Preferred Retinal Locus—Hand Coordination in a Maze-Tracing Task

Abstract

Purpose.: Fine manual tasks require coordination of vision, eye movements, and motor control. Macular scotomas from age-related macular degeneration (AMD) may adversely affect this coordination. The purpose of this research was to find whether the preferred retina locus for fixation (fPRL) also guided the hand in performing fine manual tasks and how the fingers, fPRL, and scotomas interacted in task performance.

Methods.: Subjects with bilateral macular scotomas from AMD and normally sighted controls traced an irregular “maze” line pattern with the index finger while viewing their hand and the maze in a scanning laser ophthalmoscope (SLO). Video images from the SLO showing the fingers and maze on the retina during the task were analyzed to produce retinal maps showing the scotoma and bivariate ellipses of fPRL and fingertip retinal positions.

Results.: Fingertip retinal ellipses surrounded and were approximately centered on the fPRL ellipses. Fingertip retinal bivariate area was positively correlated with fPRL bivariate area and the percent time the fPRL was on the maze was correlated with visual acuity. Maze-tracing accuracy was positively correlated with saccade rate for scotoma subjects.

Conclusions.: Concentric overlap of fPRL and fingertip retinal ellipses indicates that it is the fPRL that guides the hand in the maze-tracing visuomotor task, just as the fovea guides the fingertip for visually normal subjects. It is likely that factors other than fPRL and scotoma characteristics contribute to poorer maze-tracing performance by scotoma subjects in comparison with controls.

Introduction

As we perform everyday visuomotor tasks, eye movements place the image of objects relevant to the task on the fovea before the hand reaches the object. Thus, foveal vision provides information for the planning and execution of hand movements on a “just-in-time” ¹ basis. This sequence of eye and hand movements has been found in a wide variety of manual tasks, including copying patterns of colored blocks, ² making tea, ³ playing cricket, ⁴ building Lego models, ⁵ washing hands, ⁶ and making a sandwich. ¹ A similar sequence of eye and hand movements occurs when subjects trace various closed shapes, such as squares or triangles. ⁷  

There has been little investigation as to how eye-hand coordination is affected when foveal vision is lost owing to central macular scotomas. Sullivan and colleagues ^8,9 investigated the effect of central scotomas caused by Stargardt's disease with a single subject who performed various tasks, such as making a sandwich, catching a bounced ball, and bolting parts together. Eye position was tracked using a video-based head-mounted eyetracker. The authors first reported that a general “retinal region,” corresponding visually to an area below the scotoma, was used to view objects during manual tasks. ⁸ Later the authors reported three separate “retinal areas,” corresponding visually to areas below the scotoma, that were used for different tasks. ⁹  

The present study investigated eye-hand coordination of subjects with bilateral macular scotomas caused by age-related macular degeneration (AMD) and age-matched, normally sighted subjects. Most individuals with bilateral central scotomas caused by AMD use a retinal area adjacent to the scotoma for fixating objects of interest: the fixation preferred retinal locus (fPRL). It is not known if the fPRL is used to guide the hand and fingers in manual tasks as well as to fixate or whether another retinal area (or areas) assumes this function, nor is it known how manual task accuracy is affected. The present work was aimed at answering these questions, as well as investigating the interaction of the scotoma, fingertip, and fPRL in a maze-tracing task in comparison with performance by normally sighted control subjects. Subjects performed a fine manual control task in which they traced an irregular line with their index fingertip while looking into a scanning laser ophthalmoscope (SLO). 

Methods

Subjects

Ten subjects with bilateral central scotomas from AMD and 10 normally sighted control subjects participated. The mean age of scotoma subjects, 81.5 years (range: 74–88 years), was not statistically significantly different (t ₁₈ = 1.341, P = 0.197) from that of control subjects (mean age 78.6 years, range: 73–89 years). All subjects were right handed and performed tasks with that hand and their dominant eye as determined using the Dolman hole-in-card test. ¹⁰ Three of the scotoma subjects and two of the control subjects used their dominant left eye to look into the SLO while performing tasks. The mean duration of AMD for scotoma subjects was 7 years (range: 4–10 years). Scotoma subjects' mean acuity in their dominant eye measured with an illuminated Early Treatment of Diabetic Retinopathy Study (ETDRS) chart was 1.07 log MAR (logarithm of the minimum angle of resolution) (range: 0.50–1.30 log MAR). Control subjects' mean acuity in their dominant eye was 0.80 log MAR (range: −0.1–0.20 log MAR). 

