Scientific Research – Definition, Types, Steps and Examples

Scientific research is a systematic, transparent, and evidence-based process used to answer questions, test explanations, describe phenomena, or solve problems. It involves defining a researchable problem, choosing an appropriate design, collecting and analysing data, interpreting uncertainty, and communicating methods and findings so that others can evaluate, reproduce, challenge, or extend the work.

Introduction

Scientific research helps people develop reliable explanations of the natural, social, behavioural, and technological world. It can be used to test whether a treatment works, investigate why a population is changing, describe how people experience an event, predict environmental conditions, or evaluate whether a new technology performs as intended.

Scientific research is sometimes described as a fixed sequence of observation, hypothesis, experiment, and conclusion. That sequence is useful for introducing experimental reasoning, but actual research is more varied. A study may use observations, interviews, archived records, computer simulations, mathematical models, laboratory experiments, clinical trials, or existing datasets.

The defining issue is not whether every researcher performs an experiment. It is whether the inquiry uses systematic, appropriate, transparent, and critically examinable methods to produce or evaluate knowledge.

Key Takeaways

• Scientific research is a structured form of inquiry based on evidence and reasoned analysis.
• The research question determines the appropriate design, data, and analytical method.
• Not every scientific study requires an experiment or formal hypothesis.
• One study rarely settles a major scientific question; confidence develops through cumulative evidence.
• Ethics, transparency, uncertainty, and methodological limitations are part of good research.
• Open science and responsible use of digital and AI tools increasingly shape modern research practice.

What Is Scientific Research?

Scientific research is an organised process for developing, testing, refining, or applying knowledge through systematic inquiry. Researchers formulate a question, select methods appropriate to that question, gather or examine evidence, analyse the evidence, and communicate conclusions together with their uncertainties and limitations.

The term covers more than laboratory experimentation. Scientific research may involve:

• Observing naturally occurring events.
• Measuring physical or behavioural characteristics.
• Comparing groups or conditions.
• Testing an intervention.
• Analysing existing records or datasets.
• Modelling complex systems.
• Synthesising results from previous studies.
• Exploring experiences and meanings using qualitative evidence.

A study is not scientific merely because it includes numbers, technical vocabulary, laboratory equipment, or statistical software. Its claims must follow from an appropriate and transparent process.

What Makes Research Scientific?

Research is scientific when its questions and claims are investigated through methods that other informed people can inspect and evaluate.

Six qualities are especially important:

1. Systematic inquiry: Decisions follow a reasoned plan rather than an improvised search for supporting information.
2. Evidence: Conclusions are connected to observations, measurements, records, texts, models, or other relevant forms of evidence.
3. Transparency: The researcher explains how evidence was selected, generated, analysed, and interpreted.
4. Critical testability: Claims are open to examination, challenge, comparison, or revision.
5. Logical coherence: Conclusions must be consistent with the evidence and reasoning presented.
6. Cumulative contribution: The study relates to existing knowledge and allows later researchers to confirm, refine, question, or extend its findings.

These features do not make science infallible. Scientific conclusions remain open to revision when better evidence, improved measurements, alternative explanations, or new analytical methods become available. NIH describes research as an evolving process in which changes in understanding can indicate that scientific scrutiny is working as intended.

Scientific Research vs. Research vs. Scientific Method

Concept	Meaning	Example
Research	A broad process of investigating a question or gathering and interpreting information. It may be academic, historical, legal, commercial, artistic, or scientific.	Examining historical letters to understand a political decision.
Scientific research	Systematic inquiry intended to create, test, refine, or apply knowledge using transparent and critically examinable methods.	Studying whether air pollution exposure predicts changes in lung function.
Scientific method	A general reasoning framework involving questions, evidence, explanations, tests, analysis, and revision. It is not one universal procedure for every discipline.	Developing a hypothesis about plant growth, testing it, and revising the explanation.
Research methodology	The overall logic and justification for the research approach, including assumptions, design, sampling, measurement, and analysis.	Explaining why a mixed-methods design is necessary to answer both outcome and experience questions.
Research methods	The specific techniques used to collect or analyse evidence.	Experiments, interviews, surveys, observations, regression, or thematic analysis.

Research methods are the practical procedures used in a study, whereas methodology explains the broader strategy and rationale for selecting those procedures.

Characteristics of Scientific Research

It begins with a researchable problem

A broad topic is not yet a research question.

