Psychology dissertation content analysis offers a systematic way to examine meaning, patterns, and representation in text, images, audio, or video. It can answer valuable questions without recruiting new participants, but it is not simply “reading documents and finding themes.” A defensible study defines its material, sampling logic, unit of analysis, coding framework, quality procedures, and limits before making claims.
This guide explains how to design, conduct, analyse, and report content analysis for a psychology dissertation. It distinguishes qualitative and quantitative approaches, shows how to build a usable coding frame, and uses psychology-specific examples throughout.
Psychology Dissertation Content Analysis
Content analysis is a family of methods for making systematic inferences from recorded communication. The material may include interview transcripts, public health campaigns, therapy manuals, online forum posts, news reports, policy documents, photographs, videos, or open-ended survey responses. The method can examine visible features, underlying meanings, or both.
The central principle is alignment. Your research question should determine what material counts as data, how it is sampled, what becomes a coding unit, and whether the output is interpretive categories, numerical frequencies, or an integrated account.
For example, a dissertation might ask how university wellbeing webpages frame responsibility for stress. The researcher could code manifest features, such as whether pages mention workload, counselling, sleep, or financial pressure. They could also interpret latent meanings, such as whether stress is framed mainly as an individual self-management problem or an institutional issue.
Table of Contents
Content analysis is not the same as thematic analysis
Content analysis and thematic analysis can both involve coding and pattern development, but they are not interchangeable labels. Content analysis often uses a defined coding frame to classify material systematically. It may count categories, compare sources, or examine both manifest and latent content. Thematic analysis is usually organised around patterns of shared meaning across a dataset and follows its own methodological traditions.
The distinction matters because readers need to understand what you actually did. Do not call an analysis “content analysis” merely because you coded text. Likewise, do not combine terminology from incompatible approaches without explaining the logic.
| Feature | Content analysis | Thematic analysis |
|---|---|---|
| Main purpose | Classify and interpret recorded content | Develop patterns of shared meaning |
| Typical output | Categories, counts, comparisons, interpretations | Themes and an interpretive account |
| Coding frame | Often explicit and structured | May be more flexible and recursive |
| Quantification | May be central or supplementary | Usually not the analytic goal |
| Quality focus | Clear units, category rules, consistency, reflexivity | Coherence, depth, reflexivity, analytic fit |
If your intended output is a set of meaning-based themes, use the site’s guide to psychology dissertation thematic analysis. Choose content analysis when classification, systematic comparison, or the relationship between visible and underlying content is central.
Choose a coherent content analysis approach
Hsieh and Shannon describe conventional, directed, and summative approaches to qualitative content analysis. Their distinctions are useful because they make the origin and function of codes explicit. The original peer-reviewed article explains that the approaches differ in coding schemes, code origins, and threats to trustworthiness.
Conventional content analysis
A conventional approach develops categories mainly from the data. It fits a question for which theory is limited or existing categories would restrict discovery. You still begin with a defined question and sampling plan, but you avoid forcing every segment into a pre-existing model.
Example: How do first-generation university students describe moments when they felt they belonged? The researcher might analyse reflective accounts and develop categories from recurring forms of inclusion, recognition, or exclusion.
Directed content analysis
A directed approach begins with concepts or categories informed by an existing theory or framework. It is appropriate when the dissertation tests, extends, or applies an established psychological model. The researcher must remain open to content that does not fit the initial framework.
Example: A study could use self-determination theory to examine how exercise-app messages support or frustrate autonomy, competence, and relatedness. An “other” pathway and category-revision rules prevent the framework from silencing unexpected material.
Summative content analysis
A summative approach begins by identifying or counting selected words, images, or content features, then interprets their use in context. Word frequency alone is not sufficient. A term may be absent because a source uses a synonym, and identical terms can have different meanings.
Example: A dissertation could compare how often mental health campaigns use “recovery,” “resilience,” and “risk,” then examine who is assigned responsibility around those terms.
Quantitative content analysis
Quantitative content analysis converts defined content features into variables for statistical description or comparison. Categories need explicit rules, and coders should be able to apply them consistently. The research might compare the proportion of social media posts containing appearance-focused messages across account types or time periods.
Counts do not make a study automatically objective. Sampling choices, category definitions, ambiguous cases, and coder decisions still shape the evidence. Report those decisions transparently.
Build an answerable research question
A strong question names the material, psychological phenomenon, analytic purpose, and relevant context. Avoid asking what “the media” says about a topic without specifying platforms, sources, dates, languages, or populations.
