Original research is a systematic investigation that makes a new scholarly contribution. It may collect new data, analyse existing data in a new way, develop or test a method or theory, replicate an earlier study, or produce a new interpretation of primary evidence. Its exact classification can vary by discipline, institution, and journal.
Introduction
Original research is central to academic knowledge, but the term is often explained too narrowly. It is commonly described as research in which investigators collect data themselves. Although this is one important form, it is not the only way to produce an original contribution.
A researcher may also generate original knowledge by analysing an existing dataset from a new perspective, testing whether a previous result can be replicated, constructing a mathematical proof, developing an algorithm, interpreting previously neglected archival sources, or proposing and evaluating a new theoretical model.
This guide explains what original research means, what qualifies as original, how it differs from related research categories, how to recognise an original research article, and how to design and report an original study responsibly.
Key Takeaways
- Original research makes a new and defensible contribution to knowledge.
- It does not always require the researcher to collect new data personally.
- Primary research and original research overlap, but the terms are not identical.
- Replications, null findings, secondary-data analyses, theoretical work, and new interpretations can be original.
- Journal and institutional classifications differ, especially for systematic reviews, case reports, and methodological papers.
- Originality must be supported by rigorous methods, transparent reporting, ethical conduct, and appropriate evidence.
What Is Original Research?
Original research is a planned and systematic investigation in which the authors perform the intellectual and analytical work needed to produce a contribution that was not already available in the same form.
The contribution may consist of:
- New observations or measurements
- New experimental or survey data
- New qualitative evidence
- A new analysis of existing information
- A new method, instrument, model, or algorithm
- A new proof or theoretical explanation
- A replication or extension of earlier work
- A new interpretation of primary texts, artefacts, images, or archival records
- Evidence that challenges, refines, or fails to support an accepted claim
Originality therefore concerns the contribution, not simply the age or ownership of the data.
Core characteristics of original research
Original research normally has the following characteristics.
| Characteristic | Meaning |
|---|---|
| Clear research problem | The study addresses a defined question, gap, uncertainty, or problem. |
| Scholarly contribution | It adds evidence, analysis, interpretation, theory, method, or understanding. |
| Systematic method | The authors follow an explicit and defensible research process. |
| Appropriate evidence | Conclusions are supported by data, sources, reasoning, or formal proof appropriate to the discipline. |
| Transparency | Readers can understand what was done and how the conclusions were reached. |
| Ethical conduct | Relevant ethical, legal, privacy, consent, and integrity requirements are followed. |
| Critical interpretation | The authors explain what the findings mean without exaggerating them. |
| Acknowledged limitations | Important uncertainties, biases, and boundaries are disclosed. |
A paper does not become original merely because its topic has not appeared in exactly the same wording. Likewise, a study is not automatically unoriginal because another researcher has investigated the general topic before.
Originality, novelty, significance, and rigour
These terms are related but not interchangeable.
Originality means that the work makes a distinguishable scholarly contribution.
Novelty describes how new or different that contribution is. Novelty may involve a question, population, setting, dataset, theory, method, analysis, or interpretation.
Significance refers to why the contribution matters. A technically original finding may still have limited practical or theoretical importance.
Rigour concerns the quality and appropriateness of the research process. An unusual idea is not valuable research if it is investigated through unreliable or inappropriate methods.
Strong research aims to be original, sufficiently important, and methodologically sound. However, originality does not require a revolutionary discovery. A careful replication, improved measurement, new application, or well-designed study in an under-researched population may provide a meaningful contribution.
Original Research Versus Related Concepts
The terms original, primary, secondary, empirical, and review research are sometimes used as if they were interchangeable. They describe different features of research.
| Concept | Main meaning | Can it be original? |
|---|---|---|
| Original research | Work that makes a new scholarly contribution | Yes, by definition |
| Primary research | The researcher collects data specifically for the current investigation | Usually |
| Secondary-data research | The researcher analyses data collected previously for another purpose | Yes, when the question, method, analysis, or interpretation contributes new knowledge |
| Empirical research | Research based on observation, measurement, or experience | Often, but empirical work may also reanalyse or verify existing evidence |
| Literature review | A narrative discussion or synthesis of published literature | Usually not classified as an original research article |
| Systematic review | A review conducted through an explicit, reproducible search and appraisal method | Original scholarship, but journal and institutional classifications vary |
| Meta-analysis | Statistical synthesis of results from multiple studies | May be treated as original research or as a separate evidence-synthesis article type |
| Theoretical research | Development, testing, refinement, or comparison of theories or formal models | Yes |
| Replication study | A new test of an earlier finding or method | Yes |
| Research protocol | A plan describing research that is intended or underway | Scholarly, but it does not report completed original findings |
| Commentary or opinion | Interpretation or argument without a systematic original investigation | Usually not an original research article |
Original research versus primary research
Original research and primary research frequently overlap, but they answer different questions.
