The measurement of a patient's visual acuity (VA) is a key feature of the optometric examination: a reduced acuity value relative to an accepted clinical “standard” may have pathological or diagnostic implications. The acuity parameter is also often the primary outcome variable in clinical trials or vision research studies. In addition, it is frequently taken as an indicator of visual ability in association with occupational tasks including the statutory licensing of automobile drivers and airline pilots.

The familiar format of the clinical VA test has remained virtually unchanged since its introduction nearly 150 years ago (Snellen, 1862: as cited in Bennett).1 Typically, a chart comprising upper-case letters arranged in rows of decreasing size from top to bottom is read by the patient to the best of their ability from a fixed viewing distance. However, it has come to be realised that the combination of an irregular progression of letter sizes down the chart, the differing number of letters comprising each chart line, and the mixed relative legibility of the letters chosen, taken together compromise the accuracy of conventional VA determination: the typical test circumstances also prohibit statistical analysis of VA data.2

Thirty years ago a revised design of letter chart was suggested3 in an attempt to standardise the VA test: a geometric inter-line size progression was introduced (0.10 log unit); five letters were presented on each line; the letter set was restricted to ten letters chosen for their similar legibility;1,4 the smallest line was deliberately beyond the anticipated visual resolution limit to avoid truncation of the VA scale; and a strict protocol was devised to define the test end point. The design innovations of the Bailey-Lovie logarithm of the Minimum Angle of Resolution (logMAR) chart enhanced test accuracy and repeatability,5 and additionally facilitated parametric statistical analysis of VA data. Together with these improvements in technique, the test procedure long familiar to patient and practitioner could be retained; i.e., a criterionfree multiple forced-choice psychophysical paradigm.

For the literate patient, the letter-based acuity test procedure is quick and easy to perform. However, for illiterate or mute subjects and for research applications (including frequently-repeated vision assessments) an alternative symbol to alphabet letters becomes necessary. A popular choice under these circumstances is the four- (or less frequently eight-) orientation Landolt broken ring, a test stimulus which has also been adopted as a standard acuity optotype1 and with a clinical history nearly as long as the Snellen VA test format.6 Using this test optotype the patient is instructed to identify the location of the break in the ring symbol (i.e., at the top, bottom, left or right). The chart layout can be exactly the same as for conventional letter testing and, once the task is understood, VA determination can usually be readily accomplished. An additional benefit in the vision research environment is that, especially in combination with the logMAR chart design format, inter-subject and repeat testing reliability can be enhanced due to the absence of relative legibility issues encountered with the use of letters.

An important distinction must however be made between these two superficially similar acuity testing formats. The detection of the break in a Landolt ring test stimulus is a “resolution” task: the basic ring stimulus is visible, it is the location of the detail (the break) that must be reported. On the other hand, when dealing with a chart comprising lines of letters (even though the patient might be unaware of the extent of the available selection) the task is one of discriminating which letter is under scrutiny; i.e., a “recognition” task. The visual tasks are different and this might influence the VA result obtained; compensatory cognitive processes may be available to a subject reading individual chart letters that are not relevant or useful when that same subject is attempting to locate the position of the gap in a broken-ring stimulus. For this reason, the suggestion7 that an empirical size-scaling of test letters might produce equivalent legibility to Landolt ring optotype is unlikely to be universally successful or applicable.

Investigation of this topic is not new. For example, in recent times test-chart luminance,8 stimulus conditions in association with statistical modelling of VA measures,9 and lenticular10 and macular11 pathologies have all been shown to influence the relative (resolution versus recognition) acuity results. However, the availability of a contemporary acuity test and protocol (i.e., the logMAR chart format) facilitates a direct investigation of this issue in a clinical population. We were interested to assess the magnitude and direction of any inter-format acuity difference that might be uncovered using normally sighted non-pathological subjects; also to assess whether any difference is likely to be of clinical (or statistical) relevance.

Mean age of the subject group (n=300) was 23.3 years (standard deviation [SD] 7.5), range 16.0 to 39.8 years. Mean age of the repeat-test sub-group (n=50) was 22.3 years (SD 6.7), range 16.1 to 36.5 years.

