See Through Their Eyes

What the curve shows

Acuity here means grating acuity: the finest stripe pattern a child reliably looks at, counted in cycles per degree (cpd), meaning the pairs of light and dark stripes that fit in one degree of what the eye sees (Neijzen et al., 2025). The chart plots, for each age, the two published limits of the typical range. The middle of that range, derived from the published limits, climbs from about 0.5 cpd at half a month to 5.3 cpd at 5.6 months and 15.8 cpd at 3 years. Most of the rise comes early: the authors describe a rapid growth phase in about the first 8 months, followed by a less steep curve (Neijzen et al., 2025).

What the typical range means

Each age has two numbers, a slower end and a faster end, called 95% prediction limits. They mark the range that should hold the scores of about 95 out of every 100 healthy children tested the same way (Neijzen et al., 2025). A prediction limit describes how much individual children differ, so it stays wide however many children are tested. A confidence interval is something else: it shows how precisely a study has pinned down an average, and it gets narrower as more children are tested. A child near either end of the band is still within the typical range, and the band cannot tell you whether any one child's vision is healthy.

The band is wide. At about 3 months the limits are 1.5 and 6.7 cpd, so the faster end sees stripes more than four times finer than the slower end, and at every age the upper limit is between about 3.9 and 4.9 times the lower one (Neijzen et al., 2025, Table A1).

How the picture uses the table

The slower end, middle and faster end settings in the viewer use the lower limit, the middle and the upper limit for the age you choose. The middle is not published as a number. It is derived here as the geometric middle of the two limits, the square root of their product. That is where the authors' own fitted average falls, because they built each limit by adding and subtracting the same amount on a logarithmic scale (Neijzen et al., 2025). Between two published ages the page interpolates on the same scale and labels the value as interpolated. Outside 0.5 to 36 months it shows nothing, because the table says nothing there (Neijzen et al., 2025).

Other studies sit beside the curve, not on it

Other studies measured acuity with other tests, and a different test can give a different number for the same baby, so their results are never joined to this curve. The newborn value comes from printed cards held 38 cm away, with its own rule for deciding that a baby has seen the stripes, and is about 0.8 to 1.2 cpd depending on the testing procedure (Brown et al., 2018). Contrast sensitivity curves for 1-, 2- and 3-month-olds gave high-frequency cutoffs of 2.4, 2.8 and 4.0 cpd, which the authors call estimates of acuity (Banks & Salapatek, 1978). A study of 600 children aged 3 to 36 months held striped paddles at distances of up to 229 cm and reported values rising from about 2 cpd in its 3-month group to about 24 cpd in its 24-month group (Elgohary et al., 2017). Its oldest groups sit at the top of what that test can measure, so only the first two years are used here. Typical values can also differ between populations and testing teams. A 2021 study of infants in southern China, using the same Teller cards, found binocular means of 3.7, 7.4 and 12.0 cpd at 12, 24 and 36 months, significantly lower than northern Chinese norms (Xiang et al., 2021). The meta-analysis left that study out because its results diverged significantly from the other studies (Neijzen et al., 2025).

20/20 and 30 cycles per degree

Eye charts call it 20/20 vision when a person can recognize letters whose parts are about 1 minute of arc wide, one sixtieth of a degree. A striped pattern whose light and dark stripes are each that wide has 30 stripe pairs per degree, so 30 cycles per degree is the usual stand-in for 20/20 (Ashraf et al., 2025). The two are not the same measurement. Grating acuity is the finest stripe pattern a person can detect, and letter acuity is the smallest letter a person can recognize (Neijzen et al., 2025). Treat 30 cycles per degree as a rough landmark, not an exact conversion. Healthy young adults often do better (Ashraf et al., 2025), and by 3 years the faster end of the typical range is already about 31 cycles per degree (Neijzen et al., 2025).