- Primary research asks: Who collected the data and for what study?
- Original research asks: Does the work make a new scholarly contribution?
A newly conducted interview study is both primary and original research when it answers a research question through an appropriate analysis.
By contrast, a researcher may download an existing national survey dataset and use it to examine a question the original investigators did not address. The data are secondary to the new researcher, but the analysis and conclusions may be original.
Original research versus a review article
An ordinary narrative review summarises, explains, or discusses existing publications. It normally does not report a newly conducted investigation and is usually classified separately from original research.
A systematic review is more complicated. It uses an explicit research question, eligibility criteria, reproducible searches, study selection, critical appraisal, and structured synthesis. Some journals treat systematic reviews or meta-analyses as original research, while others classify them under a separate review category.
A mixed-methods study of clinical journal editors found no uniform definition. Most participating editors regarded systematic reviews as original in at least some circumstances, but their reasons and article classifications differed (Krnic Martinic et al., 2019). Researchers should therefore check the rules of the relevant institution, assessment, funder, database, or journal.
Original research versus a research article
“Research article” is a broad publishing label. It may refer to an article reporting original findings, but databases and publishers do not use article labels consistently.
Do not rely only on a database tag. Read the abstract and methods to determine whether the authors conducted a systematic investigation and produced their own analysis or findings.
What Counts as Original Research?
A project normally counts as original research when it asks a meaningful question and uses appropriate methods or reasoning to produce evidence, analysis, or interpretation that was not already available in that form.
The following cases illustrate how this principle works.
| Project | Original research? | Explanation |
|---|---|---|
| Conducting an experiment to test a hypothesis | Yes | The experiment produces new evidence. |
| Interviewing teachers about a new curriculum | Yes | The study generates and analyses new qualitative data. |
| Analysing an existing census dataset to answer a new question | Usually yes | The dataset is existing, but the research question and analysis can be original. |
| Repeating an earlier experiment with a new sample | Yes | Replication provides a new test of the claim. |
| Summarising five articles without a systematic method | Usually no | This is closer to a narrative literature summary. |
| Conducting a systematic review | It depends on classification | It is systematic scholarship, but institutions and journals classify it differently. |
| Developing and evaluating a new algorithm | Yes | The method and its evaluation may constitute an original contribution. |
| Producing a new mathematical proof | Yes | Originality can be formal and theoretical rather than empirical. |
| Reinterpreting unpublished archival documents | Yes | The selection, analysis, and interpretation can produce new historical knowledge. |
| Writing an opinion based only on personal belief | No | Opinion alone is not a systematic investigation. |
| Reporting a null or negative result | Yes | The direction of the result does not determine originality. |
| Writing a study protocol | Not completed research | The protocol describes planned methods but does not yet present findings. |
| Creating a questionnaire without validating or applying it | Possibly incomplete | Instrument development may be original, but its design and validation must be demonstrated. |
| Generating an essay entirely from AI output | No | Automated text generation is not evidence of a defensible scholarly investigation. |
Can secondary data produce original research?
Yes. Existing data can support original research when the new study makes a distinct analytical or interpretive contribution.
Examples include:
- Testing a new hypothesis using a public health dataset
- Combining several datasets to examine a previously unstudied relationship
- Applying a new statistical model to archived observations
- Reanalysing interview transcripts using a different theoretical framework, where consent and data-use conditions permit
- Investigating historical change through digitised government records
- Evaluating algorithmic bias using an established benchmark dataset
- Using satellite archives to measure environmental change
The researcher must still explain the origin, quality, limitations, permissions, and suitability of the data. Existing data do not remove the need for a research design.
Does replication count as original research?
Yes. A replication is a new investigation of an existing claim. It may repeat the original methods closely, test the finding in a different population, use improved measurement, or examine whether the effect survives under changed conditions.