The fPRLs were measured for scotoma subjects' dominant eyes by asking the subjects to fixate a cross in the SLO. Subjects were instructed to “Look at the center of the cross as accurately as possible just as though you were looking at a small, far-away object.” Thirty-second fixation trials were conducted until two artifact-free trials were recorded. The densest aspect of the scotoma in the dominant eye was mapped using a 9053 photopic td, 0.5° diameter, 200 msec (6 video frames) stimulus. Fixation of the nondominant eye was checked in the SLO to be certain it was not foveal. The fixation retinal locus of control subjects' dominant eye was also measured. For both scotoma and control subjects, the fPRL (or foveal fixation) bivariate ellipse area and fPRL eccentricity were calculated as described previously. ¹¹  

Potential subjects were excluded from participation if they took medications that could affect arm or hand movements, if they had neuromuscular disorders that affected arm or hand movement, or if they failed to pass the Folstein Mini Mental test. All subjects provided written informed consent by means of a consent form approved by the Kansas City Veterans Administration Medical Center Internal Review Board. The research reported here adhered to the tenets of the Declaration of Helsinki. 

Apparatus

The apparatus used for recording the retinal positions of the hand and objects has been generally described previously. ¹² For the present work, the subject rested his or her hand on a horizontal liquid crystal display (LCD) video monitor on a table adjacent to the right side of the SLO (Fig. 1). The subject's hand and the LCD monitor display were imaged by a charge-coupled device camera above the table top. The video signal from the camera was processed electronically and input to the SLO acousto-optic modulator, which reproduced the camera image in the SLO laser-beam raster. Subjects looking into the SLO saw their hand and images on the LCD monitor in the SLO while performing the manual task. The SLO output video image showed the hand and objects on the subject's retina while the subject performed the task (Fig. 2). To improve visibility of the fingertip in SLO images, a small (<1-mm diameter) infrared light-emitting diode (LED) was attached to the distal end of the subject's index finger. 

Figure 1.

 

Modified SLO used for experiments. Subject looks into SLO with dominant eye and rests hand on LCD monitor presenting a maze generated by computer at right. Video camera (arrow) images hand and maze in SLO laser-beam raster using electronics on right.

Figure 2.

 

SLO image of hand and maze on retina. Subject traced the maze with the right index finger. A small infrared LED was attached to the index fingernail. Scale shows visual angle in degrees.

Procedure

Subjects traced crooked-line “mazes” with the tip of their right index finger while looking into the SLO. Each maze was made of eight line segments, two each of 22, 44, 66, and 88 mm (i.e., in the ratio of 1:2:3:4) in length and 11 mm in width. In the SLO laser-beam raster, the maze lines subtended 1.9° in width and 3.8°, 7.6°, 11.4°, and 15.2° in length. All mazes began with one of the shortest line segments marked with a black dot (Figs. 3B–D). At the beginning of each trial, only the short starting segment was visible and the subject was instructed to place his or her index fingertip on the black dot (Fig. 3A). The experimenter then presented the remaining portions of the maze all at once, forming the entire maze (Fig. 3B). Subjects were instructed to trace the maze with their index finger as quickly and accurately as possible. A different, novel maze was used for each maze-tracing trial. SLO video images were recorded on digital versatile disc (DVD) throughout each trial. Subjects traced two to three practice mazes and then four to six additional mazes (repeats were sometimes necessary because of poor images). To investigate the role of visual memory in tracing, subjects also traced six “disappearing” mazes. A maze was presented and the subject was encouraged to inspect it carefully and try to remember its shape. The maze was then removed and the subject instructed to trace its pattern from memory. Initially, 10 of the subjects were allowed to inspect the maze for as long as they wished before it was removed and the duration of inspection recorded. No subject inspected mazes for more than 10 seconds; consequently, in all subsequent trials the maze was presented for 10 seconds and then removed. 

Figure 3.

 

Mazes for tracing as they appeared on the horizontal video monitor. (A) Short starting segment on which subject places index fingertip. (B) Rest of maze appears when trial begins. (C) and (D) Other maze shapes. Calibration bars show visual angle subtended in the SLO laser-beam raster.