• Broad topic: Social media and students
• Research problem: Uncertainty about whether late-night social-media use is associated with sleep quality
• Research question: Among first-year university students, is late-night social-media use associated with self-reported sleep quality during the examination period?

A researchable question specifies what will be examined and helps determine what evidence is needed.

It builds on previous knowledge

Researchers review relevant theories, findings, methods, and unresolved disagreements before designing a study. This helps them:

• Avoid unnecessary duplication.
• Identify gaps.
• Refine concepts.
• Select appropriate methods.
• Anticipate limitations.
• Explain how the study will contribute.

A literature review should not be a collection of summaries. It should show how the evidence relates to the new research problem.

It uses methods appropriate to the question

No single research method is best for every problem.

An experiment may be appropriate for estimating a causal treatment effect. Interviews may be more appropriate for understanding how participants experience that treatment. A longitudinal design may be needed to study change over time, while a cross-sectional survey may be sufficient to estimate current prevalence.

It makes concepts observable or analysable

Researchers must define what important concepts mean within the study.

For example, “academic success” might be operationalised as:

• Examination score
• Grade-point average
• Course completion
• Retention
• A validated learning measure

Different operational definitions can produce different findings. Researchers should therefore explain and justify their measurements.

It manages uncertainty

Scientific evidence is rarely perfect. Samples differ from populations, measurements contain error, models depend on assumptions, and several explanations may fit the same observations.

Good research does not hide this uncertainty. It reports limitations, confidence intervals or other uncertainty measures where appropriate, alternative interpretations, missing information, and the conditions under which conclusions may not apply.

It is open to correction

Scientific knowledge is provisional rather than arbitrary. Conclusions are accepted in proportion to the strength, consistency, and quality of the evidence.

A result may later be:

• Confirmed.
• Refined.
• Restricted to particular conditions.
• Reinterpreted.
• Contradicted.
• Replaced by a better explanation.

What Is the Purpose of Scientific Research?

Scientific research can pursue several purposes at the same time.

To explore

Exploratory research investigates an insufficiently understood issue. It helps identify important concepts, patterns, mechanisms, or questions.

Example: Interviewing teachers to explore how generative AI is changing assessment practices.

To describe

Descriptive research documents characteristics, frequencies, distributions, or changes without necessarily explaining why they occur.

Example: Estimating the prevalence of food insecurity among postgraduate students.

To explain

Explanatory research investigates why or how something happens. It may examine causal mechanisms, contributing factors, or theoretical relationships.

Example: Testing whether reduced sleep helps explain the relationship between night-shift work and impaired attention.

To predict

Predictive research estimates an unknown or future outcome from available information.

Example: Developing a model to predict which patients are at increased risk of hospital readmission.

A predictive model can perform well without establishing that its predictors cause the outcome.

To evaluate

Evaluation research assesses the design, implementation, outcomes, cost, or impact of a programme, policy, intervention, or technology.

Example: Evaluating whether a tutoring programme improves course completion and whether the improvement justifies its cost.

To develop or apply knowledge

Some studies seek fundamental understanding, while others address a defined practical need.

The OECD distinguishes basic research, applied research, and experimental development when classifying research and development activity.

Types of Scientific Research

Scientific research should not be classified using one undifferentiated list. A study can belong to several categories simultaneously.

For example, a project may be applied, explanatory, quantitative, longitudinal, and quasi-experimental.

Classification by intended use

Type	Main purpose	Example
Basic research	Develop understanding without requiring an immediate practical application.	Investigating how memory consolidation occurs during sleep.
Applied research	Address a defined practical problem.	Testing a sleep intervention for hospital staff.
Translational research	Move discoveries between foundational research, practical testing, and implementation.	Adapting a laboratory biomarker for clinical diagnosis.
Evaluation research	Judge the implementation, effectiveness, impact, or value of a programme or policy.	Assessing a school nutrition programme.
Action research	Investigate and improve practice through iterative cycles involving practitioners or participants.	Teachers testing and refining a classroom feedback strategy.

Classification by research objective

Type	Central question
Exploratory	What may be happening, and which concepts or questions matter?
Descriptive	What exists, how often, for whom, or under what conditions?
Explanatory	Why or how does it happen?
Correlational	How are variables associated?
Causal	Does changing one factor produce a change in another?
Predictive	How accurately can an unknown or future outcome be estimated?
Methodological	How well does a measurement, procedure, model, or research method perform?