Replication may confirm, weaken, refine, or contradict an earlier conclusion. All of these outcomes can be informative. The contribution lies in providing an independent test, not in guaranteeing a different result.
Are negative or null findings original?
Yes. A result remains original when it comes from a valid new investigation, even when a predicted relationship is absent or an intervention does not work.
Researchers should not change the research question after seeing the outcome, hide prespecified analyses, or describe an inconclusive result as proof that no effect exists. Null and negative findings require the same attention to design quality, statistical precision, measurement, uncertainty, and limitations as positive findings.
Is a dissertation original research?
Many doctoral dissertations, and some master’s dissertations, are expected to make an original contribution. However, the degree of originality required varies by programme and qualification.
A dissertation is not automatically original because it is long or independently written. Its originality must come from its question, evidence, method, analysis, theoretical development, or interpretation.
An undergraduate dissertation may demonstrate independent research competence without being expected to transform the field. Students should follow their department’s learning outcomes and assessment criteria.
Is a case study original research?
A systematically designed case study can be original research. It may investigate an organisation, policy, event, community, person, programme, or bounded situation using interviews, records, observations, documents, or multiple forms of evidence.
In medicine, an individual case report may present an original observation but is often published under a separate article category. Researchers should distinguish a methodological case study from a clinical case report and check applicable reporting guidance.
Types and Examples of Original Research
Originality appears differently across fields. The research method should fit the question rather than an assumed universal model.
Experimental research
Experimental research deliberately changes an independent variable or condition and examines its effect on an outcome.
Examples include:
- Comparing two instructional approaches
- Testing a material under different temperatures
- Evaluating an intervention through a randomised trial
- Measuring how interface design affects task completion
Experimental research is valuable for causal questions, but poor randomisation, weak controls, inadequate measurement, attrition, or low statistical precision can limit conclusions.
Observational research
Observational researchers measure characteristics, behaviours, exposures, or outcomes without assigning the conditions being studied.
Examples include:
- A cohort study following participants over time
- A cross-sectional survey of student wellbeing
- Observation of communication in classrooms
- Analysis of routinely collected health records
Observational evidence can reveal patterns and associations, but causal claims require careful handling of confounding, selection, measurement, and temporal order.
Qualitative research
Qualitative research investigates meaning, experience, interpretation, interaction, culture, or process through non-numerical evidence.
Common approaches include:
- Interviews
- Focus groups
- Ethnography
- Participant observation
- Discourse analysis
- Qualitative case studies
- Document analysis
Originality may come from identifying a previously unrecognised process, developing a conceptual explanation, documenting an under-represented perspective, or applying a framework in a revealing way.
Mixed-methods research
Mixed-methods research deliberately integrates quantitative and qualitative evidence.
For example, a researcher might:
- Compare student outcomes statistically.
- Interview students and teachers.
- Integrate the results to explain why the programme produced different experiences or outcomes.
Simply placing a survey and several interviews in the same project does not create a strong mixed-methods design. The researcher should explain why integration is needed and how the findings inform one another.
Secondary-data research
Secondary-data research uses information originally collected by governments, organisations, researchers, platforms, archives, or administrative systems.
Examples include:
- Labour-market analysis using government datasets
- Public-health research using hospital records
- Political research using election data
- Environmental research using remote-sensing archives
- Linguistic research using an existing text corpus
Originality depends on the new question, analytical strategy, combination of sources, interpretation, or application.
Methodological research
Methodological research develops, compares, validates, or improves research methods.
Examples include:
- Creating and validating a measurement scale
- Comparing missing-data techniques
- Developing a sampling procedure
- Evaluating the reliability of an automated classification system
- Improving a laboratory or imaging protocol
A method should be evaluated against relevant alternatives, assumptions, criteria, or performance measures rather than presented as original merely because it is newly proposed.
Computational and data-science research
Computational originality can involve:
- A new algorithm
- A new software architecture
- A new benchmark
- A reproducible data pipeline
- A new simulation
- A model evaluated on an appropriate task
- An analysis revealing a previously unknown pattern
Researchers should report data provenance, preprocessing, evaluation design, baselines, code or pseudocode, parameter choices, computational resources where material, and limitations of generalisation.
