Abstract
Purpose: :
This study investigated computational methods that can be used for comprehensive reading rehabilitation of the visually impaired, both for the diagnosis/evaluation of reading performance problems and for training in reading rehabilitation. The overall research objective is to provide easy to use and accurate computational methods for reading rehabilitation to improve everyday tasks associated with reading.
Methods: :
132 older adults age 65 to 91 years (mean = 75.6) completed the MNRead test. 121 subjects returned a second time more than 1 month later to repeat the MNREAD test. Reading performance is a growth process (reading rate typically increases as a function of increasing letter size) and from studies of growth processes in both animate and inanimate systems, three commonly studied growth functions were used: 1) Logistic, 2) Gompertz, and 3) Weibull. The test/retest variance of the reading performance attributes (positive slope, critical print size, maximum reading rate, and falloff of reading rate at large text size) found by Weibull model were compared to the variability found by five reading rehabilitation experts doing a visual inspection method of the reading rate function.
Results: :
Visual acuity (ETDRS) ranged from -0.15 to 0.64 logMAR (median = 0.15), contrast sensitivity (Pelli-Robson) ranged from 1.05 to 2.05 logcontrast (median = 1.74), and Humphrey visual field testing confirmed subjects had normal age-related visual function. Weibull regression had the highest frequency of data set convergence at 248/253 (98%), followed by Logistic (92%), and then Gompertz (91%) and the Weibull model gave lower residual error than either Logistic or Gompertz models. Successive shortening of the regressed data set length gave poorer fitting error and changes regression parameter values, which suggests that reading rate functions may give nonsigmoidal responses at the largest print sizes. A nonmonotonic Weibull model that was sensitive to declining reading rate at large print size improved regression on 22% of the subjects. Test-retest variability for the Weibull function was similar or significantly (p<0.05) less than the variability of measuring reading performance attributes by visual inspection of the reading rate function.
Conclusions: :
The Weibull model provides the best computational method for modeling the reading rate function. The falloff at the end of the maximum reading rate plateau in a subgroup of the subjects is similar to the positive and then negative reading rate slopes at different print sizes found in some older adults with visual field loss. The assumption that visual field loss is the cause of falloff at large prints sizes may need to be evaluated.
Keywords: reading • aging: visual performance • low vision