Share this article on:

Face Recognition in the Elderly


doi: 10.1097/01.opx.0000180764.68737.91
Articles: Original Article

Purpose. The purpose of this study was to assess face recognition ability in a large sample of elders (n = 572, mean age = 78.1 years) and to identify factors that affect performance.

Methods. Face recognition was measured by presenting standardized faces of varying sizes to simulate normal-sized faces at different viewing distances. Subjects were asked to identify the name of the person and their facial expression. Threshold equivalent viewing distance (EVD) was calculated. High- and low-contrast acuity, contrast sensitivity, low-contrast/low-luminance acuity, disability glare, stereoacuity, and visual field measures (with and without an attentional task) were also measured. These vision measures, along with demographic information (age, sex, education) and cognitive status, were included in a multiple regression analysis to determine which factors predicted task performance.

Results. This cross-sectional sample of elders showed significant declines in face recognition with age. Mean threshold EVD ranged from 8.0 m for participants ≤70 years of age to 2.2 meters for those over 85 years. Multiple regression analysis revealed that age, sex, years of education, spatial vision, and cognitive status were all significant predictors of face recognition, accounting for approximately 46% of the variability. Spatial vision (high-contrast acuity) and age were the best predictors. Although each spatial vision measure was significantly correlated with face recognition, adding low-contrast or contrast sensitivity measures to the regression analysis explained no more variance than age and high-contrast acuity alone.

Conclusions. The marked decline in face recognition ability in elders is related to declines in spatial vision and cognitive status. All spatial vision measures have similar predictive ability for face recognition.

Smith Kettlewell Eye Research Institute, San Francisco, California (LAL, GHP, MES, JAB); and the School of Optometry, University of California, Berkeley, Berkeley, California (GHP, MES)

Received February 17, 2005; accepted May 25, 2005.

The ability to recognize faces is an important skill for the elderly. Previous research has investigated age-related decrements in face recognition, but most have focused on the effect of aging on memory for faces,1–3 finding significant deficits with advancing age.

Other studies have emphasized vision-related factors. For example, Owsley et al.4 found that contrast discrimination thresholds for faces were approximately 0.30 log units worse in elderly than in young observers. Owsley and Sloane5 examined the relationship between contrast sensitivity, high-contrast acuity, and contrast thresholds for real-world target identification (including faces) for subjects ranging in age from 20 to 77 years. Age and contrast sensitivity (6 cycles/degree) were statistically significant predictors of face identification. However, once these factors were taken into account, acuity added no more predictive power to the model.

Studies investigating the relationship between contrast sensitivity and face recognition in low-vision patients (primarily with a diagnosis of age-related maculopathy [ARM]) have had variable results. Some studies provided support for the predictive ability of contrast sensitivity in face recognition,6 whereas others suggest that other vision measures are more strongly associated with face recognition.7,8 Specifically, Bullimore et al.,7 using a face size threshold task, found that word reading acuity had a higher association with face recognition than letter acuity, grating acuity, or contrast sensitivity. Tejeria et al.8 showed that distance acuity was most highly related to recognition of familiar faces, whereas continuous word reading acuity had the highest association with discrimination of facial expression.

This study differs from previous work on face recognition in that our participants are part of a large, population-based sample of elders. Another unique aspect of this study is the large number of individuals over the age of 85, a largely unstudied but rapidly growing age group. Although most participants have relatively good high-contrast acuity, they show a broad range of visual abilities when nonstandard vision measures are assessed.9,10 The goal of the current study was to focus on vision-related factors affecting face recognition in this large sample of elders.

Back to Top | Article Outline



Face recognition was assessed at the second measurement period of the Smith-Kettlewell Eye Research Institute (SKI) study, an ongoing, longitudinal assessment of health and functioning11 and vision function9,10 of an elderly population in Marin County, California. Face recognition data were available for 572 of 596 individuals who completed the second round of testing. The mean age of the elderly participants in this cross-sectional sample was 78.1 ± 7.7 years (range, 64–102 years) and 56% were female.

