Learning to identify near-acuity letters, either with or without flankers, results in improved letter size and spacing limits in adults with amblyopia

Abstract

Amblyopia is a developmental abnormality that results in deficits for a wide range of visual tasks, most notably, the reduced ability to see fine details, the loss in contrast sensitivity especially for small objects and the difficulty in seeing objects in clutter (crowding). The primary goal of this study was to evaluate whether crowding can be ameliorated in adults with amblyopia through perceptual learning using a flanked letter identification task that was designed to reduce crowding, and if so, whether the improvements transfer to untrained visual functions: visual acuity, contrast sensitivity and the size of visual span (the amount of information obtained in one fixation). To evaluate whether the improvements following this training task were specific to training with flankers, we also trained another group of adult observers with amblyopia using a single letter identification task that was designed to improve letter contrast sensitivity, not crowding. Following 10,000 trials of training, both groups of observers showed improvements in the respective training task. The improvements generalized to improved visual acuity, letter contrast sensitivity, size of the visual span, and reduced crowding. The magnitude of the improvement for each of these measurements was similar in the two training groups. Perceptual learning regimens aimed at reducing crowding or improving letter contrast sensitivity are both effective in improving visual acuity, contrast sensitivity for near-acuity objects and reducing the crowding effect, and could be useful as a clinical treatment for amblyopia.

Conflict of interest statement

Competing Interests: The authors have declared that no competing interests exist.

Figures

Figure 1. Training data for individual observers.

The top row shows the data for the flanked letter training group while the bottom row shows the data for the isolated letter training group. Each observer (except for JHS who was trained for 26 sessions) participated in 10 sessions of training. In all panels, each unfilled symbol represents the performance for a block of 100 trials. Filled symbols on the leftmost and rightmost edge of each panel represent the data during pre-test and post-test. Error bars represent ±1 s.e.m. Linear regression function was used to fit each set of data. The slope of this function, if different from 0 (p-value given in each panel), implies significant improvement. The slopes of the linear function for observers SDW and RE in (b) were negative, thus the p-value for improvement was listed as “NAN". Note also the change in scale on the ordinate for observers RE and LA in (b). The slope of the regression line (m), the t-statistic in calculating the significance and the p-value are given in each panel.

Figure 2. Proportion of correct responses in identifying flanked letters as a function of center-to-center letter separation in trigrams, for the task of measuring the spacing limit, is plotted for each individual observer.

Letter separations are specified as multiples of the x-height. Unfilled symbols represent pre-test results and filled symbols represent post-test results. The smooth curve drawn through each data-set represents a cumulative-Gaussian function fitted to the data, from which we define the spacing limit as the letter separation that yields 0.52 on the cumulative function. The rightmost points (for a separation of ∞) represent performance for identifying single (unflanked) letters. The two data points are offset slightly to avoid clutter. Error bars represent ±1 s.e.m.

Figure 3. Proportion of correct responses in all three letters in trigrams, presented at different letter position left and right of fixation, for the task of assessing the visual span.

Data are plotted for each individual observer. Unfilled symbols represent pre-test results and filled symbols represent post-test results. The smooth curve drawn through each data-set represents a split-Gaussian function fitted to the data. The size of the visual span, akin to the measurement of the area under the curve, was quantified by first converting each proportion-correct value (from the fitted curve) to bits of information transmitted, then summing up these values across all letter positions (values plotted in Fig. 4d).

Figure 4. Comparisons of the post- and pre-test performance for four untrained visual tasks.

a. The letter size limit (acuity) in degrees of visual angle. b. The letter spacing limit (defined as the letter separation that yielded 52% on each fitted function in Fig. 2), converted to degrees of visual angle by multiplying the estimate with letter size. c. Contrast threshold for identifying single letters. d. The size of the visual span in bits of information transmitted. In each panel, the dashed line represents the 1∶1 line and the light gray shaded region represents improvement. Each symbol represents data for one observer, with red representing strabismic amblyopes and green representing non-strabismic amblyopes. Filled bowtie symbols represent observers trained on the flanked letter task and unfilled circular symbols represent observers trained on the isolated letter task.

Figure 5. Post-pre ratios and difference comparisons for the four untrained visual tasks between the two training groups.

a. Flanked letter training. b. Isolated letter training. Post-pre ratios were calculated for letter size limit (size), spacing limit (spacing) and contrast threshold for identifying single letters (contrast). Post-pre differences were calculated for the size of the visual span (vspan). Small unfilled symbols represent individual observers data with red representing strabismic amblyopes and green representing non-strabismic amblyopes. Black filled symbols represent the group-averaged value, with error bars representing the 95% confidence intervals. For comparison, the improvements due to training were included as blue lines (dark blue dotted line: ratio calculated based on the expected values for the first and the last block of trials derived from the linear functions fitted to the training data; light blue dashed line: ratio calculated based on the empirical performance for the first and the last block of trials). The ratio plotted for the training data was the pre-post ratio, instead of the post-pre ratio, as the performance accuracy was higher after training than before.