Useful question forms include:
- How is adolescent anxiety framed in publicly available school guidance documents?
- What categories of coping are represented in student reflections about assessment stress?
- How often do selected fitness posts contain appearance, health, or performance messages?
- How do portrayals of memory loss differ between two types of public information resource?
- To what extent do counselling-training cases reflect the components of a specified therapeutic framework?
Use the research-question guide to test clarity, scope, feasibility, ethics, and design alignment. A question about prevalence needs a sampling strategy that supports frequency claims. A question about meaning needs enough contextual material for interpretation.
Define the dataset and sampling frame
Your dataset is not “everything available online.” Define a bounded universe of potentially eligible material. State the source, date range, language, format, geographic or organisational setting, and inclusion and exclusion criteria. Explain how each boundary supports the question.
For a study of anxiety information on university webpages, the sampling frame might include all publicly accessible student-support pages indexed under a defined set of institutions on a specific date. Exclusions might cover duplicate pages, staff-only resources, news articles, or pages with no student-facing mental health content.
Select material systematically
A census includes every eligible item in the bounded frame. Probability sampling supports inference to a larger frame when a complete list exists. Purposive sampling selects information-rich cases for a clear analytic reason. Stratified sampling can preserve important groups, such as public and private providers or early and later time periods.
Convenience sampling may be feasible, but it limits claims. The first 100 search-engine results, for example, reflect ranking systems and search personalisation rather than a neutral population. If you use a platform search, document the date, search terms, filters, account state, and collection procedure.
Read the dedicated psychology dissertation sampling guide when defining the population, sampling unit, inclusion rules, and claim boundaries.
Separate sampling units from coding units
The sampling unit is the item selected for the study, such as a webpage, advert, transcript, episode, or post. The context unit is the surrounding material needed to interpret a segment. The coding unit is the smallest portion assigned a category, such as a sentence, speaking turn, image, paragraph, or whole document.
These units can differ. A sampled video may be the sampling unit, a full scene the context unit, and each spoken statement or visual shot a coding unit. Defining them prevents a coder from switching unpredictably between words, sentences, and whole documents.
Address ethics, copyright, and privacy
Public availability does not remove ethical responsibility. Online material may contain sensitive disclosures, searchable quotations, usernames, images, or details that make people identifiable. A dissertation should consider users’ expectations, vulnerability, platform context, consent, quotation traceability, and possible harm.
Seek the required institutional review before collecting or analysing data. Decide whether to paraphrase rather than reproduce searchable text. Store source files and coding data securely, minimise identifiers, and explain any use of copyrighted images or documents. The site’s psychology dissertation ethics guide provides a fuller risk-and-safeguard workflow.
Create a coding frame that can answer the question
A coding frame is the structured set of categories and rules used to classify content. Each category should have a name, definition, inclusion rule, exclusion rule, and at least one example. Categories at the same level should be conceptually comparable.
Begin with a small, varied subset of the data. For inductive work, note recurring distinctions and compare candidate categories. For deductive work, translate theoretical concepts into observable indicators. Hybrid frames can combine concept-driven categories with data-driven additions, but the process must be documented.
| Codebook field | Purpose | Illustrative entry |
|---|---|---|
| Category | Names the feature | Individual coping advice |
| Definition | States its meaning | Action a person is advised to take alone |
| Include | Defines positive cases | Breathing, sleep, scheduling, self-monitoring |
| Exclude | Separates neighbours | Professional treatment or institutional change |
| Unit | Fixes coding granularity | One recommendation statement |
| Decision rule | Handles ambiguity | Code each distinct action once per statement |
Make categories exhaustive and distinct where required
If each coding unit must receive one category, the options should cover all relevant cases and be mutually exclusive. Add a defined “not classifiable” category rather than forcing uncertain material. If multiple codes can apply, say so and specify whether repeated instances are counted.
Hierarchical frames can organise broad dimensions and narrower subcategories. For example, “source of responsibility” might include individual, family, institution, health service, community, and shared responsibility. Do not create so many narrow codes that each applies once and the comparison becomes meaningless.
Distinguish manifest and latent content
Manifest content is directly observable, such as the presence of a helpline, emotion word, diagnostic label, or image type. Latent content is an interpretation of underlying meaning, assumptions, or framing. Both can be valuable, but latent coding needs stronger contextual rules and reflexive explanation.
For instance, an image of a person meditating is manifest content. Interpreting a campaign as individualising responsibility for distress is latent. The second claim should draw on several features and their context, not one isolated image.