Theoretical and mathematical research
Not all original research requires participants or empirical datasets.
Original contributions can include:
- A new theorem or proof
- A theoretical model
- A logical analysis
- A formal framework
- A new solution to a mathematical problem
- A derivation showing that an assumption leads to unexpected consequences
The standards of evidence are disciplinary. Formal validity, internal consistency, explanatory value, relationship to prior theory, and the importance of the problem may matter more than data collection.
Humanities research
Humanities scholars can produce original research through systematic engagement with primary materials such as:
- Manuscripts
- Historical documents
- Legal texts
- Artworks
- Films
- Music
- Speeches
- Correspondence
- Material artefacts
- Digital archives
Originality may lie in discovering a source, constructing a new interpretation, challenging an established reading, comparing materials in a new way, or developing a theoretical argument grounded in textual or historical evidence.
Practice-based and design research
In art, architecture, design, media, and related fields, the production of an artefact may form part of the research.
The artefact alone is not necessarily the complete contribution. Researchers normally need to articulate:
- The research question
- The context and prior practice
- The creative or design process
- The evidence generated through practice
- The critical reflection
- The resulting contribution to knowledge or practice
How to Identify an Original Research Article
An original research article normally states a research question, explains how the investigation was conducted, presents the authors’ own analysis or findings, and interprets what those findings contribute.
Use the following seven checks.
1. Read the title and article-type label
Look for terms such as:
- Original article
- Research article
- Original research
- Empirical study
- Clinical investigation
- Full paper
Treat labels as clues rather than proof because publishers classify articles differently.
2. Examine the abstract
An original-study abstract usually contains some combination of:
- Background or context
- Objective or research question
- Design or methodology
- Sample, dataset, materials, or sources
- Results
- Conclusion
A review abstract will normally describe searching, selecting, and synthesising previous studies rather than recruiting participants or analysing a primary dataset. However, a secondary-data study may analyse an existing dataset and still be original.
3. Look for a methods section
The paper should explain how evidence was produced or analysed.
Depending on the discipline, this may appear under headings such as:
- Methods
- Methodology
- Materials and Methods
- Research Design
- Data and Methods
- Experimental Procedure
- Corpus and Approach
- Sources and Analytical Framework
Humanities and theoretical papers may integrate methods into the introduction or analysis rather than provide a separate IMRaD section.
4. Identify the evidence
Ask what the authors actually examined:
- Participants
- Experiments
- Surveys
- Interviews
- Observations
- Records
- Existing datasets
- Documents
- Textual corpora
- Artefacts
- Images
- Simulations
- Algorithms
- Formal propositions
5. Find the authors’ results or analysis
The paper should present more than a summary of other publications. Look for tables, figures, themes, models, proofs, observations, performance results, or interpretive findings generated through the study.
6. Check the contribution statement
The introduction or discussion should explain what the work adds. A credible contribution is usually specific rather than a vague claim that the topic has “never been studied.”
7. Verify the publication type and journal rules
Consult:
- The journal’s article-type description
- Database indexing labels
- The author guidelines
- Course or assignment instructions
- Institutional definitions
This is particularly important for systematic reviews, protocols, case reports, brief communications, methodological notes, and data papers.
How to Conduct Original Research
Original research begins with a defensible question and proceeds through a transparent sequence of design, ethical preparation, evidence collection or selection, analysis, interpretation, and reporting.
Step 1: Define the intended contribution
State what the project is expected to add.
Possible contributions include:
- New evidence
- A new explanation
- A replication
- A new population or context
- A new method
- An improved measure
- A theoretical development
- A new interpretation
- A more reliable estimate
- Evidence that challenges an assumption
Do not define originality only as “nobody has used these exact words before.”
Step 2: Review the existing literature
Use appropriate scholarly databases, library catalogues, citation indexes, repositories, and reference lists.
The review should identify:
- What is already known
- How it is known
- Where findings disagree
- Which methods have been used
- Which populations or contexts are missing
- Which assumptions remain untested
- What limitations repeatedly appear
A research gap should arise from critical engagement with evidence, not merely from failing to find an identical title.
Step 3: Formulate the research question
A strong research question should be:
- Clear
- Researchable
- Focused
- Ethically acceptable
- Feasible with available time and resources
- Connected to a meaningful scholarly problem
Quantitative studies may also specify hypotheses. Qualitative, exploratory, theoretical, and interpretive projects may use objectives or guiding questions instead.