The study adhered to the tenets of the Declaration of Helsinki and was approved by the Smith-Kettlewell Eye Research Institute and the California Pacific Medical Center's Institutional Review Board (IRB). After an explanation of the test procedures, possible risks, and discomforts, each participant signed the IRB-approved consent form.

Back to Top | Article Outline


Face recognition was measured using a variant of a technique described by Bullimore et al.,7 in which standardized faces of varying sizes were used to simulate normal-sized faces at different viewing distances. Test stimuli were taken from the frequently used “Pictures of Facial Affect” set.12 Four models (2 males, 2 females) displaying 4 possible expressions (happy, sad, angry, or afraid) were used for this study. No “person”/”expression” combination was used more than once and each “person” was presented with only 3 expressions to create the 12-slide set on which each participant was tested. Faces were cropped at the inner hairline to prevent recognition on the basis of distinctive hairline, color, or style. Photographs of the 4 faces in neutral poses, labeled with a name (i.e., Betty, Alice, Sam, or George) as well as a list of the 4 possible expressions, were continuously displayed. The angular size of each of the 4 neutral expression photographs was 7.2 × 10.0° (to approximate the average size of an adult face at 1 m), and the names and expressions were displayed in large print (Palatino, 60-point font at a distance of 1 m). Participants were free to look back and forth between the test face and the neutral poses, associated names, and list of 4 expressions.

Face stimuli were projected on a screen at a viewing distance of 1 m using a slide projector. Six simulated or equivalent viewing distances (EVD) ranging from 0.75 to 24 m were used. The angular size of the projected image corresponding to EVD = 0.75 m was 9.1 × 14.3°. Each consecutive EVD corresponded to a doubling of its predecessor (i.e., halving of angular size). Two images were presented at each size for a total of 12 presentations. Participants viewed the faces through their habitual distance correction and a +1.00-D lens.

Participants were initially shown the panel of neutral faces with associated names and were told that they would be shown a slide of one of the models displaying one of the 4 listed emotional expressions. Their task was to identify who was pictured and which expression was depicted. They were told that the neutral faces would be present throughout the experiment but that they would only have 10 seconds to view each test face. No repetition of slides was allowed, participants were asked to guess if they were unsure, and no feedback was given. Two slides were presented at each EVD size; presentations continued until the participant viewed all 12 slides (or missed both the names and expressions of 2 slides of the same size).

Threshold EVD was calculated for each individual participant using probit analysis. The primary face recognition measure of interest was “person and expression” EVD (i.e., the participant had to correctly identify both the person and expression). In addition to this measure, “person only” and “expression only” EVDs were calculated separately for each individual. The average error of the “person and expression” threshold estimate was 0.28 log units. However, most participants (73%) had well-behaved psychometric functions, and the median error of the threshold EVD estimate was 0.07 log units.

Participants who were unable to identify any of the face stimuli (n = 5: 0.8% of the sample) were arbitrarily assigned an EVD = 0.50 m (-0.30 log meters). Those who were able to correctly identify all (n = 4: 0.7% of the sample) were assigned an EVD of 24.0 m (1.38 log meters).

Back to Top | Article Outline

Vision Measures

The test battery has been described in detail elsewhere.9 All participants were tested under binocular viewing conditions with habitual correction. The luminance for all vision tests was ∼150 cd/m2. All acuity tests were scored as number of letters read correctly and converted to logarithm of the minimum angle of resolution (logMAR).