Calculated21 mean sphero-cylindrical monocular spectacle prescription of the entire subject group was -0.25 -0.04×70°; for the repeat-test group it was -0.35 -0.10×76.9°.

It was possible to record a sighting preference without equivocation for each subject. Overall 66% of persons indicated a right-preferred (dominant) eye, and 68% in the repeat-test group: both of these proportions are in close agreement with a published general population estimate of approximately 70%.22

Initial analysis of the letter and Landolt logMAR acuity data indicated that subject's gender, the laterality of the dominant eye (i.e., either left or right), and the order of testing (ie, whether the letter or the Landolt chart was used first) were all non-statistically significant (and non-interactive) features (P>0.5). Consequently, the grouped VA data of all 300 subjects was used in all subsequent analyses, unless indicated otherwise in the text.

Mean monocular (dominant eye) letter acuity was -0.119 logMAR (SD 0.070), equivalent to Snellen 6/4.6. Mean monocular Landolt acuity was slightly worse, being -0.078 log - MAR (SD 0.074) or Snellen 6/5.0. As might be anticipated, inter-chart correlation of VA measurement was high (Pearson r=0.89, P<0.001). Further investigation of this relationship23 determined that the mean paired (letter minus Landolt) acuity difference was -0.041 logMAR units (SD 0.034): this difference was statistically significant (P≤0.0001). The calculated 95% limits of agreement (LoA) for this acuity difference (=SDX 1.96)23 were ±0.067 logMAR units: because a single optotype (letter or Landolt ring) on either chart was equivalent to 0.02 logMAR units, this means that when comparing acuity scores on the two charts a difference (better/worse) of 3 optotype can be taken as equivalent. These limits are illustrated in figure 1, where paired acuity difference (letter minus Landolt logMAR) is plotted against mean acuity. Although many data points are coincident, it is evident that the distribution is even across the plot area. This apparent uniformity of distribution was confirmed for these data by the value of the correlation coefficient (r=0.14, P=0.1): no statistically significant association existed between acuity level and paired acuity (letter versus Landolt) difference. What is demonstrated, however, is that slightly higher acuity values will tend to be recorded with the letter compared to the Landolt broken ring version of the logMAR test chart.

Figure 1.

LogMAR visual acuity limits of agreement plot, letter chart versus Landolt ring chart. The dependent variable is paired acuity difference [letter minus Landolt] and the independent variable is mean acuity: the ordinate is scale-marked in divisions of 0.02 logMAR units (ie, equivalent to one chart optotype). The inter-chart mean difference (0.041 logMAR units in the direction of the letter chart) is indicated by the continuous horizontal line across the centre of the plot; the pair of dashed horizontal lines mark the upper and lower 95% limits of agreement (LoA), the width being ±0.067 logMAR units or ±3.3 chart symbols. The three short vertical solid bars at the extreme right-hand side of the plot delineate the calculated 95% confidence intervals associated with the mean and the upper/lower limits, and are indicative of the precision linked to the estimates derived from the clinical subject group. N=300 (note that many data points are coincident).

(0.09MB).

It was important to establish test repeatability values for each logMAR chart type: substantial variation of VA measurements on one or both of the test charts would not only introduce comparative chart measurement difficulties but also have implications for clinical usefulness. In other words, only if the acuity recorded on either chart at a second occasion differed from the original VA by more than the repeatability value could it be regarded as a real acuity change. Using the data of the test-retest patient group, the calculated mean paired (original minus repeat) difference for the letter chart was 0.000 logMAR units (SD 0.021, P=1.00); similarly, for the Landolt chart the calculated mean paired difference was -0.008 logMAR units (SD 0.025, P=0.05). The calculated 95% LoA are thus 0.042 logMAR units (or ±2.1 letters) for the letter chart and 0.049 logMAR units (or ±2.4 rings) for the Landolt chart. Test variability was slightly greater for the Landolt as compared to the letter chart, but only by a fraction (0.3) of a single optotype: however, importantly, both charts show test repeatability values that are within the ±0.067 logMAR units (±3.3 chart symbols) inter-chart 95% LoA mentioned above.