Step 4: Select the research design
Choose a design capable of answering the question.
Possibilities include:
- Experiment
- Survey
- Cohort study
- Case-control study
- Case study
- Ethnography
- Interview study
- Focus-group study
- Archival research
- Corpus analysis
- Secondary-data analysis
- Simulation
- Algorithm evaluation
- Theoretical modelling
- Mixed-methods design
The design should follow from the question. A convenient method is not necessarily an appropriate method.
Step 5: Define the sample, sources, variables, and analysis
Specify:
- Target population or source universe
- Inclusion and exclusion criteria
- Sampling or source-selection strategy
- Variables, concepts, outcomes, or themes
- Measurement tools
- Data-quality checks
- Planned analysis
- Missing-data handling
- Sensitivity or robustness analyses
- Criteria for interpreting results
Qualitative studies should explain recruitment, sampling logic, data generation, researcher positioning where relevant, analytical procedures, and how interpretations will be supported.
Step 6: Address ethics, permissions, and data protection
Obtain the required approval before recruiting participants or accessing restricted data.
Consider:
- Informed consent
- Participant risk
- Privacy and confidentiality
- Vulnerable groups
- Data security
- Anonymisation or pseudonymisation
- Copyright and licensing
- Data-use agreements
- Cultural and community responsibilities
- Conflicts of interest
- Appropriate compensation
- Plans for sharing or retaining data
Ethical approval is not a retrospective administrative label. When approval is required, researchers should not begin first and seek permission later.
Step 7: Prepare a protocol or preregistration
A protocol records the planned question, design, methods, outcomes, and analysis. Preregistration creates a time-stamped record before the relevant data or results are examined.
Preregistration helps readers distinguish planned confirmatory analyses from later exploratory work. It does not prevent justified changes, but changes should be documented and explained (Center for Open Science, n.d.).
A Registered Report goes further: the research question and method are reviewed before results are known, and qualifying studies may receive in-principle acceptance.
Step 8: Pilot the procedures
A pilot can reveal problems with:
- Recruitment
- Instructions
- Survey wording
- Interview guides
- Equipment
- Timing
- Coding procedures
- Software
- Data formats
- Missing responses
- Measurement reliability
A pilot is not a substitute for the main study. Researchers should state whether pilot observations were included in the final analysis and why.
Step 9: Collect or assemble the evidence
Follow the approved protocol as closely as possible and maintain documentation.
Useful records include:
- Recruitment logs
- Laboratory notebooks
- Data dictionaries
- Interview notes
- Codebooks
- File-naming conventions
- Version histories
- Data-cleaning decisions
- Deviations from the protocol
- Software and package versions
- Reasons for exclusions
Step 10: Analyse the evidence
Apply the planned methods and verify the analysis.
Good practice includes:
- Checking data quality
- Testing relevant assumptions
- Keeping raw and processed data separate
- Reviewing code or calculations
- Reporting exclusions
- Distinguishing confirmatory from exploratory analyses
- Assessing uncertainty
- Conducting appropriate sensitivity checks
- Looking for alternative explanations
- Preserving an auditable workflow
Do not choose an analysis solely because it produces the preferred result.
Step 11: Interpret the findings cautiously
Discuss:
- What the findings show
- What they do not show
- How they relate to previous evidence
- Plausible alternative explanations
- The effect of bias or uncertainty
- The boundaries of generalisation
- Practical or theoretical implications
- Questions that remain unanswered
Statistical significance alone does not establish importance, causality, validity, or generalisability.
Step 12: Report and share the research transparently
Prepare the manuscript according to the study design and target publication.
Where ethically and legally possible, provide:
- Protocol or preregistration
- Analysis code
- Data dictionary
- Materials
- Measures
- Supplementary results
- Data-access statement
- Conflict-of-interest statement
- Funding information
- Contributor roles
- Ethical approval information
- Deviations from the original plan
Sensitive participant data should not be made openly available merely to appear transparent. Controlled access, metadata, synthetic data, or justified restrictions may be more appropriate.
