Description
Determines empirically when additional qualitative data stop yielding new codes or category properties, justifying sample size with evidence instead of convention.
Glaser and Strauss defined theoretical saturation as the point where new data add no new properties to categories. In practice 'we reached saturation' is often asserted without evidence. Guest, Bunce and Johnson found that in a relatively homogeneous sample about 92% of codes emerged within the first 12 interviews; Guest, Namey and Chen (2020) proposed a transparent method using a base size (for example 4 interviews), a run length (for example 2 further interviews) and a new-information threshold (for example 5% or less) to declare saturation. Hennink and colleagues distinguished code saturation, hearing it all, from meaning saturation, understanding it all, which takes longer. Saturation detection tracks new codes per interview, applies these thresholds and reports them, separating code from meaning saturation.
Attributes
- Discipline
- Research Design & Rigor
- Loss type
- Vestigial
- Stage
- Discovery
- Data
- Mixed
- Effort
- Hours
Steps
- Define the saturation type you need: code saturation for descriptive inventories, or meaning or theoretical saturation for explanatory work.
- Code interviews in the order they were conducted and record, for each interview, the number of new codes added to the codebook.
- Set parameters in advance: a base size (for example the first 4 interviews), a run length (for example 2 or 3 interviews) and a new-information threshold (for example 5% or less of the base codebook).
- After each run, compute the ratio of new codes in the run to the base codebook size; declare code saturation when the ratio falls at or below the threshold.
- For meaning saturation, track whether each important code's properties and dimensions are still changing, and continue sampling until the main codes are fully understood.
- Check saturation within each important subgroup separately (for example each region), since saturation across a pooled heterogeneous sample can hide gaps.
- Report the saturation curve, parameters and subgroup results in the methods section.
Sources
- Rocking Qualitative Social Science
- Sage Handbook of Qualitative Data Analysis (Flick)
- Qualitative Research from Start to Finish (Yin)
- The Inquiry Series, Ch 26 — Grounded Theory — Generating Theory from Data
Also known as
- Data saturation assessment
- Theoretical saturation
- Code saturation analysis
- Saturation curve
Categories
- Research Operations
- Research Design & Rigor