Figure 2
From: Improvement in reading performance through training with simulated thalamic visual prostheses
Population Reading Accuracy and Speed by Viewing Condition, Over Time. Reading accuracy and speed for the six subjects as a population are shown as observed data (circles), standard deviation (filled areas), and exponential fits (curved lines), broken down by pattern and pooled across all font sizes (A) Reading Accuracy. Session number progresses horizontally, while mean reading accuracy is shown vertically. Improvement in accuracy is initially quite rapid for all patterns although saturation effects become apparent for the Easy/P2000 trace after only 10 sessions, reflected in the fitted time constant τ of 3.4 sessions (95% confidence bounds of 2.4 to 4.3). While the Medium/P1000 trace appears to be asymptoting by the end of data collection with τ = 7.1 sessions (5.9 to 8.4), the Hard/P500 trace suggests that improvement would continue through additional training with τ = 14 sessions (11 to 17). (B) Reading speed. Improvement in reading speed is not as rapid as with accuracy, and has not obviously saturated for any pattern, reflected in the fitted time constant of τ = 10 sessions (5 to 15) for the Easy/P2000 pattern, τ = 12 sessions (8 to 16) for the Medium/P1000 pattern, and τ = 45 sessions (6 to 85) for the Hard/P500 pattern. At the end of training, the reading speed for the Hard/P500 pattern surpasses that for the untrained speed, measured on the first day when the subjects were naive, for the next easiest pattern, Medium/P1000. Similarly, the trained speed for the Medium/P1000 pattern surpasses that for the untrained speed of the next, Easy/P2000.
