tesifyUK
Start free
How it worksFAQBlogHelpContactLog in

Category: Psychology

  • How to Write Hypotheses for a Psychology Dissertation: 12 Worked Examples (2026)

    How to Write Hypotheses for a Psychology Dissertation: 12 Worked Examples (2026)

    A psychology dissertation hypothesis is a single, testable sentence that names your variables, states the relationship or difference you expect, and can be shown to be wrong by the data you are about to collect. Write the alternative and the null as a pair, decide whether the prediction is directional, operationalise every term, and give each hypothesis exactly one planned analysis. Everything else in the introduction is there to earn the right to that sentence.

    This guide is for UK undergraduate and MSc psychology students on BPS-accredited programmes, where the empirical project is normally quantitative and the marker will read your hypotheses before anything else in the results chapter. It works through the steps in the order they should be done, with twelve worked examples across the designs undergraduates actually run.

    Step 1: Start from the research question and name its shape

    Every quantitative psychology hypothesis has one of three shapes, and the shape decides both the wording and the statistical test. Write your research question as a sentence and look at the verb.

    • Difference: do groups or conditions differ on an outcome?
    • Association: do two measured variables move together?
    • Prediction: do several variables together predict an outcome?

    If the question cannot be forced into one of these shapes, it is not yet operationalised and no hypothesis can be written from it. The mapping from shape to test is set out in our guide to choosing the right statistical test for a psychology dissertation; this article stops at the sentence the test will evaluate.

    Output of this step: your research question, labelled difference, association or prediction.

    Grad Coach on the anatomy of a research hypothesis; the psychology-specific steps below build on it.

    Step 2: Write the alternative and the null as a pair

    The alternative hypothesis (H1) states the effect you expect. The null hypothesis (H0) states that the effect is absent in the population: no difference, no association, no prediction. The statistical test evaluates the null, which is why UK psychology departments still expect both to be written out even though APA style does not require the null to be stated in the introduction.

    Write them in this form:

    H1: Participants who revise by self-testing will recall significantly more word pairs than participants who revise by rereading.

    H0: There will be no significant difference in word-pair recall between the self-testing and rereading conditions.

    Two things to notice. The null is not the opposite of the alternative; it is the absence of the effect. And “significantly” is doing work: it signals that the claim will be evaluated statistically, at a level (conventionally .05) you will name in the methods chapter.

    Output of this step: a numbered H1 and H0 for each research question.

    Step 3: Decide whether the prediction is directional

    A directional (one-tailed) hypothesis says which way the effect will go: higher, lower, positive, negative. A non-directional (two-tailed) hypothesis says only that a difference or association will exist. The choice is made from the literature, not from confidence: use a directional hypothesis when prior studies consistently point one way, and a non-directional one when the evidence is mixed or the question is new.

    The choice has a statistical consequence. A one-tailed test puts the whole rejection region on one side, so it is more powerful for the predicted direction and blind to the other. If you specify “higher” and the data go the other way, you have not found a result; you have found nothing. Markers check that the direction in your hypothesis matches the tail you reported, and a two-tailed p value under a one-tailed hypothesis reads as not having understood the distinction.

    Two hand-drawn bell curves in a notebook, one shaded at a single tail and one at both, illustrating directional and non-directional hypotheses
    One tail or two is decided by the literature before data collection, and it must match the p value you report.

    Output of this step: each H1 marked directional or non-directional, with one sentence of literature-based justification.

    Step 4: Operationalise every term

    “Stress” is a construct. “Score on the Perceived Stress Scale” is a variable. A hypothesis is only testable when every construct in it has been replaced by the measure you will actually use, and the measure has a level of measurement your test can handle. For each hypothesis, state:

    1. The independent variable or predictor, its levels or its scale.
    2. The dependent variable or outcome and the instrument that produces it.
    3. Whether the design is between-participants or repeated measures.