Babies cannot read an eye chart or tell us what they see, so researchers infer what a baby sees from what a baby does (Adolph & Schmuckler, 2024). The oldest tool is preferential looking. In the late 1950s Fantz showed that infants prefer to look at patterned displays rather than blank ones (Fantz, 1958; Kiorpes, 2016), and infants under 5 days old looked more at black-and-white patterns than at plain colored surfaces (Fantz, 1963). To turn this into a measurement, the stripes get finer from card to card. In the acuity card procedure, an adult observer who does not know which side the stripes are on watches the baby through a small peephole and judges whether the baby saw them (McDonald et al., 1985; Joo et al., 2020). The finest stripes the baby reliably looks toward give the score (Chang & Borchert, 2024).

Other methods do not need a baby to look at all. In a sweep VEP (visual evoked potential), the stripes become finer over about 10 seconds while electrodes record the brain's electrical response at the scalp, and acuity is estimated from where the response fades to zero (Norcia & Tyler, 1985; Norcia et al., 2015). In habituation tests, a baby is shown one picture until she loses interest, then it is paired with a new one, and looking longer at the new one suggests she can tell them apart (Fantz, 1964).

Each method has limits. A card score measures detecting stripes, which is not the same as recognizing letters (Neijzen et al., 2025). Some researchers also argue that looking time is a lean measure of what babies do and know, and that richer behavior should be recorded too (Adolph & Schmuckler, 2024).

One acuity-card trial, in four steps

Details differ between clinics and studies. One card set had 17 cards, with stripes from 0.23 to 38 cycles per centimeter in half-octave steps (Joo et al., 2020). Studies report test distances of 55 cm (Joo et al., 2020; Chang & Borchert, 2024) and 38 cm (Tartarella et al., 2025; Brown et al., 2015). The game above copies none of these. Its cards run from 0.25 to about 39 cycles per degree, each about 1.4 times finer than the last, which is close to a half-octave step but is a design choice for the game and not a claim about any real card set. Cycles per degree (cpd) counts pairs of light and dark stripes within one degree of the visual field, so a larger number means finer stripes.

What contrast is

Contrast is how much lighter the light parts of a pattern are than the dark parts. For stripes it is usually measured as Michelson contrast: the brightest luminance minus the darkest, divided by their sum, over one repeat of the pattern (Brown et al., 2015). Luminance is the amount of light coming from a surface. A contrast of 0.5, or 50%, means the light stripes are three times as bright as the dark ones. A contrast of 1% means they are only about 2% brighter.

The lowest contrast a person can detect is the threshold, and the reciprocal of the threshold is the sensitivity (Pelli & Bex, 2013). A threshold of 50% is a sensitivity of 2, and a threshold of 1% is a sensitivity of 100. A higher sensitivity means a person can see fainter things.

Newborns

In a study of 119 healthy newborns tested within their first days, a very coarse striped pattern had to reach a contrast of about 0.5 before the babies reliably looked at it, which means the light stripes were about three times as bright as the dark ones, a contrast sensitivity of about 2 (Brown et al., 2015). A second newborn study found median thresholds of 33% and 46% for similar coarse stripes, depending on the testing procedure (Brown et al., 2018). Stripes this coarse are well below the finest stripes a newborn can resolve, so the result is about faintness, not fineness (Brown et al., 2015).

Adults and 3-month-olds

Adults need far less contrast. Across many kinds of targets, healthy adults reach a threshold of about 1% contrast (Pelli & Bex, 2013), and with a brain-response measure under ideal laboratory conditions, adults' thresholds for coarse patterns were as low as 0.22 to 0.32% (Norcia et al., 1990). At 3 months, the best contrast sensitivity infants reach is about 15 (range 10 to 25), roughly one fiftieth of the adult value, which corresponds to a threshold of about 7%, between 4 and 10% (Brown & Lindsey, 2009).

These numbers come from different targets and methods, so they do not line up exactly. That is why the lab above marks only "about 1% or better" for adults and draws no adult curve.