Briefly, the vision measures used in the current analyses included: 1. High- and low-contrast distance acuity (HCA and LCA) were obtained using Bailey-Lovie charts13,14 at 3-m viewing distance. 2. Contrast sensitivity (CS) was measured with the Pelli-Robson Chart15 at 3-m viewing distance. The Pelli-Robson chart was scored letter by letter and converted to log CS.16 3. Low-contrast/low-luminance near acuity (SKILL Dark Acuity [SKD]) was measured using the “dark” side of the Smith-Kettlewell Institute Low Luminance (SKILL) Card17 at a test distance of 40 cm. This chart consists of black letters on a dark gray background (approximately 15% Weber contrast). 4. Low-contrast near acuity in disability glare (Glare) was measured with the Berkeley Glare Test18 at 40 cm. Participants read the chart with the 3300 cd/m2 surrounding glare. Those who were unable to read the chart at this glare level (n = 48, 8.4%) were asked to read the chart with a lower glare setting (800 cd/m2). For these individuals, a correction value of 25 letters was subtracted from the glare acuity score before converting to logMAR. 5. Standard and attentional fields were measured with a modified Synemed perimeter. This test and the analysis technique have been described previously.9 Briefly, visual fields were measured using bright targets without (standard) and with (attentional) the addition of a central distracting task (count the number of central target flashes while completing the field test). Standard field score was percentage of locations missed (corrected for false-positive responses). Attentional field was scored the same way, except that the difference between reported and actual number of central flashes was also taken into account. Attentional field (percentage of locations missed) was subtracted from the standard field score to yield a measure of the impact of attention on field performance. 6. Stereoacuity was measured with the Frisby Test,19 which consists of 3 plates of different thickness. At the 40-cm viewing distance, these plates yield disparities of 340, 170, and 85 arc seconds. Performance was measured as a categorical variable (i.e., number of plates passed = 0, 1, 2, or 3).

Other demographic information (age, sex, years of education) and an estimate of cognitive status were available for each participant. The cognitive status measure used for the current study was the Mental Alternation Test (MAT).20,21 In the MAT, participants are asked to alternate between numbers and letters of the alphabet in order (i.e., 1-A, 2-B, 3-C, and so on). The MAT is scored as the number of letters and numbers correct in 30 seconds. This test is similar to the Trail-Making Test (TMT), a measure of cognitive flexibility (i.e., the ability to maintain and switch between multiple conceptual representations of knowledge).22 The MAT was selected for the current study, because unlike the TMT, the MAT does not require visual search or line drawing on the part of the participant. In a study by Billik et al.,21 MAT scores for normal geriatric controls (mean age 77.3 ± 7.9 years) ranged from 11 to 44 (mean MAT = 28.1 ± 7.7).

As part of the test battery, participants completed a modified version of the Visual Activities Questionnaire (VAQ).23 Subjective experience of face recognition difficulty was assessed using 2 items. Participants were asked to provide a frequency response (never, rarely, sometimes, often, or always) to the following 2 statements: “I have trouble recognizing people's faces when they are across the room.” “I have trouble recognizing people's faces in dimly lit places.”

Back to Top | Article Outline

Data Analysis

Data were analyzed using SPSS 6.1 for Macintosh. Descriptive statistics (means, medians, and standard deviations for continuous variables; frequency distributions for categorical variables) were computed and examined. Univariate analyses included analysis of variance (ANOVA), Student's t test, and Pearson r correlations for continuous data. Where categorical data are presented (e.g., sex or stereoacuity), Spearman rho (rank order) correlations are reported.

Preliminary analyses using stepwise regressiona were used to determine which spatial vision measure(s) were significant independent predictors of face recognition. Additionally, to address the possibility that predictor variables could differentially affect face recognition, depending on participants' age or sex, tests of the interaction of each predictor variable with age and sex were included in preliminary models. Predictor variables or interaction effects associated with face recognition at a significance level of 0.15 were included in a standard multiple regression model.b

The main dependent variable under investigation is “person and expression” EVD, but when relevant, “person only” and “expression only” data are also reported. Because the face recognition distribution expressed in meters was positively skewed, all statistical analyses were performed using log meter EVD.

