Chapter Three of a mass communication project in Nigeria is built on one of two designs: a content analysis of media output, or a survey of an audience. Each has its own population, sampling rule, instrument and reliability test, and most returned chapters fail because a survey template was used to describe a content analysis. This guide writes the chapter for both designs, section by section, with the sentences a panel in Nsukka, Lagos or Zaria expects to read.
The generic Chapter Three, with its Taro Yamane calculation and questionnaire validation, is covered in the guide to writing Chapter Three section by section. That template fits the audience survey half of this discipline and fails the other half, because a newspaper does not fill in a questionnaire. What follows keeps the departmental section order and shows, at each section, what changes when the units you study are stories, broadcasts or posts rather than people.
Step 1: Choose the design from the research questions, not from habit
Expected output: one sentence naming the design and the reason.
Read your research questions. If they ask what the media carried (how much coverage, in what frame, with what prominence), the design is content analysis. If they ask what an audience knows, prefers, believes or does, the design is a survey. If they ask both, the project is a mixed design and Chapter Three has two of everything, written in parallel. The sentence to write:
The study adopted the content analysis design because the research questions concern the volume, prominence and direction of coverage of the 2023 general election in selected Nigerian newspapers, which can only be answered by examining the newspapers themselves.
Berelson’s 1952 book Content Analysis in Communication Research and Holsti’s 1969 Content Analysis for the Social Sciences and Humanities are the two primary sources Nigerian departments expect for the design’s definition; both are catalogued in Crossref through their contemporary reviews, and citing one of them is enough.
Step 2: Define the population in the units the design studies
Expected output: a population sentence with a count, and a stated unit of analysis.
For a survey the population is people, counted from a source you can cite. For a content analysis the population is editions, programmes or posts, and it is countable from the calendar. Three national dailies studied over the six months from 1 October 2022 to 31 March 2023 have a population of three titles times 182 publishing days, which is 546 editions; state the number. Then state the unit of analysis, which is the thing you will code: the news story, the editorial, the front page, the tweet. A common fault is a population of newspapers with a unit of analysis never stated, so the panel cannot tell whether the sample is 60 editions or 600 stories.
| Design | Population | Unit of analysis | Sampling frame |
|---|---|---|---|
| Content analysis, print | All editions of the selected titles in the study period | News story, editorial, feature, photograph | Publishing calendar of each title |
| Content analysis, broadcast | All bulletins or programme episodes in the period | News item, segment | Broadcast schedule; recordings or transcripts obtained |
| Content analysis, social media | All posts by named accounts or under a named hashtag in the period | Post, comment | Archived posts retrieved on a stated date |
| Audience survey | Residents, students or listeners of a named place or station | Individual respondent | Enrolment list, population projection, or listener estimate with its source |
Step 3: Sample editions with a rule you can name
Expected output: a sampling paragraph that names the technique and its source.
Newspaper content varies by day of the week, so a random draw of dates can over-sample Sundays. The constructed-week technique fixes this: draw one Monday, one Tuesday and so on at random from the study period until each weekday is represented, and repeat for as many constructed weeks as the period needs. Riffe, Aust and Lacy tested this against random and consecutive-day sampling in 1993 (Journalism Quarterly, 70(1), 133 to 139) and found the constructed week efficient, which is the citation the panel expects beside the technique. For a six-month study, two constructed weeks per title (14 editions each, 42 in total) is a defensible sample; state the number of stories those editions yielded as well, since the story is the unit.
For a survey, the sampling section is the ordinary one: population figure with its source, sample size by formula, and a multi-stage procedure from state to local government to ward to household or hall. The worked Yamane calculation and the proportional allocation are in the generic guide already linked; do not re-derive them here, only apply them.
Step 4: Build the coding sheet as your instrument
Expected output: a coding sheet in the appendix and a category paragraph in the chapter.