    Use validated instruments wherever they exist, because they bring published reliability with them and remove the measurement question from your hypothesis. Which scales an undergraduate can use without paying or registering is set out in our comparison of psychology scales and their licences.

    Output of this step: a variables table with one row per hypothesis: IV, DV, instrument, level of measurement, design.

    Step 5: One hypothesis, one planned analysis

    Each hypothesis should map to exactly one analysis you commit to before you see the data. Number them H1, H2 and, where one question has sub-parts, H2a and H2b. Three or four hypotheses is normal for an undergraduate project; eight is a sign that the study is really three studies.

    The rule exists because of a named problem. Norbert Kerr’s 1998 paper in Personality and Social Psychology Review gave it the name HARKing, hypothesising after the results are known: presenting a post-hoc finding as if it had been predicted. Undergraduate projects are exposed to it because the temptation arrives exactly when the planned analysis returns nothing. The defence is to write the hypotheses down, date them, and keep the document. Many UK departments now encourage students to preregister an undergraduate project on the Open Science Framework; even an emailed copy to your supervisor does the job.

    If you later run an unplanned analysis, report it as exploratory and say so. That is not a weakness; it is the distinction the marker is looking for.

    Output of this step: a numbered list of hypotheses, each with its named test, saved and dated before data collection.

    Step 6: Twelve worked examples across undergraduate designs

    Each example gives the research question, the alternative hypothesis, the null, and the analysis it commits you to. Adapt the wording; keep the structure.

    Design H1 H0 Planned analysis
    Two independent groups Students who revise by self-testing will recall more word pairs than students who reread. Recall will not differ between conditions. Independent-samples t test, one-tailed
    Two repeated conditions Reaction times will be slower after 30 minutes of smartphone use than before. Reaction times will not differ before and after. Paired-samples t test, one-tailed
    Three independent groups Recall will differ across silent, instrumental-music and lyrical-music conditions. Recall will not differ across the three conditions. One-way ANOVA, two-tailed
    Repeated measures, three levels Perceived task difficulty will differ across low, medium and high cognitive-load conditions. Perceived difficulty will not differ across load conditions. Repeated-measures ANOVA
    2 × 2 factorial The effect of time pressure on accuracy will depend on task type. There will be no interaction between time pressure and task type. Two-way ANOVA (interaction term)
    Association, continuous Higher Perceived Stress Scale scores will be associated with poorer sleep quality on the Pittsburgh Sleep Quality Index. Stress and sleep quality will not be associated. Pearson’s r, one-tailed
    Association, ordinal Self-rated social media use and body-image concern will be positively associated. The two rankings will not be associated. Spearman’s rho
    Prediction, several predictors Trait anxiety and rumination will together predict sleep quality. Neither predictor will account for variance in sleep quality. Multiple regression
    Prediction, one predictor over another Rumination will predict sleep quality after controlling for trait anxiety. Rumination will add no variance once anxiety is entered. Hierarchical regression
    Two categorical variables Choice of revision strategy will be associated with year of study. Strategy and year will be independent. Chi-square test of independence
    Non-parametric difference Median wellbeing scores will be higher in students who exercise three or more times a week. Median wellbeing will not differ by exercise frequency. Mann–Whitney U
    Null-effect prediction Font style will not affect recall accuracy. Recall accuracy will differ between font styles. Equivalence test or clearly reported effect size with confidence interval

    The last row is deliberate. A study can predict no effect, but the ordinary null-hypothesis test cannot confirm one, so the hypothesis has to be paired with an equivalence test or an effect size and confidence interval that the discussion interprets honestly.

    Step 7: If your project is qualitative, do not write a hypothesis

    Interview and focus-group projects analysed thematically do not test predictions; they explore experience. They need research aims and open research questions, not H1 and H0, and forcing a hypothesis onto them signals a misunderstanding of the design. Write “This study aimed to explore how final-year students describe…” and stop. Mixed-methods projects carry hypotheses for the quantitative strand only.