Newborns are not color-blind. In one study, 74% of newborns told a large red patch from grays of every brightness, but only 14% did so with blue, 36% with green and 25% with yellow (Adams et al., 1994). Their color vision is poor: colors need to be strong, large and of the right hue to be noticed (Skelton et al., 2022).

Color vision depends on three types of cone cell and on channels that compare their signals. The red-green channel develops first, the blue-yellow channel about 4 to 8 weeks later, and by 3 months both are active (Skelton et al., 2022). Even then, faint colors are hard for babies to detect, and sensitivity to faint colors does not reach adult levels until late adolescence (Skelton et al., 2022). By 4 months, infants appear to respond to color in categories. In a study of 179 infants about 5 months old, recognition memory grouped fully saturated colors into red, yellow, green, blue and purple (Skelton et al., 2017). Infants from 3 months did not treat a color as new when only the lighting changed, a sign of at least rudimentary color constancy, which is still maturing at 2 to 4 years (Skelton et al., 2022).

Most of this research comes from a narrow range of populations (Skelton et al., 2022). Young babies are also less sensitive to contrast in general, which may help explain why color looks weak to them (Brown et al., 1995).

How the picture handles color

The picture at the top of this page follows this research in two ways. At each age it blurs color detail more than brightness detail, and it removes only the colors that a baby of that age could not tell from gray, hue by hue. A newborn's picture therefore keeps faint color in large areas of strongly saturated red or green, such as the caregiver's top or the rug when they fill enough of the view, and loses blue. Small colored things, such as the ball, lose their color first.

Four findings sit behind these rules. First, newborns told colored lights from white only when the lights were highly saturated. Saturation here is measured as excitation purity, which is how far a color lies from white, from 0% for white to 100% for a pure single-wavelength color. In 180 newborns shown 16 degree lights, discrimination from white appeared only above purities of 41% for green, 47% for red and 65% for yellow, against under 1% in adults (Adams & Courage, 1998). Second, brain responses to purely colored patterns, with no change in brightness, were missing in most infants before about 5 weeks in a study of 8 infants (Morrone et al., 1990) and before 7 to 8 weeks in a study of 13 infants (Morrone et al., 1993). When responses first appeared, they needed patterns 20 times coarser than the finest brightness patterns that gave responses (Morrone et al., 1990). Third, color sensitivity has further to grow than brightness sensitivity. From 8 weeks to adulthood, the brain's response to red-green gratings rose in sensitivity by 0.91 log units, about 8 times, against 0.64 log units, about 4 times, for brightness gratings (Kelly et al., 1997). Even so, results in babies 2 to 8 weeks old are consistent with working red-green mechanisms (Allen et al., 1993). Fourth, in 172 people aged 3 months to 86 years, the smallest color change that could be detected shrank by a factor of about 1.8 with each doubling of age until the late teens, with fitted turning points between about 12 and 22 years depending on the color direction and the curve fitted; from 3 months to adolescence it fell about 30-fold (Knoblauch et al., 2001).

The numbers in the rule are rendering choices built on these findings. They are not a measurement of how color looks to a baby. The newborn picture is not black and white: large areas of strong red and strong green keep some color, while blue is lost. Red is not special: in the newborn study, green needed about as little saturation as red. Only large areas keep any: with the caregiver about 1 m away her red top keeps a faint red, and with her 3 m away the rug, now filling more of the view, keeps a faint green. The rendering settings include a switch that leaves color unchanged, so you can see the effect of the blur alone.

Age by age: what research has shown

Each line says what the study tested, because a result is only as wide as its test.

Seeing depth takes more than acuity, and acting safely on what you see takes more than seeing. This topic follows three threads: depth from two eyes, motion and approaching objects, and a baby's first experience of edges and falls. Ages are weeks or months after birth.

