Enugu, Nigeria
Nigeria
Enugu State
Nigeria

What Are Validity And Reliability In Research? And How Do You Show Them In Your Project

Picture this. You have just handed your supervisor the Chapter Three of your final year project, and she looks up from the pages and asks one question: "How do you know your instrument is valid and reliable?" Your mind goes blank. You wrote the questionnaire yourself, you collected the data, the numbers look fine, but you have no answer ready. If that scene sounds familiar, this article is for you, because validity and reliability are the two questions every examiner asks, and they are much simpler than they sound once someone explains them in plain language.


Academic Integrity & Digital Research Tools

Writing a high-quality Project Thesis requires more than just data; it requires precision. To ensure your Research Methodology meets international standards, we recommend using advanced AI Plagiarism Checkers and Grammar Correction Software before submission.

Professional Proofreading and Formatting

Many academic institutions now use automated AI Detection Tools. If you are preparing your Statement of Purpose (SOP) or Manuscript, investing in professional Academic Proofreading Services can help you maintain a high Similarity Index score while ensuring your research remains authentic and impactful.


Validity and reliability are the two tests of whether your research can be trusted. Reliability asks whether your instrument gives consistent results. If you measured the same thing again under the same conditions, would you get the same answer? Validity asks whether your instrument measures what it claims to measure. Are you really measuring student motivation, or are you accidentally measuring something else, like attendance or fear of failing? You need both. A study that is consistent but wrong is no better than a study that is right by accident, and your examiner will check for both before anything else.

The weighing scale test: the easiest way to remember the difference

Think of a bathroom weighing scale. If you step on it three times and it reads 68 kg, 68 kg and 68 kg, the scale is reliable. It gives the same result every time. But if your real weight is 63 kg, the scale is not valid, because it is consistently wrong. It is reliable without being valid.

Now imagine a scale that reads 60 kg, then 65 kg, then 70 kg for the same person on the same morning. That scale is neither reliable nor valid. You cannot trust any number it gives you.

That is the rule to carry into your project: an instrument can be reliable without being valid, but it cannot be valid without being reliable. If your questionnaire gives wildly different answers each time the same person fills it in, it cannot be measuring anything real. Reliability is the foundation, validity is the building on top of it.

What is reliability, and how do you show it?

Reliability is about consistency. There are four common types, and you only need the ones that fit your instrument.

Test-retest reliability

Give the same questionnaire to the same people on two occasions, usually two or three weeks apart, and compare the scores. If the scores are very similar, the instrument is stable over time. This works best when what you are measuring does not change quickly, such as an attitude or a trait. It does not work well for knowledge, because people remember the questions and may answer differently the second time.

Internal consistency

This is the type most undergraduate projects report. It asks whether the items in one scale hang together and measure the same thing. The usual statistic is Cronbach's alpha, and the rule of thumb examiners accept is that an alpha of 0.70 or above shows acceptable internal consistency, while 0.80 to 0.90 is good. You calculate it in SPSS or a similar package after your pilot study, and you report the figure in Chapter Three. For a fuller walkthrough of the tool itself, see How to Use SPSS for Data Analysis in Your Final Year Project.

Inter-rater reliability

If two or more people are observing, coding or marking the same thing, do they agree? This matters when your data depends on human judgement, for example two observers rating classroom behaviour or two coders sorting interview responses into themes. You report the level of agreement, often with Cohen's kappa, and you improve it by training your raters and writing a clear coding guide before you start.

Parallel forms reliability

This applies when you have two versions of the same test, such as two sets of exam questions meant to be equally difficult. If the same students score similarly on both versions, the forms are equivalent. Most final year projects never need this one, but you should know it exists.

What is validity, and how do you show it?

Validity is about accuracy, and it comes in several kinds. Unlike reliability, there is no single number that proves validity. You build a case for it with evidence, piece by piece.