Classification by type of data

Quantitative research

Quantitative research represents information numerically and uses mathematical or statistical analysis.

It is useful for:

• Estimating frequencies or averages.
• Comparing groups.
• Testing specified associations.
• Estimating effects.
• Evaluating predictions.
• Quantifying uncertainty.

Qualitative research

Qualitative research analyses meanings, experiences, language, interactions, documents, observations, or visual material.

It is useful for:

• Understanding how people interpret an experience.
• Examining processes in context.
• Developing concepts or theories.
• Investigating why an intervention succeeds or fails.
• Exploring issues that cannot be adequately reduced to numerical variables.

Qualitative research can be scientific when its sampling, data collection, analysis, reflexivity, and interpretation are systematic and transparent. The University of Sheffield describes qualitative research as an approach for investigating nuance and depth while requiring cautious interpretation and acknowledgement of bias and limitations.

Mixed-methods research

Mixed-methods research intentionally integrates quantitative and qualitative evidence.

For example, a researcher might:

1. Measure whether an intervention changes attendance.
2. Interview participants to understand why it did or did not work.
3. Integrate both forms of evidence in the interpretation.

Simply placing a survey and several interviews in the same project does not automatically create a strong mixed-methods design. The researcher must explain how the two components relate.

Classification by researcher intervention

Design	Researcher intervention	Typical inference
Experimental	The researcher assigns or manipulates an exposure or intervention.	Can support causal conclusions when design and implementation are strong.
Quasi-experimental	An intervention or exposure is examined without full random assignment.	Can estimate causal effects under defensible assumptions.
Observational	The researcher observes exposures and outcomes without assigning them.	Describes patterns and associations; causal interpretation requires additional assumptions and methods.
Natural experiment	External events or rules produce exposure differences that may approximate an experiment.	May support causal inference when the allocation process is plausibly independent of other causes.

Classification by source of evidence

• Primary research: The study generates or collects original evidence.
• Secondary research: The study analyses data collected previously.
• Evidence synthesis: The study systematically identifies and combines existing studies.
• Theoretical research: The study develops, compares, or refines concepts and explanatory structures.
• Computational research: The study uses algorithms, simulations, code, or computational models.
• Methodological research: The study develops or evaluates measurements, analytical procedures, or research designs.

Classification by time

• Cross-sectional research: Evidence is examined at one time or over a short defined period.
• Longitudinal research: The same units or populations are studied across time.
• Prospective research: Outcomes are observed after study entry.
• Retrospective research: Existing records are used to examine events that have already occurred.

The Scientific Research Process

Scientific research usually involves defining a problem, reviewing existing evidence, formulating questions, designing a study, addressing ethics, collecting and analysing evidence, interpreting uncertainty, reporting results, and enabling further scrutiny. The process is iterative: later findings often lead researchers back to earlier assumptions, questions, or methods.

Step 1: Define the research problem

Identify a genuine uncertainty, contradiction, practical need, or theoretical problem.

A useful problem statement explains:

• What is not adequately known.
• Why the gap matters.
• Who or what is affected.
• What type of evidence could reduce the uncertainty.

Step 2: Review existing evidence

Search systematically enough to understand:

• Current explanations.
• Important findings.
• Disagreements.
• Common methods.
• Measurement limitations.
• Populations already studied.
• Remaining gaps.

A research gap should not be claimed merely because the researcher has not found a study. Search strategy and source quality matter.

Step 3: Formulate the research question and objectives

A strong research question is:

• Clear
• Focused
• Researchable
• Significant
• Ethically answerable
• Feasible with available time and resources

Objectives specify what the study will do. They are commonly written with verbs such as estimate, compare, explore, evaluate, explain, develop, validate, or test.

Step 4: Develop hypotheses where appropriate

A hypothesis is a testable expectation about a relationship, difference, mechanism, or outcome.

Example:

Students who use social media for more than two hours after 10 p.m. will report lower average sleep-quality scores than students who use it less frequently.

Not every study needs a formal hypothesis. Exploratory qualitative research, descriptive surveys, discovery research, and some model-development studies may be guided by questions or objectives instead.

Step 5: Select the research design

The design should match the question.