The instrument of a content analysis is the coding sheet, and Chapter Three describes it the way a survey chapter describes a questionnaire. Each research question needs at least one content category, and each category needs mutually exclusive, exhaustive values. The standard set in Nigerian mass communication projects:
| Category | What it records | Typical values |
|---|---|---|
| Frequency | How many stories on the subject | Count per edition and per title |
| Prominence | Where the story was placed | Front page, back page, inside page; lead or non-lead |
| Story type | The genre | Straight news, feature, editorial, opinion, photograph |
| Direction or slant | The tone toward the subject | Favourable, unfavourable, neutral |
| Frame | The dominant interpretive frame | Conflict, human interest, economic consequence, responsibility, morality |
| Source | Who was quoted | Government official, party official, security agency, citizen, expert |
| Size | Space given | Column centimetres or word count |
Write one operational definition per value in the appendix so that a second coder could apply it without you. Where the frames come from a theory, name it here as the source of the categories and leave the argument for Chapter Two: Entman’s 1993 clarification of framing (Journal of Communication, 43(4), 51 to 58) and McCombs and Shaw’s 1972 agenda-setting study (Public Opinion Quarterly, 36(2), 176 to 187) are the two usually cited, and Katz, Blumler and Gurevitch’s 1973 paper (Public Opinion Quarterly, 37(4), 509 to 523) anchors an audience survey on uses and gratifications. All four were checked against Crossref on 30 August 2026.
Step 5: Establish validity and reliability the content-analysis way
Expected output: a face-validity sentence and an intercoder reliability coefficient with its formula.
Validity of a coding sheet is established the same way as a questionnaire’s: the supervisor and two other lecturers review the categories against the research questions. Reliability is different. A content analysis does not compute Cronbach’s alpha; it computes agreement between two coders who code the same subsample independently. The simplest coefficient, and the one most Nigerian departments accept, is Holsti’s: reliability equals 2M divided by (N1 plus N2), where M is the number of coding decisions on which the two coders agree and N1 and N2 are the decisions each coder made. Coding 40 stories on seven categories gives 280 decisions each; 252 agreements gives 2 × 252 ÷ 560 = 0.90. Lombard, Snyder-Duch and Bracken’s 2002 review of reliability reporting in mass communication (Human Communication Research, 28(4), 587 to 604) is the standard citation for the reporting rule, and Krippendorff’s 2004 paper in the same journal (30(3), 411 to 433) is the source if your supervisor asks for a coefficient that corrects for chance. Report the coefficient per category, not only overall, and say what you did about the category that scored lowest.
For a survey, the pilot and alpha are the ordinary route, and the answer to what a good Cronbach’s alpha is covers the thresholds; a mass communication questionnaire about media use often has knowledge items and preference items in one instrument, so compute the alpha per section.
Step 6: Say how the material was obtained and coded
Expected output: a method-of-data-collection paragraph that a reader could repeat.
Name where the editions came from (the newspaper’s own archive, the departmental library, the National Library, an online archive read on a stated date), who coded them (you alone, or you and one trained coder), over what period, and how disagreements were resolved. For broadcast studies, say whether you worked from recordings, station transcripts or monitoring notes, and for social media studies give the retrieval date, because a post count changes daily. Panels ask about this section because it is the one that decides whether the study could be repeated.
Step 7: Match the analysis to the data you will actually have
Expected output: a method-of-data-analysis paragraph naming the tables and the test.
Content analysis produces counts, so the analysis is frequency and percentage tables by title and by category, and where a hypothesis compares titles (whether prominence differed significantly between the three newspapers), the test is chi-square on the cross-tabulation. Say so in one sentence and give the significance level. Survey data goes through the mean-score and chi-square route of any Nigerian questionnaire project, and which test fits which question is set out in the answer to which statistical test a final year project should use. Whichever design, Chapter Four then presents one table per research question in the order of the questions, following the guide to writing Chapter Four.
Step 8: State the ethical position the design raises
Expected output: two sentences.
