The CSF can be used to predict an individual’s ability to detect and recognize or categorize targets: sensitivity to high spatial frequencies is directly related to target detection and fine discriminations; but, sensitivity to lower spatial frequencies may be the basis for much of our ability to recognize and categorize objects [26–28]. Object perception involves both holistic (Gestalt-like) and analytic (bottom-up) processes (see ); but regardless of one’s preference for one approach or the other, the visual input to both must be filtered through the ST-CSF.
Figure 2 shows that females and males have similar sensitivities at low spatial frequencies; but with increasing spatial frequency, males’ higher sensitivity becomes more and more apparent. (An earlier study that used a much more restricted range of spatial and temporal frequencies, found a similar pattern of male–female differences ) The higher sensitivity of males at middle and high spatial frequencies may be common to mammals: the same pattern of sensitivities has been found, using behavioral techniques, in hooded rats .
The ST-CSF is probably determined at an early cortical level, whereas object recognition is a higher cortical function. We assert that primary visual cortex (V1) is the major locus for the ST-CSF: most of its neurons are narrowly tuned to specific stimulus sizes and orientations. In fact their receptive fields are best described as responding to a narrow band of spatial frequencies [46, 47], and these receptive fields are orientation specific, as is the ST-CSF – an adapting grating is maximally effective when its orientation is the same as that of the test gratings . Furthermore, an adapting grating is effective when it is presented to one eye and the test grating to the other  – V1 is the first level at which the neurons are binocular, with matching receptive fields in each eye. But the detection capabilities of V1 neurons that we have just described should be tempered: the spatial tuning of a V1 neuron’s responses changes dynamically according to the prevailing stimulus environment [48, 49].
We can only speculate why there is the sex difference that we report here. Are there plausible differences in cortical receptive fields that might account for the difference? One possibility is that females' receptive fields are slightly larger than those of males. Or, males might have greater intra-cortical inhibitory feedback that might increase the degree to which their receptive fields are tuned to finer patterns. Additionally, human males, like rats, may also have 20% more neurons in their visual cortex [16, 17]; this alone might increase sampling of the visual image and increase signal-to-noise ratios.
Finally, our data show that any appreciable temporal variation in the stimulus (e.g., 4 Hz and greater modulations) shifts the ST-CSF towards lower spatial frequencies – there is a drop in sensitivity at high spatial frequencies and a considerable improvement in sensitivity at lower spatial frequencies (see Figure 3). Furthermore, all the CSFs shift by almost the same amount, and this shift is not very different between males and females. (There is, of course an absolute difference between the sexes, in that males are generally more sensitive.)
Thus, the canonical static CSF (approximated by our 1-Hz condition) may not be the relevant one in most circumstances – the most relevant may be the shape we obtain when gratings are temporally modulated (see Figure 3). The more usual real-world condition is that the retinal image is far from stationary – either the eye or the entire body moves, or objects in the environment move. In fact it may be ecologically useful that the curves shift in this way: object categorization depends on low spatial frequencies; it is probably more useful to recognize that a moving object is a cheetah without necessarily identifying which cheetah based on fine details of number of spots. Incidentally, even when the world is static and the observer is fixating as steadily as possible, there is still an appreciable temporal modulation of the retinal image: under best fixation conditions, when standing upright without any restraints, there is a drift away from the fixation point of approximately 0.7 deg/s . This implies a 7-Hz modulation rate for a grating (or component of an image) of 10 cy/deg.
There is some evidence from psychophysics with infants that some of the sex effects may have a maturational component that has been speculatively linked to early testosterone surges in males .
We find that male acuities exceed those of females both for essentially static and temporally modulated stimuli. Similarly, it has been reported, based on large samples and measured with static and moving stimuli, that males had significantly better acuities under all conditions [3, 4]; we have found no studies that show the reverse. The standard Snellen acuity limit is taken as 1 arcmin, which is the width of the line making an optotype for 6/6 (20/20) resolution. Our acuity values (except those for females at a temporal modulation of 24 Hz) imply a bar width less than 1 arcmin; furthermore, the difference between geometric means for males and females is significant (see above).
There are two issues about our acuity findings: why are our acuity values better than the standard 1 arcmin, and why are males better than females? For the first issue we will consider the impact of psychophysical measurement techniques. For the second we will consider a range of possible optical factors. Anticipating the discussion, we can state that none of these factors change our findings that males do better than females across the spatio-temporal domain, including acuity.
Possible psychophysical reasons for acuities better than the 1 arcmin standard include: (a) our use of forced-choice psychophysics, and (b) our use of grating stimuli. Grating acuity is often assumed to be superior to optotype acuity . However, there is no clear evidence for the superiority of grating measures when testing normal adults: (i) gratings (specifically, Teller Acuity Cards) may underestimate acuity as measured with standard letter charts (Snellen optotypes) ; (ii) others have reported few or no differences between the two acuity measures when stimuli are viewed foveally; however grating acuity is increasingly superior with increasing eccentricity of viewing – apparently Snellen acuity deteriorates more rapidly across the visual field . We are left with the conclusion that our better than standard acuity values are probably due to the psychophysical algorithm we used: stimulus contrasts are changed from trial to trial until we reach a stringent confidence criterion.
