Freiburg vision test (FrACT): optimal number of trials?

Visual acuity is a psychophysical threshold that we want to determine as precisely and efficiently as possible. The Freiburg Vision Test FrACT employs the automated Bayesian "Best PEST" algorithm for this purpose: the next optotype size is always selected to be at threshold based on the information acquired so far, thereby maximizing information gain.

We assessed the test-retest Limits of Agreement (LoA, Bland & Altman 1986) across 6 to 48 trials in 2 × 78 runs involving 26 participants; visual acuity (in part artificially reduced) ranged from 1.22 to -0.59 LogMAR.

LoA exhibited a steep decline from ± 0.46 LogMAR at six trials to ± 0.17 at 18 trials; with more trials, LoA showed less change, reaching ± 0.12 LogMAR at 48 trials. LoA did not significantly change over the wide acuity range assessed here.

These findings suggest that 18 trials represent an efficient balance between precision and burden on the participant and examiner. This observation holds for the eight response alternatives used in this study (8 Landolt C orientations) and is anticipated to apply to the ten Sloan letters as well. With only four choices (e.g., tumbling E), more trials will be necessary.

What is known When assessing visual acuity, a tradeoff between precision and effort is necessary. What is new A run length of 18 trials is a good compromise between effort and precision for an 8-alternative task (the Landolt C). With 18 trials a 95% confidence interval of ± 0.17 LogMAR for test-retest is found. The test-retest precision is independent of the acuity level over the 1.5 LogMAR range studied here.