Back to Top | Article Outline


The mean high contrast distance acuity for the sample was 0.14 ± 0.21 logMAR (range: −0.18 to 1.18: Snellen: 20/13 to 20/300), and mean log contrast sensitivity was 1.44 ± 0.28 (range: 0.0 to 1.90). As previously reported,9,10,24 the population from which the SKI Study is drawn is a highly educated group of elders. In the current sample, 76.2% reported receiving more than 12 years of education. Table 1 presents the characteristics of this sample by age group.

Back to Top | Article Outline

Face Recognition, Demographic and Vision Measures: Univariate Analyses

Figure 1 presents the mean face recognition EVDs for identifying “expression only,” “person only,” and “person and expression” for participants grouped by age. The error bars indicate ± one standard error of mean (SEM). Significant declines in face recognition are evident. Clearly, the ability to identify emotional expression was an easier task than person identification for many of the elderly participants in this study.

When elders were required to correctly identify both the expression and name of the person depicted in the slide, performance declined dramatically. For the oldest age group (>85 year olds), “person and expression” EVD was, on average 0.24 log meters (1.74 m) compared with a mean of 0.82 log meters (6.61 m) for the 64 to 70 year olds. The significant declines in “person and expression” face recognition with age were confirmed using a one-way analysis of variance (F[4, 567] = 60.49, p < 0.001). Tukey test for pairwise comparisons25 was used to determine whether the differences between the 5 age groups were statistically significant. This analysis revealed that face recognition performance was significantly different (p < 0.05) for all age groups except the 71- to 75-year-old versus 76- to 80-year-old groups.

For comparison, face recognition ability was assessed using the same task in a convenience sample of 10 younger subjects (mean age = 32.9 ± 4.8 years) with very good acuity (−0.13 ± 0.08 logMAR: approximately 20/15) and normal Pelli-Robson CS (1.87 ± 0.05 log). “Person and expression” face recognition in this nonrandom sample was 17.1 ± 6.6 m (1.19 ± 0.21 log meters).

Table 2 presents bivariate correlations between “person and expression” EVD and the vision and demographic measures examined in this study. Correlations between independent variables and “expression only” or “person only” EVD are not presented but were virtually identical.

As evident by the statistically significant correlations presented in Table 2, each of the measured variables was associated with face recognition. Poorer vision (as measured by any of the vision measures), lower MAT scores (i.e., poorer cognitive status), and fewer years of education were each related to poor face recognition task performance. There was also a small but statistically significant correlation of sex with face recognition (Spearman rho = 0.11). This indicates that face recognition performance was slightly better in females than males (means = 0.64 and 0.58 log meters, respectively; t[570] = −2.35, p < 0.02).c

Back to Top | Article Outline

Association Between Subjective and Objective Measures of Face Recognition

Figure 2 presents the mean “person and expression” EVD for participants grouped according to their answers to the modified VAQ items assessing face recognition difficulties. Clearly, there is a relationship between participants' self-reported difficulties and function measured in the current study. Separate one-way analyses of variance determined that the decrements in “person and expression” EVD with self-report were significant for both face recognition questions (i.e., difficulty recognizing faces: “across the room” [F(4, 566) = 18.92, p < 0.001], and “in dimly lit places” [F(4, 567) = 21.79, p < 0.001]). Tukey tests for pairwise comparisons revealed that individuals in the “never,” “rarely,” or “sometimes” self-report categories had significantly better “person and expression” EVDs than those reporting “always” or “often” (p < 0.05) for both questions. Additionally, for the question regarding difficulty recognizing faces in dimly lit places, significant differences in face recognition performance were also obtained for the “rarely” versus “sometimes” and “often” versus “always” self-report category.