Optical factors are known to modify the CSF, especially at higher spatial frequencies (e.g. [37, 55]). These effects must be ruled out before asserting that the sex differences we report for ST-CSFs are due to neural differences. We consider only those factors for which separate male and female values are available, or can be inferred plausibly (unfortunately, there is a dearth of direct information and many values must be estimated). The factors are: (i) retinal illuminance, (ii) angular subtense of cones, and (iii) densities of the photopigments in the cones. We restrict discussion to these factors because they are the only ones for which there are sex data, or plausible estimates of sex differences.
(i) Retinal illuminance
This clearly affects the CSF (e.g., [56, 57]), especially when it falls below approximately 100 Td (a measure of retinal illuminance). Retinal illuminance depends on both the eye’s light-gathering capacity, described by its f-stop (focal length/pupil diameter), and the luminance of the stimulus. Pupils of males may be slightly larger than those of females by a small but statistically non-significant amount [58, 59]. For our male participants we estimate the average pupil size to be 3.1 mm for our stimulus luminance of 55 cd/m2 (approximately 400 Td); we base this on published measures of mean pupil size for different stimulus luminances . For females we scale the male value by the ratio of the dark-adapted diameters for individuals aged 16–19 years – by which ages the dark-adapted diameters have reached their asymptote : the resulting estimate is 3.0 mm, and the ratio of male-to-female diameters is 1.03. However f-stop depends not only on pupil diameter; it also depends on focal length (f.l.).
In lieu of direct measures from both sexes, we estimated f.l. in two ways. Firstly, by assuming that the length of the vitreal chamber closely approximates f.l. – strictly, f.l. is the distance between the nodal point of the eye’s optics and the retina. The relevant nodal point for the eye’s optics lies approximately 0.28 mm in front of the rear surface of the lens ; adding this extra distance to vitreal length produces only a small percentage change. From a large sample, mean vitreal length for males: 16.36 mm; mean for females: 16.07 mm . The difference is small and not statistically significant (based on a t-test that we performed: t = 0.86). The ratio of male-to-female measures is 1.02. The calculated f-stops are: males, 5.3; females 5.4.
Secondly, if the male eye differs only by a scale factor from that of female, the f-stops must be identical. Mean axial lengths of male eyes are consistently longer than those of females: range across all the cited studies was from 23.73 to 25.54 mm; for males, mean length was 24.28 mm, and 23.74 mm for females [63–65]. The percentage change in the ratio of male to female lengths ranged from 1 to 4% across studies; mean of the ratios, male/female, (across studies) was 1.023. This value for the ratio is very similar to that for vitreal length (see above). Given the close similarity of these ratios of measures for pupil, axial length and vitreal length, it is reasonable to conclude that the eyes of the two sexes differ only by a scale factor and therefore f-stops must be the same, approximately 5.4 under our viewing conditions.
Clearly retinal illuminances are almost the same and cannot account for sex-related degradation of the ST-CSF. But even if retinal illuminances are the same, there may still be sex differences due to differences in the abilities of individual receptors to trap photons and resolve spatial differences in stimuli.
(ii) Angular subtense of cones
The amount of light funneled into a single cone depends on the area of the cone’s inner segment through which light must pass to reach the outer segment that contains the photopigment. Specifically, this depends on the angle subtended at the nodal point by the diameter of the base of the cone’s inner segment. In the very central retina the cones are as closely packed as possible (hexagonally close-packed) [66, 67]. Thus, cone-spacing is effectively the same as cone diameter. This measure defines the retinal mosaic that samples the visual image, which limits acuity . Direct studies that measured cone spacing and acuity in the same participants show that acuity is better in those who have finer receptor mosaics (i.e., smaller cone diameters) [69, 70]. Because, to a first approximation, male and female eyes are very similar in focal length, any differences in cone diameter should affect spatial resolution. The diameter of the inner segment is approximately 2 μm [71–73]; for this diameter, the subtense, using the above estimates of focal length, is approximately 0.5 arcmin. This cone diameter would permit detection of gratings of 60 cy/deg, which is better than the 20/20 standard of 1 arcmin: at 60 cy/deg a single cycle has a width of 1 arcmin; the width of a dark bar is, therefore, 0.5 arcmin. However cone diameter depends crucially on precisely which cones are measured: in the very center of the fovea cones can be as narrow as 1 μm, implying even better resolution. Clearly, a major limiting factor for acuity is diameter of the cone’s inner segment. However, we have no evidence that females’ cones are larger than those of males, which might have accounted for our reliable difference in acuity.
Apertures of the cones, however, are not the only relevant factor: receptor pooling must also be considered. It is generally accepted that the diameters of the centers of neuronal receptive fields determine local retinal spatial resolution. Based on the above cone diameters, the centers of the relevant neurons would have to be driven by single cones, which is the case for the midget ganglion cells that predominate in the central retina . To account for sex-related differences in acuity, females would have to have many fewer midget-system cells.