    A printed variables table with one row per hypothesis, annotated in pencil beside a laptop
    One row per hypothesis: variable, instrument, level of measurement, design, test. The table is written before the data exist.

    Step 8: Put the hypotheses where the marker expects them

    In a UK psychology dissertation the hypotheses close the introduction, immediately after the rationale that justifies them. They are restated at the head of each results section so the reader can see which analysis answers which prediction, and they are revisited in the first paragraphs of the discussion, one by one, with a verdict. Tense shifts as you go: future in the introduction if your department writes the proposal-style introduction, past in the results and discussion. The per-chapter rule, and the UK departments that disagree about it, is in our guide to which tense to use in each dissertation chapter.

    Output of this step: hypotheses appearing three times, in the same numbering, with consistent wording.

    Step 9: Report the verdict in the right words

    Hypotheses are supported or not supported. They are not proved, confirmed or rejected, and the null is retained rather than accepted. The sentence markers want reads: “H1 was supported: participants in the self-testing condition recalled significantly more word pairs than those in the rereading condition, t(58) = 3.52, p = .001, d = 0.91.” The effect size is what turns a verdict into a finding, and a non-significant result reported with its effect size and confidence interval is a legitimate outcome for an undergraduate project, provided the discussion treats it as evidence rather than failure. The mechanics of building the results chapter from the hypotheses rather than from the SPSS output are in our guide to writing the results chapter.

    Step 10: Cite the methods literature correctly

    Most UK psychology departments use APA style, currently the seventh edition, although the British Psychological Society’s own guidance for accredited programmes accepts alternatives, so check your handbook. A worked reference-list entry for the paper behind Step 5:

    Kerr, N. L. (1998). HARKing: Hypothesizing after the results are known. Personality and Social Psychology Review, 2(3), 196–217.

    In-text: (Kerr, 1998). If your reference list is the part of the dissertation that never matches the handbook, Tesify’s automatic bibliography formats APA and Harvard entries from a DOI and keeps the list in order as the chapters grow.

    A checklist before you show your supervisor

    1. Every hypothesis is one sentence and names both variables.
    2. Each H1 has an H0, and the null is the absence of the effect, not its opposite.
    3. Direction is stated where the literature supports it, and the tail matches the test.
    4. Every construct has been replaced by a named measure with a stated level of measurement.
    5. Each hypothesis has exactly one planned analysis, and the list is dated.
    6. The number of hypotheses fits the sample you can recruit; our guide to sample size for an undergraduate dissertation explains why four hypotheses on forty participants is a power problem.
    7. The wording is identical in the introduction, results and discussion.

    If the topic is settled but the question still will not sharpen into a testable sentence, the four narrowing moves in our guide to turning a dissertation topic into a research question usually get it there in an afternoon.

    When the hypotheses are written and the introduction still is not

    Tesify can build the introduction and methods sections around your own hypotheses, working from your variables table, your instruments and your planned analyses, so the rationale leads to the prediction the way a marker expects. The hypotheses, the design and the reasoning stay yours and are 100% written by you; what you get is the chapter moving.

    Frequently asked questions

    How many hypotheses should a psychology dissertation have?

    Usually two to four, each tied to one planned analysis. More than that on an undergraduate sample spreads statistical power too thin and usually means the project has not been narrowed enough.

    Do I have to write the null hypothesis?

    Most UK psychology departments expect it, even though APA style does not require it in the introduction. Write both, number them, and keep the null as the absence of the effect rather than the opposite prediction.

    Should my hypothesis be directional?

    Only when previous studies consistently point one way. If the evidence is mixed or the question is new, use a non-directional hypothesis and a two-tailed test, and make sure the p value you report matches.

    Can a hypothesis predict no effect?

    It can, but an ordinary significance test cannot confirm it. Pair the prediction with an equivalence test or report the effect size with a confidence interval and interpret it in the discussion.