Analysis

SLO video images on DVD were converted to avi files, which, in turn, were converted to 15 frames per second jpg images using custom Matlab programs. The positions of the fingertip and a retinal landmark, such as a blood vessel, were measured on every other jpg frame (i.e., at 133-msec intervals). The outline of the maze on the retina was also measured on a single frame. Four tracing trials on different mazes were measured for each subject for “normal” (always present while tracing) and “disappearing” (drawn from memory) mazes. 

Retinal positions of the fingertip on each measured SLO video frame were used to calculate a P = 0.68 fingertip retinal bivariate normal ellipse. The fingertip retinal bivariate ellipse and the fPRL (or foveal fixation area) P = 0.68 bivariate ellipses were plotted on an SLO image of the subject's retina. The normal foveal fixation area based on measurements of retinal fixation positions relative to the optic disk of 50 visually normal subjects was also plotted. ¹¹ The percentage of time the fingertip was in the subject's scotoma and the percentage of time the fPRL or fovea was on the maze were calculated, as well as the time to complete tracing. 

Maze-tracing accuracy was quantified as illustrated in Figure 4. The distance the fingertip advanced along the maze (arrows in Fig. 4) was measured for each maze traced. Tracing accuracy was then calculated as the total length the fingertip advanced along the maze divided by the maze length (dotted line Fig. 4). 

Figure 4.

 

Method of quantifying maze-tracing accuracy. Arrows indicate length fingertip advanced along the maze when fingertip was within the maze lines. Dotted line indicates total maze length. Maze tracing accuracy = (total length fingertip advanced/maze length) × 100. Fingertip position: grey line.

Statistical comparisons were carried out using parametric statistical methods (e.g., t tests) when the data were normally distributed and met equal variance criteria. The critical value (α) for multiple t-test comparisons of correlation coefficients was corrected using a sequential Bonferroni method. ^13,14 Nonparametric statistics (e.g., Mann-Whitney rank sum test) were used for data that did not meet normal distribution or equal variance criteria. 

Results

Retinal Locus of Fingertip and fPRL or Fovea During Maze Tracing

Examples of SLO images of a control subject and a scotoma subject tracing mazes are shown in Figures 5A and 5C, respectively. (All SLO images are oriented so that up was visually up for the subject and left was visually left; in other words, the images are a projection of the retina onto the visual field.) Retinal maps based on analysis of tracing four mazes are shown in Figures 5B and 5D. Individual fingertip retinal positions are shown for the control subject (Fig. 5B, black circles). The P = 0.68 bivariate ellipse calculated from the fingertip positions is shown as a white dashed line. The control subject's fixation P = 0.68 bivariate ellipse was quite small and centered within the fingertip bivariate ellipse. The normal foveal fixation area ¹¹ is indicated by a white, solid line P = 0.90 bivariate normal ellipse. The subject's fixation bivariate ellipse is within the normal foveal fixation area. 

Figure 5.

 

( A) SLO video image showing the hand, fingertip with small infrared LED attached, and maze on the retina of a control subject as he traced the maze. Small cross: Subject's foveal fixation position. Calibration bars show visual angle for all images. (B) Maze-tracing retinal map of the same subject shown in (A). Small black circles: Individual measured fingertip positions from four maze tracings. Dashed line: P = 0.68 fingertip bivariate ellipse. Large solid-line ellipse: Control subject fixation retinal locus. Small solid-line ellipse: Subject's fixation locus based on 30-second fixation trial. (C) Same as (A), but with a scotoma subject. Black dot: Subject's fPRL bivariate ellipse centroid. (D) Maze-tracing retinal map for subject shown in (C). fPRL: Solid black ellipse with black dot centroid. Fingertip retinal locus: White dashed-line ellipse with white circle centroid. Normal foveal fixation area: Solid white line ellipse with open square centroid. Dense central scotoma: Solid gray line.

A maze-tracing retinal map for a scotoma subject is shown in Figure 5D. The normal foveal fixation area is within the dense scotoma mapped with the SLO. The subject's fPRL was visually inferior to the scotoma and was within the fingertip bivariate ellipse. The subject's fingertip bivariate ellipse is considerably larger than that seen for the control subject. Overall, scotoma subjects' mean fPRL P = 0.68 bivariate ellipse areas were more than six times larger than control subjects' mean foveal fixation areas (4.79 deg² vs. 0.76 deg²), a statistically significant difference (t ₁₈ = 3.264, P = 0.004). As found previously, ¹¹ the log of fPRL bivariate ellipse area was positively correlated with fPRL eccentricity (r = 0.737, P = 0.015). Scotoma subjects' fingertip retinal bivariate ellipse areas were approximately 2.5 times larger than those of controls (62.9 deg² vs. 24.8 deg²), a statistically significant difference (t₁₈ = 4.828, P < 0.001). 