• Use an experiment when an intervention can be assigned ethically and causality is central.
• Use an observational design when manipulation is impossible or inappropriate.
• Use qualitative methods when experience, meaning, context, or process is central.
• Use mixed methods when neither numerical nor qualitative evidence alone is sufficient.
• Use evidence synthesis when the objective is to evaluate the existing body of research.

Step 6: Address ethics, governance, and feasibility

Before data collection, determine:

• Whether ethics committee or institutional review board approval is required.
• How informed consent will be obtained.
• Whether participants may experience harm, pressure, or unfair exclusion.
• How privacy and confidentiality will be protected.
• Whether vulnerable groups require additional protections.
• How data will be stored, accessed, retained, and destroyed.
• Whether the project has adequate time, expertise, equipment, and funding.

In the United States, federally conducted or supported human-subjects research may fall under the Common Rule and related agency regulations. Medical research involving human participants should also follow applicable ethical frameworks, with the 2024 Declaration of Helsinki now replacing its earlier versions as the current official WMA text.

Step 7: Plan sampling and measurement

Define:

• Target population
• Eligibility criteria
• Sampling frame
• Sampling method
• Intended sample size
• Recruitment procedure
• Variables or qualitative domains
• Measurement instruments
• Timing of measurements

A large sample does not correct a biased sampling process. Sample size and representativeness are separate issues.

Researchers should also consider:

• Reliability of measurements.
• Evidence of validity.
• Respondent burden.
• Cultural or linguistic suitability.
• Instrument permissions.
• Missing-data risk.
• Calibration and quality control.

Step 8: Develop a protocol and analysis plan

The protocol records what the study intends to do before the results are known.

It may include:

• Questions and hypotheses.
• Study design.
• Variables and outcomes.
• Sampling and recruitment.
• Data-collection procedures.
• Exclusion criteria.
• Data-cleaning rules.
• Analytical methods.
• Ethical procedures.
• Data-management plan.
• Dissemination plan.

Preregistration creates a time-stamped record of selected research plans before data collection or analysis. It can help readers distinguish planned confirmatory analyses from later exploratory analyses, but it does not make a weak design valid.

Step 9: Pilot and collect evidence

A pilot study or pretest can reveal:

• Ambiguous questions.
• Technical failures.
• Recruitment difficulties.
• Unexpected costs.
• Excessive participant burden.
• Data-format problems.
• Procedures that cannot be implemented consistently.

During data collection, researchers should document deviations, equipment settings, recruitment numbers, missing observations, protocol changes, and quality checks.

Step 10: Analyse the evidence

Analysis should follow the design and data.

Quantitative analysis may include:

• Data checking and visualisation.
• Descriptive statistics.
• Estimation and confidence intervals.
• Hypothesis tests.
• Regression or multivariable modelling.
• Prediction assessment.
• Sensitivity or robustness analysis.

Qualitative analysis may include:

• Familiarisation with the material.
• Coding.
• Category or theme development.
• Comparison across cases.
• Reflexive interpretation.
• Examination of contradictory cases.
• Documentation of analytical decisions.

Researchers should not choose only the analyses that produce preferred results.

Step 11: Interpret the findings

Interpretation asks what the results mean rather than merely repeating them.

Researchers should consider:

• Whether the evidence answers the question.
• The magnitude and practical importance of findings.
• Uncertainty.
• Alternative explanations.
• Bias and confounding.
• Measurement limitations.
• Generalisability or transferability.
• Consistency with previous research.
• Which conclusions are supported and which remain speculative.

Association alone does not establish causation. A statistically detectable result is not automatically large, useful, clinically important, or theoretically meaningful.

Step 12: Report, share, and enable further research

A complete report explains:

• Why the study was conducted.
• What was done.
• What was found.
• How uncertainty was assessed.
• What the findings may mean.
• What limitations apply.
• How the work relates to previous evidence.

Where ethically and legally possible, researchers may also share protocols, instruments, code, metadata, analysis outputs, and appropriately protected data.

Research reporting guidelines provide structured checklists or diagrams for particular study designs. The EQUATOR Network maintains a searchable collection of reporting guidance, including the updated CONSORT 2025 guideline for randomised trials.

Worked Example of Scientific Research

Suppose a university wants to know whether late-night smartphone use is related to poor sleep among first-year students.

Research problem

Many students report tiredness, but the university does not know whether late-night smartphone use is independently associated with sleep quality.

Research question

Among first-year students, is smartphone use after 10 p.m. associated with sleep quality during the examination period?

