4.2.3.2 - Coding in content analysis
Coding turns complex communication into analysable categories, so the quality of the findings depends on the rules used to classify each unit. You will define units of analysis, build mutually clear categories, record frequencies and test agreement between coders. The lesson also examines how a coding frame can improve consistency while stripping away context or reflecting the assumptions of its designers.
What coding does
Content analysis studies communication material such as articles, adverts, interviews, videos, speeches, posts or images. Coding is the part of content analysis where the researcher decides exactly what features of the material will be recorded.
Coding
Coding is the process of applying labels or categories to units of material so that patterns in communication can be recorded, counted or interpreted systematically.
The key idea is that coding turns messy qualitative material into organised evidence. A raw comment such as "These exam tips made me feel calmer, but I still worry about timing" is not yet a number. A researcher can code it for features such as:
- emotional tone: mixed;
- mention of anxiety: present;
- mention of coping or reassurance: present;
- practical concern: exam timing.
That does not mean the original meaning disappears. It means the researcher has created a clear way to compare many pieces of material without relying only on vague impressions.
Code
A code is a short label attached to a feature of the material, such as "supportive", "threat", "recovery", "aggression" or "stereotype".
Coding category
A coding category is a defined class into which units of material are placed. In a quantitative content analysis, categories are usually designed so the researcher can tally frequencies.
Coding frame
A coding frame is the full set of coding categories, rules and examples used by coders when analysing the material.
For AQA, the most important skill is to explain how coding is made systematic: the researcher decides the unit of analysis, creates clear categories, applies the same rules to all material and records counts or classifications in a consistent way.
This lesson is narrower than the earlier lesson on content analysis as a method. Here, the focus is not "What is content analysis?" but "How does the researcher code the material well?"
Units of analysis
Before coding begins, the researcher must decide what counts as one item to be coded. This is the unit of analysis.
Unit of analysis
The unit of analysis is the portion of material that is coded as one item, such as a word, sentence, image, scene, comment, article or whole video.
Different units answer different research questions. A researcher investigating threat-related language might code individual words. A researcher investigating whether characters with mental health difficulties are shown sympathetically may need to code whole scenes or whole films.
| Unit of analysis | Example coding decision | Strength | Limitation |
|---|---|---|---|
| Word | Count each use of "dangerous", "unstable" or "violent" in newspaper articles. | Precise and easy to tally. | Can miss context, such as whether the word is rejected or challenged. |
| Phrase or sentence | Code each sentence as supportive, stigmatising or neutral. | Keeps more meaning than single words. | Longer sentences may contain more than one idea. |
| Image or shot | Code whether an advert image shows a parent doing paid work, housework or leisure. | Useful for visual material. | Coders need rules for ambiguous images. |
| Scene | Code each film scene as threat-focused, recovery-focused or neutral. | Retains context and interaction. | Takes longer and involves more judgement. |
| Whole post/article/video | Code each complete item for its overall portrayal. | Good for overall tone or message. | Can hide variation within the item. |
A good unit is neither so tiny that it strips away meaning nor so broad that it hides important differences. The unit should match the aim.
Example. Dr Morgan wants to compare how two news websites portray teenagers in crime stories. If the aim is to count labels such as "gang", "victim" and "suspect", the word or phrase may be the right unit. If the aim is to judge the overall portrayal of teenagers in each story, the whole article may be the better unit.
The researcher should also decide whether a unit can receive more than one code. For example, one social media post might be both "supportive" and "recovery-focused". If the study requires one category only, the coding frame must say which category takes priority.
Building coding categories
Coding categories need to be clear enough for another coder to apply. A category such as "negative" is usually too vague on its own. A category such as "threat portrayal: the item endorses describing the person or group as dangerous, violent, unpredictable or a risk to others" is much more usable. A quoted label that the item rejects is excluded.
Operationalisation
Operationalisation means defining an abstract idea in a measurable or observable way.
In coding, operationalisation means turning ideas such as "stigma", "gender stereotype" or "aggression" into rules that coders can use.
| Vague category | Better operationalised category |
|---|---|
| Negative language | Words or phrases that describe the person or group as dangerous, irresponsible, weak or blameworthy. |
| Aggression | Any spoken threat, insult, physical hit, push, thrown object or deliberate damage to property. |
| Gender-role portrayal | Record the stated gender and activity of each character, or code an explicit statement that an activity is suitable only for one gender. A single character doing childcare does not by itself establish a stereotype. |
| Recovery message | The item explicitly mentions treatment, support, coping, improvement or returning to everyday life. |
Strong coding categories should be:
- relevant to the research question;
- clear enough that coders know what counts;
- non-overlapping if only one category can be chosen;
- exhaustive enough for the task, often including "other" or "unclear" if real material may not fit neatly;
- supported by examples, especially for borderline cases.
Non-overlapping categories
Non-overlapping categories are categories that are defined so the same unit does not fit more than one category, unless the coding frame deliberately allows multiple codes.
Here is a small coding frame for analysing newspaper headlines about teenagers:
| Category | Code | Count if the headline... | Do not count if... |
|---|---|---|---|
| Threat | T | describes teenagers as dangerous, violent, criminal or a risk to others. | it reports that teenagers are victims of threat. |
| Responsibility | R | describes teenagers as helping, leading, achieving or taking positive action. | it only reports that an adult helps teenagers. |
| Vulnerability | V | describes teenagers as victims, at risk, exploited or needing support. | it uses emotional language but gives no risk or victim meaning. |
| Neutral | N | reports a teenager-related event without threat, responsibility or vulnerability language. | another category clearly applies. |
This frame allows multiple T, R and V codes when a headline contains more than one portrayal. N is used only when none applies. For a one-choice study, the researcher must instead add an explicit priority rule before counting.
Notice the "do not count if" column. It is easy to write categories; it is harder, and more useful, to write exclusion rules. This is what prevents coders from guessing when material is ambiguous.
Pre-set and revised categories. Researchers may begin with pre-set categories based on previous research or a clear aim. They may also pilot the coding frame on a small sample and revise unclear categories before the main analysis.
From words to counts
Coding often converts qualitative material into quantitative data. The original material may be words, images or video, but the coded output can be a frequency, percentage, rank, score or category count.
Frequency count
A frequency count is the number of times a category, code or event occurs.
The simplest approach is a tally chart. Imagine a researcher codes 20 public posts about exam stress. The unit of analysis is one complete post. Coders apply one main code to each post.
Continue this lesson.
You’re previewing NovaLearn without an account. Create a free account to read the rest.
Create free account