Retinal Location of the Fovea and fPRL in Relation to the Fingertip Retinal Locus

The foveal fixation bivariate ellipse was within the fingertip retinal bivariate ellipse for all 10 control subjects. Figure 6 shows maze-tracing retinal maps of four control subjects. For all control subjects, centroid-to-centroid distances of the fixation and fingertip ellipses ranged from 0.12° to 2.71°. Fixation bivariate ellipses were all within the P = 0.90 normal foveal fixation area or adjacent to it. The retinal position of the fingertip was relatively precise with a mean P = 0.68 bivariate ellipse area of 24.8 deg² (range: 10.2 deg²–53.8 deg²). 

Figure 6.

 

Maze-tracing retinal maps of four control subjects (retinal image omitted). Arrows point to the fixation bivariate ellipse. Dashed ellipse: P = 0.68 bivariate ellipse of retinal fingertip positions for four maze-tracing trials. Solid ellipse with square centroid: Normal fixation area. Dark circle: Optic disk. Numbers to lower right of each panel are the distance between the fixation ellipse centroids and the fingertip ellipse centroids. Scale in degrees for all plots shown in (A).

Figure 7 shows examples of retinal maps of four scotoma subjects with various distances of fPRL ellipse center to fingertip ellipse center. The fPRL bivariate ellipse was within the fingertip bivariate retinal ellipse for 8 of 10 scotoma subjects. One subject's fPRL was 95% within the fingertip bivariate ellipse and one subject's fPRL was outside the fingertip bivariate ellipse. The mean vector distance of the fPRL centroid to the fingertip ellipse centroid was 2.187° (range: 0.5°–6.75°). Thus, on average, the fingertip retinal image stayed relatively close to the fPRL during maze tracing. 

Figure 7.

 

Maze-tracing retinal maps of four scotoma subjects. Absolute vector distance between the fPRL centroid and fingertip retinal bivariate ellipse centroid in degrees shown in lower right of each panel. (A) Smallest fPRL-to-fingertip bivariate ellipse centroid distance. (B,C) Intermediate fPRL-to-fingertip centroid distances. (D) Largest fPRL-to-fingertip centroid distance. fPRL: Small solid line ellipse with black dot centroid. Fingertip bivariate retinal ellipse: Dashed ellipse with white dot centroid. Control subject fixation area: Solid line ellipse with white square centroid. Optic disk: grey circles. Scale in degrees for all plots shown in (B).

The relative positions of the fingertip ellipse centroids and the foveal fixation or fPRL centroids are shown in Figure 8. For scotoma subjects, the median vector distance of the fPRL centroid to the fingertip centroid (2.02°) was significantly longer than the median vector distance from the foveal fixation centroid to the fingertip for control subjects (0.712°) (Mann-Whitney rank sum, T = 137.0, P = 0.017). For both scotoma and control subjects, mean fingertip position relative to the fPRL or fovea was slightly to the left of the fPRL or foveal fixation centroid (Fig. 8, filled symbols). 

Figure 8.

 

Plot of fingertip bivariate ellipse centroids relative to the fPRL or foveal fixation centroids (0, 0 coordinate). Filled symbols: Mean positions.

Maze-Tracing Performance

Control subjects maintained their foveas and fingertips on the mazes more accurately than did scotoma subjects. Figure 9 shows examples among the best and among the worst tracing by control subjects (Figs. 9A and 9B) and by scotoma subjects (Figs. 9C and 9D). In Figure 9A, the control subject kept both his fingertip and fovea on or very near the maze line throughout the trial. The control subject whose maze trace is shown in Figure 9B also maintained his fovea on or near the maze but his fingertip often left the maze. In addition, for this subject, both foveal and fingertip positions were much more variable than those shown in Figure 9A. Figure 9C shows a relatively good maze trace by a scotoma subject. In this trace, the fingertip and fPRL stayed close together and close to or on the maze for much of the trace. A worst-case trace for a scotoma subject is shown in Figure 9D. Neither the fingertip nor the fPRL were often on the maze and the positions of both appear almost random. Control subjects kept their foveas on the maze 64.7% of the time compared with 36.8% of the time for scotoma subjects, a statistically significant difference (t ₁₈ = 4.321, P < 0.001). Control subjects' fingertips also stayed on the maze longer than did scotoma subjects' fingertips, producing significantly more accurate maze tracing of 81.5% accuracy as compared with 48.3% accuracy for scotoma subjects (t ₁₈ = 3.847, P = 0.001). Mean tracing completion times for control and scotoma subjects were 16.7 seconds and 22.5 seconds, respectively, not a statistically significant difference (t ₁₈ = 1.725, P = 0.102). Scotoma subjects' fingertips were in their scotoma an average of 47.8% of the time. The amount of time the fingertip was in the scotoma was not correlated with scotoma area (r = 0.146, P = 0.490) or with tracing accuracy (r = 0.208, P = 0.564). 