Pilot the coding process before full analysis
Piloting tests the codebook, units, workflow, and data-management system. Select material that includes typical, difficult, and boundary cases. Apply the frame, record uncertainties, revise definitions, and repeat until the rules are usable.
A pilot is not evidence that the categories are permanently fixed. In qualitative content analysis, revision may continue as interpretation develops. In a confirmatory quantitative study, changes after formal coding begins require version control, recoding rules, and transparent reporting.
Decide whether more than one coder is appropriate
Multiple coders are common in structured quantitative content analysis and some qualitative traditions, but they are not a universal quality requirement. The decision should fit the approach. A reflexive interpretive study may prioritise documented researcher positioning and analytic dialogue. A classification study making frequency comparisons may need independent double-coding and a chance-corrected agreement statistic.
Cofie and colleagues advise treating intercoder work as a process that includes codebook development, training, discussion, and transparent reporting, not as a single coefficient. Agreement should be assessed on an appropriate subset selected before outcome comparisons. Explain how disagreements were handled and whether resolved codes or initial independent codes entered the analysis.
Code systematically and preserve an audit trail
Create an immutable source archive and work on copies. Assign stable item and segment identifiers. Record the codebook version, coder, date, decision notes, and any changes. Keep a decision log for difficult cases and a reflexive journal for assumptions that may influence interpretation.
Software can organise material, retrieve coded segments, and export matrices. It cannot decide what a category means or whether an inference is defensible. Spreadsheet software may be adequate for a small structured dataset. Qualitative analysis software may help with larger text, image, audio, or video collections. Choose tools after defining the analytic procedure.
Follow the site’s data-collection guide for file naming, secure storage, version control, and analysis-ready data preparation.
Analyse categories without overclaiming
Qualitative interpretation
Examine what each category contains, how categories relate, where cases differ, and what contextual conditions shape meaning. Use selected extracts or image descriptions as evidence. Discuss negative cases that complicate the main interpretation. The final account should answer the question, not merely list code frequencies.
Quantitative description and comparison
Report item counts, category frequencies, percentages, or rates only with clear denominators. A post can contain several coded statements, so the number of coded instances is not necessarily the number of documents. Avoid treating non-independent segments as independent observations in statistical tests.
When comparing sources or periods, decide the level of analysis. For example, calculate the proportion of posts containing at least one stigma-reducing message rather than testing every sentence as if it came from a different source. Use effect estimates and uncertainty where appropriate, and interpret small or selective samples cautiously.
Integrate counts and meaning carefully
Counts can show distribution, while close interpretation explains form and context. The two should address related questions. If “self-care advice” appears in 70% of sampled pages, examine what actions are recommended, whose circumstances are assumed, and which structural pressures are omitted.
The data-analysis guide explains how to map questions to outputs, preserve raw data, handle exploratory decisions, and report uncertainty.
Evaluate quality and trustworthiness
Quality comes from a coherent chain between the question, dataset, coding decisions, evidence, and claims. Hsieh and Shannon emphasise that different approaches carry different threats to trustworthiness. APA’s qualitative reporting standards also encourage researchers to explain their design, data sources, analytic process, and researcher role clearly.
| Quality question | Useful evidence | Weak substitute |
|---|---|---|
| Is the dataset appropriate? | Bounded frame and justified sample | Large item count alone |
| Are categories usable? | Definitions, examples, pilot revisions | Code names without rules |
| Are interpretations grounded? | Contextual extracts and negative cases | Researcher assertion |
| Was coding consistent? | Documented procedure suited to approach | Unexplained percentage agreement |
| Are claims proportionate? | Limits tied to frame and units | General claims about all media |
Reflexivity is especially important for latent and theory-directed coding. State relevant expectations, positions, and interpretive commitments. Reflexivity does not remove subjectivity; it makes the researcher’s role available for evaluation.
Report content analysis across dissertation chapters
Introduction and literature review
Define the psychological problem, synthesise relevant research, identify the gap, and explain why recorded content can answer the question. Do not use method convenience as the main rationale.
Methodology
Name and justify the content analysis approach. Describe data sources, frame, dates, sampling, units, collection, ethics, codebook development, pilot work, coding procedure, researcher role, software, agreement or dialogue procedures, and analysis plan. The explanation should be detailed enough for readers to understand the decisions.
Results or findings
Organise findings around research questions or analytic categories. Give denominators for counts and contextual evidence for interpretations. Tables should clarify category distributions, not reproduce the entire codebook.