    What is the difference between a research question and a hypothesis?

    A research question asks; a hypothesis predicts a specific answer that data can contradict. Quantitative projects need both, with the hypothesis derived from the question. Qualitative projects need only the question.

    What happens if my hypothesis is not supported?

    Nothing bad, provided you report it honestly with an effect size and discuss power, measurement and design. A non-significant result with a competent discussion earns the same marks as a significant one; a quietly re-analysed one does not.

    Can I change my hypothesis after collecting data?

    No. Changing the prediction to fit the result is HARKing. You can add an exploratory analysis and label it as such, but the original hypotheses stay in the dissertation with their verdicts.

    Where do the hypotheses go in the dissertation?

    At the end of the introduction, restated at the start of each results section, and revisited one by one at the start of the discussion, with identical wording and numbering in all three places.

  • Which Psychology Scale Can You Actually Use in an Undergraduate Dissertation? Licences Compared (2026)

    Which Psychology Scale Can You Actually Use in an Undergraduate Dissertation? Licences Compared (2026)

    Students choose a measure because it appeared in a paper they liked, build the questionnaire, submit the ethics form, and only then find the scale costs £2 per response or is restricted to registered practitioner psychologists. By that point the ethics application has to be resubmitted and a fortnight has gone.

    The table below sorts the measures UK undergraduates most often want by what it actually takes to use them legitimately. Licence terms do change, so verify on the publisher’s own site before you commit — but the broad tiers below have been stable for years and will tell you immediately whether a scale is realistic for a student project with no budget.

    The comparison table

    Scale Measures Items Cost to a student What you must do
    DASS-21 Depression, anxiety, stress 21 Free Nothing. Freely available from UNSW; cite the source.
    Perceived Stress Scale (PSS-10) Perceived stress 10 Free for non-commercial research Nothing beyond citing Cohen et al.
    GAD-7 Generalised anxiety 7 Free Reproduce without alteration; cite Spitzer et al.
    PHQ-9 Depression symptoms 9 Free Reproduce without alteration; cite Kroenke et al.
    Rosenberg Self-Esteem Scale Global self-esteem 10 Free Credit the Morris Rosenberg Foundation.
    UCLA Loneliness Scale (v3) Loneliness 20 Free for research Cite Russell (1996).
    IPIP personality scales Big Five traits 20–300 Free (public domain) Nothing. Fully unrestricted.
    Big Five Inventory (BFI-2) Big Five traits 60 (short forms exist) Free for non-commercial research Check the author’s stated terms.
    WEMWBS / SWEMWBS Mental wellbeing 14 / 7 Free for non-commercial use Register with the University of Warwick first.
    IPAQ Physical activity 7 (short form) Free Follow the published scoring protocol.
    Maslach Burnout Inventory Occupational burnout 22 Paid per response Purchase a licence through the publisher.
    STAI State and trait anxiety 40 Paid per response Purchase a licence through the publisher.
    Beck Depression Inventory-II Depression severity 21 Paid, restricted Qualification-restricted; usually unavailable to undergraduates.
    CD-RISC Resilience 25 / 10 Paid (student rates exist) Apply to the copyright holders.

    The three tiers, and what they mean for your timetable

    Tier 1: use immediately, no permission needed

    DASS-21, PSS-10, GAD-7, PHQ-9, the Rosenberg Self-Esteem Scale, the UCLA Loneliness Scale and the IPIP item pools are all freely usable in student research. You do not need to email anyone, you do not need to wait, and you can build the questionnaire this afternoon.

    These cover an enormous share of realistic undergraduate research questions: stress, anxiety, low mood, self-esteem, loneliness, personality. If your topic can be addressed with a Tier 1 measure, use one. There is no additional credit for choosing an expensive scale, and there is real cost in the delay.