Back to Top | Article Outline

Predictors of Face Recognition Performance: Multivariate Analysis

Examination of Table 2 shows that spatial vision measures (HCA, LCA, CS, SKD, and Glare) have approximately the same correlation with face recognition (Pearson rs = 0.49 to 0.56). However, even higher correlations were obtained between these various spatial vision measures (Pearson rs = 0.72 to 0.92). To avoid multicollinearity in the multivariate analysis, stepwise linear regression was performed to determine which of the spatial vision measures (HCA, LCA, CS, SKD, glare) was the best independent predictor of face recognition performance when age and sex were taken into account. The results of both backward and forward stepping models yielded identical results: HCA was identified in all stepwise models, so it was included in the final standard multiple regression analysis. However, it is important to note that multiple Rs obtained in regression models substituting LCA, CS, or SKD were not significantly different from that using HCA. Because these low-contrast variables have slightly higher correlations with age (Pearson rs = 0.51, 0.54, and 0.64, respectively) than HCA (r = 0.49), it is not surprising that the latter was selected in a stepwise regression that included age.

Standard multiple regression revealed that age, sex, years of education, cognitive status (MAT), high-contrast acuity (HCA), and stereoacuity were significant independent predictors of “person and expression” face recognition (multiple r = 0.679, R2 = 0.467, R2[adj] = 0.455, F[6, 536] = 76.45, p < 0.0001). Table 3 presents the unstandardized coefficients (B), the standardized coefficients (β), p values, and the squared semipartial correlation (sr2) associated with each independent variable. In standard multiple regression, sr2 is a measure of the amount of unique variance explained by each independent variable.

Very similar results were obtained with multiple regression models using “expression only” or “person only” face recognition as the dependent variable, except that cognitive status did not attain statistical significance in the “person only” regression analysis.

Back to Top | Article Outline

Face Recognition in Elders with Good versus Poor Acuity

Figure 3 presents mean “person and expression” EVD as a function of age group for our elderly participants with different levels of HCA (i.e., ≤0.0 logMAR [better/equal to 20/20]: black bars; 0.02 to 0.30 logMAR [20/21 to 20/40] gray bars; >0.30 logMAR [worse than 20/40] white bars). A 4 (age group) × 3 (HCA category) between-subjects analysis of variance revealed significant main effects of age group (F[3, 453] = 9.89, p < 0.001) and HCA category (F[2, 453] = 24.87, p < 0.001), but the interaction of age and HCA was nonsignificant (F[6, 453] = 0.58, p > 0.70).

Stepwise multiple linear regression analysis conducted on the subset of participants with HCA better or equal to 20/40 (≤0.30 logMAR) revealed that LCA was the best independent predictor of face recognition. LCA, along with age, sex, MAT, and years of education, accounted for approximately 29% of the variability in face recognition (multiple r = 0.546, R2 = 0.298, R2[adj] = 0.291, F[5, 462] = 39.29, p < 0.001). Again, much of the predictive ability was accounted for by variance shared by the independent variables. Age made the strongest unique contribution to the model (8.7%), with only 2.8% attributable to LCA alone.

Back to Top | Article Outline


Face recognition task performance showed a significant decline with age. These decrements in face recognition are even seen in elders with acuity better than or equal to 20/20. This is consistent with the results of the multiple regression, which revealed that age, years of education, cognitive status, and acuity were all significant independent predictors of face recognition performance. However, together these variables only account for ∼46% of the total variability. What else could be accounting for the rest? The face recognition task was part of a large battery of vision tests used in the SKI study, so it was necessary to limit the time spent on each task to avoid tiring out these elderly observers. The use of fewer trials (only 2 per stimulus size) in this study may have made the measures more variable than had a full psychophysical test procedure been used. Furthermore, the unusual faces (black and white, hairline cropped, relatively short duration, and so on) could also have increased variability.

Self-reported difficulty in recognizing faces was also consistent with task performance decrements in the current study, but the correlations between self-report and the face recognition tasks were fairly low. This is similar to the results reported by Szlyk et al.26 who assessed the relationship between self-perceived difficulty and performance of relevant tasks in 62 individuals with retinitis pigmentosa. These authors reported a correlation of 0.45 between self-reported difficulty (i.e., “recognizing faces across the street”) and ability to recognize facial expressions from a distance of 20 feet.