Face validity

Does the instrument look, on the surface, as if it measures what it claims to measure? This is the weakest form of validity, but it is where you start. If your questionnaire on reading habits asks mostly about television, something is wrong before you even begin.

Content validity

Does the instrument cover every important part of the topic and nothing irrelevant? A questionnaire on job satisfaction should cover pay, working conditions, supervision and promotion, not just pay. You establish content validity by having experts, usually your supervisor and one or two lecturers in the field, review every item and confirm that nothing important is missing and nothing off-topic has crept in.

Construct validity

Does the instrument really measure the idea it claims to measure? This is the hardest kind to prove and the one examiners care about most. You build evidence for it by checking whether your scores behave the way theory says they should: for example, your new measure of study motivation should correlate with hours spent studying, but not with something unrelated like height. If you adopted an existing scale, cite the study that validated it and say whether it was tested on people like your own respondents.

Criterion validity

Does the instrument predict something it should predict? An aptitude test has criterion validity if it predicts first year grades. A screening questionnaire has it if its results match a proper diagnosis. In most undergraduate projects this one is optional, but it is worth mentioning if your instrument has an obvious outcome to predict.

Validity of the whole study, not just the instrument

So far we have talked about the validity of your instrument. Your examiner will also ask about the validity of your study as a whole, and two terms matter here.

Internal validity asks whether you can fairly claim that your independent variable caused the change you observed. Could something else explain the result? In an experiment, threats include events that happen between measurements, students maturing or getting tired, and participants dropping out. Random assignment and a control group are the strongest protections. If you are writing up your variables properly, our guide to What Are Variables in Your Research? explains the independent and dependent variables these threats act on.

External validity asks whether your results apply beyond your own sample. If you surveyed only students in one department of one university, can you honestly generalise to all Nigerian undergraduates? Usually not, and the honest move is to say so in your limitations rather than pretend otherwise. Your choice of sampling method shapes this directly, so see How to Choose a Sampling Technique for Your Final Year Project before you defend your sample.

How to show validity and reliability in Chapter Three

This is the part students get wrong most often. Writing "the instrument was valid and reliable" is a claim, not evidence. Your examiner wants to see what you actually did. Here is a practical order to follow.

  1. Use an established instrument where you can. If a validated scale already exists for your topic, adopt or adapt it, cite the original validation study, and say whether it was used with people like your respondents. Examiners trust recognised tools more than a questionnaire written from scratch in an afternoon.
  2. Define your concepts before you write a single item. Write down exactly what you mean by each variable and how you will measure it. Vague ideas like "academic success" become measurable when you define them as final year GPA or exam score. This one step protects both validity and reliability.
  3. Get your instrument reviewed by experts. Ask your supervisor and at least one other lecturer to read every item for clarity, relevance and coverage. Record who reviewed it and what they asked you to change. That record is your content validity evidence.
  4. Run a pilot study. Test the questionnaire or interview guide on a small group, usually 5 to 10 people who resemble your target population but will not be in the main sample. Time it, watch where respondents hesitate, and fix unclear wording. Report what changed after the pilot. A pilot study is also how you get the data to calculate Cronbach's alpha before the main collection.
  5. Standardise how you collect data. Give every respondent the same instructions, in the same order, in the same setting where possible. Small differences in how you administer an instrument add random error and weaken reliability.
  6. Report the numbers, not just the claim. In Chapter Three, write the actual reliability figure from your own data, for example "Cronbach's alpha for the 12-item motivation scale was 0.83, indicating good internal consistency." If you adapted a published scale, report both the published alpha and your own.
  7. Name the threats you could not remove. If your sample was a convenience sample, say so and explain what it means for generalisation. Examiners reward honesty far more than they reward a perfect-looking study, because no undergraduate study is perfect.

What if your project is qualitative?

Reliability and validity are quantitative ideas, and forcing them onto interviews or focus groups does not quite work. A second interview with the same person is never the same conversation, so "getting the same result twice" is the wrong test. Qualitative researchers usually use a different framework, called trustworthiness, proposed by Lincoln and Guba in 1985. It has four criteria.