Possible design

A prospective observational study could measure smartphone use and sleep over four weeks.

Population and sample

• Population: All first-year students at the university.
• Sample: Students recruited using stratified sampling across faculties.
• Inclusion criteria: Enrolled first-year students aged 18 or older.
• Exclusion criteria: Defined only where scientifically and ethically justified.

Variables

• Exposure: Minutes of smartphone use after 10 p.m.
• Outcome: Sleep duration and a validated sleep-quality score.
• Potential confounders: Employment, caffeine intake, stress, chronic illness, course workload, and living arrangements.

Data collection

Participants might consent to provide:

• Weekly questionnaires.
• Device-based screen-time summaries.
• Sleep diaries.
• Optional wearable-device measurements.

Analysis

Researchers could estimate the association between late-night smartphone use and sleep outcomes while adjusting for selected confounding variables.

Interpretation

Even if heavier use is associated with poorer sleep, the observational design would not by itself prove that smartphone use caused the difference. Poor sleepers may use their phones more, or an unmeasured factor may influence both behaviours.

Follow-up research

The results could inform:

• A qualitative study of students’ night-time routines.
• A randomised behavioural intervention.
• A longitudinal study across the academic year.
• A replication at other universities.

This example demonstrates that scientific research is a connected programme of questions rather than a single study that permanently settles the issue.

How to Choose a Scientific Research Method

Ask the following questions in order.

1. What kind of answer is needed?

• Frequency or prevalence → descriptive quantitative study
• Experience or meaning → qualitative study
• Association → correlational or observational design
• Causal effect → experiment or defensible quasi-experiment
• Prediction → predictive modelling
• Existing overall evidence → systematic review or meta-analysis
• Instrument quality → measurement or validation study

2. What evidence can answer the question?

Determine whether the study needs:

• Numbers
• Text
• Images
• Observations
• Biological samples
• Sensor data
• Documents
• Existing datasets
• Experimental outcomes
• Multiple forms of evidence

3. Can the relevant factor be manipulated?

An experiment may be inappropriate when manipulation is:

• Unethical
• Impossible
• Too expensive
• Too slow
• Likely to distort the phenomenon

4. What level of inference is justified?

A design should be judged by the conclusion it is expected to support.

A cross-sectional survey may estimate prevalence but cannot usually establish temporal order. A randomised experiment may strengthen causal inference but still have limited real-world generalisability.

5. What constraints apply?

Consider:

• Time
• Budget
• Sample access
• Researcher expertise
• Equipment
• Software
• Ethics approval
• Participant burden
• Data sensitivity
• Required statistical power
• Institutional policies

How Is the Quality of Scientific Research Evaluated?

Research quality is multidimensional. A study can be strong in one area and weak in another.

Quality criterion	Central question
Construct validity	Do the concepts and measures represent what the study claims to investigate?
Internal validity	Is the observed result plausibly attributable to the proposed explanation rather than bias, confounding, or procedural differences?
External validity	To which people, settings, conditions, or times might the findings apply?
Reliability	Does the measurement or procedure produce sufficiently consistent results?
Credibility	Are qualitative interpretations well supported by the evidence and analytical process?
Transparency	Are the methods and decisions described clearly enough to evaluate?
Robustness	Do reasonable alternative assumptions or analyses lead to similar conclusions?
Ethical acceptability	Were participants, communities, animals, data, and researchers treated responsibly?
Reproducibility	Can the reported analysis be repeated using the same data, code, and procedures?
Replicability	Do new studies addressing the same claim obtain sufficiently consistent findings?

No checklist can replace expert judgement. Quality criteria must be interpreted in relation to the study’s purpose and discipline.

Reproducibility and Replicability

Reproducibility generally concerns obtaining consistent results using the original data, computational procedures, code, or analytical conditions. Replicability generally concerns obtaining consistent evidence in a new study addressing the same scientific question. Terminology varies among disciplines, so authors should define how they use the terms.

Under the National Academies’ framework:

• Reproducibility focuses on computational consistency using the same input data and analytical steps.
• Replicability focuses on consistency across studies that collect or use new evidence.

Neither concept means that every number must be identical. Differences can arise from sampling variation, context, implementation, measurement, or genuine differences in the phenomenon.