Although direct comparisons between the performance of our elderly participants and that of the convenience sample of younger adults should be met with extreme caution, there are sizable differences between the 2 groups. The EVDs reported in the current study were markedly shorter than those reported for normal elderly controls in the study by Bullimore et al.7 However, there are some notable procedural differences that may have led to the disparate results.

Our participants were taken from a population-based study; the sample included individuals with eye disease as well as those with good ocular health, and eye health was not assessed in this study. It is possible that there was significant eye disease even in our 64- to 70-year-old group, reducing their performance relative to the healthy, aged controls of Bullimore et al.7 However, given the fact that there was no difference in the face recognition performance of 64- to 70-year-old participants with better than 0.00 logMAR (20/20 acuity) versus those with 0.02 to 0.30 logMAR (20/21 to 20/40 acuity), it is unlikely that this can account for the difference between this study and that of Bullimore et al.7

Because the goal of the SKI Study is aimed at assessing current vision status in this elderly population, habitual correction rather than best-corrected vision was used. As mentioned previously, the use of fewer trials in our “screening” study may have made the measures more variable than had a full psychophysical test procedure been used. Additionally, Bullimore et al.7 cropped the photos to remove the hair outline to reduce identification based on hairstyle, whereas our faces were cropped at the inner hairline in an effort to make it even less likely that the hair would be used as a cue to identification. All these factors could account for the differences between the 2 studies.

It has been suggested that CS plays an important role in face and object recognition.5 On the other hand, Bullimore et al.7 found that word reading acuity was most highly correlated with face recognition in their sample of ARM patients. Furthermore, the correlation between contrast detection and face recognition was not statistically significant. These authors suggested that the difference might be partly the result of the use of a different task: Owsley and Sloane5 assessed face recognition with a contrast discrimination task, whereas the EVD face recognition task is more consistent with acuity measurement. Furthermore, there were differences in acuity levels of participants in the 2 studies. Although Owsley and Sloane5 did not exclude subjects on the basis of ocular diagnosis, their range of acuities was quite small (i.e., 20/15 to 20/50). The acuities of the ARM participants of Bullimore et al.7 ranged from 20/40 to 20/400 (0.30 to 1.30 logMAR) acuity. With a wider range of acuities, one might expect to see greater impact of acuity on face recognition. This is supported by the current study. Our participants had acuities ranging from -0.18 to 1.18 logMAR (20/13 to 20/300), although very few had acuities worse than 20/100 (n = 16, 2.8% of the sample). When we restricted our analysis to individuals with HCA ≤0.30 logMAR (i.e., better than or equal to 20/40), LCA was the spatial vision measure that best predicted face recognition performance.

Our results were similar to those of Tejeria et al.8 with ARM patients: Once age and HCA were included in regression model, CS (or any other low-contrast variable) provided no additional explanatory power. Again, we must reiterate that the multiple Rs obtained in regression models substituting LCA, CS, or SKD were not significantly different from that using HCA. That is, any one of these spatial vision measures yields approximately equivalent predictive ability.

West et al.27 found that both HCA and CS (measured with the Pelli-Robson chart) were significant independent predictors of face recognition performance in their large, population-based sample (Salisbury Eye Evaluation: SEE Study). What accounts for the difference? For one thing, the face recognition task included in the SEE study was different from the current task. Participants were simultaneously shown 4 faces in different poses (approximately the size of our EVD = 3 m). Three depicted the same individual, the fourth was a different face, and the task was to indicate this “odd man out.” Face recognition was scored as number correct out of 15 possible. The SEE study also used best-corrected acuity and had a much larger sample size than the current study (2520 vs. 572). The additional power in a larger sample size would increase the chances of obtaining a significant effect. A larger percentage of our face recognition participants were over the age of 80 (38% compared with 11% for SEE Study). Age is such a strong independent predictor of face recognition ability and is more highly correlated with CS and the low-contrast measures (i.e., shares more variability) than with high-contrast acuity. The inclusion of a wider range of ages in the current study might make it less likely that a significant independent contribution of low-contrast or CS measures would be revealed by the stepwise regression analysis.