(iii) Cone photopigment density
Once a photon is funneled into a cone, the probability that it will stimulate the cone depends on the total number of photopigment molecules in that receptor; this number is determined by the length of the pigment-containing outer segment and the density of molecules – i.e., absorbance. At least for density, there are no sex-related differences: transverse absorbances (i.e., perpendicular to long axes of the cones) are the same for both L- and M-cones in humans . Thus, to account for acuity differences, the lengths of the outer segments would have to be quite different for males and females; the ratios (M/F) of acuities for each temporal rate (see Figure 4) range from 1.10 to 1.16 with a mean of 1.14. Unfortunately, we do not have any relevant anatomical data.
In conclusion, we find no obvious optical explanation for the sex differences we find (Figure 2). Sex differences in eye size, f-stop, and focal length are approximately 1%. Yet acuity of males is some 5 cy/deg higher than that of females (Figure 4), which is approximately a 10% difference. Furthermore, we cannot find any explanation of the acuity differences based on properties of the cones; the diameters of the cones’ inner segments (cone entrance pupil) and the lengths of the photopigment-containing outer segments do not seem to be different – the differences would have to be large to account for 10% differences in acuity.
Receptive fields and sex-differences
We emphasize that the sex-differences we find are important, especially for understanding the embryogensesis and maturation of the primary visual centers. To provide V1 neurons with elongated and oriented receptive fields, the axonal inputs from the lateral geniculate nucleus of the thalamus (LGN) must be quite specific. Receptive fields of LGN neurons are circular with opposed responses from center and surround (lateral inhibition). To assemble a neuron with an elongated and oriented receptive field, as in V1, that neuron must receive inputs from a specific, orderly array of LGN neurons, neurons that have similar receptive fields; in its simplest form, similar LGN neurons would have receptive fields located in a row across the relevant retinal area [76, 77]. While it is generally accepted that this model is too simple, something not unlike it must be the case . Thus, to account for our male–female differences, either the properties of the LGN neurons must differ, or the ways in which specific sets of LGN neurons are connected to any given V1 neuron must differ. Having ruled out many of the factors that could account for the sex differences in CSFs, we argue that the differences are most likely due to androgen effects at the cortex. This could be directly at the cortex or could affect projections from LGN to V1 – there are many V1 efferents back to the corresponding LGN regions, perhaps even more than the direct LGN-V1 afferents .
It seems reasonable to conclude that male–female differences in basic sensory capacities are adaptive. Otherwise why is this pattern found across sensory modalities? Specifically, in our case, why is there so heavy an involvement of the sex chromosomes in visual functions?
A plausible reason for sex-differences in spatio-temporal resolution stems from the period when hominids “descended from the trees” and ventured onto the savannas of Africa. Dwellers in forested regions have limited distances over which they must detect and identify objects, be they predators, enemies, or food. On open plains, however, the views are much longer. This places a premium on acuity, for early detection; we reiterate that we find the largest effect sizes at the high spatial frequencies and that our extrapolated acuity data show that males have a 10% advantage in acuity.
The sex-differences in vision might relate to different roles of males and females of early hunter-gatherers; males, being generally larger and more powerful, would have to detect possible predators or prey from afar and also identify and categorize these objects more easily. It is noteworthy that sensitivities to low spatial frequencies is enhanced by temporal modulation; in the real world, retinal images are rarely stationary – objects move and the observer moves.
Evidence for the “hunter-gatherer hypothesis” can be found in studies of visuo-spatial abilities of existing hunter-gatherers: a large meta-analysis of such studies showed that in general males performed better than females . Furthermore, there are significant sex differences in “near-vision” and “far-vision”: males are generally better for accurately perceiving and estimating sizes of targets in far-space . In monkeys there are different populations of cortical neurons dealing with eye-hand coordination and perception of objects in near- or far-space . And this perceptual dichotomy may be related to different neuronal populations associated with the “ventral and dorsal pathways” from primary visual cortex to the higher cortical areas that process visual information. The ventral pathway is from primary visual cortex to infero-temporal cortex and deals with vision-for-identification; the dorsal pathway to the parietal lobe deals more with spatial localization, or vision-for-action [82, 83]. Parenthetically, however, we should note that this dorsal/ventral dichotomy is not quite as clear as originally posited (e.g. [84, 85]).
Much of the research on sex-effects among hunter-gatherers implicitly assumes that the differences are congenital. However, it has been reported that there is a significant sex-by-age interaction in infants, indicating important maturational factors : females have higher sensitivity at the peak of the CSF at six months, but not at four and eight months – it seems that the sex-related differences we find (overall male superiority) may not be true from birth.
The hunter-gatherer hypothesis correctly predicts that adult males will perform better for targets in far-space – the hunter must perceive and correctly aim at more distant targets – while females will be better for near-space – arguing that they are the gatherers and foragers for nearby foods . Our findings seem to fit this model: males indeed are much more sensitive to high spatial frequencies. However, the sensitivity difference at low spatial frequencies is relatively small and indeed females may do better for static or slowly moving targets (see Figure 2), which would accord with attending to nearby, stationary objects.