Plain-text format for structured data, organized as nested key-value pairs and lists. Readable by humans, parseable by every modern programming language. Handles hierarchical structures like transcripts with speakers, timestamps, and coding tags without flattening them into rows.
Definition: Plain-text format for structured data, organized as nested key-value pairs and lists. Readable by humans, parseable by every modern programming language. Handles hierarchical structures like transcripts with speakers, timestamps, and coding tags without flattening them into rows.
JSON stores data as nested objects (key-value pairs) and arrays (ordered lists), letting one record carry a full hierarchy: a single interview can hold metadata, speaker turns, timestamps, code tags, and quote-level annotations in one file. CSV cannot do this without flattening, which destroys context.
For research exports, JSON wins on three counts: nothing is forced into a single shape, every field has a name (no positional guessing), and every modern programming language has a built-in parser. The trade-off is that JSON is less convenient to open in a spreadsheet; for spreadsheet-friendly views, export both formats.
Look for "newline-delimited JSON" (NDJSON or JSONL) when datasets are large: each line is one record, which lets you stream and filter without loading everything into memory.
This term is referenced in the following articles: