Skip to content

Staircase Experiments

Experiments

Staircase procedures automatically adjust the difficulty of trials based on a participant's performance, making them well suited for threshold estimation and signal-detection tasks. In Testable, staircase parameters are defined in the **`staircase`** column.

You prepare a sequence of trials in ascending difficulty order (easiest to hardest). The staircase procedure tracks correct and incorrect responses and adjusts which difficulty level to present next by moving up (harder) or down (easier) according to defined rules.

Place the staircase parameters on the first trial of the sequence:

staircase = A: start 3, down 1, up 2, correct 2, booster 2, incorrect 1, reversal 3

On subsequent trials in the same staircase, just reference the label:

staircase = A
ParameterMeaning
start NStart at difficulty level N (1-indexed, row position in staircase)
down NMove this many steps down (easier) after a correct trial
up NMove this many steps up (harder) after an incorrect trial
correct NNumber of consecutive correct responses required to step down
incorrect NNumber of consecutive incorrect responses required to step up
booster NQuick step-down size used after the first correct response
reversal NStop after N reversals (direction changes)
trials NStop after N total trials regardless of reversals

A reversal occurs whenever the direction of adjustment changes (from going up to going down, or vice versa). Specifying reversal 3 terminates the staircase after 3 direction changes, which produces stable threshold estimates.

Use trials N if you want a fixed number of trials regardless of performance.

You can run several independent staircases in the same experiment by giving them different labels:

staircase = A: start 3, down 1, up 2, correct 2, booster 2, incorrect 1, reversal 3
staircase = B: start 1, down 1, up 2, correct 2, booster 1, incorrect 1, reversal 4

Group multiple staircases together by appending a :group suffix:

staircase = A:group1
staircase = B:group1

Testable interleaves the trials from grouped staircases, alternating between them throughout the session.

Staircase experiments produce standard result files. The difficulty level presented on each trial is recorded alongside the response, allowing you to compute the threshold from the reversal points.