No. It is affirming the consequent. Q could be true for some other reason.

Level 2: Part A Style

Practice Q1. Which of the following is a testable, falsifiable hypothesis?

(a) Some plants may respond to music in ways science cannot yet detect. (b) Tomato seedlings grown under 16 h of light per day reach a greater height at 21 days than those grown under 8 h. (c) Life is best explained by a vital force. (d) Gene X has some role in development.

Answer: (b)

Option (b) names the independent variable (photoperiod), the dependent variable (height at 21 days) and the direction of the effect, so a result could contradict it.

(a) explicitly places its claim beyond detection. (c) makes no measurable prediction. (d) is so vague that any result fits it.

Practice Q2. A survey records the daily screen time and the sleep duration of 2,000 students on a single day. This is best described as a:

(a) randomised controlled trial (b) cohort study (c) cross-sectional study (d) case–control study

Answer: (c) cross-sectional study

Exposure (screen time) and outcome (sleep) are measured together at one time point, with no assignment by the researcher and no follow-up. It cannot show which came first, so it cannot by itself support a causal claim.

Practice Q3. Across 50 towns, the number of ice-cream vans is positively correlated with the number of drowning incidents in a summer month. The most reasonable interpretation is:

(a) ice cream causes drowning (b) drowning increases ice-cream sales (c) a third variable, such as hot weather bringing more people to the beach, is linked to both (d) the correlation must be a calculation error

Answer: (c)

Hot weather and crowded beaches raise both ice-cream demand and the number of swimmers, so the two counts rise together without either causing the other. This is confounding. (d) is wrong because a confounded correlation is still a real correlation.

Practice Q4. A researcher reasons: “If this compound is an inhibitor of enzyme E, adding it will lower the activity of E. Adding it did not lower the activity of E. Therefore it is not an inhibitor of E under these conditions.” This reasoning is:

(a) invalid, because it affirms the consequent (b) valid, an example of modus tollens (c) inductive, so the conclusion is only probable (d) invalid, because it denies the antecedent

Answer: (b)

“If P then Q; not Q; therefore not P” is modus tollens, a valid deductive form. The words “under these conditions” matter: the conclusion depends on the assumptions behind the test (the compound dissolved, the concentration was high enough, the assay worked). A failed prediction refutes the hypothesis together with those assumptions.

Level 3: Multi-Step (Part C depth)

Practice Q5. A study reports that children who were breastfed for longer score higher on a reading test at age 10. The authors conclude that longer breastfeeding improves reading ability. Consider the statements:

  1. Mothers’ education could be a confounder, if it influences both breastfeeding duration and children’s reading.
  2. Because breastfeeding comes before the reading test, temporality is satisfied.
  3. Temporality being satisfied is enough to establish causation.
  4. Randomising mothers to breastfeeding durations would remove confounding but raises ethical problems.

Which statements are correct?

(a) 1 and 2 only (b) 1, 2 and 4 only (c) 2 and 3 only (d) 1, 2, 3 and 4

Answer: (b) 1, 2 and 4 only
  • 1 is correct. A variable linked to both the exposure and the outcome is a potential confounder.
  • 2 is correct. The exposure precedes the outcome.
  • 3 is wrong. Temporality is necessary but not sufficient; confounding and chance remain.
  • 4 is correct. Random assignment breaks the link with confounders, but assigning infant feeding against a mother’s choice raises obvious ethical problems. That is why such questions are mostly studied observationally, with statistical adjustment for known confounders.

Practice Q6. Read the description and answer the two parts.

A student wants to know whether a fungal extract slows bacterial growth. She sets up four flasks of the same nutrient broth with the same inoculum: (A) broth + extract dissolved in ethanol, (B) broth + the same volume of ethanol only, (C) broth only, (D) broth + a known antibiotic. Growth is measured as optical density after 18 h at 37 °C.

(i) Identify the independent variable, the dependent variable and two controlled variables. (ii) What does each of flasks B, C and D add to the experiment?

Answer

(i) Independent: presence of the fungal extract. Dependent: optical density at 18 h (a measure of bacterial growth). Controlled: broth composition and volume, inoculum size, temperature, incubation time, ethanol volume.

(ii)

  • B (vehicle control): shows the effect of ethanol alone. Without it, a fall in growth in A could be due to the solvent, not the extract.
  • C (untreated or negative control): shows normal growth, so the culture and medium are working.
  • D (positive control): shows that the assay can detect inhibition. If D did not reduce growth, a negative result in A would be uninterpretable.

The meaningful comparison for the extract is A vs B, since they differ in only one thing. Controls are covered in more depth on the next page.

Practice Q7. Classify each study on two axes (purpose: basic or applied; design: experimental or observational).

  1. Mapping where a transcription factor binds in the genome of unmodified yeast cells.
  2. Randomly assigning farm plots to two irrigation schedules to find which gives the higher wheat yield.
  3. Recording the bird species seen at 30 lakes each month for five years.
Answer
  1. Basic, observational. The aim is understanding gene regulation, and nothing is assigned or manipulated in the cells described.
  2. Applied, experimental. The aim is a farming decision, and the researcher assigns the treatment at random.
  3. Basic (descriptive), observational and longitudinal. No treatment is assigned; the same lakes are followed over time.

Quick Revision Table

IdeaKey point
ResearchSystematic, documented, reproducible search for knowledge
Basic vs appliedUnderstanding vs practical problem; classify by stated aim
Experimental vs observationalWho assigns the exposure; only random assignment supports causation by design
Cohort vs case–controlExposure → outcome (forward) vs outcome → exposure (back)
StepsProblem → review → hypothesis → design → collect → analyse → interpret → report
Good hypothesisTestable, falsifiable, specific, parsimonious
\( H_0 \)No effect; you reject it or fail to reject it, never “prove” it
VariablesIndependent (changed), dependent (measured), controlled (constant), confounding (varies with both)
Valid formsModus ponens, modus tollens
Invalid formsAffirming the consequent, denying the antecedent
CausationTemporality is essential; consider confounding, reverse causation and chance

Key Takeaways

What’s Next