Figure 9.

 

Example maze traces showing control subjects' best (A) and worst (B) performance and scotoma subjects' best (C) and worst (D) performance. Scale in degrees for all plots shown in (C).

Correlation of Maze-Tracing Performance with Retinal Functional Geography

The influence of retinal functional geography on maze-tracing performance was explored by correlation analysis. Retinal functional geographic measures were as follows: (1) fPRL bivariate ellipse area, (2) fPRL eccentricity, (3) scotoma area, and (4) visual acuity. Maze-tracing performance measures were as follows: (1) tracing accuracy, (2) fingertip retinal bivariate area, (3) percentage of time the fingertip was in the scotoma, (4) percentage of time the fPRL was on the maze, (5) tracing time, and (6) fPRL-fingertip delay time (Table 1). Two of 24 possible correlations were statistically significant after sequential Bonferroni correction. First, the log of fingertip retinal bivariate ellipse area was positively correlated with log fPRL bivariate ellipse area (r = 0.801, P = 0.005) (Fig. 10). Thus, the larger the subject's fPRL (i.e., the less stable fixation), the larger was the fingertip retinal ellipse area (i.e., the less precise fingertip control). Second, the percentage of time the fPRL was on the maze was correlated with visual acuity (r = 0.806, P = 0.005) (Fig. 11). (It was assumed that visual acuity measured with an ETDRS chart reflected fPRL acuity.) This plot shows that the poorer the acuity, the less time the fPRL spent on the maze. 

Table 1.

 

Correlation coefficients and t test probability (in parentheses) for the four retinal functional geographic measures with each of the six maze-tracing performance measures.

	Retinal Functional Geographic Measures
Performance Measures	fPRL BA	fPRL ecc	Scotoma Area	fPRL Acuity
Tracing accuracy	0.054 (0.882)	0.371 (0.291)	0.268 (0.454)	0.581 (0.078)
Fingertip BA	0.801 (0.005)	0.666 (0.036)	0.012 (0.974)	0.465 (0.175)
% fingertip in scotoma	0.314 (0.376)	0.192 (0.595)	0.146 (0.688)	0.325 (0.360)
% fPRL on maze	0.293 (0.411)	0.407 (0.243)	0.260 (0.243)	0.806 (0.005)
Tracing time	0.326 (0.357)	0.230 (0.522)	0.423 (0.223)	0.307 (0.389)
fPRL-fingertip Δt	0.229 (0.524)	0.220 (0.541)	0.021 (0.953)	0.117 (0.748)

 Bold type indicates correlation coefficients found to be statistically significant after sequential Bonferroni correction. ecc, eccentricity; BA, bivariate area.

Figure 10.

 

Regression plot of log fingertip retinal bivariate ellipse area versus log fPRL bivariate ellipse area.

Figure 11.

 

Regression plot of percentage of time the fPRL was on the maze being traced versus visual acuity.

fPRL and Fovea Lead Time

Many previous studies (see Land ¹⁵ for a review) have found that in normal vision the fovea first moves to relevant task objects and the hand follows. This is also true for the maze-tracing task. Figures 12A and 12B show the relative positions of a subject's fovea and fingertip in the horizontal (Fig. 12A) and vertical (Fig. 12B) directions. The fovea is seen to precede the fingertip to a given position most of the time. For example, the horizontal line in Figure 12A represents a fixed horizontal position in the SLO image of the maze at 3° relative to the fingertip starting position. It is seen that the fovea (black line) reaches this position first and is then followed a short time later by the fingertip. The fPRL also leads the fingertip as shown in Figures 11C and 11D. The amount of fPRL lead appears to be greater than for the control subject and, in addition, appears more irregular relative to the fingertip. 