Discussion
Interpret the pattern in relation to theory and prior evidence. Discuss what the sampling frame, source context, coding approach, and researcher position allow you to claim. Avoid generalising from selected content to audience effects unless audience responses were studied.
Common content analysis mistakes and repairs
Calling any coding exercise content analysis
Repair: Name the specific approach and explain its assumptions, category logic, and output.
Sampling material opportunistically
Repair: Define the eligible universe, retrieval procedure, dates, and selection method before analysing patterns.
Using vague or overlapping categories
Repair: Add definitions, inclusion and exclusion rules, examples, and boundary-case decisions. Pilot again.
Reporting frequency as importance
Repair: Treat frequency as distribution within the sampled material. Consider salience, context, absence, and the consequences of less frequent representations separately.
Forcing data into a theory
Repair: Retain an explicit route for uncategorised or contradictory material and explain revisions to the directed frame.
Using quotations without analysis
Repair: Explain what each extract demonstrates, how it relates to the category, and whether other cases vary.
Claiming reliability from one unexplained number
Repair: Report coder preparation, the double-coded subset, statistic and rationale, disagreements, revisions, and final coding procedure.
A practical content analysis workflow
- Write a focused question and define the intended inference.
- Select and justify the qualitative, quantitative, directed, conventional, or summative approach.
- Define the sampling frame, eligibility rules, retrieval method, and dates.
- Set sampling, context, and coding units.
- Complete ethics and data-management review before collection.
- Build an initial coding frame from theory, data, or both.
- Pilot typical and difficult cases, then revise rules.
- Code systematically while preserving versions and decisions.
- Analyse distributions, meanings, exceptions, and context at the correct unit.
- Report the evidence chain, limitations, and researcher role transparently.
Psychology dissertation content analysis checklist
- The research question names the material and analytic purpose.
- The approach is named and justified.
- The sampling frame, dates, and selection process are reproducible.
- Sampling, context, and coding units are defined.
- Ethics, privacy, copyright, and quotation traceability are addressed.
- Categories have definitions, boundaries, and examples.
- Pilot revisions and codebook versions are documented.
- Coder procedures fit the methodological approach.
- Counts have correct denominators and units.
- Interpretations use contextual evidence and negative cases.
- Claims remain within the sampled content and design.
- Methods and findings follow appropriate reporting guidance.
Frequently asked questions
Can a psychology dissertation use only existing documents?
Yes. Documents, webpages, media, transcripts, images, or other recorded content can form a complete dataset when they directly answer the question. You still need a justified frame, systematic selection, ethics review, and a transparent analysis.
Is content analysis qualitative or quantitative?
It can be qualitative, quantitative, or combine interpretive and numerical elements. State which tradition you are using and ensure the coding, quality procedures, and claims match it.
How many documents do I need?
There is no universal number. The required material depends on the frame, unit, heterogeneity, question, analytic depth, and comparison plan. Justify adequacy in relation to the intended inference rather than citing a generic minimum.
Do I need two coders?
Not automatically. Independent coding is often important for structured classification studies, while some interpretive approaches use reflexivity and analytic dialogue instead. Follow the logic of the chosen method and explain the decision.
Can I count words and call it content analysis?
Word counts may form part of summative or quantitative content analysis, but they need contextual interpretation, defined retrieval rules, and a defensible link to the research question. Frequency alone rarely establishes meaning.
Can content analysis use social media data?
Yes, but public access does not remove privacy, consent, vulnerability, platform, or quotation risks. Seek institutional guidance, minimise identifiable data, and avoid reproducing searchable disclosures without strong justification.
Conclusion
Strong content analysis depends on visible research logic. Define the material, choose a coherent approach, specify units, build and pilot a clear coding frame, preserve an audit trail, and keep claims within the evidence. Counts should retain their denominators, interpretations should retain their context, and ethical safeguards should match the material rather than its apparent availability.
If you need support, use feedback to test alignment, category definitions, sampling logic, and reporting clarity while keeping authorship and analytic decisions your own. Ethical dissertation support should strengthen your reasoning, not replace your judgement or fabricate analysis.
References and further guidance
- Hsieh and Shannon: Three Approaches to Qualitative Content Analysis
- Elo and Kyngäs: The Qualitative Content Analysis Process
- Kleinheksel and colleagues: Demystifying Content Analysis
- Cofie, Braund, and Dalgarno: Intercoder Reliability Guidance
- APA Style: Qualitative Research Reporting Standards
- Assarroudi and colleagues: Directed Qualitative Content Analysis