Depth from two eyes and from one

Stereopsis is depth perception from the slightly different views of the two eyes. The difference between the two views is called disparity. When stereopsis starts depends on how it is tested. Over half of infants 5 to 10 weeks old made eye movements that followed a changing disparity (Seemiller et al., 2018). In preferential-looking tests, a response to crossed disparity became significant at 8 weeks in one study and a response to uncrossed disparity at 13 weeks in another, and tests that simply recorded looking time found responses later, at 12 to 15 weeks (Kavšek, 2013; Kavšek & Braun, 2016). In the classic study, stereopsis was first shown at a mean age of 16 weeks, and infants reached a fine stereoacuity of 1 minute of arc or better by a mean of 21 weeks, a far faster course than the improvement of acuity (Held et al., 1980).

Depth cues that need only one eye have been studied too. Motion parallax, depth from relative motion, as when you move your head, appeared by about 16 weeks in 11 infants followed from 8 to 29 weeks (Nawrot et al., 2009). Pictorial cues, such as relative size and overlap, are reported at 3 to 5 months in some looking studies, and at about 6 months in studies that controlled for simple two-dimensional differences (Kavšek et al., 2012). The picture at the top of this page has one view, so it cannot show stereopsis.

Motion and approaching objects

Cortical responses that signal the direction of a moving pattern appeared in fewer than a quarter of infants under 7 weeks old and in 80% or more at 11 to 13 weeks (Braddick et al., 2005). Looming is the growing image of something moving toward you. In a classic study of 24 infants aged 2 to 11 weeks, a shadow that grew symmetrically, as an approaching object's would, led to an avoidance response and upset, whereas shrinking or lopsided shadows did not (Ball & Tronick, 1971). Newer work with newborns finds sensitivity without such a reaction. Newborns only a few hours old are sensitive to matches between a growing image and a rising sound (Orioli et al., 2018a), and they look preferentially at trajectories directed toward their own bodies (Orioli et al., 2018b). The authors of the second study did not look for defensive reactions, and note that newborns might not categorize approaching stimuli as possible dangers (Orioli et al., 2018b). So the sources agree that babies are sensitive to looming from the first weeks, but not on what babies do about it.

Edges, crawling and falls

The visual cliff is a glass-topped table. On one side a checkerboard lies flush under the glass, and on the other it lies about a meter below, which looks like a drop-off (Kretch & Adolph, 2013). Eleanor Gibson devised it to test infants' responses to an apparent drop-off, and at first took avoidance of the deep side as a sign of depth perception that was innate, a view she later changed (Adolph & Schmuckler, 2024).

Seeing a drop is not the same as being wary of it. Babies with little crawling experience show no wariness of heights, and wariness is related to locomotor experience (Dahl et al., 2013). Earlier falls do not account for it: novice crawlers and walkers go over an edge whatever falls they have had before (Adolph & Hoch, 2019). What infants learn is specific to the posture they have practiced. Experienced 12-month-old crawlers refused to crawl over risky drop-offs, while novice walkers of the same age stepped repeatedly over the edge (Kretch & Adolph, 2013). Experienced 18-month-old walkers refused to walk over them but used other ways to get down (Kretch & Adolph, 2013). Nine-month-olds who avoided reaching over risky gaps while sitting fell into them when they reached while crawling (Adolph, 2000).

Falls are common while babies learn to walk. Twelve- to 19-month-olds in free play averaged 2,368 steps and 17 falls per hour (Adolph et al., 2012). In a case-control study in the United Kingdom of 610 children under 5 with medically attended stair-fall injuries, parents of the injured children were more likely than parents of matched children not to have stair gates, or to leave them open (Kendrick et al., 2016). That is an association between homes and injuries. The studies described in this topic did not test whether better vision lowers the risk of a real fall, so this page does not claim that it does.

Vision changes quickly in infancy, but it does not stop changing there (Leat et al., 2009). Several abilities are still developing in school-age children (Leat et al., 2009). Researchers measure them with tasks, such as saying which way a gap faces or where a pattern is. A child's score depends on how well the eyes and brain work and also on how well the child keeps attention on the task (Dekker et al., 2020). For that reason, studies often disagree about the age at which children match adults. Where they disagree, the list below gives a range. Each line names an ability, explains it in plain words, gives the age the source reports and says how it was measured. The list runs from the earliest age to the latest. Ages are years after birth.