A modest but statistically significant association between coarse stereoacuity and face recognition performance was found in the current study. Although it is plausible that stereoacuity may play some role in recognition of real, 3-dimensional faces, it is difficult to assign a direct role in the recognition of 2-dimensional images. It is possible, however, that stereoacuity has an indirect role: Individuals with unequal vision in the 2 eyes will have poor stereoacuity. In those eyes, there may be a failure of binocular summation and, in some cases, even binocular inhibition, both of which are likely to reduce CS and resolution, particularly for low-contrast targets.28,29 Because CS and low-contrast acuity is related to face recognition, we hypothesize that those with no binocular summation or with binocular inhibition will have poorer face recognition performance. Thus, the association with stereoacuity could reflect unequal vision in the 2 eyes, reducing CS/low-contrast resolution and thus face recognition.

Clearly this experimental paradigm (i.e., 2-dimensional, black and white photos, cropped hairlines, lack of familiarity, and other important cues used to identify individuals such as movement patterns and voice) is a far cry from “real-world” face recognition. However, this is an early attempt to quantify face recognition problems experienced by elderly individuals and identify the factors that play a role. Face recognition ability is important in the social life of elders with “normal” vision, as well as those with low vision.

Back to Top | Article Outline


The authors thank Bebe St. John, Ruth Youngquist, and the volunteers at The Buck Center for Research in Aging. Funding for this study was provided by the National Institute of Health (EY09588 to JAB) and Smith-Kettlewell Eye Research Institute. These results were presented, in part, at the Association for Research in Vision and Ophthalmology, 2001 in Ft. Lauderdale, Florida.

