> For the complete documentation index, see [llms.txt](https://api-docs.speechace.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://api-docs.speechace.com/interpreting-results/interpreting-all-pause-list.md).

# Interpreting all pause list

The `all_pause_list[]` field reports every pause Speechace detected in the speaker's audio. It appears in the `fluency` node of a [Score Text](https://api-docs.speechace.com/api-reference/score-text.md) response when you set `include_fluency=1`.

## What it contains

* Each entry is a `[begin, end]` tuple of **extent markers** for one pause, in **units of 10 ms**, so a pause's duration in milliseconds is `(end - begin) × 10`. These are the same 10 ms extents used elsewhere in the response (see [Getting word timestamps in audio](https://api-docs.speechace.com/interpreting-results/getting-word-timestamps-in-audio.md)).
* The list is **exhaustive**. It contains every pause, down to very short 10 ms gaps.
* It includes both **unfilled** pauses (silence) and **filled** pauses (fillers such as "um" or "uh"). A pause can be either.

Because the list is exhaustive, it mixes trivial gaps with meaningful ones, so don't present it to a learner as-is. Filter and segment it first.

## Filter for pauses of a meaningful length

Keep only pauses long enough to matter. Convert each entry with `(end - begin) × 10` ms, then bucket by length. A common split:

| Category | Duration    |
| -------- | ----------- |
| Short    | 300–500 ms  |
| Medium   | 500–1000 ms |
| Long     | ≥ 1000 ms   |

## Segment to find where a pause is (in)appropriate

A pause is only a disfluency if it falls where a fluent speaker wouldn't pause. Apply segmentation to find the legitimate points for short, medium, and long pauses, for example using the prompt's punctuation or text segmentation, then judge each detected pause against those boundaries. A long pause at a sentence boundary is natural; the same pause mid-phrase is a disfluency.

## Present feedback

Once the pauses are filtered and segmented, you can give the learner targeted feedback. Rather than showing a raw list of every gap, highlight the places where disfluencies or inappropriate pausing occurred, such as mid-phrase hesitations, overlong stops, or excessive fillers.
