> ## Documentation Index > Fetch the complete documentation index at: https://docs.voxworks.ai/llms.txt > Use this file to discover all available pages before exploring further. # Attempts and step levels > How many times the survey interviewer re-asks, reads back, clarifies and nudges before moving on, and how quickly it takes turns: the survey-wide floor in the Agent section, per-question overrides, each attempt cap, the Eagerness, Idle and Thinking levels, and the FAQ and behaviour sets. ## Two kinds of setting * **Attempt caps** decide *how persistent* the interviewer is — how many times it re-asks, reads back, clarifies or nudges before it records a question as skipped and moves on. * **Step levels** decide *the pace of the conversation* — how quickly a pause ends the respondent's turn, how long a silence is tolerated, and how much the interviewer reasons about each reply. Both are set once for the whole survey and can be overridden on individual questions. *** ## Where they're set | Level | Where | Applies to | | ----------------- | ---------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------- | | **Survey** | **Agent** section → **Attempts** group (*"The floor every question inherits unless it states its own."*) and **Step levels** group | Every question that doesn't set its own. | | **Question** | The question's **Behaviour** panel | That question only. | | **FAQ set** | **FAQs** section → step levels | The interviewer's answers to your FAQs. | | **Behaviour set** | **Behaviours** section → step levels | The busy and stop closing lines. | See [Agent section](/surveys/agent), [FAQs](/surveys/faqs) and [Behaviours](/surveys/behaviours) for those sections' other settings. *** ## Attempt caps Each cap is a whole number from 0 to 50. | Cap | Hint in the Agent section | What it limits | Default | | -------------- | --------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------- | | **Ask** | *"Read-outs of the question."* | How many times the question is asked. Silence, "say that again" and a turn that captured nothing each use one. When it runs out, the question is skipped. | 5 | | **Confirm** | *"Confirm read-back turns."* | How many read-backs and "did you say…?" checks the question can spend. When it runs out, the question is skipped. See [Answer confirmation](/surveys/answer-confirmation). | 5 | | **Clarify** | *"Off-list turns before a skip."* | How many times the interviewer asks again after an answer it couldn't use — off the list, unclear, or the wrong shape of number. | 5 | | **Don't know** | *"Nudges; 0 accepts the first don’t-know."* | How many times a "don't know" is nudged before the question is skipped. | 1 | | **Decline** | *"Nudges; 0 accepts the first refusal."* | How many times a refusal of the question is nudged before it's skipped. | 1 | | **Acks** | *"Listening turns that do not spend an ask."* | How many bare "okay"s and "uh-huh"s, said while the question is still being read, are let pass without using an **Ask**. | 3 | | **Spell** | Question only — First name | How many times the interviewer asks for or checks a spelling. | 5 | **Spell** only appears in the **Behaviour** panel of a [First name](/surveys/question-types/first-name) question; there's no survey-wide Spell cap. ### Choosing caps The defaults are generous. Many surveys run better with lower numbers — two asks, two confirms and two clarifies keeps the interview moving when a respondent is struggling with a question, rather than asking five times. * **Don't know** and **Decline** at 1 give one gentle nudge, which recovers a surprising number of answers without pressuring anyone. Set them to 0 on sensitive questions — income, health, voting — so a "don't know" or "I'd rather not say" is accepted straight away. * **Required** questions always nudge a refusal at least once. If a question is required and its **Decline** is 0, the question inspector warns *"Required questions nudge a refusal at least once, but decline nudges is 0 in Behaviour. Raise it to 1 or turn Required off."* and Readiness warns that the question *"is required but accepts the first refusal (attempts.decline is 0)."* The interview behaves as if **Decline** were 1. * **All caps at 0** makes a question say its prompt once: a clearly heard answer is still recorded, but nothing is re-asked, read back or nudged — useful for a notice authored as a question (with **Statement prompt** on). ### Overriding on a question In the question's **Behaviour** panel: * **Inherit survey default** on (the default) — the fields show the survey's caps, greyed out: *"The survey’s own Behaviours, from the Agent section. Nothing is stored on this question."* Change the survey's caps later and this question follows. * **Inherit survey default** off — the fields become editable, and this question keeps its own numbers (*"Nudges before a don’t-know or a refusal is recorded. 0 accepts the first one."*). Later changes to the survey's caps no longer affect it. A question that has its own caps stays pinned to them until you turn **Inherit survey default** back on. *** ## Step levels | Level | Hint in the Agent section | Lower means | Higher means | | ------------- | --------------------------------------- | --------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------- | | **Eagerness** | *"How quickly a pause ends the turn."* | The interviewer waits longer through pauses before treating the respondent as finished. | The interviewer replies as soon as the respondent pauses. | | **Idle** | *"Silence tolerated before a prompt."* | A shorter silence before the interviewer prompts again. | A longer silence before prompting. | | **Thinking** | *"Reasoning effort on instruct calls."* | Faster replies, less deliberation over what an answer meant. | More deliberation over difficult answers, at the cost of a slightly longer pause. | Each can be set to low, medium or high. ### How levels are resolved * **On a question**, each level is **Template default** or low / medium / high. Left on **Template default**, the question follows the survey's **Step levels** in the Agent section. * **On the survey** (Agent section), each level is **Template default**, **Low**, **Medium** or **High**. Left on **Template default**, the built-in levels apply: **Eagerness** high, **Idle** medium, **Thinking** low. * **First name** and **Phone number** questions wait longer through pauses by design, because names are spelled letter by letter and numbers are read in groups. Leave their **Eagerness** on **Template default** unless you have a reason to change it. ### Recommended settings | Situation | Setting | | ------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------ | | Most surveys | Leave **Idle** at medium survey-wide. Low can re-prompt people who are simply taking a few seconds to weigh up a rating. | | Open [Long text](/surveys/question-types/long-text) questions | **Eagerness** low, so a pause mid-thought isn't taken as the end of the answer. | | A postcode or other multi-digit [Number](/surveys/question-types/number) | **Eagerness** medium. | *** ## The FAQ and behaviour sets FAQ answers and the busy and stop closing lines have **their own** step levels, set in the **FAQs** and **Behaviours** sections. They don't follow the survey's step levels, so slowing the questionnaire down for your audience doesn't slow down a one-line answer to *"who is calling?"*. * **FAQ set** — *"The gates every FAQ answer above is spoken under — its own, not the survey's. An answer is one line said over a question that is still open, so it waits briefly and hands straight back."* * **Behaviour set** — *"The gates the busy and stop closes are spoken under — their own, not the survey's. The reason is already recorded and the call is ending, so the line is delivered and the wait after it is short."* In these two sets an unset level reads **Survey default**. When unset, **Eagerness** is high (*"Unset here: high."*) and **Idle** is low (*"Unset here: low."*); **Thinking** follows the survey's. *** ## Next Steps * [Agent section](/surveys/agent) — where the survey-wide caps and levels live. * [Question settings](/surveys/question-settings) — the Behaviour panel in context. * [Answer confirmation](/surveys/answer-confirmation) — what the Confirm cap counts.