Lori A. Lott

Smith-Kettlewell Eye Research Institute

2318 Fillmore Street

San Francisco, CA 94115


Back to Top | Article Outline


1. Smith AD, Winograd E. Adult age differences in remembering faces. Dev Psychol 1978;14:443–4.
2. Ferris SH, Crook T, Clark E, McCarthy M, Rae D. Facial recognition memory deficits in normal aging and senile dementia. J Gerontol 1980;35:707–14.
3. Bartlett JC, Leslie JE, Tubbs A, Fulton A. Aging and memory for pictures of faces. Psychol Aging 1989;4:276–83.
4. Owsley C, Sekuler R, Boldt C. Aging and low-contrast vision: face perception. Invest Ophthalmol Vis Sci 1981;21:362–5.
5. Owsley C, Sloane ME. Contrast sensitivity, acuity, and the perception of 'real-world' targets. Br J Ophthalmol 1987;71:791–6.
6. Alexander MF, Maguire MG, Lietman TM, Snyder JR, Elman MJ, Fine SL. Assessment of visual function in patients with age-related macular degeneration and low visual acuity. Arch Ophthalmol 1988;106:1543–7.
7. Bullimore MA, Bailey IL, Wacker RT. Face recognition in age-related maculopathy. Invest Ophthalmol Vis Sci 1991;32:2020–9.
8. Tejeria L, Harper RA, Artes PH, Dickinson CM. Face recognition in age related macular degeneration: perceived disability, measured disability, and performance with a bioptic device. Br J Ophthalmol 2002;86:1019–26.
9. Haegerstrom-Portnoy G, Schneck ME, Brabyn JA. Seeing into old age: vision function beyond acuity. Optom Vis Sci 1999;76:141–58.
10. Brabyn J, Schneck M, Haegerstrom-Portnoy G, Lott L. The Smith-Kettlewell Institute (SKI) longitudinal study of vision function and its impact among the elderly: an overview. Optom Vis Sci 2001;78:264–9.
11. Reed D, Satariano WA, Gildengorin G, McMahon K, Fleshman R, Schneider E. Health and functioning among the elderly of Marin County, California: a glimpse of the future. J Gerontol A Biol Sci Med Sci 1995;50:M61–9.
12. Ekman P, Friesen WV. Pictures of Facial Affect. Palo Alto, CA: Consulting Psychologists Press, 1975.
13. Bailey IL, Lovie JE. New design principles for visual acuity letter charts. Am J Optom Physiol Opt 1976;53:740–5.
14. Ferris FL III, Kassoff A, Bresnick GH, Bailey I. New visual acuity charts for clinical research. Am J Ophthalmol 1982;94:91–6.
15. Pelli DG, Robson JG, Wilkins AJ. The design of a new letter chart for measuring contrast sensitivity. Clin Vis Sci 1988;2:187–99.
16. Elliot DB, Bullimore MA, Bailey IL. Improving the reliability of the Pelli-Robson contrast sensitivity test. Clin Vis Sci 1991;6:471–5.
17. Haegerstrom-Portnoy G, Brabyn J, Schneck ME, Jampolsky A. The SKILL Card. An acuity test of reduced luminance and contrast. Smith-Kettlewell Institute Low Luminance. Invest Ophthalmol Vis Sci 1997;38:207–18.
18. Bailey IL, Bullimore MA. A new test for the evaluation of disability glare. Optom Vis Sci 1991;68:911–7.
19. Frisby JP. The Frisby stereotest: amended instructions. Br Orthopt J 1980;37:108–112.
20. Jones BN, Teng EL, Folstein MF, Harrison KS. A new bedside test of cognition for patients with HIV infection. Ann Intern Med 1993;119:1001–4.
21. Billick SB, Siedenburg E, Burgert W III, Bruni-Solhkhah SM. Validation of the Mental Alternation Test with the Mini-Mental State Examination in geriatric psychiatric inpatients and normal controls. Compr Psychiatry 2001;42:202–5.
22. Kortte KB, Horner MD, Windham WK. The trail making test, part B: cognitive flexibility or ability to maintain set? Appl Neuropsychol 2002;9:106–9.
23. Sloane ME, Ball K, Owsley C, Bruni JR, Roenker DL. The Visual Activities Questionnaire: developing an instrument for assessing problems in everyday visual tasks. In: Noninvasive Assessment of the Visual System, Technical Digest Series, vol 1. Washington, DC: Optical Society of America, 1992:26-9.
24. Lott LA, Schneck ME, Haegerstrom-Portnoy G, Brabyn JA, Gildengorin GL, West CG. Reading performance in older adults with good acuity. Optom Vis Sci 2001;78:316–24.
25. Keppel G. Design and Analysis: A Researcher's Handbook, 2nd ed. Englewood Cliffs, NJ: Prentice Hall, 1982.
26. Szlyk JP, Seiple W, Fishman GA, Alexander KR, Grover S, Mahler CL. Perceived and actual performance of daily tasks: relationship to visual function tests in individuals with retinitis pigmentosa. Ophthalmology 2001;108:65–75.
27. West SK, Rubin GS, Broman AT, Munoz B, Bandeen-Roche K, Turano K. How does visual impairment affect performance on tasks of everyday life? The SEE Project. Salisbury Eye Evaluation. Arch Ophthalmol 2002;120:774–80.
28. Pardhan S, Gilchrist J. The importance of measuring binocular contrast sensitivity in unilateral cataract. Eye 1991;5:31–5.
29. Pardhan S. Binocular performance in patients with unilateral cataract using the Regan test: binocular summation and inhibition with low-contrast charts. Eye 1993;7:59–62.

a Identical results were obtained using forward and backward elimination models. Cited Here...

b None of the interactions terms was statistically significant, so the final regression analysis includes no interactions. Cited Here...

c In the context of relatively large samples such as this, a statistically significant 0.06 log unit difference illustrates the important distinction between “statistical” and “clinical” or “practical” significance. Cited Here...


aging; face recognition; acuity; contrast sensitivity; low-contrast vision

© 2005 American Academy of Optometry