These repeatability data are summarised in figure 2 (letters) and figure 3 (Landolt rings). As found for the main group, the distribution was uniform in each case (letters: r=0.04, P=0.7; Landolt: r=0.02, P=0.8).

Figure 2.

LogMAR visual acuity limits of agreement plots for repeat testing: figure details as per figure 1. Repeat testing with letter chart: mean difference [original minus repeat] is 0.000 logMAR units, with LoA of ±0.042 logMAR units or ±2.1 chart letters.

(0.08MB).

Figure 3.

Repeat testing with Landolt ring chart: the bias is -0.008 logMAR units, with LoA of ±0.049 logMAR units or ±2.4 chart rings. N=50 in each case (note that in each plot several data points are coincident).

(0.08MB).

Analysis of these new monocular recognition (letter) and resolution (Landolt) logMAR acuity data indicates: (i) a high correlation (0.89) was present between the acuity measurements obtained with the two charts; (ii) a higher mean VA (-0.041 logMAR units) was recorded with letters compared to Landolt optotype, with calculated 95% LoA equivalent to ±3.3 chart optotype; (iii) test repeatability for either chart type was smaller than the between chart 95% LoA, being equivalent to ±2.1 letters and ±2.4 Landolt rings, respectively.

Viewed from the clinician's perspective, although the inter-chart acuity difference was statistically significant, can this distinction be regarded as clinically relevant?

Fifty-five years ago Sloan24 considered that binocular acuity recorded with her chosen letter set was clinically equivalent to that recorded with Landolt symbols. A more recent investigation9 using the logMAR chart design format has reported that binocular acuity determined with Sloan letters was 0.038 logMAR units (1.9 letters) better than that recorded with Landolt rings; in the same investigation9 binocular VA obtained with British Standard letters was 0.005 logMAR units (0.25 letters) better than that with Landolt symbols.

While acknowledging that the monocular acuity testing in the work reported herein might produce an inter-study difference in the absolute values obtained, nevertheless it is hard to disagree with the conclusion of Raasch et al.9; namely, that acuity measured with letters (British Standard or Sloan sets) will tend to be slightly better and slightly more repeatable than that obtained with Landolt symbols.

Photopic chart (background) luminance in the present investigation and in the earlier study9 was similar and approximately equal to 160 cd/m2, this being the level advised by British Standard BS4274:19684 for sight testing. At this luminance value Sheedy et al.8 have also found letter acuity to be superior to Landolt ring acuity (whereas below approximately 120 cd/m2 the situation was reversed).

All of these results were obtained upon normally sighted subjects tested under ideal conditions. Do the same conclusions apply if subjects have compromised or pathologicallyimpaired vision? The “resolution versus recognition” issue has been investigated in young amblyopic individuals and normally sighted persons with simulated cataract, ametropia and inappropriate refractive correction;25 as well as in older persons with actual cataracts,10 and patients with macular holes before and after surgery.11 The consensus of these studies is that compared to letter acuity, the Landolt VA deficit increases and the associated variability of the results widens under adverse viewing conditions. This conclusion might be regarded as somewhat paradoxical, given that the identification of a letter is regarded as a more demanding or complex visual task than that of locating the break in a ring.24 It has also been observed that Landolt acuity is less compromised by changes in physical conditions such as test chart luminance compared to letter VA.8 However, it has long been appreciated that the legibility of individual letters varies considerably.1,24 And therein might lie the answer to the paradox:11 under non-optimum viewing conditions, particular constructional components of an otherwise indistinct letter might be distinguished, enabling cognitive recognition of that letter, a benefit denied the observer of a uniform stimulus configuration such as the Landolt ring. Extending this reasoning further, perhaps it is the case that Landolt acuity reflects an individual's retinal-based resolution ability; additional cognitive input refines the process and enhances the visual percept. Such considerations highlight the acknowledgement by many clinicians that letter and Landolt ring charts may have particular and appropriate applications depending upon the visual debility or the characteristics of the patient group under investigation.

Consequently, an extra informative dimension is provided by the comparative VA outcomes of subjects with compromised visual abilities. However, with normally sighted subjects tested under optimum conditions, we conclude that on the basis of the new data presented here an assumption of clinical equivalence between letter and Landolt VA is justified.