How to Write an Original Research Article
Many empirical original-research articles follow the IMRaD structure: Introduction, Methods, Results, and Discussion. However, this is a convention rather than a universal definition of original research.
| Section | Main purpose | Questions to answer |
|---|---|---|
| Title | Identify the study accurately | What was studied, in whom or what, and through which broad design? |
| Abstract | Summarise the complete study | Why, how, what was found, and what does it mean? |
| Introduction | Establish the problem and contribution | What is known, what remains unresolved, and what does this study ask? |
| Methods | Enable evaluation and, where possible, replication | What design, sample, evidence, procedure, and analysis were used? |
| Results | Present findings without selective omission | What did the investigation find? |
| Discussion | Interpret findings in context | What do the results mean, how do they compare, and what are the limitations? |
| Conclusion | State the supported answer | What can reasonably be concluded? |
| Declarations | Provide transparency | What were the ethics, funding, conflicts, data availability, AI use, and contributions? |
| Supplementary material | Preserve necessary detail | What additional instruments, code, data, analyses, or documentation support the report? |
Writing the introduction
The introduction should move from the research problem to the specific study.
A useful sequence is:
- Establish the academic or practical problem.
- Summarise the most relevant existing knowledge.
- Identify the unresolved issue.
- Explain why the issue matters.
- State the research question, objective, or hypothesis.
- Indicate the contribution without exaggerating it.
Avoid claiming that “no research exists” unless a defensible search supports that conclusion.
Writing the methods
The methods should be detailed enough for readers to evaluate the study and, where applicable, reproduce or adapt the procedure.
Include relevant information about:
- Design
- Setting
- Participants, data, texts, or materials
- Sampling and eligibility
- Measures or instruments
- Procedures
- Ethics and consent
- Data preparation
- Analytical methods
- Software and versions
- Protocol deviations
Use a reporting guideline suited to the study. The EQUATOR Network provides structured guidance for designs including randomised trials, observational studies, qualitative studies, diagnostic studies, case reports, protocols, and systematic reviews.
Writing the results
Report findings in a logical order linked to the research questions.
Good results sections:
- Include relevant findings even when they are null
- Avoid duplicating every number in both text and tables
- Report uncertainty where appropriate
- Identify analysis populations and missing observations
- Distinguish planned and exploratory findings
- Use tables and figures to clarify rather than decorate
- Avoid introducing unsupported causal language
Writing the discussion
A strong discussion normally addresses:
- The principal findings
- Their relationship to prior research
- Possible explanations
- Theoretical or practical implications
- Strengths
- Limitations
- Boundaries of generalisation
- Future research priorities
- A proportionate conclusion
Do not hide limitations at the end in a single sentence. Explain how each important limitation could affect interpretation.
Advantages and Limitations of Original Research
Advantages
| Advantage | Why it matters |
|---|---|
| Produces new knowledge | It can answer questions not resolved by existing literature. |
| Gives control over design | Researchers can align methods and measurements with the question. |
| Tests existing claims | Replication and extension strengthen or correct the evidence base. |
| Supports theory development | Findings can refine concepts, explanations, and models. |
| Addresses specific contexts | Research can examine neglected populations, settings, or materials. |
| Develops research capacity | Conducting a study builds methodological, analytical, and scholarly skills. |
| Creates reusable resources | Data, code, instruments, and methods may support later research. |
Limitations and challenges
| Challenge | Practical consequence |
|---|---|
| Time and cost | Recruitment, fieldwork, analysis, and approval can exceed the available schedule. |
| Ethical and legal obligations | Some questions or data cannot be investigated without safeguards or permissions. |
| Sampling limitations | A convenient or small sample may restrict generalisation. |
| Measurement error | Weak instruments can undermine otherwise original work. |
| Researcher bias | Design, interpretation, and reporting decisions can influence results. |
| Data-access restrictions | Existing records or sensitive datasets may be inaccessible. |
| Reproducibility problems | Incomplete documentation makes verification difficult. |
| Publication bias | Positive or striking findings may be favoured over null results. |
| Technical complexity | Advanced analysis may require specialised expertise. |
| Classification differences | A journal or university may classify the same design differently. |
Originality cannot compensate for a design that is unable to answer the question.
Common Mistakes in Original Research
Treating an unstudied local setting as sufficient originality
Changing only the name of the university, city, company, or country may not produce a meaningful contribution. Explain why the setting tests a theory, fills an evidence need, challenges generalisation, or has distinctive characteristics.