A content analysis of published material raises no consent issue, and say so rather than pasting a consent paragraph from a survey template; what it raises is fair attribution, so cite every title and edition. A survey needs the ordinary consent and confidentiality sentences. Where the topic touches journalists’ conduct, the Nigerian Press Council’s Code of Ethics for Nigerian Journalists, which opens with editorial independence and accuracy and fairness, is the reference document, and it was online at presscouncil.gov.ng on 30 August 2026; the National Broadcasting Commission’s site was under maintenance the same day, so if your topic needs the Nigeria Broadcasting Code, obtain a copy early and cite the edition.
A worked audience-survey section for comparison
Topic: Influence of social media on news consumption among undergraduates of the University of Nigeria, Nsukka. The design sentence names the survey; the population is the enrolment figure from the university’s records office with the year; the sample is by Yamane at the 0.05 level and allocated proportionally across faculties; the instrument is a structured questionnaire in three sections (bio-data, platform use, news-consumption behaviour) with four-point Likert items; validity is by three lecturers; reliability is a pilot of 30 students in another faculty with alpha per section; analysis is mean scores against the 2.50 criterion and chi-square for the hypotheses. One caution for the background of that project: the Reuters Institute’s 2025 Digital News Report for Nigeria records Facebook at 65%, WhatsApp at 53%, YouTube at 49% and X at 49% for news use, but states that its respondents are English-speaking online news users who are younger, more educated and more urban than the population, and that the findings should not be taken as nationally representative. Quote the figures with that caveat attached, or the panel will attach it for you.
Draft both designs and keep only the one your questions need
Tesify drafts Chapter Three from your research questions, so a content analysis gets a population of editions, a constructed-week sampling paragraph, a coding-sheet description and an intercoder reliability sentence, while a survey gets the questionnaire route, and neither borrows the other’s template. The automatic bibliography formats Berelson, Holsti, Riffe and the reliability papers in APA or your department’s style. There is a free plan and no card is required.
Frequently asked questions
What research design is used for a mass communication project in Nigeria?
Content analysis when the research questions are about media output, a survey when they are about an audience, and a mixed design when they are about both. The design is chosen from the questions, and Chapter Three is written differently for each.
What is the population in a content analysis project?
The editions, programmes or posts in the study period, counted from the publishing or broadcast calendar. Three dailies over 182 days is a population of 546 editions. The unit of analysis, usually the story, is stated separately.
How do I sample newspapers for a content analysis?
By constructed week: draw one of each weekday at random from the period until every weekday is represented, and repeat as needed. Cite Riffe, Aust and Lacy (1993) for the technique and state how many editions and stories the sample yielded.
What is the instrument in a content analysis?
The coding sheet, with one or more content categories per research question and mutually exclusive values for each category. Its operational definitions go in the appendix, and it is described in Chapter Three as a questionnaire would be.
How is reliability tested in a content analysis?
By intercoder reliability, not Cronbach’s alpha. Two coders code the same subsample independently and agreement is computed, most simply with Holsti’s formula, 2M divided by N1 plus N2. Report the coefficient per category.
What is a good intercoder reliability score?
Nigerian departments generally accept 0.80 and above on Holsti’s formula. Below that, revise the operational definitions of the weakest category, retrain, and code the subsample again before the main coding.
Which statistical test is used in a content analysis project?
Frequencies and percentages answer the research questions, and chi-square on a cross-tabulation tests hypotheses that compare titles, periods or story types, at the 0.05 level.
Does a content analysis need ethical clearance or consent?
Published material needs no respondent consent, and the chapter should say so plainly. Fair attribution of every title and edition is the obligation. A survey component needs the ordinary consent and confidentiality statements.
Can I cite the Reuters Digital News Report for Nigeria in my project?
Yes, with its own caveat: the report states that its Nigerian sample is of English-speaking online news users who are younger, better educated and more urban than the population, and that the findings are not nationally representative.