A failed replication also does not automatically prove misconduct. It may reveal:

• A context-dependent effect.
• An imprecise original estimate.
• Differences in methods.
• Measurement problems.
• Analytical errors.
• Incomplete reporting.
• Random variation.
• A claim that does not generalise.

Ethics in Scientific Research

Research quality and research ethics are connected. A poorly designed study may expose participants to burden without a reasonable prospect of producing useful knowledge.

Core ethical considerations

Researchers should consider:

• Scientific value.
• Fair participant selection.
• Informed and voluntary consent.
• Proportionate risks and benefits.
• Privacy and confidentiality.
• Secure data handling.
• Additional protection for vulnerable participants.
• Community involvement where relevant.
• Management of conflicts of interest.
• Responsible animal care.
• Environmental and societal consequences.
• Fair attribution and authorship.
• Responsible communication of results.

Ethics requirements depend on jurisdiction, institution, discipline, participant group, and data type. Researchers should not assume that anonymous, online, publicly available, or secondary data are automatically exempt from ethical review.

Research Integrity and Misconduct

Research integrity means conducting and communicating research honestly, accurately, transparently, and responsibly.

The US Office of Research Integrity defines research misconduct as:

• Fabrication: Making up data or results.
• Falsification: Manipulating materials, procedures, data, or results so the research record is inaccurate.
• Plagiarism: Using another person’s ideas, processes, results, or words without appropriate credit.

The federal definition does not include honest error or differences of scientific opinion.

Other harmful practices may include:

• Selective outcome reporting.
• Undisclosed removal of inconvenient data.
• Inappropriate image manipulation.
• Misleading statistical presentation.
• Gift or ghost authorship.
• Duplicate publication.
• Citation manipulation.
• Failure to disclose conflicts of interest.
• Breach of participant confidentiality.
• Use of unverified or fabricated references.

Not every poor practice meets a formal legal definition of misconduct, but it can still weaken the reliability and trustworthiness of research.

Peer Review and the Development of Scientific Knowledge

Peer review is the evaluation of a manuscript, proposal, or research output by people with relevant expertise.

Reviewers may assess:

• Importance of the question.
• Appropriateness of the design.
• Accuracy of the analysis.
• Consistency between evidence and conclusions.
• Reporting completeness.
• Ethical issues.
• Engagement with existing literature.

Peer review can identify problems, but it is not a certificate of truth. Reviewers may miss errors, disagree, or lack access to raw data and code.

Scientific confidence develops through a wider process:

1. Initial investigation
2. Methodological criticism
3. Peer review
4. Independent analysis
5. Replication
6. Evidence synthesis
7. Theoretical integration
8. Application across settings
9. Continuing revision

Preprints allow rapid communication but should be clearly distinguished from final peer-reviewed publications. NIH advises that unreviewed results should be communicated with appropriate context.

Open Science and Recent Research Practices

Open science refers to practices that make scientific knowledge, methods, evidence, and participation more accessible, transparent, reusable, and inclusive while respecting ethical, legal, security, and privacy constraints.

UNESCO’s Recommendation on Open Science encourages openness throughout the scientific process and includes publications, data, software, hardware, infrastructure, and engagement with society.

Open-science practices may include

• Open-access publication.
• Public protocols.
• Preregistration.
• Registered Reports.
• Preprints.
• Shared research materials.
• Open-source code.
• Data repositories.
• Persistent identifiers.
• Open peer review.
• Transparent reporting of deviations.
• Publication of negative or inconclusive findings.

FAIR data

FAIR means that data and associated metadata should be:

• Findable
• Accessible
• Interoperable
• Reusable

FAIR does not mean that every dataset must be publicly downloadable. Sensitive human data may need controlled access while still having findable metadata, clear governance, and documented reuse conditions. The FAIR principles were originally published by Wilkinson and colleagues as guidance for scientific data management and stewardship.

Preregistration

Preregistration records selected hypotheses, methods, outcomes, and analyses before researchers observe the relevant results.

It can:

• Separate confirmatory from exploratory analysis.
• Make undocumented outcome switching easier to identify.
• Encourage more complete planning.
• Preserve evidence of original intentions.

It cannot:

• Rescue a poorly designed study.
• Eliminate all researcher judgement.
• Prevent justified protocol changes.
• Guarantee that results will replicate.

Researchers should transparently document and justify necessary deviations.

Digital Tools Used in Scientific Research

Digital tools support different stages of research, but software does not determine whether a study is scientifically valid.

