Balloon risk game: method

What it measures

Behavioural risk taking when each step adds a small reward but also the risk of losing everything gained on that trial. The adjusted average pumps is the standard index.

Procedure

30 balloons in sequence. Each pump (button, or the Space or P key) adds 5 points to the current balloon and inflates it; 'Collect' (button, or Enter or C) adds the balloon's points to the total and shows the next balloon. Each balloon bursts at a whole-number pump drawn uniformly from 1 to 128, generated on the server from a per-run seed (the same distribution as the original 1/128, 1/127, … schedule); reaching that pump bursts the balloon and its points are lost. No practice and no time limit; leaving the tab does not end the run. The page sends each balloon's pumps and whether it burst; the server regenerates the burst points from the seed and checks every balloon.

Parameters

Scoring

Primary metric: adjusted average pumps = mean number of pumps on balloons that were collected (not burst), to one decimal. Neither higher nor lower is 'better'. Also reported: total points (5 × pumps on collected balloons), balloons burst and collected, total pumps, and the adjusted average for each third of the run (balloons 1–10, 11–20, 21–30). A descriptive style label is our own and not from the literature: under 16 'Careful', 16 to under 32 'Steady', 32 to under 48 'Bold', 48 or more 'Daredevil' (also used when every balloon burst). A published reference value is shown next to the result: 128-pump balloons in Lejuez et al. (2002), n = 86, mean 29.4, SD 13.0.

Valid-run rules

A run must have exactly 30 balloons; a burst must happen exactly at the seeded burst point, and a collected balloon must have fewer pumps than its burst point, otherwise the run is rejected. A run with 4 or more balloons collected exactly one pump before their burst point is flagged ('exact cash-outs', a sign that the hidden schedule was read) and left out of comparisons. Comparisons with other players use first, unflagged runs and appear once 30 such runs exist.

Differences from the classic task

Lejuez et al. (2002) used 90 balloons in three colours with different burst limits and paid real money (5 cents per pump); their main score came from the 30 balloons with the 128-pump limit, which is the condition used here. This version has only that condition, points instead of money, and runs online without supervision.

Limitations

Hypothetical points may change risk taking compared with real money. The burst points are sent to the browser for the animation, so a technically skilled player could read them (this is flagged when it shows in the choices). Links between BART scores and real-life risk behaviour are small. Players learn the burst distribution over the run, and knowing the expected-value curve changes choices. Not a personality or clinical measure.

Languages and review status

Language-independent stimuli, instructions in all site languages.