  • Credibility: do your findings reflect what your participants actually meant? You build it with member checking, taking your findings back to participants to confirm them, and with peer debriefing, asking a colleague to challenge your interpretation.
  • Transferability: have you described your setting and participants in enough detail that a reader can judge whether the findings fit their own context?
  • Dependability: is your process consistent and documented well enough that another researcher could follow your decisions? A clear audit trail of how you coded is the evidence here. Our guide to How to Analyse Qualitative Data in Your Final Year Project shows how to keep that record.
  • Confirmability: can a reader see that your findings came from the data, not from your own opinions? Quotes linked to themes, and an honest account of your assumptions, are how you show it.

If your project is qualitative, frame your quality section in these terms. Reporting a Cronbach's alpha for interview data will only confuse your examiner.

Common mistakes to avoid

Claiming validity and reliability without evidence. This is the single most common error in Chapter Three. A sentence that says the instrument "was validated" tells the examiner nothing. Show the expert review, the pilot results and the alpha value.

Reporting only the published alpha. Reliability belongs to scores from your particular sample, not to the instrument in general. Always report the figure from your own data alongside the published one.

Confusing the two terms. Saying your study is reliable when you mean it is accurate, or valid when you mean it is consistent, is an easy way to lose marks in a defence. Keep the scale analogy in mind and you will not mix them up.

Piloting on your main sample. People who answer the pilot questionnaire should not also answer the final one. Their answers will be shaped by having seen the questions before.

Hiding limitations. A study with no stated limitations looks careless, not strong. Name the threats you could not remove and explain what they mean for your conclusions.

Frequently asked questions

Can an instrument be valid but not reliable?

No. If an instrument gives different results every time you use it, it cannot be measuring anything real, so it cannot be valid. Reliability is necessary for validity, but it is not enough on its own.

What is a good Cronbach's alpha value?

As a rule of thumb, 0.70 and above is acceptable, 0.80 to 0.90 is good, and above 0.90 may mean some items are repeating each other. Always check your department's guide, because some programmes set their own threshold.

Do I need to report validity and reliability for secondary data?

Yes, but differently. You cannot pilot a government dataset, so instead you report where the data came from, how it was collected, and any known limitations of the source. Your choice of design shapes what you can claim, so our guide to How to Choose the Right Research Design for Your Final Year Project is a useful companion here.

How many people do I need for a pilot study?

For an undergraduate project, 5 to 10 people who resemble your target population is usually enough to catch unclear wording and timing problems. The aim is to test the instrument, not to test your hypothesis.

Your Chapter Three checklist

Before you submit, run through this list. If you can tick every box, your validity and reliability section will stand up in a defence.

  • You can explain the difference between validity and reliability in one sentence each.
  • You named the instrument, cited its source, and said whether it was validated before.
  • You described the expert review and what changed because of it.
  • You ran a pilot study and reported what you changed afterwards.
  • You reported a reliability figure from your own data, with the threshold you judged it against.
  • You named at least one threat to validity that you could not remove.
  • If your study is qualitative, you used credibility, transferability, dependability and confirmability instead of alpha values.

Validity and reliability sound like textbook jargon until your supervisor asks about them, and then they become the most practical part of your whole project. They are simply the questions any careful reader would ask: did you measure the right thing, and would you get the same answer if you measured it again? Answer those two questions with evidence, not with a single sentence, and Chapter Three becomes one of the strongest parts of your project instead of one of the weakest. For a deeper, fully referenced treatment of the same ideas, the Grad Coach open educational resource on validity and reliability is an excellent next read, and Research Prospect's guide to reliability and validity gives further worked examples of each type.

Check This Out:
Absu Library And Information Science Project Topics And Materials
Be the First to Share On Social


1GB data
1GB data
1GB data

8