Figure 12.

 

Plots of fovea (or fPRL) and fingertip positions versus time. (A) Control subject, horizontal positions. (B) Control subject vertical positions. (C) Scotoma subject, horizontal positions. (D) Scotoma subject, vertical positions.

To quantify these apparent differences, cross-correlation analyses were performed for four mazes for each subject. The peak cross-correlation coefficient corresponds to the mean amount of time the fovea or fPRL led the fingertip. A unitary fPRL-fingertip lag time was calculated using a method ¹⁶ in which horizontal and vertical fingertip positions as a function of time were interleaved. The same interleaving was done for fPRL horizontal and vertical positions and the cross-correlation was then performed with the interleaved data. The resulting mean fPRL lead time was 0.334 second (± 0.112) and 0.153 (± 0.031) for the fovea, a statistically significant difference (t ₁₈ = 4.912, P < 0.001). 

Saccades and Fixations

Saccades during maze tracing were measured from shifts in positions of retinal vessel landmarks. Saccades identified by retinal velocities of 5° per second or greater corresponded well to saccades observed in SLO videos; consequently, this threshold velocity was used in identifying all saccades. With this threshold saccade velocity criterion, saccades of 0.7° amplitude and larger were identified. Horizontal and vertical fPRL (or foveal) positions were plotted as a function of time showing the saccades identified by the software. These plots were inspected to make sure no smooth pursuit movements were incorrectly counted as saccades. Mean saccade rates for control subjects (1.60 saccades per second) and scotoma subjects (1.47 saccades per second) were not significantly different (t = 0.685, P = 0.602). Mean saccade amplitude of controls, 2.13°, was not significantly different from that of scotoma subjects, 2.03° (t = 0.482, P = 0.635). Mean saccade velocity was also not significantly different for the two groups (control: 13.96°/s, scotoma: 13.65°/s) (t = 0.235, P = 0.817). For scotoma subjects, the log of maze-tracing accuracy was positively correlated with the log of saccade rate (r = 0.679, t = 0.031); subjects with higher saccade rates tended to trace mazes with greater accuracy. Maze-tracing accuracy of control subjects was not correlated with saccade rate. 

Fixations during maze tracing were identified when the vector velocity calculated from two successive X, Y retinal landmark positions was 1° per second or less. This criterion agreed well with fixations identified from observations of the SLO video. Figure 13 shows all fixations of 8 of the 10 control subjects (Fig. 13A) and 8 of the 10 scotoma subjects (Fig. 13B) who traced the maze shown with SLO images adequate for analysis. Control subjects made fewer fixations than scotoma subjects and their fixations were more confined to the maze than were those of scotoma subjects. 

Figure 13.

 

Plots of fixation positions on a maze traced by 80% of the subjects. (A) Control subjects. (B) Scotoma subjects.

Long-term Visual Memory in Maze Tracing

The position of the fingertip on the maze could be guided by real-time visual feedback of fingertip position. It is also possible that subjects form a visual memory or “model” of a maze before tracing it and use that memory (at least in part) to guide the fingertip. To investigate these possibilities, subjects were asked to trace 6 “disappearing” mazes in which the maze was visible initially then removed from view before tracing began. The maze was presented for 10 seconds and the subject instructed to scan the maze and try to remember it. Tracing accuracy was calculated for four disappearing mazes for each subject. To estimate maze-tracing accuracy for random fingertip movements, four subjects (two scotoma subjects and two controls) were asked to make up an imaginary eight-segment maze similar to the ones they had previously traced and trace it in the SLO. For analysis, an arbitrary maze was selected and the accuracy of tracing the imaginary maze was calculated relative to the arbitrary maze. The calculated accuracy data were normally distributed, consequently a two-way repeated measures ANOVA was performed with Student-Newman-Keuls (S-N-K) post hoc tests. The results are shown in Figure 14. Disappearing mazes were traced with lower accuracy than normal, continuously visible mazes for both control and scotoma subjects. 

Figure 14.

 

Mean percent tracing accuracy for continuously visible mazes (Normal), mazes that disappeared before tracing (Disappear), and tracing imaginary 8-segment mazes (Random). Asterisks indicate statistically significant differences (S-N-K, P < 0.001).