Research stage	Tool categories
Literature discovery	Bibliographic databases, citation indexes, scholarly search engines
Reference management	Citation libraries and bibliography managers
Protocol development	Collaborative documents, registration platforms, project-management systems
Data collection	Electronic forms, survey platforms, laboratory systems, sensors, mobile applications
Quantitative analysis	Statistical software, programming languages, notebooks
Qualitative analysis	Coding, annotation, transcription, and retrieval software
Reproducibility	Version control, repositories, computational environments, workflow tools
Visualisation	Statistical graphics, dashboards, geographic-information systems
Collaboration	Shared repositories, issue trackers, communication platforms
Dissemination	Journals, repositories, preprint servers, data archives

Researchers should select tools according to:

• Methodological suitability.
• Data security.
• Auditability.
• File and export formats.
• Accessibility.
• Licensing.
• Long-term preservation.
• Collaborator needs.
• Institutional policy.
• Reproducibility requirements.

Artificial Intelligence in Scientific Research

Artificial intelligence can assist with literature discovery, coding, translation, transcription, data classification, pattern detection, simulation, writing support, and administrative tasks. Researchers remain responsible for verifying every output, protecting confidential material, disclosing relevant use, and ensuring that AI does not replace scientific judgement or accountable authorship.

Potential uses

AI may help researchers:

• Generate search terms.
• Screen or organise records under human supervision.
• Improve code documentation.
• Detect patterns in large datasets.
• Support image or signal analysis.
• Transcribe interviews where consent and privacy permit.
• Translate or edit language.
• Check internal consistency.
• Create preliminary visualisations.
• Automate repetitive workflow steps.

Important risks

AI systems can:

• Invent references.
• Misstate study findings.
• reproduce hidden biases.
• Generate insecure or incorrect code.
• Expose confidential data.
• Hide uncertainty behind fluent language.
• Produce text that cannot be traced to reliable evidence.
• Encourage superficial rather than critical reading.

Responsible-use principles

1. Verify factual claims against original sources.
2. Never cite a reference that has not been checked.
3. Do not upload confidential manuscripts, participant data, or proprietary material to an unsuitable system.
4. Retain a human decision-maker for substantive scientific choices.
5. Document AI-assisted data processing sufficiently for evaluation.
6. Follow institutional, funder, journal, and disciplinary policies.
7. Disclose use when required or when it materially affected the work.
8. Do not list an AI tool as an author.

ICMJE states that AI tools should not be listed as authors because they cannot accept responsibility for accuracy, integrity, and originality. It also requires authors to remain accountable for AI-assisted content and to describe AI use in methods when it contributed to data collection, analysis, or figure generation.

Journal policies are not identical and continue to evolve. A 2026 Nature Methods editorial similarly emphasises disclosure, human responsibility, and the risk of hallucinated scientific references.

How Scientific Research Is Used Across Disciplines

Health and medicine

Researchers use laboratory studies, observational epidemiology, diagnostic studies, clinical trials, qualitative research, evidence synthesis, and implementation research.

Natural sciences

Research may involve controlled experiments, field observations, measurement, mathematical theory, remote sensing, and simulation.

Social and behavioural sciences

Methods include surveys, experiments, interviews, ethnography, longitudinal studies, administrative-data analysis, and mixed-methods research.

Engineering

Engineering research may investigate materials, systems, algorithms, prototypes, safety, performance, optimisation, and real-world implementation.

Computer science

Research may involve theoretical proof, algorithm design, benchmark evaluation, experiments, simulation, user studies, and system deployment.

Environmental science

Environmental research frequently integrates field measurements, satellite observations, laboratory tests, modelling, historical records, and policy analysis.

The diversity of these approaches illustrates why scientific research cannot be reduced to laboratory experimentation alone. Scientific methods are often interconnected and used in combination.

Advantages of Scientific Research

Systematic problem solving

Scientific research converts broad uncertainty into questions that can be investigated using explicit methods.

Transparent evaluation

Documented procedures allow others to examine how conclusions were reached.

Cumulative knowledge

New studies can compare, reproduce, refine, or challenge earlier findings.

Estimation of uncertainty

Well-designed research identifies what is known, how precisely it is known, and what remains uncertain.

Practical application

Scientific evidence can inform medicine, engineering, education, public policy, environmental management, and technology.

Error correction

Replication, criticism, improved measurement, and new evidence can reveal errors and lead to stronger explanations.

