COM5101 Chap.4 Research Traditions in Communication
Research Traditions in Communication
Research design begins with inference
Quantitative, qualitative and computational approaches are not rival brands. They answer different questions by standardising, contextualising or scaling evidence. Choose among them by asking what the conclusion must establish.
Estimating a pattern, reconstructing a meaning process and tracing millions of recorded events require different bargains.
Standardisation enables comparison
Surveys, experiments and structured content analysis create common measures across cases. This supports estimation and formal comparison, but only after sampling, operation and causal design have been assessed separately.
A large sample cannot repair a measure that consistently captures the wrong construct, and a precise association does not establish a causal direction.
Interpretation preserves process
Interviews, observation and close analysis can show how participants understand a message, negotiate a relationship or produce a category. The researcher participates in interpretation.
Rigour therefore depends on transparent case selection, engagement with alternative readings and a traceable route from material to conclusion rather than on pretending interpretation disappeared.
Computational methods add scale and transformation
Computational communication research often works with large, complex and naturally occurring records.
Algorithms can classify, cluster or model these materials, but their output does not interpret itself. Training data, platform selection and model choices become parts of the measurement process. Statistical precision and hermeneutic interpretation can complement one another when their roles are stated.
Integration needs a handoff
A study is not mixed simply because it uses more than one tool.
One method should change the design, measure or interpretation of the next. Interviews might define categories for structured coding; a large pattern might identify contrasting cases for close analysis. Name that handoff so the reader can see why the combination is stronger.
What this chapter covers
- 01
Inference first
- 02
Quantitative comparison
- 03
Qualitative interpretation
- 04
Computational scale
- 05
Sampling and causation
- 06
Integrated design
Worked example · free
Study disagreement in remote teams
- 2Identify the target inference.
- 2Assign each method a job.
- 2Connect the methods.
- 2Limit the conclusion.
Key terms
- Quantitative Design
- A design that standardises observations to estimate patterns, differences or relationships across cases.
- Interpretive Design
- A design that reconstructs how meaning is produced in context through transparent analysis of situated material.
- Computational Method
- An algorithmic approach used to process or model large, complex or naturally occurring communication records.
- Method Integration
- A design in which one method informs the sampling, measurement or interpretation of another.
Research Traditions in Communication FAQ
Should the research question or method come first?
The research question and required inference should come first. Choose a method because it can produce the evidence needed for that conclusion. Starting with a fashionable tool often leaves a mismatch, such as using message volume to answer a question about meaning or interviews to estimate population prevalence.
What makes qualitative research rigorous?
Rigour comes from transparent selection, sustained engagement with the material, a traceable analytic process and serious treatment of alternative interpretations. The researcher does not disappear from analysis. Instead, the account makes interpretive decisions visible so a reader can evaluate how the conclusion was reached.
When is a design genuinely integrated?
Methods are integrated when one changes the other. Interviews may define categories for coding, a trace pattern may select cases for close analysis, or an experiment may test a mechanism discovered through observation. Merely placing a survey and interview side by side does not show why their combination improves inference.
Can one dataset support every research question?
No. A dataset can be strong for estimating a recorded pattern and weak for reconstructing private meaning or causal direction. Evaluate sampling, measurement and design separately. Name the inference the material supports, then identify the additional evidence required for a stronger conclusion.
Assessment move
Turn every method name into an inference sentence: use this design to learn this kind of thing because it preserves or standardises this feature. Then add the principal limitation. Practise with one research question and design three versions, quantitative, interpretive and computational, making the conclusion different each time. This shows that method choice is not a cosmetic change.
In the paper, organise methodological critique around the claim each design can support rather than around generic advantages and disadvantages. In a presentation, visualise the handoff between methods so integration is visible. Build one sequential design each week.
Start with the phase that discovers categories or estimates a pattern, then specify exactly what it passes to the next phase: a codebook, a set of contrasting cases, a hypothesis or an unexplained result. If nothing passes across, the design is parallel rather than integrated. Review verbs after reviewing methods.
Highlight estimates, interprets, predicts and causes in different colours, then check whether the design earns each verb. This quick audit catches the moment a precise estimate turns into an unsupported causal story or a deep interpretation turns into a claim about prevalence.