Random tracing accuracy, estimated by tracing imaginary mazes, was 16.2%, close to disappearing maze accuracies of scotoma subjects (18.5%) and control subjects (23.7%). Thus, it appears that long-term visual memory contributes little to tracing performance and that continuous visual feedback is the major eye-hand coordination mechanism. Although maze-tracing accuracy was low when tracing mazes that had disappeared, subjects often reproduced the maze shape or part of it, as illustrated in Figure 15. 

Figure 15.

 

Examples of maze shapes that were traced after the maze shown was no longer visible. (A) Control subject. (B) Scotoma subject. Dark line: Fingertip position. Open circles: Fovea or fPRL positions.

Discussion

The primary goals of the present study were to determine what retinal area(s) subjects with bilateral macular scotomas used to perform a tracing task and to explore the interactions of the scotomas, fPRLs, and fingertips in performing the task. It was found that the fPRL guided the fingertip in maze tracing for scotoma subjects, just as the fovea guided the fingertip for visually normal subjects. Maze-tracing performance was uniformly poorer for scotoma subjects in comparison with controls. This performance deficit does not appear to be primarily a result of retinal functional geography, as there were relatively few correlations of performance measures with retinal functional geography. Tracing accuracy of scotoma subjects was, however, positively correlated with saccade rate. In addition, the fPRL moved in advance of the fingertip by a significantly longer interval than did control subjects' foveas. When tracing mazes that disappeared after free viewing, both control and scotoma subjects performed with close to random accuracy, indicating that long-term visual memory of the maze contributed little to tracing accuracy. These results are discussed in detail as follows. 

Retinal maze-tracing maps showed that fingertip retinal positions surrounded the retinal fixation area of all control subjects and the fPRL of 9 of the 10 scotoma subjects. Further, fingertip bivariate ellipse centroids were within 1.08° of control subjects' fixation ellipse centroids and within 2.6° of scotoma subjects' fPRL ellipse centroids (with one exception). Thus, it appears that the fovea and the fPRL provide the visual information that guides the fingertip during maze tracing, and there is not a separate “maze-tracing PRL” for this visuomotor task. In their case study of a patient with Stargardt's disease, Sullivan et al. ^8,9 did not report whether the subject's fixation corresponded to any of the other reported “retinal areas”; however, many more subjects with Stargardt's disease and more direct measurements of retinal position would be needed to establish that different “retinal areas” are used for different manual tasks and their relationship to the fixation retinal locus. It is conceivable, however, that if subjects with central scotomas from bilateral AMD were to make a sandwich or catch a ball, a different retinal area than that used to fixate might be used. 

The only maze-tracing retinal map of a subject whose fPRL was not within the fingertip retinal position is shown in Figure 16. The fingertip retinal area was the second largest recorded (83.4 deg²) and was mostly in the subject's scotoma. There was some overlap of the fingertip ellipse and the normal fixation ellipse. When asked to fixate, this subject used a well-defined fPRL, as shown in Figure 16. The map suggests that the subject may have been trying to keep the fingertip retinal image near to his (nonfunctional) fovea. It is also possible that there was some residual visual function near the fovea that was undetected in mapping the scotoma. 

Figure 16.

 

Maze-tracing retinal map of a scotoma subject whose fPRL was not within the fingertip bivariate retinal area. fPRL: Small solid line ellipse with black dot centroid. Fingertip bivariate retinal ellipse: Dashed ellipse with white dot centroid. Normally-sighted subjects fixation area: Solid line ellipse with white square centroid. Optic disk: Dark gray circle.

Only 2 of 24 possible correlations of maze-tracing performance with retinal functional geography were significant: fingertip retinal bivariate ellipse area was positively correlated with fPRL bivariate area, and the percentage of time the fPRL spent on the maze was correlated with visual acuity. In a previous study, ¹⁷ we also found few correlations between reach-to-grasp performance and retinal functional geography. Although more subjects might reveal additional correlations, the present results suggest that retinal functional geography may not be the primary determinant of visuomotor performance. Factors such as motivation and practice may be more important determinants of visuomotor performance than retinal functional geography. For example, among the subjects we tested, a few had hobbies that involved eye-hand coordination, such as scroll-saw work, and their maze-tracing performance was usually better than others. This observation suggests the possibility that training regimes may substantially improve visuomotor performance. 