Assuming that new data automatically create new knowledge
A new survey can reproduce an already answered question without improving theory, measurement, or context. Originality requires an analytical purpose.
Using “first study” claims without evidence
Statements such as “This is the first study ever conducted” are difficult to verify. More defensible wording includes:
- “We found no previous study that examined…”
- “Existing studies have not addressed…”
- “This study extends prior work by…”
State the databases, dates, and search boundaries when the originality claim is important.
Confusing complexity with originality
A complicated model is not automatically more original than a clear, well-justified analysis. Unnecessary complexity can reduce transparency and increase error.
Changing hypotheses after seeing the results
Exploratory findings can be valuable, but they should be labelled as exploratory. Presenting a post-hoc pattern as a prespecified prediction can mislead readers.
Ignoring contradictory or null results
Selective reporting distorts the evidence. Report relevant outcomes and analyses even when they do not support the expected conclusion.
Using inappropriate methods to appear innovative
A fashionable method should not be selected unless it fits the research question, data, assumptions, and available expertise.
Overstating causality or generalisation
A cross-sectional association does not normally establish cause and effect. Findings from one setting or sample should not be presented as universal without justification.
Beginning data collection before approval
When ethical or institutional approval is required, starting early can expose participants to risk and may make the data unusable.
Failing to preserve an audit trail
Unrecorded cleaning, coding, exclusion, and model-selection decisions make the research difficult to verify.
Original Research in Modern Academic Practice
Modern original research is increasingly evaluated not only by what was discovered but also by how transparently, reproducibly, and responsibly the work was conducted.
Reporting guidelines
Reporting guidelines provide structured lists of information that should appear in reports of particular research designs. They improve readers’ ability to understand, evaluate, reproduce, and synthesise research.
Examples include:
- CONSORT for randomised trials
- STROBE for observational studies
- PRISMA for systematic reviews
- COREQ and SRQR for qualitative research
- CARE for case reports
- STARD for diagnostic-accuracy studies
- TRIPOD for prediction models
- ARRIVE for animal research
- SQUIRE for quality-improvement studies
A reporting checklist does not repair a weak design. It helps authors report the design and conduct transparently.
Preregistration and Registered Reports
Preregistration records the intended hypotheses, outcomes, design, and analysis before the relevant results are known. Its purpose is not to ban exploratory research but to distinguish planned tests from analyses developed after inspecting the data.
Registered Reports add peer review before results are available. This shifts part of the publication decision toward the importance of the question and quality of the method rather than the direction of the outcome.
Data management and sharing
Researchers should decide early:
- What data will be produced
- How files will be named and documented
- Where data will be stored
- Who will have access
- How privacy will be protected
- Which formats support long-term use
- Whether and how data can be shared
- How long records must be retained
The FAIR Principles encourage data and metadata to be findable, accessible under appropriate conditions, interoperable, and reusable (Wilkinson et al., 2016).
FAIR does not mean that every dataset must be publicly downloadable. Ethical, legal, contractual, cultural, security, and confidentiality restrictions may require controlled access or non-disclosure.
For NIH-supported research in the United States, the 2023 Data Management and Sharing Policy remains in force, while a revised 2026 plan format applies to relevant awards. Researchers should always check the current requirements of their own funder and institution.
Code, materials, and reproducible workflows
When possible and appropriate, preserve:
- Analysis scripts
- Software environment information
- Package versions
- Data dictionaries
- Survey instruments
- Interview guides
- Laboratory protocols
- Stimuli
- Search strategies
- Coding frameworks
- Decision logs
Version-control systems and research repositories can help document changes and preserve materials. Restricted or copyrighted resources should be handled according to applicable licences and agreements.
Contributor roles and authorship
Authorship should reflect genuine intellectual responsibility and contribution. The CRediT taxonomy provides 14 roles, including conceptualisation, methodology, investigation, formal analysis, data curation, software, validation, visualisation, supervision, and writing.
CRediT describes contributions; it does not by itself determine who qualifies as an author. Researchers should also follow the authorship policy of the journal, institution, and discipline.