Limitations of Scientific Research

Measurement is imperfect

Many concepts cannot be observed directly and must be represented through imperfect indicators.

Samples may not represent populations

Findings can be distorted by non-response, selection bias, attrition, exclusion criteria, or narrow study settings.

Causal inference is difficult

Observed relationships may reflect confounding, reverse causation, selection effects, or measurement error.

Ethical restrictions limit possible studies

Researchers cannot deliberately expose people to serious harm merely to obtain stronger causal evidence.

Findings may depend on context

Results can differ across cultures, institutions, technologies, time periods, implementation conditions, and populations.

Research requires resources

High-quality studies may require funding, specialist skills, equipment, long follow-up periods, secure infrastructure, or access to rare populations.

Publication systems can distort the evidence base

Positive, novel, or statistically detectable results may be more likely to be submitted, accepted, or publicised than negative or inconclusive findings.

Science does not remove judgement

Researchers make decisions about questions, measurements, models, thresholds, interpretations, and priorities. Transparency and critical scrutiny make these judgements more visible; they do not eliminate them.

Common Mistakes in Scientific Research

Mistake	Why it is harmful	Better practice
Starting with a broad topic rather than a researchable problem	Produces unfocused data and unclear conclusions	Define the precise uncertainty and intended contribution
Choosing a method before defining the question	Creates a mismatch between evidence and objective	Let the question guide the design
Assuming every study needs a hypothesis	Forces confirmatory language onto exploratory research	Use questions, objectives, or hypotheses as appropriate
Treating association as causation	Ignores alternative explanations	Match causal language to the design and assumptions
Using an easy sample without discussing bias	Limits generalisability	Explain sampling limitations and improve coverage where possible
Measuring a construct with an unsuitable instrument	Weakens construct validity	Use justified, tested, and context-appropriate measures
Changing outcomes after seeing results without disclosure	Encourages selective reporting	Preregister where appropriate and label exploratory changes
Reporting only p-values	Conceals magnitude and uncertainty	Report estimates, uncertainty, and practical importance
Removing inconvenient observations without a rule	Can bias findings	Establish and justify exclusion procedures
Ignoring missing data	May distort conclusions	Examine patterns and use appropriate methods
Hiding null or contradictory findings	Produces an incomplete evidence base	Report relevant outcomes transparently
Claiming a study “proves” a broad theory	Overstates limited evidence	State exactly what the design and data support
Trusting AI-generated references	Risks fabricated or inaccurate citations	Verify every source in the original publication
Treating peer review as proof	Overstates the reliability of one publication	Evaluate methods and the broader evidence base

Scientific Research Planning Template

Use the following template before preparing a full proposal.

Research problem

What is not yet adequately known, understood, predicted, or solved?

Significance

Why does the problem matter academically, practically, socially, or methodologically?

Existing evidence

What is already known, and what limitations or disagreements remain?

Research question

What precise question will the study answer?

Objectives

What will the study explore, describe, compare, test, estimate, develop, or evaluate?

Hypotheses

Are formal predictions appropriate? If so, state them before analysing the relevant results.

Study design

Which design is best suited to the question, and why?

Population or material

Who or what will be studied?

Sampling

How will cases, participants, records, documents, observations, or datasets be selected?

Concepts and variables

How will the central concepts be defined and measured?

Data collection

What procedures, instruments, sources, or technologies will be used?

Analysis

How will the evidence be organised, analysed, compared, and interpreted?

Quality safeguards

How will reliability, validity, credibility, bias, confounding, reflexivity, and uncertainty be addressed?

Ethics and governance

What approval, consent, privacy, security, and community considerations apply?

Transparency

Will the protocol, preregistration, instruments, data, code, or materials be shared? What restrictions are necessary?

Limitations

What can the design establish, and what will it be unable to establish?

Dissemination

How will results be communicated to academic, participant, policy, professional, or public audiences?

Conclusion

Scientific research is a disciplined way of developing and evaluating knowledge. Its strength comes not from a single fixed method but from the alignment of questions, evidence, design, analysis, ethics, and transparent reasoning.

A well-conducted study does not claim more than its evidence can support. It explains its assumptions, reports uncertainty, acknowledges limitations, and creates a record that other researchers can examine and extend. Scientific knowledge becomes more dependable through this continuing process of investigation, criticism, replication, synthesis, and revision.