Scotoma subjects' maze-tracing accuracy was uniformly poorer than that of control subjects. Fixation stability and eccentricity of the fPRL and scotoma size do not appear to be responsible for this deficit, as they were not significantly correlated with tracing accuracy. Tracing accuracy was also uncorrelated with time the fingertip was in the scotoma. Perhaps this is because the fPRL moves ahead of the fingertip, providing visual information about where next to move the fingertip to the motor system, so it might not matter if the fingertip then enters the scotoma. Another possibility is that a more proximal portion of the finger that is not in the scotoma serves as a “marker” for the fingertip. Maze-tracing accuracy was also uncorrelated with percentage of time the fPRL was on the maze. This implies that it is not necessary to keep the fPRL on the maze to be accurate. Tracing accuracy, however, was positively correlated with saccade rate for scotoma subjects but not controls; scotoma subjects who made more saccades per second tended to have greater tracing accuracy. This may indicate that scotoma subjects who have higher saccade rates are more skilled in visual search and better able to locate the fingertip and maze in the presence of a scotoma. 

In the present study, the mean fPRL-fingertip lead time was 0.334 second compared with a fovea-fingertip lead time of 0.153 second for controls. These lead times are shorter than the 0.7 to 0.8 second reported by Pelz et al. ⁵ for normally sighted subjects. The task in that study ⁵ of building a structure using Legos was substantially different from tracing a pattern with the fingertip and could possibly be responsible for the longer hand-eye delay times. Scotoma subjects' mean fPRL lead time was significantly longer than the foveal lead time. Thus, on average, scotoma subjects' fingertips took somewhat more than twice as long to “catch up” to the fPRL position than control subjects' fingertips took to catch up to the foveal position. This increased delay might be because of increased time required to process degraded visual information from the fPRL about the maze line. 

The present results are in general agreement with Gowen and Miall's study ⁷ of tracing shapes with a pen, in that subjects in the present work also made a series of saccades of approximately 2° amplitude in advance of the moving finger. Control and scotoma subject saccade rates in the present study (1.60 s⁻¹ and 1.47 s⁻¹ respectively) were approximately half that found by Gowen and Miall (3.13 s⁻¹). ⁷ Their study ⁷ also found more frequent fixations at the corners of traced triangles and squares, whereas the present study found no evidence of greater numbers of fixations of either the fovea or fPRL on the corners of the mazes, as shown in Figure 13. 

The present study found that attempting to trace a maze from memory resulted in tracing accuracies approximately the same as for random fingertip movements. Thus, it seems that inspecting a complex geometric visual pattern, such as a maze, produces little long-term visual memory that can be used to guide manual interaction with the pattern. Using a virtual-reality task of moving blocks, however, Brouwer and Knill ¹⁸ found that the remembered location of a block was used to plan movements to it. In addition, Hayhoe et al. ¹ noted that before making a sandwich, subjects scanned the scene, possibly to build up a representation of objects in the scene to use in the task. They ¹ concluded, however, that short-term visual representations of the scene were more predominant than long-term representations. Despite the inability to trace mazes from memory, subjects frequently traced the approximate maze shape (Fig. 15) but in the wrong size and position (i.e., with the wrong scale). This suggests that maze shape is available in long-term memory but not the scale. If some type of scale, such as a square that outlined the maze, was provided after the maze disappeared and before tracing began, tracing accuracy might improve. Another approach to investigate the role of visual-spatial memory in maze tracing might be for only two or three segments of the maze to disappear rather than the whole maze and measure the tracing accuracy of the segments that disappeared. This might also elucidate the role of short-term memory in tracing. 

In summary, we have found that subjects with bilateral macular scotomas from AMD use their fPRL to guide their hand in a fine manual task and that they perform more poorly than age-matched, normally sighted control subjects. Poorer task performance appears to be largely unrelated to scotoma size, acuity, fPRL eccentricity, or fPRL stability. It may be of interest to investigate the relationship of task performance to nonvisual and retinal factors, such as subject demographics, including type and amount of visual rehabilitation training. Further, it is possible that factors such as practice and motivation could help to overcome degraded visual information from a PRL and obscurations by a scotoma when performing manual tasks. Because fine manual control is important in many daily activities, such as food preparation, household repairs, and sorting pills, the authors believe it would be useful to explore the effects of training on task performance and on retinal functional geography. 

Footnotes

 Disclosure: G.T. Timberlake, None; E. Omoscharka, None; S.A. Grose, None; R. Bothwell, None

Footnotes

 Supported by Department of Veterans Affairs, Rehabilitation, Research, and Development Service Grant C6218R.