    Tier 2: free but gated

    WEMWBS is the one most UK students hit. It is free for non-commercial research, but the University of Warwick requires you to register your use before you receive the licence. The process is straightforward and usually resolves in days rather than weeks, but it is a step, and it must be complete before your ethics submission — most UK ethics committees expect evidence of permission attached to the application.

    Several other measures sit here: free in principle, conditional on a form or an author’s email in practice. Budget a week and do it early.

    Tier 3: paid or restricted — usually the wrong choice

    The Maslach Burnout Inventory, the State-Trait Anxiety Inventory, the Beck Depression Inventory-II and CD-RISC are commercially licensed. Costs are typically charged per administration, which means a study of 120 participants becomes a real invoice that no undergraduate research budget will cover.

    The BDI-II carries a further barrier: publisher qualification levels restrict purchase to users with specified clinical training. An undergraduate cannot ordinarily buy it, and a supervisor generally will not buy it on your behalf for a student project.

    Practical substitutions that keep your design intact:

    • Want burnout? The Copenhagen Burnout Inventory is free and well validated. The Oldenburg Burnout Inventory is another free option.
    • Want anxiety? GAD-7 instead of the STAI, unless your design genuinely requires the state/trait distinction — and few undergraduate designs do.
    • Want depression? PHQ-9 or the depression subscale of the DASS-21 instead of the BDI-II.
    • Want resilience? The Brief Resilience Scale is free and only six items.

    In every case, write one sentence in your methodology explaining the substitution on validity grounds rather than cost grounds: “The Copenhagen Burnout Inventory was selected as it captures personal, work-related and client-related burnout as separate dimensions and is freely available for research use.” That is an academic justification, and it is also true.

    Free does not mean unrestricted

    Four rules apply even to Tier 1 measures, and breaking them undermines your results chapter.

    1. Do not change the wording. A scale’s validation applies to the exact items as published. Rewriting “over the last two weeks” as “recently” means you are no longer using the validated instrument and cannot cite its psychometric properties.
    2. Do not drop items to shorten it. If a 21-item measure feels long, use a published short form, which has been separately validated. An ad hoc subset has no established reliability.
    3. Do not change the response format. Converting a 4-point scale to a 5-point scale invalidates the published norms and scoring cut-offs.
    4. Use the published scoring rules. The DASS-21 requires its subscale scores to be doubled to compare against DASS-42 severity bands. Students forget this constantly and report severity categories that are simply wrong.

    Where minor adaptation is genuinely unavoidable — changing “at work” to “on placement”, for instance — report it explicitly, and expect to report your own reliability figures for the adapted version. That means calculating internal consistency for your sample, and knowing what counts as defensible; the thresholds are set out in the guide to what an acceptable Cronbach’s alpha is for a dissertation.

    The clinical-scale problem your ethics committee will raise

    PHQ-9 and GAD-7 are free, which makes them tempting. They are also clinical screening tools, and administering them to a non-clinical student sample creates two obligations that ethics committees at UK universities will insist on.

    First, a distress protocol. If a participant scores in the severe range, or endorses item 9 of the PHQ-9 concerning thoughts of self-harm, what happens? You need a written procedure, and signposting to named support services — university counselling, the Samaritans, NHS 111 — presented at the end of the survey.

    Second, a clear disclaimer. Participants must be told the questionnaire is not a diagnosis and that a high score does not mean they have a disorder. Anonymous survey designs cannot follow up individuals, and your information sheet must be honest about that.

    Some departments simply prohibit undergraduate use of clinical screening tools with the general public. Check yours before designing around one. The broader question of what needs review and when to submit is covered in the guide to whether an undergraduate dissertation needs ethics approval, and it is worth reading before you finalise your measures rather than after.

    How many scales should one dissertation include?

    Two or three. Rarely more.

    Each additional measure lengthens the survey, raises dropout, and multiplies the correlations you will be tempted to run — which inflates your false positive rate and invites a marker to ask why you tested twenty relationships to report the three that reached significance.