Digital research tools
Common tool categories include:
| Research task | Examples |
|---|---|
| Literature discovery | Library databases, citation indexes, Google Scholar, discipline-specific databases |
| Reference management | Zotero, EndNote, Mendeley |
| Surveys and data capture | Qualtrics, REDCap, institutional survey systems |
| Quantitative analysis | R, Python, SPSS, Stata, SAS |
| Qualitative analysis | NVivo, ATLAS.ti, MAXQDA |
| Reproducible documents | Quarto, R Markdown, Jupyter notebooks |
| Version control | Git and institutional repositories |
| Project and preregistration records | Open Science Framework and discipline-specific registries |
| Data preservation | Institutional or recognised disciplinary repositories |
Tool selection should be based on security, accessibility, reproducibility, institutional approval, data sensitivity, and methodological suitability rather than popularity alone.
Artificial Intelligence and Original Research
Artificial intelligence can assist parts of a research workflow, but it cannot assume scholarly accountability. Human researchers remain responsible for the accuracy, originality, ethics, reasoning, analysis, citations, and final manuscript.
The ICMJE’s January 2026 recommendations expanded its guidance on AI in publishing. Its position includes the principle that AI systems cannot qualify as authors and that human authors remain accountable for submitted content (ICMJE, 2026).
Potentially appropriate uses
Subject to institutional, funder, ethics, and journal rules, AI tools may assist with:
- Brainstorming search terms
- Improving language and readability
- Producing preliminary code suggestions
- Explaining unfamiliar technical concepts
- Creating draft data dictionaries
- Supporting transcription or classification when validated
- Identifying inconsistencies in a manuscript
- Generating test cases for software
- Helping convert documented procedures into checklists
These outputs must be checked independently.
High-risk uses
Researchers should be particularly cautious about:
- Uploading identifiable participant data
- Uploading confidential manuscripts or peer-review material
- Accepting fabricated citations
- Using unvalidated AI classifications as research evidence
- Allowing AI to choose results selectively
- Presenting generated text as independent scholarly reasoning
- Using AI to conceal plagiarism
- Relying on a model’s explanation instead of checking primary sources
- Treating AI output as an authoritative source
- Listing an AI system as an author
How to report material AI use
When AI materially affects the research method or manuscript, report enough information for readers to understand its role.
Depending on relevance, include:
- Tool and provider
- Model or version
- Access date
- Research task
- Input-data type
- Important prompts or workflow
- Settings or parameters
- Human review process
- Validation method
- Error-handling procedure
- Privacy safeguards
- Known reproducibility limitations
The appropriate location may be the methods, acknowledgements, cover letter, supplementary material, or another disclosure section specified by the journal.
Language correction may require a different level of disclosure from AI-based coding, screening, image analysis, prediction, or data generation. Always consult the applicable policy rather than assuming a universal rule.
Original Research Checklist
Before describing a project as original research, ask:
Contribution
- Is the research problem clearly defined?
- Does the project add evidence, analysis, interpretation, theory, or method?
- Is the contribution more meaningful than a minor change of wording or location?
- Is the originality claim supported by a suitable literature search?
Design
- Can the selected design answer the research question?
- Are the sample, data, materials, or sources appropriate?
- Are concepts and outcomes defined clearly?
- Is the analysis planned and justified?
- Have important sources of bias and uncertainty been considered?
Ethics and governance
- Have required approvals and permissions been obtained?
- Are consent, privacy, confidentiality, and security addressed?
- Are licences and data-use conditions respected?
- Are conflicts of interest and funding relationships disclosed?
Transparency
- Is there a protocol or preregistration where appropriate?
- Are deviations from the plan documented?
- Can readers understand how the evidence was produced and analysed?
- Are null and contradictory findings reported?
- Are data, code, materials, or access conditions described?
Reporting
- Has an appropriate reporting guideline been used?
- Are conclusions proportionate to the evidence?
- Are limitations explained honestly?
- Are contributor roles accurate?
- Is material AI use disclosed according to current rules?
- Has the target journal’s article classification been checked?
Conclusion
Original research is any systematic scholarly investigation that produces a defensible new contribution. Although many original studies collect primary data, originality can also arise from secondary-data analysis, replication, methodological development, theoretical work, formal proof, computational research, or a new interpretation of primary sources.
The decisive questions are not simply whether the data are new or whether the result is surprising. Researchers must show what the work contributes, why the methods are appropriate, how the evidence supports the conclusions, and how ethical and transparent practices were maintained. Because article classifications vary, researchers should also verify the rules of their discipline, institution, funder, and intended journal.