Contrast sensitivity, how faint a pattern can be and still be seen, is not in this list. The topic Faint things are harder to see covers it, including its development in childhood.

Sensitive periods, amblyopia and screening

A sensitive period is a time when visual experience has an unusually strong effect on development. Babies born with dense cataracts, which block patterned light from reaching the retina, show this. After surgery between 1 week and 9 months of age, their acuity (sharpness of vision) was at first no better than a newborn's and then improved quickly (Maurer et al., 1999). Different abilities have different sensitive periods (Lewis & Maurer, 2005).

Amblyopia, often called lazy eye, is a reduction in acuity in one eye, or less often both, that comes from abnormal processing of images in the brain (Cruz et al., 2023). Pooled estimates put it near 1.75% of people (Hashemi et al., 2018). Treatment is most successful in young children, though improvement can still be expected in older children and teenagers (Cruz et al., 2023).

The US Preventive Services Task Force recommends at least one vision screening for children aged 3 to 5 years (Grossman et al., 2017). The American Academy of Pediatrics advises first trying screening with a measuring device between 12 months and 3 years (Donahue et al., 2016).

This page is for teaching, not diagnosis. A parent with a concern should ask an eye care professional.

How the eye grows

The eye grows quickly early in life, from about 16.2 mm long around the time of a full-term birth to about 21.8 mm at 3 years (Groot et al., 2022), and to an average of 23.1 mm by 9 years (Tideman et al., 2018). On average, infants were farsighted at 3 months and less so at 9 months, as growth of the eye reduced the farsightedness, a process called emmetropization (Mutti et al., 2005).

Myopia and time outdoors

Myopia, or nearsightedness (trouble seeing distant things clearly), often starts in the school years. In a trial in Guangzhou, China, schools were assigned by chance either to add a 40-minute outdoor class to each school day or to carry on as usual. After three years, 30.4% of the first-graders without myopia at the start had developed it in the schools that added outdoor time, against 39.5% in the other schools (He et al., 2015). A larger trial in Shanghai also found less myopia onset where schools added outdoor time (He et al., 2022). A meta-analysis (a study that combines the results of several studies) of seven such trials agreed (Mei et al., 2024), and an expert review concluded that randomized trials have confirmed the protective effect (Morgan et al., 2021). A 2016 analysis projected that about half of the world's people would have myopia by 2050 (Holden et al., 2016).

Crossing a road asks a child to judge how fast a vehicle is coming and how much time is left. In one laboratory study, children aged 6 to 11 and adults watched a simulated car approach on a screen and said whether its image was growing. Noticing that growth is called looming detection. All three groups of children needed more growth than adults to notice it (Wann et al., 2011).

When the car image also shifted sideways, children could reliably detect approach only below about 25 miles per hour (mph), while adults managed close to 40 mph (Wann et al., 2011). This was a short screen display viewed with one eye, so it does not measure real crossings.

Timing also matters. On virtual roads, children chose gaps as large as adults' but entered them later, so they had less time to spare (Plumert et al., 2004). Entry timing reached adult-like levels by 14 years (O'Neal et al., 2018).

A popular claim says that children's field of view is one third narrower than an adult's. Tests that map the visual field do show smaller fields in younger children (Patel et al., 2015), but we found no measurement that supports that figure. Early laboratory tests found little evidence that children's peripheral detection (noticing things away from where you look) is poorer relative to their central vision (David et al., 1986).

Perception is one contributor among many. Timing, attention and a child's temperament also matter (O'Neal et al., 2018; Remington et al., 2014; Shen et al., 2015). The lab studies how children cross streets in virtual reality (see Research Areas). In one of its studies, 485 children aged 7 to 9 made 21 virtual street crossings, and the link between temperamental fear (a child's general tendency to be fearful) and collisions was weaker for children with higher visual discrimination scores, a test of visual perception skill rather than of acuity (Davis et al., 2026).