    A clean undergraduate design looks like: one predictor scale, one outcome scale, a short demographics block, total completion time under eight minutes. That structure gives you a clear hypothesis, a defensible analysis, and a realistic response rate. Once you know your variables and their measurement level, the decision tree in choosing the right statistical test for a psychology dissertation will tell you what analysis your design actually supports.

    Where to host the questionnaire

    Your licence governs the scale, but your university governs the data. Most UK institutions provide a licensed platform — Qualtrics, JISC Online Surveys or Microsoft Forms — and require you to use it, because free consumer tools frequently store data outside approved jurisdictions and fail institutional UK GDPR requirements. Using an unapproved platform is a common reason for an ethics application to be returned. The practical differences between the platforms available to UK students are set out in the comparison of which survey platform you can actually use for a UK student dissertation.

    Reporting your measures in the methodology chapter

    For each scale, four sentences is the standard. Name it with its citation and item count. State the response format and an example item. Give the scoring direction and range. Report published reliability and then your own sample’s reliability.

    A worked example: “Perceived stress was measured using the 10-item Perceived Stress Scale (Cohen, Kamarck and Mermelstein, 1983). Items are rated on a 5-point scale from 0 (never) to 4 (very often); a sample item is ‘In the last month, how often have you felt that you were unable to control the important things in your life?’. Four items are reverse scored, giving a total from 0 to 40 with higher scores indicating greater perceived stress. Cronbach’s alpha was .85 in the present sample, comparable to the .78 reported by the original authors.”

    That paragraph does everything a marker needs and takes ten minutes to write.

    The verdict

    For a UK undergraduate psychology dissertation with no budget and a fixed deadline, build your design around Tier 1 measures — DASS-21, PSS-10, the Rosenberg Self-Esteem Scale, the UCLA Loneliness Scale and IPIP personality items. They are free, immediate, well validated and widely published, so you will have comparison data for your discussion chapter. Reach for a Tier 2 measure such as WEMWBS only when it genuinely fits your question better, and start the registration the day you decide. Treat Tier 3 as closed unless your supervisor has an existing departmental licence and has said so in writing.

    Once your measures are settled, the methodology chapter is largely mechanical. Tesify can turn your chosen scales, design and participant details into a properly structured methodology draft, with the measures section formatted as above and your references built as you go — leaving you free to spend the time on analysis instead of scaffolding.

    Frequently asked questions

    Do I need written permission to use a free psychology scale?

    For genuinely open measures such as DASS-21, PSS-10 and GAD-7, no — correct citation is sufficient. Gated measures such as WEMWBS require registration first, and some ethics committees ask you to attach that confirmation to your application.

    Can I put a published scale in my dissertation appendix?

    For freely licensed scales, generally yes with full attribution. For commercially licensed instruments, reproducing the full item set usually breaches copyright — include a sample item and cite the source instead. Check your own department’s guidance.

    Is it acceptable to write my own questionnaire instead?

    Only for factual or demographic items. Self-written scales measuring psychological constructs have no established validity or reliability, and markers will treat conclusions drawn from them as weakly supported. Use a validated measure wherever one exists.

    What if my sample’s Cronbach’s alpha is much lower than the published figure?

    Report it honestly and discuss it. Common causes are small samples, a population unlike the validation sample, reverse-scored items entered incorrectly, or careless responding. Check your reverse scoring first — it is the most frequent culprit.

    Can I use a scale in a language other than the one it was validated in?

    Only if a validated translation exists — cite that translation specifically. Translating a scale yourself creates an unvalidated instrument, and proper cross-cultural adaptation is a research project in its own right.

    How long should my whole questionnaire take to complete?

    Under ten minutes, and ideally around five. Dropout climbs steeply beyond that, and your information sheet must state a realistic estimate — an understated completion time is itself an ethical problem.